Documente Academic
Documente Profesional
Documente Cultură
edited by
Young C Kim
California State University, Los Angeles, USA
World Scientific
NEW JERSEY • LONDON • SINGAPORE • BEIJING • SHANGHAI • HONG KONG • TA I P E I • CHENNAI
For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance Center,
Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy is not required from
the publisher.
ISBN-13 978-981-281-929-1
ISBN-10 981-281-929-0
Printed in Singapore.
Preface
Although coastal and ocean engineering is a very ancient field with the construction
of Port A-ur near the mouth of the Nile in 3,000 BC, significant advances in this
field have been made in the last several decades. The rise of interest in this field can
be seen from the number of attendees by academics and practitioners in interna-
tional conferences. The first International Conference on Coastal Engineering was
held in Long Beach, California in 1950 with less than 100 people. When the same
conference was held in San Diego, California, in 2006, over 1000 delegates attended.
In the last several decades, the world has seen significant coastal and ocean engi-
neering projects, one of which is the Delta Project in the Netherlands. This project
was designed to shorten and strengthen the total length of coast and dykes washed
by the sea by closing off the sea arms in the Delta region. Other noteworthy coastal
engineering projects include the Kansai Airport Project in Japan and, in recent
years, the construction of mobile barriers at inlets to regulate tides in the Venice
Lagoon known as the Venice Project. Interest in coastal and ocean engineering
has arisen in recent years as humankind experiences coastal disasters that derive
from coastal storm, hurricane and coastal flooding and seismic activities such as
tsunamis, and the impacts of climate change which result in sea-level rise. The
tsunami activity in Sumatra in December 2004 affected countries throughout the
Indian Ocean and resulted in the loss of thousands of lives. Hurricane Katrina in
New Orleans also claimed many lives with property damage exceeding $63 billion.
Global warming and sea-level rise will affect shoreline retreats, inundate low coastal
areas, damage coastal structures, and accelerate beach erosion. The need for better
understanding of our coastal and ocean environment has risen considerably in
recent years.
This handbook contains a comprehensive compilation of topics that are
the forefronts of many technical advances in ocean waves, coastal and ocean
engineering. It represents the most comprehensive reference available on coastal
and ocean engineering to date, and it also provides the most up-to-date technical
advances and latest research findings on coastal and ocean engineering. More than
70 internationally recognized authorities in the field of coastal and ocean engineering
contributed papers on their areas of expertise to this handbook. These interna-
tional luminaries are from highly respected universities and renowned research and
consulting organizations from all over the world.
v
August 25, 2009 18:3 9.75in x 6.5in b684-fm FA
vi Preface
Preface vii
Young C. Kim
Los Angeles, California
January 2008
August 25, 2009 18:3 9.75in x 6.5in b684-fm FA
Contents
Preface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii
1. Wave Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
R. G. Dean and T. L. Walton
2. Wavemaker Theories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
R. T. Hudspeth and R. B. Guenther
3. Analyses by the Melnikov Method of Damped Parametrically
Excited Cross Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
R. B. Guenther and R. T. Hudspeth
4. Random Wave Breaking and Nonlinearity Evolution Across
the Surf Zone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Y. Goda
ix
August 25, 2009 18:3 9.75in x 6.5in b684-fm FA
x Contents
Contents xi
xii Contents
Contributors
Jonathan C. Allan
Coastal Field Office
Oregon Department of Geology and Mineral Industries
Newport, Oregon
jonathan.allan@dogami.state.or.us
William Allsop
Technical Director
HR Wallingford
Wallingford, UK
w.allsop@hrwallingford.co.uk
Elena Sanchez Badorrey
Associate Professor
CEAMA — Universidad de Granada
Granada, Spain
elenasb@ugr.es
Asuncion Baquerizo
Associate Professor
CEAMA — Universidad de Granada
Granada, Spain
Jurjen A. Battjes
Emeritus Professor
Environmental Fluid Mechanics Section
Delft University of Technology
Delft, The Netherlands
j.a.battjes@tudelft.nl
Michael J. Briggs
Research Hydraulic Engineer
Coastal and Hydraulics Laboratory
U.S. Army Engineer Research and Development Center
Vicksburg, Mississippi
michael.j.briggs@erdc.usace.army.mil
xiii
August 25, 2009 18:3 9.75in x 6.5in b684-fm FA
xiv Contributors
Tom Bruce
School of Engineering and Electronics
University of Edinburgh
Edinburgh, UK
tom.bruce@ed.ac.uk
Subrata Chakrabarti
Joint Professor, Civil and Mechanical Engineering
University of Illinois at Chicago
Chicago, Illinois
chakrab@aol.com
Peter J. Cowell
Associate Professor
School of Geosciences
Institute of Marine Science
University of Sydney
Sydney, Australia
Daniel T. Cox
Professor
School of Civil and Construction Engineering
Oregon State University
Corvallis, Oregon
dtc@oregonstate.edu
Robert G. Dean
Graduate Research Professor of Coastal Engineering, Emeritus
Department of Civil and Coastal Engineering
University of Florida
Gainesville, Florida
dean@coastal.ufl.edu
Pierre Debaillon
Research Hydraulic Engineer
Centre d’Etudes Techniques Maritimes Et Fluviales (CETMEF)
Compiegne, France
pierre.debaillon@equipement.gouv.fr
Martijn P. C. de Jong
Formerly at Environmental Fluid Mechanics Section
Delft University of Technology
Presently at Delft Hydraulics
Delft, The Netherlands
August 25, 2009 18:3 9.75in x 6.5in b684-fm FA
Contributors xv
Lesley Ewing
Senior Coastal Engineer
California Coastal Commission
San Francisco, California
lewing@coastal.ca.gov
Leopoldo Franco
Professor of Coastal Engineering
Department of Civil Engineering
University of Rome 3
Rome, Italy
leopoldo.franco@uniroma3.it
Panagiota Galiatsatou
Research Associate
Department of Civil Engineering
Aristotle University of Thessaloniki
Thessaloniki, Greece
pgaliats@civil.auth.gr
Yoshimi Goda
Professor Emeritus
Yokohama National University
Adviser to ECHO Corporation
Tokyo, Japan
goda@ecoh.co.jp
Ronald B. Guenther
Professor Emeritus
Department of Mathematics
Oregon State University
Corvallis, Oregon
guenther@math.orst.edu
John Rong-Chung Hsu
Professor
Department of Marine Environment and Engineering
National Sun Yat-sen University
Kaohsiung, Taiwan
Honorary Research Fellow
School of Civil and Resource Engineering
University of Western Australia
Nedland, Australia
jrchsu@mail.nsysu.edu.tw
August 25, 2009 18:3 9.75in x 6.5in b684-fm FA
xvi Contributors
Robert T. Hudspeth
Professor and Director, Emeritus
Coastal and Ocean Engineering Program
Oregon State University
Corvallis, Oregon
robert.hudspeth@orst.edu
Hwung-Hweng Hwung
Professor of Hydraulic and Ocean Engineering
Director of Tainan Hydraulics Laboratory
Department of Hydraulic and Ocean Engineering
National Cheng Kung University
Tainan, Taiwan
hhhwung@mail.ncku.edu.tw
Masahiko Isobe
Professor and Special Adviser to the President
Department of Sociocultural Environmental Studies
Graduate School of Frontier Sciences
The University of Tokyo
Chiba, Japan
isobe@k.u-tokyo.ac.jp
Mamta Jain
Coastal Engineer
Halcrow Inc.
Tampa, Florida
mjain@halcrow.com
Bradley D. Johnson
Coastal and Hydraulics Laboratory
U.S. Army Engineer Research and Development Center
Vicksburg, Mississippi
Shohachi Kakuno
Professor and Vice President
Department of Civil Engineering
Osaka City University
Osaka, Japan
kakuno@ado.osaka-cu.ac.jp
J. William Kamphuis
Professor of Civil Engineering, Emeritus
Department of Civil Engineering
Queen’s University
Kingston, Ontario, Canada
kamphuis@civil.queensu.ca
August 25, 2009 18:3 9.75in x 6.5in b684-fm FA
Contributors xvii
Akira Kimura
Professor
Department of Social Systems Engineering
Tottori University
Tottori, Japan
kimura@sse.tottori-u.ac.jp
Nobuhisa Kobayashi
Professor and Director
Center for Applied Coastal Research
University of Delaware
Newark, Delaware
nk@coastal.udel.edu
Yukio Koibuchi
Assistant Professor
Department of Sociocultural Environmental Studies
Graduate School of Frontier Sciences
The University of Tokyo
Chiba, Japan
koi@k.u-tokyo.ac.jp
Paul D. Komar
Professor of Oceanography
College of Oceanic and Atmospheric Sciences
Oregon State University
Corvallis, Oregon
pkoma@coas.oregonstate.edu
Andreas Kortenhaus
Leichtweiss-Institute for Hydraulics
Technical University of Braunschweig
Braunschweig, Germany
kortenhaus@tu-bs.de
Nicholas C. Kraus
Senior Scientist
Coastal and Hydraulics Laboratory
U.S. Army Engineer Research and Development Center
Vicksburg, Mississippi
nicholas.c.kraus@erdc.usace.army.mil
Alberto Lamberti
Professor
Department of Civil Engineering
University of Bologna
Bologna, Italy
alberto.lamberti@unibo.it
August 25, 2009 18:3 9.75in x 6.5in b684-fm FA
xviii Contributors
Fang-Chun Lee
Department of Marine Environment and Engineering
National Sun Yat-sen University
Kaohsiung, Taiwan
Jiin-Jen Lee
Professor of Civil and Environmental Engineering
Sonny Astani Department of Civil and
Environmental Engineering
University of Southern California
Los Angeles, California
jjlee@usc.edu
Yu-Cheng Li
Professor
School of Civil Engineering
Dalian University of Technology
Dalian, China
liyuch@dlut.edu.cn
Miguel A. Losada
Professor
Research Group on Environmental Flux Dynamics
CEAMA — Universidad de Granada
Granada, Spain
mlosada@ugr.es
Etienne P. D. Mansard
Executive Director
Canadian Hydraulics Centre
National Research Council Canada
Ottawa, Ontario, Canada
etienne.mansard@nrc-cnrc.gc.ca
Ashish J. Mehta
Professor of Coastal Engineering
Department of Civil and Coastal Engineering
University of Florida
Gainesville, Florida
mehta@coastal.ufl.edu
Michael D. Miles
Canadian Hydraulics Centre
National Research Council Canada
Ottawa, Ontario, Canada
August 25, 2009 18:3 9.75in x 6.5in b684-fm FA
Contributors xix
Nobuhito Mori
Associate Professor
Disaster Prevention Research Institute
Kyoto University
Kyoto, Japan
nobuhito.mori@hy5.ecs.kyoto-u.ac.jp
Tad S. Murty
Adjunct Professor
Department of Civil Engineering
University of Ottawa
Ottawa, Ontario, Canada
tadmurty@gmail.com
Ioan Nistor
Assistant Professor
Department of Civil Engineering
University of Ottawa
Ottawa, Ontario, Canada
inistor@uottawa.ca
Younes Nouri
Department of Civil Engineering
University of Ottawa
Ottawa, Ontario, Canada
Miquel Ortega
Associate Professor
CEAMA — Universidad de Granada
Granada, Spain
miguelos@ugr.es
Hocine Oumeraci
University Professor
Leichtweiss-Institute for Hydraulic Engineering
and Water Resources
Technical University of Braunschweig
Braunschweig, Germany
h.oumeraci@tu-bs.de
Dan Palermo
Assistant Professor
Department of Civil Engineering
University of Ottawa
Ottawa, Ontario, Canada
palermo@eng.uottawa.ca
August 25, 2009 18:3 9.75in x 6.5in b684-fm FA
xx Contributors
Andres Payo
Graduate School of Science and Technology
University of Kumamoto
Kumamoto, Japan
Krystian W. Pilarczyk
(Former) Manager, Research and Development
Hydraulic Engineering Institute
Rykswaterstaat
Delft, The Netherlands
HYDROpil Consultancy
Zoetermeer, The Netherlands
krystian.pilarczyk@gmail.com
Panayotis Prinos
Professor of Hydraulic Engineering
Department of Civil Engineering
Aristotle University of Thessaloniki
Thessaloniki, Greece
prinosp@civil.auth.gr
Tim Pullen
Senior Engineer
HR Wallingford
Wallingford, UK
tap@hrwallingford.co.uk
Alexander B. Rabinovich
P.P. Shirshov Institute of Oceanology
Russian Academy of Sciences
Moscow, Russia
Department of Fisheries and Oceans
Institute of Ocean Sciences
Sidney, B.C., Canada
abr@iki.rssi.ru.
Roshanka Ranasinghe
Associate Professor
UNESCO-IHE/Delft University of Technology
Delft, The Netherlands
r.ranasinghe@unesco-ihe.org
Juan Recio
Leichweiss-Institute for Hydraulic Engineering
and Water Resources
Technical University of Braunschweig
Braunschweig, Germany
August 25, 2009 18:3 9.75in x 6.5in b684-fm FA
Contributors xxi
Julie D. Rosati
Research Hydraulic Engineer
Coastal and Hydraulics Laboratory
U.S. Army Corps of Engineers
Mobile, Alabama
julie.d.rosati@erdc.usace.army.mil
Peter Ruggiero
Assistant Professor
Department of Geosciences
Oregon State University
Corvallis, Oregon
ruggierp@science.oregonstate.edu
Murat Saatcioglu
Professor
Department of Civil Engineering
University of Ottawa
Ottawa, Ontario, Canada
murat.saatcioglu@uottawa.ca
Juan M. Santiago
Associate Professor
CEAMA — Universidad de Granada
Granada, Spain
santi@ugr.es
Holger Schuttrumpf
Professor and Director
Institute of Hydraulic Engineering
and Water Resources Management
RWTH — Aachen University
Aaachen, Germany
schuettiumpf@iww.rwth-aachen.de
Richard Silvester
Professor Emeritus
School of Civil and Resource Engineering
University of Western Australia
Nedland, Australia
rsilvest@bigpond.net.au
August 25, 2009 18:3 9.75in x 6.5in b684-fm FA
xxii Contributors
Marcel J. F. Stive
Professor and Director
Delft Water Research Centre
Department of Hydraulic Engineering
Delft University of Technology
Delft, The Netherlands
m.j.f.stive@tudelft.nl
Kyung-Duck Suh
Professor
Department of Civil and Environmental Engineering
Seoul National University
Seoul, Korea
kdsuh@snu.ac.kr
Shigeo Takahashi
Executive Researcher and Director
Tsunami Research Center
Port and Airport Research Institute
Yokosuka, Japan
takahashi s@pari.go.jp
Klemens Uliczka
Research Hydraulic Engineer
Federal Waterways Engineering and Research Institute (BAW)
Hamburg, Germany
klemens.uliczka@baw.de
Jentsje van der Meer
Principal
Van der Meer Consulting
Heerenveen, The Netherlands
jm@vandermeerconsulting.nl
Marc Vantorre
Professor
Division of Maritime Technology
Ghent University
Ghent, Belgium
marc.vantorre@ugent.be
August 25, 2009 18:3 9.75in x 6.5in b684-fm FA
Contributors xxiii
Todd L. Walton
Director
Beaches and Shore Resource Center
Florida State University
Tallahassee, Florida
twalton@mailer.fsu.edu
Xiuying Xing
Graduate Research Assistant
Sonny Astani Department of Civil and Environmental Engineering
University of Southern California
Los Angeles, California
Harry Yeh
Professor
School of Civil and Construction Engineering
Oregon State University
Corvallis, Oregon
harry@engr.orst.edu
Dong Hoon Yoo
Professor
Department of Civil Engineering
Ajou University
Suwon, Korea
dhyoo@ajou.ac.kr
Sung Bum Yoon
Professor
Department of Civil and Environmental Engineering
Hanyang University
Ansan, Korea
sbyoon@hanyang.ac.kr
Melissa Meng-Jiuan Yu
Department of Marine Environment and Engineering
National Sun Yat-sen University
Kaohsiung, Taiwan
Barbara Zanuttigh
Assistant Professor
Department of Civil Engineering
University of Bologna
Bologna, Italy
barbara.zanuttigh@mail.ing.unibo.it
August 25, 2009 18:3 9.75in x 6.5in b684-fm FA
The Editor
xxv
This page intentionally left blank
July 31, 2009 8:18 9.75in x 6.5in b684-ch01 FA
Chapter 1
Wave Setup
Robert G. Dean
Department of Civil and Coastal Engineering
University of Florida, Gainesville, FL, USA
dean@coastal.ufl.edu
Todd L. Walton
Beaches and Shores Resource Center
Florida State University, Tallahassee, FL, USA
twalton@fsu.edu
Wave setup is the increase of water level within the surf zone due to the transfer of
wave-related momentum to the water column during wave-breaking. Wave setup
has been investigated theoretically and under laboratory and field conditions,
and it includes both static and dynamic components. Engineering applications
include a significant flooding component due to severe storms and oscillating water
levels that can increase hazards to recreational beach goers and can contribute to
undesirable oscillations of both constructed and natural systems including harbors
and moored ships. This chapter provides a review of the knowledge regarding wave
setup and presents preliminary recommendations for design. It will be shown that
wave setup is not adequately understood quantitatively for engineering design
purposes.
1.1. Introduction
Wave setup was brought to the attention of coastal engineers and scientists in the
1960s (i.e., see Ref. 1, p. 245) after the initial theoretic developments of Longuet-
Higgins2 and Longuet-Higgins and Stewart3,4 along with limited field observations
and laboratory studies supported the existence of wave setup, the magnitude of
which was observed to be in the order of 10–20% of the incident wave height. It
was noted in early field observations that water levels on the beach were higher
than those recorded by a tide gauge at the end of a pier suggesting a wave setup
physically forced by wind waves and swell.
1
July 31, 2009 8:18 9.75in x 6.5in b684-ch01 FA
Fig. 1.1. Definition sketch. Energy and momentum are transferred from winds to waves in the
generating area. The waves convey energy and momentum to the surf zone where the waves break.
Upon breaking, the energy is dissipated and the momentum is transferred to the water column
resulting in longshore and onshore forces exerted on the water column.
Wave setup is the additional water level that is due to the transfer of
wave-related momentum to the water column during the wave-breaking process.
As waves approach the shoreline, they convey both energy and momentum in the
wave direction. Upon breaking, the wave energy is dissipated, as is evident from the
turbulence generated; however, momentum is never dissipated but rather is trans-
ferred to the water column resulting in a slope of the water surface to balance the
onshore component of the flux of momentum (see Fig. 1.1). If waves are irregular,
in addition to a steady wave setup, the setup includes a dynamic component that
oscillates with the wave group period and there may be a weak resonance within
the nearshore amplifying this oscillating component. These have been termed infra-
gravity waves and are more dominant for narrow banded spectra both in frequency
and in directional spreading. The oscillatory component is denoted “dynamic wave
setup” in this chapter.
This chapter discusses the significance of wave setup to coastal engineering
design, provides a review of the classical linear wave theory of wave setup, reviews
results from laboratory and field studies, summarizes results and recommends pre-
liminary design approaches for the static component. To provide a “look ahead,”
we will see that the phenomenon of wave setup is not yet adequately understood
for satisfactory engineering calculations and that the effects of profile slope are very
July 31, 2009 8:18 9.75in x 6.5in b684-ch01 FA
Wave Setup 3
significant. The interested reader is also referred to an earlier review article on wave
setup by Holman.5
Wave setup (both static and dynamic components) is relevant to a number of engi-
neering applications. The contributions of wave setup under extreme storm events
can be substantial, adding several feet to the elevated water levels. The interaction
of wave setup with vegetation differs from wind surge and thus it is important to
differentiate the two components, for example in ascertaining the benefits of wet-
lands in reducing wave setup. Finally, the oscillating component of wave setup is
relevant to beach safety in some locations and to many natural and constructed
coastal systems that have the capability to resonate including harbors and moored
ships.
Standard terminology defines the water level in the absence of wave effects as “still
water level,” whereas wave setup will cause a departure from the still water level
and this water level including the effects of the waves is the “mean water level.”
As implied, the mean water level is determined as the average of the fluctuating
water level over a suitable time frame usually taken as a number of multiples of
the short wave period, say the spectral peak. In considering wave setup, often the
location of interest is that of the maximum wave setup at the shoreline. This raises
the question of whether wave setup is defined at elevations above the maximum
rundown, say on the beach face where the water is present over only a portion of
the wave period. Since wave setup is defined as the mean water level, over what
period should the water surface be averaged on the beach face which is “wetted”
over only a portion of the wave period? If the time average is over only the portion
of the period that water is present, in the upper limit, the maximum setup will be
the maximum runup. For purposes here, wave setup will usually be defined only for
conditions where water is present over a full wave period.
When calculating wave runup on a structure such as a levee or revetment, the
question arises whether it is appropriate to first calculate wave setup and then add
the wave runup which is usually empirically based on model results. In the more
recent empirical results (e.g., the TAW method, see Ref. 6), the runup is expressed
as a proportion of the significant wave height at the base of the steeper slope (e.g.,
at a revetment or levee). The wave runup determined in the model on which the
method was based generally included some wave setup (or setdown) seaward of the
toe of the slope and included wave setup landward of the toe of the slope. Thus, in
the application of interest, the most appropriate approach is to calculate and include
wave setup at the toe of the slope; however, recognizing that the measured landward
runup includes setup, no additional setup should be added explicitly landward of
the toe of the slope.
July 31, 2009 8:18 9.75in x 6.5in b684-ch01 FA
Wave Setup 5
where u2 and w2 represent the time averages of the square of the first-order hori-
zontal and vertical wave velocities evaluated at the mean water surface, respectively.
Equation (1.4) is a type of a Bernoulli equation for unsteady flows which, when eval-
uated at the break point and considering no wave setup in deep water to evaluate
the constant, C = 0, the setup is negative (setdown) and given by
Hb2 kb
ηb = − , (1.5)
8 sinh 2kb hb
where Hb is the breaking wave height and kb is the wave number at breaking. For
shallow water conditions and depth limited breaking (Hb = κ(hb + η b )), Eq. (1.5)
yields
κHb
ηb = − . (1.6)
16
As an example, for a κ value of 0.78, the wave setdown is approximately 5% of the
breaking wave height.
With the seaward boundary condition now established, for the case of shallow
water wave-breaking and the consideration of depth limited breaking across the surf
zone, the wave setup is
κHb 3κ2 /8
η=− + (hb − h) . (1.7)
16 (1 + (3κ2 /8))
It is noted that in the above equation, the bottom shear stress has been taken as
zero and that a shoreward directed bottom shear stress on the water column as
would occur due to undertow would increase the wave setup. As examples, the ratio
of wave setup to breaking height at the still water line (h = 0) and at the location
of maximum wave setup (η = − h) for a κ value of 0.78 are
η(h = 0, κ = 0.78) 5κ
F0 |κ=0.78 ≡ = = 0.198 (1.8)
Hb 16(1 + (3κ2 /8)) κ=0.78
and
η(h = − η, κ = 0.78) 3κ2 5κ
Fmax |κ=0.78 ≡ = F0 1+ = = 0.244 .
Hb 8 κ=0.78 16 κ=0.78
(1.9)
It is seen that the wave setup is strongly dependent on the value of the breaking ratio
κ which will be shown to decrease with decreasing beach slope. Figure 1.2 presents
the ratios, F0 and Fmax versus κ. It is useful to relate κ in an approximate manner to
beach slope. Although there is not a one-to-one correspondence, Fig. 1.3 is based on
the Dally et al.8 wave-breaking model and provides an approximate correspondence
between uniform profile slope and the associated κ value. It is evident that the
Dally et al. model provides reasonable κ values for smaller beach slopes (say less
than about 0.06), but the κ values are too large for steeper slopes.
July 31, 2009 8:18 9.75in x 6.5in b684-ch01 FA
0.4
Fo
Fmax
0.3
Fo and Fmax
0.2
0.1
0.0
0.0 0.2 0.4 0.6 0.8 1.0
Kappa
2.0
1.5
Kappa
1.0
0.5
0.0
0.00 0.02 0.04 0.06 0.08 0.10
Profile Slope
Fig. 1.3. Relationship between profile slope and κ (kappa) value. Based on Dally et al.8 wave-
breaking model.
July 31, 2009 8:18 9.75in x 6.5in b684-ch01 FA
Wave Setup 7
1.2
Ratio of Nonlinear to Linear Momentum Flux
1.0
0.8
.0
5
.7
b =1
=0
50
b
/H
H
5
0.
H/
0.6
.2
b=
b =0
H
H/
H
H/
0.4
0.2
0.0
2 3 4 5 6 2 3 4 5 6 2 3 4 5 5 2 3 4 5 6
10-3.000 10-2.000 10-1.000 100.000 101.000
h/Lo
Fig. 1.4. Ratio of nonlinear to linear wave momentum flux, Sxx , for forty stream function wave
combinations.12
July 31, 2009 8:18 9.75in x 6.5in b684-ch01 FA
at breaking). The reduction in Sxx for a wave of given height and for shallow water
conditions is 16.7%.
η = F Hb , (1.10)
η max = F (H1 + H2 ) ,
(1.11)
ηmin = F (H1 − H2 ) .
July 31, 2009 8:18 9.75in x 6.5in b684-ch01 FA
Wave Setup 9
It can be shown that the average wave setup depends on the ratio H2 /H1 as shown
in Table 1.1. The fourth column presents the ratios of the maximum dynamic wave
set amplitude to the average wave setup component.
In the case with H2 = H1 , the dynamic wave setup displacement from the mean
setup equals 57% of the average wave setup (Table 1.1, Column 4).
In the above, we have examined the dynamic wave setup for the case of a simple
bichromatic wave system in which the difference in frequencies of the two com-
ponents was fairly small. For the case of a wave spectrum, the situation is much
more complex with, for the case of a narrow spectrum, the group envelope varying
according to the Rayleigh distribution. For the case of a wide spectrum, the dynamic
component is reduced considerably.
Having reviewed the theory of wave setup and its relationship to various factors,
the two sources available for evaluation are laboratory and field data.
conveys wave energy and momentum landward prior to transfer of the momentum
to the water column and the associated wave setup.21 Later, Stive and Wind22 con-
ducted a very detailed laboratory investigation in which they demonstrated the role
of wave nonlinearity. In this study, the momentum flux components (velocities and
pressures) were measured to the degree possible and it was found that the calcu-
lated wave setup based on nonlinear wave theories was in much better agreement
with measured wave setup than calculations based on linear wave theories. In these
comparisons, it was not necessary to introduce the roller concept.
The two laboratory studies reviewed above have focused on static setup and
it has been noted that irregular waves also produce dynamic wave setup. Hedges
and Mase23 have presented an interesting reanalysis of earlier runup laboratory
measurements by Mase24 in which irregular waves provided the forcing.b The planar
slopes represented in the data were: 1:5, 1:10, 1:20, and 1:30. Figure 1.6 presents
an example of the form in which the data were plotted where the horizontal axis is
b Walton25 was the first to analyze the Mase data to extract the static wave setup.
July 31, 2009 8:18 9.75in x 6.5in b684-ch01 FA
Wave Setup 11
as discussed earlier), and the second is that the oscillating wave setup component
should depend on the width of the input spectrum. Referring to Fig. 1.6, which is
one of several similar plots presented in the Hedges and Mase paper, since each plot
may include a mix of beach slopes, the slope dependency is not resolved in the Smean
results which of course are derived from the y-intercept of these graphs. Secondly,
these experiments were not designed to evaluate the effect of spectral width and the
spectral characteristics included in the experiments are not known. However, it is of
interest to identify the “representative” κ value and beach slope associated with a
Smean = η avg /H1/3 of 0.17. Referring to Fig. 1.2, we see that the associated κ value
is approximately 0.63 for F0 . Based on the Dally et al.8 breaking wave model, the
associated beach slope from Fig. 1.3 is 1:29 compared to the beach slopes in the
Mase experiments ranging from 1:30 to 1:5.
Hansen also noted the maximum wave setup to be approximately 50% of the signif-
icant breaking wave height. It is not clear as to the methodology utilized to obtain
η max in this field experiment.
July 31, 2009 8:18 9.75in x 6.5in b684-ch01 FA
Wave Setup 13
A wave setup field experiment was conducted as part of the Nearshore Sed-
iment Transport Study at Torrey Pines Beach, San Diego, California by Guza and
Thornton.30 The Torrey Pines Beach face was gently sloping (beach slope ≈ 0.02)
and the beach material was a moderately sorted fine grain sediment (≈ 0.1 mm).
A dual wire resistance runup meter was used for the recording and estimation of
the wave setup. It should be noted that the measurements of the wave setup were
considered to be the average runup determined from wires placed approximately
3 cm above the beach level rather than an actual water level at one location in
these experiments. Offshore pressure sensors outside the surf zone at mean depths
of 7 to 10.5 m were used for estimating wave height with recording lengths of 4096 s.
Guza and Thornton30 note specific problems in the data set, which are typical of
field measurements, i.e., the difficulty in obtaining a common datum for the off-
shore wave measurements and the beach wave setup measurements. Results of their
measurement program suggest an empirical relationship as follows:
η = 0.17Hos = 0.24Horms (1.16)
with scatter that suggests η/Hos ranging approximately from 0.05 to 0.50 for indi-
vidual experiments.
Holman and Sallenger28 conducted a field experiment for measuring wave setup
as well as other surf zone parameters at the U.S. Army Corps of Engineers field
research pier in Duck, NC, USA. Data on water level at the shoreline were col-
lected using longshore looking time lapse photography from Super-8 movie cameras
mounted on the research pier scaffolding. The beach at the experiment site had a
very steep foreshore (∼ 1 on 10) while the offshore profile slope is much milder (∼ 1
on 100). Beach material was bimodal in size with a median sand size of 0.25 mm and
a coarse fraction of 0.75 mm. Results of the experiments showed considerable scatter
and dependence on tide level. Regression lines were fit to the data (segmented by
tide levels) with results as follows for high tide and mid-tide data:
η
= 0.35ξ0 + 0.14 (high tide) , (1.17)
Hs
η
= 0.46ξ0 + 0.06 (mid-tide) . (1.18)
Hs
As most of the data fell in a range of ξ0 = 1 to 2, the maximum setup was noted
to be of the same order as the significant wave height in many of the experiments,
much higher than theoretically suggested values. Note that in terms of Horms (based
on consideration of monochromatic theory results) the setup would be much higher
than most other studies show or suggest.
Although Holman and Sallenger28 conclude from their experiments that the
setup is dependent on the Iribarren number, it is not entirely clear from their data,
especially for higher waves (i.e., see Fig. 1.4, Ref. 28). An additional problem that
must be considered when computing the Iribarren number for real beaches and
irregular waves is how to define beach slope. It should be noted that video camera
(visual) approaches estimate setup via the measurement of the water surface ele-
vation on the beach (similar to the Guza and Thornton measurements) rather than
an actual vertically fluctuating water level. The anomaly between dependence of
July 31, 2009 8:18 9.75in x 6.5in b684-ch01 FA
setup on Iribarren number as noted by Holman and Sallenger28 is likely due to the
aforementioned relationship between the wave-breaking coefficient, κ, and beach
slope.
Nielsen31 and Davis and Nielsen32 conducted a novel setup experiment on Dee
Why Beach, in New South Wales, Australia using a set of manometer tubes as shown
in Figs. 1.7 and 1.8 from Davis and Nielsen.32 The tubes were deployed throughout
the beach face and surf zone. A total of 120 setup profiles were measured in 11 days.
Wave heights Horms ranged from 0.6 to 2.6 m in height and significant wave periods
(Ts ) ranged from 5.8 to 12.1 seconds. A shoreline setup of about 40% of Horms was
found although Davis and Nielsen32 point out that there is reason to believe that
the surf zone characteristics influence the relationship between wave height and
setup magnitude, and also note a problem of defining beach slope via the Iribarren
number. Nielsen31 and Davis and Nielsen32 also observe that a major portion of the
setup occurred on the beach face as shown in Fig. 1.9.
Nielsen31 points out that previous field investigations have typically measured
the mean water level elevation on the beach as opposed to the average fluctuating
mean water level in the vertical plane (i.e., the wave setup as usually defined), and
that the two measurements are often different in part due to the beach permeability,
which in turn is related to beach material size. The issue of extracting wave setup
from runup and rundown on the beach is illustrated in Fig. 1.10.
King et al.33 collected wave setup data at Woolacombe Beach in North Devon,
U.K. which faces the North Atlantic Ocean. The beach face slope varied between
1 on 40 at high tide and 1 on 70 at mid-tide level with a tidal range of 3 m at
neap and as much as 9 m at springs. Beach face material consisted of fine sand with
Fig. 1.7. Manometer setup of Davis and Nielsen32 for measuring setup.
July 31, 2009 8:18 9.75in x 6.5in b684-ch01 FA
Wave Setup 15
Fig. 1.9. Dimensionless setup versus total depth where much of the setup occurs on the beach face
(from Ref. 32). In this figure, B and D are equal to η and h as used in this chapter, respectively.
90% in the 0.125 mm to 0.25 mm size range. Pressure transducers were utilized to
collect wave and setup information at various stations across the beach and also
in a longshore direction to assess the spatial variability of the mean setup. Both
tripod mounted and buried pressure transducers were utilized. The buried pressure
transducers were 50 to 80 cm below the beach surface and were protected by a
porous cover. Instruments collected pressure data which were then transformed to
water level data over 4096 second intervals. Data did not include sampling in very
shallow water and the maximum wave setup was estimated by extrapolating the
July 31, 2009 8:18 9.75in x 6.5in b684-ch01 FA
Fig. 1.10. Illustration of differences between mean water level (MWL) shoreline and mean water
line on beach.
water surface from the most shoreward water stations. Wave setup estimated from
the data showed the wave setup to be roughly:
with most of the values of η/H orms between 0.11 and 0.15. The authors do not
speculate as to why such low values of setup (compared to analytical results) were
found in this measurement program.
Yanagishima and Katoh34 discuss field measurements of mean water level near
the shoreline on the Pacific Coast of Japan as measured by an ultrasonic wave gage
mounted on a pier where the mean depth of water was ∼ 0.4 m. The setup was
determined via a multiple regression approach on 1305 sets of (20 minute records)
data taking into account astronomical tide, wind setup, and atmospheric pressure
head components of mean water level. Their data included 91 records in which the
offshore wave height was above 3 m. Yanagishima and Katoh’s34 regression analysis
suggested the following relationship:
0.2
Los
η = 0.0520H os , (1.20)
Hos
which can be formulated in terms of Iribarren number for their beach slope (1 on
60) to the following:
Yanagishima and Katoh34 noted reasonable agreement with the theory of Goda35
(to be discussed later). Even higher values of setup would be expected on the beach
face in accord with theory and findings of other researchers.
Greenwood and Osborne36 conducted field measurements on a Georgian Bay
Beach, in Lake Huron, Ontario, Canada. Lake Huron has no measurable tide and
the beach profile at the site had a slope of 0.015 with a steeper sloped (0.031 to
0.047) inshore bar. Setup was measured using surface piercing resistance wire wave
staffs with the shoreward most gage being in approximately 0.4 m of water depth.
Measured setup values were found as follows:
Wave Setup 17
It is again noted that even higher values of setup would be expected on the beach
face in accord with theory and experience of other researchers.
Further work by Hanslow and Nielsen37,38 utilized the manometer tube
deployment shown in Fig. 1.7 on three additional beaches (Seven Mile, Palm, and
Brunswick) in New South Wales, Australia. With beach face slopes ranging from
0.03 to 0.16 and mean grain sizes of swash zone beach material ranging from 0.18
to 0.5 mm, shoreline beach setup was measured using 20 minute record averages.
Using the data from these three beaches as well as earlier measurements at Dee
Why Beach (see Refs. 31 and 32), linear least square relationships were fit to the
data as follows:
or
η = 0.040 Hos L0 = 0.048 Horms L0 with R = 0.77 , (1.24)
where a somewhat higher value of explained regression was noted using wave height
and wave period. Data and regression lines for these two relationships are shown
in Figs. 1.11 and 1.12. The improvement in fit due to inclusion of the deep water
wavelength is not evident visually.
A significant finding of these studies was that a major portion of the setup
occurred on the beach face (see Fig. 1.9). Further measurements on wave setup at
two river entrances is also discussed in Hanslow and Nielsen37 and Dunn et al.39
with the result that the wave setup at river entrances was found to be (somewhat
surprisingly) negligible.
Lentz and Raubenheimer40 report on a field experiment at the U.S. Army Field
Research Pier in Duck, NC, USA where 11 pressure sensor gages and 10 sonar
altimeters extended across the surf zone from 2 to 8 m of water depth. Close
agreement with Longuet-Higgins radiation stress theory for wave setup was noted
Fig. 1.11. Empirical relationship between setup and wave height (from Ref. 38).
July 31, 2009 8:18 9.75in x 6.5in b684-ch01 FA
Fig. 1.12. Empirical relationship between setup and wave parameters (from Ref. 38).
although the lack of setup measurements in shallow water (< 2 m) did not allow
conclusions regarding the maximum setup that might be expected on the beach.
Raubenheimer et al.41 report on a second field experiment at the U.S. Army
Field Research Pier in Duck, NC, USA where 12 buried pressure sensor gages were
employed across the surf zone from the shoreline to 5 m of water depth. Again good
agreement with Longuet-Higgins and Stewart42 radiation stress theory was noted
by integration of the cross-shore momentum equation to estimate the wave setup
for water depths greater than 1 m but the theory was found to under-predict wave
setup in shallow water (h < 1 m). The lack of setup measurements on the beach face
did not allow conclusions regarding the maximum setup that might be expected on
a beach although an empirical equation was provided to estimate wave setup at the
SWL line as follows:
η SWL −1/3
= 0.019 + 0.003βf , (1.25)
Hos
where βf is the average slope across the surf zone. Raubenheimer et al.41 suggest
that theory under-predicts the setup by a factor of 2 for water depths less than 1 m.
Stockdon et al.43 using video shoreline water level time series determined wave
setup and wave runup results during 10 diverse field experiments (four from Duck,
NC, USA; four from West Coast beaches in California/Oregon, USA; and two from
Terschelling, The Netherlands). These wave setup results were analyzed to provide
empirical parameterizations for wave setup under many natural beach conditions
as follows:
η = 0.385βf H0 L0 , (1.26)
which, assuming that tan βf ≈ βf can be expressed in terms of the Iribarren
number as
η
= 0.385ξ0 (1.27)
H0
July 31, 2009 8:18 9.75in x 6.5in b684-ch01 FA
Wave Setup 19
where H0 is the effective deep water wave height, L0 is the deep water wave length
associated with the peak spectral period and βf is the average slope over a depth
range defined in terms of the standard deviation of the water surface displacement.
It should again be noted that the video camera (visual) approach estimates setup
via the mean of measurements of the water surface elevation on the beach rather
than the mean of fluctuating water levels at one location.
Results from nine of the field experiments presented here have been analyzed to
determine the average ratio of wave setup at the still water line to significant wave
height and its associated standard deviation. The ratios at the still water line were
determined to be 0.191±0.100. Several caveats apply to these results. In cases where
the beach slope and/or the deep water wave steepness was incorporated into the
expression presented, these were taken as 0.01 and 0.04, respectively. Some of the
published expressions were in terms of the breaking wave height and some in terms
of the deep water wave height and no attempt was made to differentiate between
breaking and deep water wave heights. The Holman and Sallenger results were not
included in these results as they appeared to be anomalously high. Finally, the wave
setup ratio at the intersection of the mean water line intersection with the beach
profile would be greater than the average ratio (0.191) above. Also, although not
examined in detail here, the dynamic wave setup which increases with energetic
narrow spectra, would also contribute to the total wave setup.
It is relevant to note that results from field measurements are often not con-
sistent, possibly due to:
Several sources of wave setup recommendations are available; two are reviewed
here. The U.S. Army Corps of Engineers 1984 Shore Protection Manual (SPM)
presents a graphical method to calculate wave setup at mid-depth of the surf zone.
This method, developed for irregular waves, is presented in Fig. 1.13 in which the
normalized setup has been multiplied by a factor of 2 to transfer approximately the
results to the still water shoreline. The effect of beach slope and deep water wave
steepness in Fig. 1.13 are evident.
July 31, 2009 8:18 9.75in x 6.5in b684-ch01 FA
Fig. 1.13. Nondimensional wave setup versus deep water wave steepness and profile slope by
the 1984 Shore Protection Manual recommendations as incorporated in Appendix D of FEMA44
Guidelines. Note that the normalized setup has been multiplied by a factor of 2 to transfer the
setup from the mid-depth of the surf zone as it appears in SPM to the approximate still water
level contour. Note: S in this figure is equal to η in this chapter.
Fig. 1.14. Nondimensional wave setup by Goda versus deep water wave steepness and relative
water depth within the surf zone. Profile slope = 1:100.
Goda35 has presented guidance for static and dynamic wave setups due to
irregular waves. The guidance for static setup and a profile slope of 1:100 is shown
in Fig. 1.14. The effects of various deep water wave steepness values are illustrated
in Fig. 1.14.
July 31, 2009 8:18 9.75in x 6.5in b684-ch01 FA
Wave Setup 21
η/H0
The reviews of theory, laboratory and field data, and published guidance for engi-
neering applications presented here have identified static and dynamic wave setup
components as contributing to the deviation from still water level in the surf zone
and their relevance to engineering design. Examination of the static wave setup
has reinforced the effect of beach slope on wave setup. The theory presented here
does not account for the onshore bottom stress acting on the water column due to
undertow.
The available field measurement results exhibit a wide range of wave setup to
wave height ratios. Some of this variability is undoubtedly due to the effect of
profile slope, which is not accounted for explicitly in some of the analyses and
part is due to the effect of wave-breaking in depths greater than the shallow water
limit.
Design methodology should account for the static and dynamic wave setup com-
ponents. In determining the wave setup to include in design, the characteristics of
the particular application of interest should be compared with those of the various
field and laboratory experiments available including those referenced here. The dom-
inant role of beach slope should be recognized. The preliminary results presented
here of η/Hs = 0.191 ± 0.100 may serve as a useful guide for the static wave setup
component.
It is hoped that further research with improved instrumentation, modern sur-
veying techniques, and more diverse field site studies will help to clarify both the
static and dynamic wave setup components for future design applications.
July 31, 2009 8:18 9.75in x 6.5in b684-ch01 FA
References
Wave Setup 23
23. T. S. Hedges and H. Mase, Modified Hunt’s equation incorporating wave setup,
J. Waterway Port Coast. Ocean Eng. 130(3), 109–113 (2004).
24. H. Mase, Random wave runup height on gentle slopes, J. Waterway Port Coast. Ocean
Div. 115(WW2), 649–661 (1989).
25. T. L. Walton, Interim guidance for prediction of wave run-up on beaches, Ocean Eng.
19(2), 199–207 (1999).
26. T. Saville, Experimental determination of wave set-up, Proc. 2nd Technical Conf.
Hurricanes, Miami Beach, FL, National Hurricane Research Project Report No. 50,
U.S. Department of Commerce Washington, D.C. (1961), pp. 242–252.
27. R. Dorrestein, Wave set-up on a beach, Proc. 2nd Technical Conf. Hurricanes, Miami
Beach, FL, National Hurricane Research Project Report No. 50, U.S. Department of
Commerce, Washington, D.C. (1961), pp. 230–241.
28. R. A. Holman and A. H. Sallenger, Setup and swash on a natural beach, J. Geophys.
Res. 90(C1), 945–953 (1985).
29. U. A. Hansen, Wave setup and design water level, J. Waterway Port Coast. Ocean
Div. 104(WW2), 227–240 (1978).
30. R. T. Guza and E. B. Thornton, Wave set-up on a natural beach, J. Geophys. Res.
96(C2), 4133–4137 (1981).
31. P. Nielsen, Wave setup: A field study, J. Geophys. Res. 93(C12), 15643–15652 (1988).
32. G. A. Davis and P. Nielsen, Field measurement of wave set-up, ASCE Int. Conf.
Coastal Engineering, Malaga, Spain (1988), Chapter 38, pp. 539–552.
33. B. A. King, M. W. L. Blackley, A. P. Carr and P. J. Hardcastle, Observations of wave
induced setup on a natural beach, J. Geophys. Res. 95(C12), 22289–22297 (1990).
34. S. Yanagishima and K. Katoh, Field observation on wave setup near the shoreline,
Proc. 22nd Int. Conf. Coastal Engineering, Vol. 1, ASCE, New York, N.Y. (1990),
Chapter 7, pp. 95–108.
35. Y. Goda, Random Seas and Design of Maritime Structures (World Scientific
Publishing Co., 2000), 443 pp.
36. B. Greenwood and P. D. Osborne, Vertical and horizontal structure in cross-shore
flows: An example of undertow and wave setup on a barred beach, Coast. Eng. 14,
543–580 (1990).
37. D. J. Hanslow and P. Nielsen, Wave setup on beaches and in river entrances, 23rd Int.
Conf. Coastal Engineering, Venice, Italy (1992), pp. 240–252.
38. D. J. Hanslow and P. Nielsen, Shoreline set-up on natural beaches, J. Coast. Res.
SI15, 1–10 (1993).
39. S. L. Dunn, P. Nielsen, P. A. Madsen and P. Evans, Wave setup in river entrances,
Proc. 27th Int. Conf. Coastal Engineering, ASCE, New York, Sydney, Australia (2000),
pp. 3432–3445.
40. S. Lentz and B. Raubenheimer, Field observations of wave setup, J. Geophys. Res.
104(C11), 25867–25875 (1999).
41. B. Raubenheimer, R. T. Guza and S. Elgar, Field observations of wave-driven setdown
and setup, J. Geophys. Res. 106(C3), 4629–4638 (2001).
42. M. S. Longuet-Higgins and R. W. Stewart, Radiation stresses in water waves: A
physical discussion with applications, Deep Sea Res. 11(4), 529–562 (1964).
43. H. F. Stockdon, R. A. Holman, P. A. Howd and A. H. Sallenger, Jr., Empirical param-
eterization of setup, swash, and runup, Coast. Eng. 53, 573–588 (2006).
44. FEMA, Guidelines and Specifications for Flood Hazard Mapping Partners,
Appendix D: Guidance for Coastal Flooding Analysis and Mapping, Map Modern-
ization Program, Washington, D.C. (2003). Also available at: www.fema.gov/fhm/
dl cgs.shtm.
July 31, 2009 8:18 9.75in x 6.5in b684-ch01 FA
Chapter 2
Wavemaker Theories
Robert T. Hudspeth
School of Civil and Construction Engineering
Oregon State University, Corvallis, OR 97331, USA
robert.hudspeth@orst.edu
Ronald B. Guenther
Department of Mathematics
Oregon State University, Corvallis, OR 97331, USA
guenther@math.orst.edu
2.1. Introduction
Wavemaker theories play several important roles in coastal and ocean engineering.
The most important role is the application to laboratory wavemakers for both
wavemaker designs and wave experiments. A second role for wavemaker theories
is to compute a scalar radiated wave potential to compute the wave-induced loads
on large solid bodies applying potential wave theory. The displacements and rota-
tions of a semi-immersed six degrees-of-freedom large Lagrangian solid body are
related to the displacements and rotations of wavemakers. The boundary between a
planar wavemaker and an ideal fluid requires special care because the fluid motion
25
July 31, 2009 8:18 9.75in x 6.5in b684-ch02 FA
is an Eulerian field with time and space as the independent variables, and the
planar wavemaker is a Lagrangian solid body with time and the wavemaker as the
independent variables. Consequently, the kinematic boundary condition will be dif-
ferent from the free surface boundary that separates two Eulerian fluid fields of air
and water. The boundary between the fluid and wavemaker separates an Eulerian
field (the fluid) from a Lagrangian body (the wavemaker), and the wavemaker
kinematic boundary condition (WMKBC) must convert the Lagrangian wavemaker
motion to a Eulerian field motion in order that the independent variables for both
dependent motion variables are equivalent. This may be accomplished by multi-
plying the Lagrangian motion of the wavemaker by the unit normal to the boundary.
Because the unit normal is a function of space and the Lagrangian wavemaker
motion is a function of time, the product will produce a motion that is a function
of both space and time that are the independent variables of the Eulerian fluid
field. Although this fact is not central to the WMBVP, it is an important con-
nection between the WMBVP and the radiation potential boundary value problem
for semi-immersed large Lagrangian solid bodies.1
The formulae for computing the two fundamental fluid unknowns for an incom-
pressible fluid of the velocity q(x, z, t) and the pressure p(x, z, t) from a scalar
velocity potential Φ(x, z, t) are given first. The classical linear WMBVP for dimen-
sionless 2D planar wavemaker is reviewed for two types of double-articulated planar
wavemakers. The sway X1 (t) displacement of a full-draft piston wavemaker and the
roll Θ5 (t) rotation of a hinged wavemaker are connected directly to the sway dis-
placement and the roll rotation of a semi-immersed large Lagrangian solid body. In
this review, integral calculus formulae for computing the integrals that are required
to compute the coefficients of the eigenseries for the fluid motion, to compute the
loads on the wavemaker and the average power required to generate the propagating
waves are replaced by generic algebraic formulae. For example, an integral equation
that is required to compute the nth eigenseries coefficient Cn for the nth eigen-
function Ψn (Kn , z/h) from a wavemaker shape function χ(z/h) may be computed
symbolically and expressed by a dimensionless algebraic formula In (α, β, b, d, Kn ),
that is given by
0
Cn = h χ(z/h)d(z/h)Ψn (z/h)d(z/h) = In (α, β, b, d, Kn ). (2.1)
−1
The coefficient in (2.1) may then be computed very efficiently by substitution into
algebraic formulae in all subsequent applications. Next a dimensionless theory for
both amplitude-modulated (AM) and phase-modulated (PM) circular wavemakers is
reviewed. Then, a dimensionless theory for double-actuated wavemakers is reviewed.
Following that, a dimensionless directional wavemaker theory for large wave basins
based on a WKBJ approximation1 is reviewed. Next, a theory for sloshing waves
due to transverse motions of a segmented wavemaker in a narrow wave channel is
reviewed. Then, 2D planar wavemakers are mapped to a unit circle by conformal
mapping and to a fixed rectangular domain by domain mapping; and both the linear
and nonlinear wavemaker solutions are computed numerically.
July 31, 2009 8:18 9.75in x 6.5in b684-ch02 FA
Wavemaker Theories 27
Two generic planar wavemaker configurations are shown in Figs. 2.1(a) and 2.1(b).
The fluid motion may be obtained from the negative gradient of a dimensional scalar
velocity potential Φ(x, z, t) according to
∂(•) ∂(•)
∇2 (•) = ex + ez .
∂x ∂z
The total pressure field P (x, z, t) may be computed from the unsteady Bernoulli
equation according to
∂Φ(x, z, t) 1
P (x, z, t) = p(x, z, t) + pS (z) = ρ − |∇Φ(x, z, t)| + Q(t) − ρgz,
2
∂t 2
(2.2b)
where Q(t) = the Bernoulli constant; and the free surface elevation η(x, t) for zero
atmospheric pressure according to
1 ∂Φ(x, η, t) 1
η(x, t) = − |∇Φ(x, η, t)| + Q(t) ;
2
x ≥ ξ(η, t); z = η(x, t).
g ∂t 2
(2.2c)
The scalar velocity potential must be a solution to the Laplace equation
∂Φ
= 0; x ≥ ξ(−h, t); z = −h. (2.3b)
∂z
DW ∂Φ ∂ξ ∂Φ ∂ξ
= + − = 0; x = ξ(z, t); −h ≤ z ≤ η(t). (2.3d)
Dt ∂x ∂t ∂z ∂z
Wavemaker Theories 29
where Re{•} means the real part of {•}; and ν = arbitrary phase angle. The
linearized WMBVP for kh = O(1) is1
∇22 ϕ(x, z) = 0; 0 ≤ x < +∞; −h ≤ z ≤ 0, (2.5a)
∂ϕ(x, z)
= 0; 0 ≤ x + ∞; z = −h, (2.5b)
∂z
∂ϕ(x, z)
− k0 ϕ(x, z) = 0; 0 ≤ x < +∞; z = 0, (2.5c)
∂z
∂
lim − iKn ϕ(x, z) = 0, (2.5d)
x→+∞ ∂x
∂ϕ(x, z) ∂ξ(z, t)
exp −i(ωt + ν) = − ; x = 0; −h ≤ z ≤ 0, (2.5e)
∂x ∂t
−iω
η(x, t) = Re ϕ(x, 0) exp −i(ωt + ν) ; x ≥ 0; z = 0, (2.5f)
g
∂Φ(x, z, t)
p(x, z, t) = ρ ; 0 ≤ x < +∞; z = 0, (2.5g)
∂t
where k0 = ω 2 /g.
Because the boundary conditions defined by (2.5b)–(2.5e) are prescribed on
boundaries with constant values of the independent variables x and z, a solution by
the method of separation of (independent) variables may be computed.1 The instan-
taneous wavemaker displacement ξ(z, t) from its mean position x = 0 is assumed
to be strictly periodic in time with period T = 2π/ω, and may be expressed by
S
ξ(z/h, t) = Re i χ(z/h) exp −i(ωt + ν)
(∆/h)
S
= χ(z/h) sin(ωt + ν). (2.6)
(∆/h)
The specified shape function χ(z/h) for the Type I wavemaker shown in Fig. 2.1(a)
is valid for either a double-articulated piston or hinged wavemaker of variable draft
and is given by the following dimensionless equation for a straight line2 :
χ(z/h) = [α(z/h) + β][U (z/h + 1 − d/h) − U (z/h + b/h)], (2.7a)
where α, β = dimensionless constants; U (•) = the Heaviside step function with two
boundary conditions given by
[S/(∆/h)]χ(z/h = −1 + d/h + ∆b /h + ∆/h) = S, (2.7b)
[S/(∆/h)]χ(z/h = −1 + d/h + ∆b /h) = Sb , (2.7c)
July 31, 2009 8:18 9.75in x 6.5in b684-ch02 FA
The coefficients α and β for the specified shape function χ(z/h) in (2.7a) may be
obtained for the Type II wavemaker shown in Fig. 2.1(b) by substituting
(S̄ + Ŝ)
χ(z/h = −1 + d/h) = S̄, (2.8b)
1 − b/h − d/h
Wavemaker Theories 31
The wave number k = 2π/λ where λ = wavelength. Because the numerical value of
kh must be computed from an eigenvalue problem in the vertical z coordinate, equiv-
alence of the eigenvalue k to the wave number 2π/λ requires a pseudo-horizontal
boundary condition of periodicity given by k = 2π/λ and ϕ(x + λ, z) = ϕ(x, z). It
is computationally efficient to normalize the eigenseries in (2.10a) according to
cosh Kn h(1 + z/h)
Ψn (Kn , z/h) = ; n = 1, 2, . . . , (2.11a)
Nn
where the nondimensional normalizing constant Nn is
2kh + sinh 2kh
0
; n = 1, (2.11b)
4kh
Nn2 = cosh2 Kn h(1 + z/h)d(z/h) =
−1 2κn h + sin 2κn h
; n ≥ 2. (2.11c)
4κn h
The eigenseries in (2.10a) may be written as an orthonormal eigenseries by
∞
and
ρω 3 S 2 Bh4
Ẇ τ = Pτ = I12 (α, β, b, d, k), (2.16c)
∆2 2kh
so that all of the average power from a wavemaker is transferred to only the
propagating eigenmode. The average energy flux in a linear wave is given by1
ρgBA2
Ėτ = CG , (2.16d)
2
Equating (2.16c) to (2.16d) gives the following transfer function for a planar
wavemaker:
A ko h
= Ψ1 (k, 0)I1 (α, β, b, d, k). (2.16f)
S kh
July 31, 2009 8:18 9.75in x 6.5in b684-ch02 FA
Wavemaker Theories 33
where the outward pointing unit normal n points from the wavemaker into the fluid,
and the pseudo-unit normal n for the rotational modes is given by
Force. For the Type I piston wavemaker of total width B, the horizontal component
of the pressure force on the fluid side only may be computed from the real part of
−b/h
where the static component of the pressure force on the fluid side only is
ρgBh2
Fs = − [1 − 2(d/h) + (d/h)2 − (b/h)2 ]. (2.18b)
2
The hydrodynamic component of F1 (t) may be separated linearly into a propagating
and an evanescent component that are related to the piston wavemaker translational
velocity and acceleration, respectively, from the real part of
where the added mass coefficient µ11 may be computed from the evanescent
eigenmodes only, and the radiation damping coefficient λ11 may be computed from
the propagating eigenmode only. The average power may be computed from
λ11 (Sω)2
−F1 Ẋ1 t = . (2.19a)
2
Equating (2.19a) to (2.16d) yields
2
A1 ρBh
λ11 = CG , (2.19b)
S1 ko h
July 31, 2009 8:18 9.75in x 6.5in b684-ch02 FA
that relates the radiation damping coefficient to the square of the ratio of the
radiated wave amplitude to the amplitude of the wavemaker displacement.
Moment. For the Type I wavemaker of width B, the dynamic pressure moment
on one side only of the wavemaker may be computed from the real part of
−b/h
z d
M5 (t) = Re iρωBh 2
Cn 1+ −
n=1 −1+d/h h h
z
× Ψn (Kn , z/h)d exp −(ωt + ν)
h
and the static component of the pressure moment on the fluid side only is
3 3
2 2 2
ρgBh3 d b d b d b
Ms = 1− +2 +3 − −3 1− .
6 h h h h h h
(2.20b)
The pressure moment M5 (t) in (2.20a) may be separated linearly into a propagating
and an evanescent component that are related to the rotational velocity and accel-
eration from the real part of1
Sω 2
M5 (t) = −Re µ55 −i
∆(1 + ∆b /∆)
Sω
+ λ55 −i exp −i(ωt + ν) ,
∆(1 + ∆b /∆)
M5 (t) = −µ55 Θ̈5 (t) − λ55 Θ̇5 (t), (2.21a)
where
I 2 (α, β, b, d, κn )
n
µ55 = ρBh4 , (2.21b)
n=2
κn h
I12 (α, β, b, d, k)
λ55 = ρωBh4 . (2.21c)
kh
Havelock5 applied Fourier integrals to develop a theory for surface gravity waves
forced by circular wavemakers in water of both infinite and finite depth. The fluid
motion may be obtained from the negative gradient of a scalar velocity potential
Φ(r, θ, z, t) according to
Wavemaker Theories 35
The total pressure field P (r, θ, z, t) may be computed from the unsteady Bernoulli
equation in polar coordinates according to
where Q(t) = the Bernoulli constant, and the free surface elevation η(r, θ, t) for zero
atmospheric pressure according to
1 ∂Φ(r, θ, η, t) 1 Q(t)
η(r, θ, t) = − |∇Φ(r, θ, η, t)|2 + ;
g ∂t 2 ρ
r ≥ b + ξ(θ, η, t); z = η(r, θ, t). (2.22d)
∂Φ
= 0; r ≥ b + ξ(θ, −h, t); 0 ≤ θ ≤ 2π; z = −h. (2.23b)
∂z
∂Φ ∂ξ 1 ∂Φ ∂ξ ∂Φ ∂ξ
+ − − = 0; r = ξ(θ, z, t); −h ≤ z ≤ η(b, θ, t). (2.23d)
∂r ∂t r2 ∂θ ∂θ ∂z ∂z
The distinction between these two types is in the azimuthal θ dependency of the
wavemaker displacement ξ(θ, z, t) from its mean position r = b, given by
mS cos mθ sin(ωt + ν) (2.23e)
ξ(θ, z, t) = χ(z/h) .
(∆/h) sin(ωt + ν + mθ) (2.23f)
Kinematic Radiation Boundary Condition (KRBC):
∂
√ lim ± iKn Φ(r, θ, z, t) = 0; r → ∞. (2.23g)
|Kn r|→+∞ ∂r
∂ϕ(r, θ, z)
= 0; b ≤ r < +∞; 0 ≤ θ ≤ 2π; z = −h, (2.25b)
∂z
∂ϕ(r, θ, z)
− ko ϕ(r, θ, z) = 0; b ≤ r < +∞; 0 ≤ θ ≤ 2π; z = 0, (2.25c)
∂z
∂
√ lim − iKn ϕ(r, θ, z) = 0, (2.25d)
|Kn r|→+∞ ∂r
∂ϕ(r, θ, z) ∂ξ(θ, z, t)
exp −i(ωt + ν) = − ; r = b; 0 ≤ θ ≤ 2π; −h ≤ z ≤ 0,
∂r ∂t
(2.25e)
−iωϕ(r, θ, z)
η(r, θ, t) = Re exp −i(ωt + ν) ; b ≤ r < ∞; 0 ≤ θ ≤ 2π; z = 0,
g
(2.25f)
P (r, θ, z, t) = {p(r, θ, z, t)} + ps (z)
= Re{−iωρϕ(r, θ, z) exp −i(ωt + ν)} − ρgz, (2.25g)
Wavemaker Theories 37
is shown in Fig. 2.2 is identical to (2.7) for a 2D planar wavemaker with the
dimension b replaced with a and the stroke S replaced with the azimuthal stroke
m S. The solution to the WMBVP (2.25) may be compactly expressed by the fol-
lowing orthonormal eigenseries:
∞
where Km (•) = the Modified Bessel (or Kelvin) function of the second kind of
order m. The coefficients Cmn may be computed by expanding the KWMBC in an
eigenseries following the procedure in (2.14) and obtaining
m Sj hω In (α, β, a, d, Kn )
Cmn = − ; n ≥ 1 and integer, (2.26f)
Kn ∆ Ln (Hm
(1)
(Kn b))
dZm (ζn ) 1
Ln (Zm (ζn )) = = {Zm−1 (ζn ) − Zm+1 (ζn )};
dζn 2
Zm (ζn ) = Jm (ζn ), Ym (ζn ), Km (ζn ), Hm
(1)
(ζn ). (2.26g)
The solution to (2.25) is given by the real part of the following eigenseries expansion:
Wavemaker Theories 39
the wavemaker surface (middle segment driven) or of the derivative of the wave-
maker surface (joint driven). A dimensional scalar spatial velocity potential ϕ(x, y, z)
may be defined by the real part of
The dimensional linear fluid dynamic pressure field p(x, y, z, t) and 3D fluid velocity
vector field may be computed from
∂Φ(x, y, z, t)
p(x, y, z, t) = ρ , (2.28b)
∂t
q(x, y, z, t) = −∇3 Φ(x, y, z, t), (2.28c)
∂ϕ(x, y, 0)
− ko ϕ = 0; x ≥ 0; −B ≤ y ≤ +B; z = 0, (2.29b)
∂z
∂ϕ
= 0; x ≥ 0; y = ±B; −h(x, y) ≤ z ≤ 0, (2.29c)
∂y
x = 0,
∂Φ(x, y, z, t) ∂ξ(y, z, t)
=− ; −B ≤ y ≤ +B, (2.29d)
∂x ∂t
−h(0, y) ≤ z ≤ 0,
∂
lim − iKn ϕ(x, y, z) = 0, (2.29e)
x→+∞ ∂x
∂ϕ(x, y, z)
= −∇2 ϕ(x, y, z) · ∇2 h(x, y); z = −h(x, y), (2.29f)
∂z
2
∂ ∂2
∇22 (•) = , (•), (2.29g)
∂x2 ∂y 2
Û(y, z)
ξ(y, z, t) = Re i [∆U (y, a)][∆U (z, b, d)] exp −i(ωt + ν) , (2.29h)
ω
Sω
Û (y, z) = Γ(y)χ(z/ho ), (2.29i)
∆/ho
∆U (y, a) = U (y + a− ) − U (y − a+ ),
(2.29j)
∆U (z, b, d) = U (z + h − d) − U (z + b),
July 31, 2009 8:18 9.75in x 6.5in b684-ch02 FA
Ψ1 (k, z/h) ; n = 1, (2.30b)
Υn (Kn , z/h) = Ψ1 (k, 0)
Ψ1 (κn , z/h); n ≥ 2, (2.30c)
ko h = Kn h tanh Kn h = 0; n = 1, 2, 3, . . . , (2.30d)
Alternatively, for wave basins with mildly sloping bottoms, the mild slope equation
may be applied according to1
Wavemaker Theories 41
where the wave group velocity CG is given by (2.16e). If the product CCG is a
constant, (2.32) reduces to the 2D Helmholtz equation (2.31). Applying the WKBJ
approximation1 for the x-dependent solution in the method of separation of variables
to (2.32) yields the following solution1 :
g
Φ(x, y, z, t) = Re i ζ(x, y)Υn (Kn , z/h) exp −i(ωt + ν)
ω
M
g
[C(x)CG (x)]x=0
i Amn Ξm (µm , y/B)
ω m=0 n=1 C(x)CG (x)
= Re
x
× Υn(Kn , z/h) exp i Qmn dξ exp − i(ωt + ν)
(2.33a)
cos µm B(y/B − 1) mπ
Ξm (µm , y/B) = ; Mm = 1 + δm0 ; µm = (2.33b–d)
Mm 2B
Qmn = Kn2 − µ2m ; m > 0; K1 = k > µ2m . (2.33e)
Amn
k S
= −Ψn (Kn , 0)In (α, β, b, d, k)
Qmn (∆/hx0 )
+
a+ /B
× Γj (qj , y/B)Ξm (µm , y/B)d(y/B). (2.34)
− a /B
−
+∞
Γj (qj , y/B) = c̃j exp i[qj B(y/B + νy )], (2.35a)
j=−∞
where the coefficients c̃j may be computed from the integral in (2.34) by
+
a+ /B
cmj = Γj (qj , y/B)Ξm (µm , y/B)d(y/B)
− a /B
−
If the full-draft piston snake wavemaker spans the entire width of the wave basin
so that a± = ±B, then (2.35b) reduces to the integral in (2.34) and
sin[qj B(νj − 1)] + (−1)m sin[qj B(νj + 1)]
4qj B
−i{cos[qj B(νj − 1)] + (−1)m cos[qj B(νj + 1)]}
cmj = . (2.35c)
((qj B − (mπ)2 )2 )(1 + δm0 )
Values for cmj for (possibly nonsymmetric) values for a± are given by Hudspeth.1
A long rectangular wave channel with a horizontal flat bottom, two rigid vertical
sidewalls, and a wavemaker may generate either 2D, long-crested progressive waves
or two types of transverse waves, viz.,
(1) sloshing waves that are excited directly by transverse motion of the wavemaker
or
(2) cross waves that are excited parametrically by the progressive waves at a
sub-harmonic of the wavemaker frequency.
The WMBVP for 3D sloshing waves is identical to (2.5) for planar 2D wave-
makers except for the KWMBC at x = 0 and an additional kinematic boundary
condition on the sidewalls of the 2D wave channel at y = ±B in Fig. 2.4. The
kinematic and dynamic wave fields may be computed from a dimensional 3D scalar
Wavemaker Theories 43
velocity potential Φ(x, y, z, t). The fluid velocity q(x, y, z, t) may be computed from
the negative directional derivative of a scalar velocity potential by
q(x, y, z, t) = −∇Φ(x, y, z, t). (2.36a)
The dimensional fluid dynamic pressure field p(x, y, z, t) may be computed from
∂Φ(x, y, z, t)
p(x, y, z, t) = ρ . (2.36b)
∂t
A spatial velocity potential ϕ(x, y, z) may be defined by the real part of
Φ(x, y, z, t) = Re{ϕ(x, y, z) exp −i(ωt + ν)}. (2.36c)
The WMBVP for sloshing waves is given by the following:
∇2 ϕ(x, y, z) = 0; x ≥ 0; −B ≤ y ≤ +B; −h ≤ z ≤ 0, (2.37a)
∂ϕ(x, y, −h)
= 0; x ≥ 0; −B ≤ y ≤ +B; z = −h, (2.37b)
∂z
∂ϕ(x, y, 0)
− ko ϕ = 0; x ≥ 0; −B ≤ y ≤ +B; z = 0, (2.37c)
∂z
∂Φ(x, y, z, t) ∂ξ(y, z, t)
=− ; x = 0; −B ≤ y ≤ +B; −h ≤ z ≤ 0, (2.37d)
∂x ∂t
∂ϕ
= 0; x ≥ 0; y = ±B; −h ≤ z ≤ 0, (2.37e)
∂y
∂
lim − iKn ϕ(x, y, z) = 0, (2.37f)
x→+∞ ∂x
x ≥ 0,
−iω
η(x, y, t) = Re Φ(x, y, 0, t) ; −B ≤ y ≤ +B, (2.37g)
g
z = 0.
A solution to (2.37) is given by the following eigenfunction expansions1 :
∂ζn (x, y)
= 0; x ≥ 0; y = ±B; −h ≤ z ≤ 0, (2.40d)
∂y
where (2.40a) is the 2D Helmholtz equation.9,10
Because the boundary conditions are prescribed on boundaries that are constant
values of (y,z), a solution to the WMBVP (2.40) may be computed by the method
of separation of variables and is given by1
Φ(x, y, z, t)
M
Ψ1 (k, z/h)
Cm1 Υm (y/B) exp iPm1 x
m=0
Ψ1 (k, 0)
Ψ1 (k, z/h)
+ Cm1 Υm (y/B) exp
g Ψ (k, 0)
1
m=M+1
= Re i exp −i(ωt + ν) ,
ω
− Ξm1 x
Ψn(κn , z/h)
+ Cmn Υm (y/B) exp
,
m=0 n=2
Ψ n (κ n 0)
− Qmn x
(2.41a)
where
cos µm B(y/B − 1)
Υm (µm , y/B) = ; Mm
2
= 1 + δm0 ; (2.41b)
Mm
mπ
µm = ,
2B
July 31, 2009 8:18 9.75in x 6.5in b684-ch02 FA
Wavemaker Theories 45
Pm1 = K12 − µ2m = k 2 − µ2m ; k > µm ; m ≤ M, (2.41c)
Ξm1 = µ2m − k 2 ; k < µm ; m > M, (2.41d)
n = 1: k = µm : Qm1 = 0
k < µm : Qm1 = µ2m − k 2 = Ξm1 > 0; m > M, (2.41g)
n ≥ 2 : Kn = iκn : Qmn = µ2m + κ2n > 0, (2.41h)
where M is the maximum integer value for m in order for µm < k. Substituting
(2.41a) into (2.40b) yields the following solutions for Cmn 1 :
ω Ψ1 (Kn , 0) +1 y 0 z z z y
Cm1 = d d U y, Ψ1 K n , Υ m µm ;
g Pm1 −1 B −1 h h h B
m≤M (2.42a)
z z
ω Ψ1 (Kn , 0) +1 y 0
z y
Cm1 = −i d d U y, Ψ1 K n , Υm µm , ;
g Ξm1 −1 B −1 h h h B
m>M +1 (2.42b)
z z
ω Ψn (Kn , 0) +1 y 0 z y
Cmn = −i d d U y, Ψn K n , Υ m µm , ;
g Qmn −1 B −1 h h h B
m ≥ 0, n ≥ 2. (2.42c)
form that may be treated as a steady flow following a Galilean transformation from a
fixed inertial coordinate system to a noninertial moving coordinate system. Domain
mapping is a transformation of the wavemaker geometry into a fixed computational
domain where a solution may be computed efficiently.
∂Φ(x, y, t)
q(x, y, t) = −∇Φ(x, y, t); p(x, y, t) = ρ . (2.43a,b)
∂t
The WMBVP and Type I wavemaker shape function are given by (2.5)–(2.7).
There are both Irregular (I) and Regular (R) points at the intersections between
the Smooth (S) and Critical (C) boundaries B1 and B2 in the WMBVP as illustrated
in Fig. 2.6 where these two boundary intersection points are identified as P1 and P2 .
The classification of the boundary points P1 and P2 in Fig. 2.6 depends on (1) the
boundary conditions ϕi (Pj ) and (2) the continuity of the boundaries Bm and their
derivatives where i, j, and m = 1 or 2. A conformal mapping of the semi-infinite
wave channel strip in the physical plane will yield a Fredholm integral equation,6,7
where these critical points may be transformed to singular points that are integrable
over a smooth continuous mapped boundary.
Wavemaker Theories 47
Fig. 2.6. Combinations of Irregular (I) and Regular (R) boundary points P1 and P2 between
Smooth (S) and Critical (C) boundaries B1 and B2 intersections in the WMBVP.8
Fig. 2.7. Mapping of the semi-infinite strip in the lower half x–y-plane in the physical z-plane to
the upper half X–Y -plane in the Z-plane.8
July 31, 2009 8:18 9.75in x 6.5in b684-ch02 FA
Fig. 2.8. Mapping of the upper half-plane in the Z-plane to the unit disk in the Q-plane.
dz C1
=√ √ (2.44a)
dZ Z +1 Z −1
z = x + iy
h
= − Log[−Z − Z 2 − 1], (2.44b)
π
where the Log[•] denotes the principal value of Log[•] and the argument of the
Log[•] is −π ≤ arg < π. Inverting (2.44b) yields
Z = X + iY = − cosh(πz/h), (2.44c)
where
! πx " ! πy "
X = − cosh cos , (2.44d)
h h
! πx " ! πy "
Y = − sinh sin . (2.44e)
h h
July 31, 2009 8:18 9.75in x 6.5in b684-ch02 FA
Wavemaker Theories 49
1 − X2 − Y 2
ξ= ,
X 2 + (Y + 1)2
(2.45c,d)
2X
ζ= ,
X 2 + (Y = 1)2
that may be transformed into the cylindrical coordinates for the unit disk in
Fig. 2.8 by
(X 2 + Y 2 − 1)2 − 4X 2 ζ 2X
r2 = ξ 2 +ζ 2 = ; = arctan = arctan .
(X 2 + (Y + 1)2 )2 ξ 1 − X2 − Y 2
(2.45e,f)
Details of the transformation are given by Hudspeth.1 A numerical solution to
the transformed WMBVP may be computed from the following Fredholm integral
equation1 :
+π
2π ∂G(r, r̂, , )
ˆ ∂ Φ̂(r̂, )
ˆ
π Φ̂(r, ) = − −π Φ̂(r̂, )ˆ
∂r̂
+ G(r, r̂, , )
ˆ
∂r̂
r̂d,
ˆ
(2.46a)
where
G(r, r̂, , )
ˆ = − ln[ρ(r, r̂, , )];
ˆ ρ2 (r, r̂, , )
ˆ = r2 − 2rr̂ cos(
ˆ − ) + r̂2 .
(2.46b,c)
July 31, 2009 8:18 9.75in x 6.5in b684-ch02 FA
Fig. 2.9. Nodal points on the unit disk in the Q-plane and the corresponding nodal points on the
wavemaker in the physical z-plane.8
z = x + iy = iz = −y + ix. (2.48a)
July 31, 2009 8:18 9.75in x 6.5in b684-ch02 FA
Wavemaker Theories 51
Fig. 2.10. WMBVP11 with the six critical boundary points at a–a0 –b–b0 –c–d.
Fig. 2.11. Rotation and translation of the physical wavemaker rectangular strip in the z-plane to
the w-plane.11
In order to map the WM geometry in the z-plane to the semi-circle in the Z-plane
shown in Fig. 2.12 as a Jacobian elliptic function, the rotated and translated strip
in the z -plane must have the dimensions of −K ≤ ξ ≤ +K and 0 ≤ ζ ≤ K , where
K = h/2 and K = 3h = 6K. This requires a coordinate amplification given by
2K
w= (x + iy )
h
2K h
= −y − + ix . (2.48c)
h 2
July 31, 2009 8:18 9.75in x 6.5in b684-ch02 FA
Fig. 2.12. Mapping of the wavemaker semi-circle in the Z-plane to the unit disk in the Q-plane.11
ĈkdZ
dw = . (2.48d)
(a − Z)(b − Z)(c − Z)(d − Z)
Z = X + iY
# $ % & ! "
sn − 2K h y + h2 , κ dn 2Kx h , k̂
! "
1 − dn2 #− 2K $y + h % , κ& sn2 2Kx , k̂
=
h 2 h
# $ % & # $ % & ! " ! "
cn − 2K h h
h y + 2 , κ dn − 2K
h y + 2 , κ , sn 2Kx
h , k̂ cn 2Kx
h , k̂
+i # $ % & ! " ,
1 − dn2 − 2K y + h2 , κ sn2 2Kx
h h , k̂
(2.48g)
where sn[•, •] in the copolar half-period trio in (2.48g) is defined in (2.48e), and
cn[•, •] and dn[•, •] are defined by cn 2 [•, •] = 1−sn 2 [•, •], dn 2 [•, •] = 1−κ2 sn2 [•, •].
The mapping from the semi-circle in the Z-plane to the unit disk in the Q-plane
July 31, 2009 8:18 9.75in x 6.5in b684-ch02 FA
Wavemaker Theories 53
Fig. 2.13. Transformed boundary conditions mapped to arcs on the perimeter of the unit disk.11
is shown in Fig. 2.12; and the mapping to the unit disk in the Q-plane is shown in
Fig. 2.13. The mapping of the Z-plane to the Q-plane is given by
the unit disk may be transformed into functions of the copolar trio of Jacobian
elliptic functions. The transformed WMBVP in circular cylindrical coordinates
is given by Hudspeth.1 A general solution to the transformed WMBVP may be
written as10
N
n ân
ϕ(R, θ) = R cos nθ + b̂n sin nθ , (2.50)
n=0
1 + δn0
where δij is the Kronecker delta function. Substituting (2.50) into the generic
boundary conditions on each of the six arcs on the perimeter of the unit disk illus-
trated in Fig. 2.13, multiplying each of these six boundary conditions by a member
of the set of the orthogonal series in (2.50), integrating over the interval of orthogo-
nality −π ≤ θ ≤ +π yields the following matrix equation for each of the coefficients
ân and b̂n
AB = H. (2.51)
July 31, 2009 8:18 9.75in x 6.5in b684-ch02 FA
Domain mapping of the WMBVP12 follows the theory by Joseph.13 The physical
fluid domain shown in Fig. 2.14 for the fully nonlinear WMBVP is mapped to a
fixed computational fluid domain, and the discretized coupled free-surface boundary
conditions are computed by an implicit Crank–Nicholson (C–N) method.14,15 At
each iteration of the C–N method, the potential field is computed by the conjugate
gradient method.15 The wavemaker motion Ξ(y/h, t) is assumed to be periodic with
period T = 2π/ω, and the WMBVP with the surface tension T̂ is given by
0 ≤ y ≤ Γ(x, t),
∇ Φ(x, y, t) = ∆Φ(x, y, t) = 0;
2
(2.52a)
Ξ(y/h, t) ≤ x ≤ L.
T̂ ∂ 2 Γ(x, t)
∂Φ(x, y, t) 1 ρ ∂x2
+ |∇Φ(x, y, t)|2 − + gΓ(x, t) = 0.
2 3/2
(2.52b)
∂t 2
∂Γ(x, t)
1+
∂x
∂Φ(x, y, t)
= 0; Ξ(Γ(x, t), t) ≤ x ≤ L; y = 0. (2.52d)
∂y
∂Φ(x, y, t)
= 0; x = L; 0 ≤ y ≤ Γ(L, t). (2.52e)
∂x
∂Φ(x, y, t) ∂Ξ(y/h, t) ∂Ξ(y/h, t) ∂Φ(x, y, t) x = Ξ(y/h, t),
=− + ; (2.52f)
∂x ∂t ∂y ∂y 0 ≤ y ≤ Γ(0, t).
July 31, 2009 8:18 9.75in x 6.5in b684-ch02 FA
Wavemaker Theories 55
The physical fluid domain shown in Fig. 2.14 may be mapped into a dimensionless
fixed rectangle of dimensions 0 ≤ ξ ≤ 1 by 0 ≤ ζ ≤ 1 by the transforms
x y Γ(x, t)
ξ= ; ζ= ; τ = ωt; γ(ξ, τ ) = , (2.53a–d)
L Γ(x, t) h
and dimensionless variables by
∇Φ(x, y, t) P (x, y, t)
q(ξ, ζ, τ ) = − ; p(ξ, ζ, τ ) = , (2.53e,f)
Aω ρA2 ω 2
Φ(x, y, t) Ξ(y/h, t) T̃
ϕ(ξ, ζ, τ ) = ; w(ζ, τ ) = ; T̂ = . (2.53g–i)
Ahω S ρALhω 2
Because ζ is a function of both x and y in (2.53b), transforming partial derivatives
with respect to x must be done with some care.12 Details of these lengthy transfor-
mations and the transformed WMBVP in the fixed mapped domain are given by
Hudspeth.1
References
1. R. T. Hudspeth, Waves and Wave Forces on Coastal and Ocean Structures (World
Scientific, Singapore, 2006).
2. R. T. Hudspeth, J. M. Grassa, J. R. Medina and J. Lozano, J. Hydraulic Res. 32, 387
(1994).
3. F. John, Commun. Pure Appl. Math. 2, 13 (1949).
4. F. John, Commun. Pure Appl. Math. 3, 45 (1950).
5. T. H. Havelock, Phil. Mag. 8, 569 (1929).
6. P. R. Garabedian, Partial Differential Equations (Wiley, Inc., New York, 1964).
7. P. M. Morse and H. Feshbach, Methods of Theoretical Physics (McGraw-Hill Book
Company, New York, 1953).
8. Y. Tanaka, Irregular points in wavemaker boundary value problem, PhD thesis,
Oregon State University (1988).
9. G. F. Carrier, M. Krook and C. E. Pearson, Functions of a Complex Variable, Theory
and Technique (McGraw-Hill Book Co. Inc., New York, 1966).
10. R. B. Guenther and J. Lee, Partial Differential Equations of Mathematical Physics
and Integral Equations (Dover Publications, Inc., New York, 1996).
11. P. J. Averbeck, The boundary value problem for the rectangular wavemaker,
MS thesis, Oregon State University (1993).
12. S. J. DeSilva, R. B. Guenther and R. T. Hudspeth, Appl. Ocean Res. 18, 293 (1996).
July 31, 2009 8:18 9.75in x 6.5in b684-ch02 FA
Chapter 3
Ronald B. Guenther
Department of Mathematics
Oregon State University, Corvallis, OR 97331, USA
guenther@math.orst.edu
Robert T. Hudspeth
School of Civil and Construction Engineering
Oregon State University, Corvallis, OR 97331, USA
robert.hudspeth@orst.edu
57
July 31, 2009 8:18 9.75in x 6.5in b684-ch03 FA
operation and the GMM. The system of nonlinear nonautonomous evolution equa-
tions determined from Hamilton’s equations of motion of the second kind must
be averaged in order to obtain an autonomous system that may be analyzed by
the GMM. Hyperbolic saddle points that are connected by heteroclinic orbits
are computed from the unperturbed autonomous system. The nondissipative per-
turbed Hamiltonian system with surface tension satisfies the KAM nondegeneracy
requirements; and the Melnikov integral is calculated to demonstrate that the
motion is chaotic. For the perturbed dissipative system with surface tension, the
only hyperbolic fixed point that survives the averaged equations is a fixed point
of weak chaos that is not connected by a homoclinic orbit; and, consequently, the
Melnikov integral is identically zero. The chaotic motion for the perturbed dis-
sipative system with surface tension is demonstrated by numerical computation
of positive Liapunov characteristic exponents. A chaos diagram of the largest
Liapunov exponent demonstrates regions in the Floquet forcing parameter space
of possible chaotic motions; and the range of values of the Floquet parametric
forcing parameter ε and of the wavemaker dissipation parameter α in the α−ε
space where chaotic motions may exist.1,3
3.1. Introduction
The cross waves shown in Fig. 3.1 are excited parametrically by progressive wave-
maker waves at a sub-harmonic of the wavemaker frequency.1 Parametrically excited
standing cross waves that oscillate in a direction transverse to the wavemaker
forcing with crests perpendicular to the wavemaker are analyzed by the gener-
alized Melnikov method (GMM) and by the Generalized Herglotz Algorithm (GHA)
extended to nonautonomous systems. The cross wave wavelengths and wave modes
possible are determined by the channel width Lc = 2(channel width)/n, where
n = the mode number and is equal to the number of half-wavelengths across the
channel. Floquet instabilities occur when the cross wave wavelengths also satisfy
the frequency dispersion equation for surface gravity waves. Energy is transferred
from the progressive wavemaker waves to the parametrically forced cross waves
through the spatial mean motion of the free surface and their growth is due to
the rate of working of the transverse stresses of the progressive wavemaker waves.
The cross wave instability is shown to be homomorphic to the Floquet oscil-
lator instability2 by constructing the neutral Floquet stability diagram shown in
Fig. 3.2. The simultaneous generation of a primary resonance (ωp : ωc = 2 : 1)
and a secondary resonance (ωp : ωc = 1 : 1) may be observed in the cross wave
data. The criteria for horseshoe maps and for homoclinic/heteroclinic orbits are
selected to test whether or not cross waves are a chaotic dynamical system. Specif-
ically, the GMM provides local criterion for the transverse intersection of stable
and unstable manifolds of the perturbed system and for the resulting chaotic
motion near the unperturbed (undamped and unforced) homoclinic/heteroclinic
orbits. In order to apply the GMM to a suspended dynamical system that will
survive the KAM averaging operation, the nonautonomous Hamiltonian is trans-
formed by three separate canonical transformations by applying the GHA for nonau-
tonomous transformations. Three canonical transformations are applied in order
to (1) eliminate cross product terms by a rotation of axes; (2) to eliminate two
degrees of freedom, and (3) to suspend the nonautonomous terms in the wave-
maker forcing component of the Hamiltonian by applying the Hamilton–Jacobian
transformation. The Liapunov characteristic exponents represent an alternative cri-
terion for diagnosing the chaotic behavior of a dynamical system by measuring
the mean rate of exponential divergence of nearby trajectories, and is computed
numerically for the perturbed dissipative Hamiltonian when the GMM fails to
predict chaos.
Fig. 3.2. Neutral Floquet stability diagram for mode 2 cross waves (o = mode 2 cross waves and
= no cross waves).1,2
July 31, 2009 8:18 9.75in x 6.5in b684-ch03 FA
The fluid is assumed to be incompressible and inviscid and the flow to be irrota-
tional. The dimensional fluid particle velocities u and the dimensional total pressure
in the fluid P are computed from
P 1 ¯ 2
u = − ∇
¯ ϕ ; = −g z + ϕ t − |∇ ϕ| ; (3.1a,b)
ρ 2
where ϕ = a velocity potential; ρ = the fluid mass density; and g = the gravita-
tional acceleration. The fluid domain is the 3D rectangular wave channel shown in
Fig. 3.3.
1 ¯ 2 (ζ − 1)
L = |∇ ϕ | − ϕt + g z dV + T dSη ; (3.2a)
V (t ) 2 S ζ
η
dSη = ζ dx dy ; ζ = [1 + |∇
¯ ϕ |2 ]1/2 . (3.2b,c)
where the generalized dissipation function per unit mass density F (Dη /Dt ) is
Dη g Dη δη
F = −α dSη ; (3.3b)
Dt κ Sη Dt ζ
The velocity potential ϕ is the field variable, and the free surface η may be
expressed in terms of ϕ by the free surface boundary conditions and the contact
line conditions at the vertical sidewalls.1,3 The velocity potential ϕ and the free
surface displacement η are assumed to be linear combinations of the progressive
wavemaker wave and the parametrically excited cross wave given by the following:
x y z t
x = ; y = ; z = ; t = √ ;
k κ κ g κ (3.7a–g)
h=κh; ξ=k; b=κb;
ωp2 = g k τλ ; τλ = 1 + (τ /λ4r ); τ = T κ 2 /g ;
(3.7h–m)
λ2r = κ /k ; ω c = g κ τ12 ;
2
τ12 = 1 + τ ;
July 31, 2009 8:18 9.75in x 6.5in b684-ch03 FA
g g
ϕp = ϕp ap ; ϕc = ϕc ac ; ηp = ap ηp ;
k κ
(3.7n–s)
La2 g
ηc = ac ηc ;
χ = aw χ; L = c .
kκ
Scaling by (3.7) yields the following dimensionless scaling parameters1,3:
ε = κ ac ; β = ωc /ωp ; γ = k aw ; Γ = ap /ac ; (3.8a–d)
and the following dimensionless field variables:
ϕ(x, y, z, t) = ϕc (y, z, t) + Γλr ϕp (x, z, t);
(3.9a,b)
η(x, y, t) = ηc (y, t) + Γηp (x, t).
The independent variation of ϕ and η vanish at arbitrary temporal values t1 and t2
in Hamilton’s principle (3.3) yielding the following scaled boundary value problem:
1
ϕxx + ϕyy + ϕzz = 0; γχ ≤ x ≤ ξ, y ≤ |b|, −h ≤ z ≤ εη, (3.10a)
λ4r
1
ϕz = −τ1 ηt + ε ϕ η
x x + ϕ η
y y ; z = εη, γχ ≤ x ≤ ξ, y ≤ |b|,
λ4r
(3.10b)
ε 1 2 1
ϕ + ϕ2y + ϕ2z − τ1 ϕt + η = τ ηxx + ηyy − αϕz ;
2 λ4r x λ4r
γχ ≤ x ≤ ξ,
(3.10c)
y ≤ |b|, z = εη
y = |b|, −h ≤ z ≤ εη, γχ ≤ x ≤ ξ
∇ϕ · n̂ = 0; (3.10d,e)
z = −h, y ≤ |b|, γχ ≤ x ≤ ξ,
γ
ϕx = − λ4r τ1 χt + γλ4r ϕz χz ; x = γχ, y ≤ |b|, −h ≤ z ≤ εη, (3.10f)
ε
plus an appropriate radiation condition at x = ξ. The free surface curvature requires
a dynamical constraint that is given either by the contact line condition ηn = 0 or
by the edge constraint boundary condition η = 0.1,3 The wavemaker perturbation
forcing parameter γ is smaller than the Floquet parametric forcing parameter ε
because experiments demonstrate that the standing cross wave amplitude becomes
larger than the wavemaker forcing amplitude as t → ∞. The parameter ordering is
γ2 γ
0 < γ 2 < εγ < ε2 < γ < ε < 1 or 0< < γ < < 1. (3.11a,b)
ε ε
The higher order terms that will be neglected are O(ε2 ), O(εγ), and O(γ 2 ); but
terms O(γ/ε) will be retained. Neglecting these higher order terms, the nondi-
mensional Lagrangian integrals may be approximated by the following linear
July 31, 2009 8:18 9.75in x 6.5in b684-ch03 FA
combination1,3 :
b
γτ1 0
Lχ = − [(ϕc + Γλr ϕp )χt ]x=0 dzdy
2ε −b −h
b
+ [εχt (ηc + Γηp )(ϕc + Γλr ϕp )]x,z=0 dy, (3.12a)
−b
b 0
Γ
Lξ = − 3 [ϕpx (ϕc + Γλr ϕp )]x=ξ dzdy
2λr −b −h
b
+ [εϕpx (ηc + Γηp )(ϕc + Γλr ϕp )]x=ξ,z=0 dy, (3.12b)
−b
b ξ
1
Lη = {τ1 (ηc + Γηp )t [(ϕc + Γλr ϕp )
2 −b 0
and the deepwater cross wave and wavemaker wave are approximated by
where the variables q(t), Q1 (t), and Q2 (t) are the generalized coordinates. The free-
surface displacement η is a solution to the linearized inhomogeneous boundary value
problem given by
where (3.15d and 3.15e) are the contact line conditions and where
τ
Lpη ηp = ηp − ηp , fp = αλr ϕpz − τ1 λr ϕpt ; z = 0, (3.15f,g)
λ4r xx
Lcη ηc = τ ηcyy − ηc , fc = αϕcz − τ1 ϕct ; z = 0. (3.15h,i)
1
ηc = (−qα + τ1 q̇) cos(y − b). (3.16)
τ12
where
bΓ2
− [ξ(Q 2
(t) + Q22 (t)) + Q1 (t)Q2 (t)]
2 1
L0 = bτ 2 2
Γ
+ 1
[ξλ4
τ
r λ (Q̇2
(t) + Q̇2
(t)) + 4τ Q̇ 1 (t)Q̇ 2 (t)]
; (3.18b)
2λ2r τλ2 1 2
bξ 2
− (q (t) − q̇ (t)) 2
2
εbΓτ1 τ εbΓ
Lε = − 3 q 2 Q̇2 − q q̇Q2 ; (3.18c)
2λr τλ 2τ1 λ3r
t Γλr τ1 Q1 Γ2 τ12 √ q q̇
Lγ = −γb cos f1 + Q1 (λr Q̇1 + τ Q̇2 ) +
2
β εβ βτλ 2β
t Γ2 τ 2 √ −q̇ 2 + q2
+ γb sin Γ2 Q21 − 2 12 (λ4r Q̇21 + λ2r τ Q̇1 Q̇2 + τ Q̇22 ) + ,
β λr τλ 2
(3.18d)
∂(L0 + Lε )
pi = ; i = 1, 2, 3, (3.20a)
∂ q̇i
εbΓ
p1 = p = bξ q̇ − qQ2 ; (3.20b)
2τ1 λ3r
bτ12 Γ2 λ2r2τ
p2 = P1 = ξ Q̇1 + 4 Q̇2 ; (3.20c)
τλ λr τλ
εbτ1 Γτ 2 bτ1 Γ λr
2 2 2
2τ
p3 = P2 = − 3 q + ξ Q̇2 + 4 Q̇1 . (3.20d)
2λr τλ τλ λr τλ
p εΓqQ2
q̇1 = + ; (3.21a)
bξ 2τ1 ξλ3r
λ4r τλ P1 − ξ P2
2τ
ετ 2 q 2
Q̇1 = − ; (3.21b)
bτ1 Γ2 λ6r ξ
2 Γλ9r ξ 2 τ1 τλ
λ4r τλ P2 − ξ P1
2τ
ετ q 2
Q̇2 = + . (3.21c)
bξτ1 Γ λ6r
2 2 2Γξλ5r τ1
Substituting (3.18), (3.20), and (3.21) into (3.19) yields the following Hamiltonian:
bΓ2 1 p2
H0 = (ξ(Q21 + Q22 ) + Q1 Q2 ) + bξq 2 +
2 2 2bξ
(P12 + P22 ) 2τ P1 P2
+ − 2 2 6 2; (3.22b)
2bβ 2 Γ2 ξ bτ1 Γ λr ξ
εΓ ετ ετ 2
Hε = qpQ2 + q 2
P2 − q 2 P1 ; (3.22c)
2τ1 ξλ3r 2τ1 Γξλ5r τ1 Γξ 2 λ9r τλ
July 31, 2009 8:18 9.75in x 6.5in b684-ch03 FA
√
bΓλr τ1 Q1 Q1 (λ2r P1 + τ P2 )
f1 +
t εβ βξλ2r
γ
cos √
β 2τ Q1 (λr P2 + τ P1 )
2
pq
Hγ = − + . (3.22d)
βξ λr τλ
2 6 2βξ
√
t (λ2r P1 + τ P2 )2 1 2 p2
−γ sin bΓ Q1 −
2 2
+ bq −
β bτ1 Γ ξ λr
2 2 2 6
2 2bξ 2
t2 t2
3 t2 b ξ
δ L dt = Di δqi dt = − (αϕz δη)dxdydt, (3.23a)
t1 t1 i=1 t1 −b 0
3
∂η ∂η
δη = δqi + δ q̇i . (3.23b)
i=1
∂qi ∂ q̇i
b ξ
∂η ∂ ∂η
Di = α −ϕz + ϕz dxdy; i = 1, 2, 3, (3.24a)
−b 0 ∂qi ∂t ∂ q̇i
2
α αεbΓ α bξ
D1 = p+ qQ2 + q, (3.24b)
τ1 2τ12 λ3r τ12
2
α 2ατ αεbΓτ 2 α bξΓ2
D2 = P1 − P2 − q +
2
Q1 , (3.24c)
λ2r τ1 ξλ6r τ1 τλ λ9r ξτλ λ2r τλ
2 2
α αεbΓτ 2α bτ Γ2 α bξΓ2
D3 = P2 + q +
2
Q1 + Q2 . (3.24d)
τ1 λ2r 2λ5r τλ λ6r τλ2 λ2r τλ
July 31, 2009 8:18 9.75in x 6.5in b684-ch03 FA
∂H ∂H
q̇i = ; ṗi = − − Di ; i = 1, 2, . . . , N. (3.25a,b)
∂pi ∂qi
∂F (u, U, t)
K(Q, P, t) = H[q(Q, P, t), p(Q, P, t), t] + ; (3.26a)
∂t
N
∂K ∂qj
Q̇i = + D̃j ; i = 1, 2, . . . , N, (3.26b)
∂Pi j=1 ∂Pi
N
∂K ∂qj
Ṗi = − − D̃j ; i = 1, 2, . . . , N, (3.26c)
∂Qi j=1 ∂Qi
(1) Rotation of axes. The cross product terms in (3.22b) may be eliminated by
that satisfy the Poisson bracket requirements in Appendix A. The rotation angle
θ = π/4 eliminates the cross product terms, and the transformed Hamiltonian is
given in component form by
H → H̃ = H̃0 + H̃ε + H̃γ (t) + O(ε2 , εγ, γ 2 ), (3.28a)
where
2
bξ 2 p̃2 bΓ ξ 2 P̃12
q̃ + + Q̃1 +
2 2bξ 2 2bβ Γ ξ
2 2
bΓ ξ 2
2
P̃22
H̃0 = + Q̃2 + , (3.28b)
2 2bβ 2 Γ2 ξ
bΓ 2
τ (P̃ 2
− P̃ 2
)
+ (Q̃2 − Q̃1 ) +
2 2 1 2
4 bτ1 Γ2 λ6r ξ 2
2
εΓ ετ (2τ + ξλ4r τλ ) 2
H̃ε = √ q̃ p̃(Q̃2 − Q̃1 ) − √ q̃ P̃1
2 2τ1 ξλ3r 2 2τ1 Γξ 2 λ9r τλ
ετ (−2τ + ξλ4r τλ ) 2
+ √ q̃ P̃2 , (3.28c)
2 2τ1 Γξ 2 λ9r τλ
bΓλr τ1 (Q̃1 + Q̃2 )
f √
1
2εβ
√ √
t
( Q̃ + Q̃ )((λ 2
r − τ )P̃ + (λ2
r + τ )P̃ )
γ cos +
1 2 1 2
β 2βξλ 2
r
√ √
τ (Q̃1 + Q̃2 )((λr + τ )P̃2 + (−λr + τ )P̃1 )
2 2
p̃q̃
H̃γ = .
− +
βξ 2 λ6 τ
r λ 2βξ
√ √
bΓ 2
((λ2
− τ )P̃ + (λ2
+ τ )P̃ ) 2
(Q̃ + Q̃ )2
− r 1 r 2
t 2 1 2
2bτ1 Γ ξ λr
2 2 2 6
−γ sin
β p̃ 2
1
+ bq̃ − 2
2 2bξ 2
(3.28d)
The first three energy square brackets [•] in (3.28b) are the action variables in the
next canonical transformation.
(2) Action/Angle Transformation. The three new canonical variables are
given by
bξ 2 p̃2 bΓ2 ξ 2 P̃12
p̂ = q̃ + , P̂1 = Q̃1 + ,
2 2bξ 2 2bβ 2 Γ2 ξ
(3.29a–c)
bΓ2 ξ 2 P̃22
P̂2 = Q̃2 + ,
2 2bβ 2 Γ2 ξ
July 31, 2009 8:18 9.75in x 6.5in b684-ch03 FA
that satisfy the Poisson brackets. The Herglotz auxiliary functions Xi are the ratios
3
∂ p̃j ∂F
q̂i = q̃j + ; i = 1, 2, 3, (3.31a)
j=1
∂ p̂i ∂ p̂i
where
2P̂1 Q̂1 Q̂1
Q̃1 = sin , P̃1 = Γβ 2bξ P̂1 cos , (3.32c,d)
bξΓ2 β β
2P̂2 Q̂2 Q̂2
Q̃2 = sin , P̃2 = Γβ 2bξ P̂2 cos , (3.32e,f)
bξΓ2 β β
2β 2 τ Q̂1 2β 2 τ
Ĥ0 = [p̂ + P̂1 + P̂2 ] + P̂1 cos 2
−
λ6r ξτ12 β λ6r ξτ12
Q̂2 P̂1 Q̂1 P̂2 Q̂2
× P̂2 cos 2
− sin 2
+ sin 2
, (3.33b)
β 2ξ β 2ξ β
July 31, 2009 8:18 9.75in x 6.5in b684-ch03 FA
Q̂2 Q̂2
P̂2 cos 2q̂ −
− cos 2q̂ +
β β
εp̂
Ĥε =
4 bξ 3 λ3r τ1
Q̂1 Q̂1
− P̂1 cos 2q̂ − − cos 2q̂ +
β β
εβτ p̂
−
2 bξ 5 λ9r τ1 τλ
Q̂ 1 Q̂
(2τ + ξλr τλ ) P̂1 cos
4 1
− cos 2q̂ +
1
β 2 β
1 Q̂1 Q̂
2
× − cos 2q̂ − + (2τ − ξλr τλ ) P̂2 cos
4
,
2 β β
1 Q̂ 1 Q̂
− cos 2q̂ +
2
− cos 2q̂ −
2
2 β 2 β
(3.33c)
√
Q̂1 Q̂2 bλr f1 τ1
P̂1 sin + P̂2 sin √
β β βε ξ
√
(λ 2
+ τ )(−2τ + ξλ 4
τ ) Q̂
r r λ
P̂
1
+
λ ξ τ
1 cos
β
t
6
r
2
λ
γ cos
β √
(λ 2
r + τ )(−2τ + ξλ 4
r τλ ) Q̂
+ P̂ cos
2
λ 6 ξ2τ
λ
2
β
r
p̂ sin(2q̂)
Ĥγ = + .
2βξ
√
−β 2 Q̂1
(λ 2
r − τ ) P̂ 1 cos
λ6τ 2 β
r 1
2
√ Q̂
γ
t + (λ 2
+ τ ) P̂ cos
2
− r 2
β
ξ
sin
β
2
Q̂ Q̂
+ P̂ sin
1
+ P̂ sin
2
1
β
2
β
+ p̂(sin q̂ − cos q̂)
2 2
(3.33d)
July 31, 2009 8:18 9.75in x 6.5in b684-ch03 FA
The nonautonomous perturbed component in (3.33d) will not survive the KAM
averaging theorem, and some of the nonautonomous terms may be suspended
by applying the following Hamilton–Jacobian nonautonomous canonical transfor-
mation.
satisfy the Poisson brackets. The Herglotz auxiliary functions Xi = Xi (p, q) satisfy
the nonzero Jacobian condition given by
∂(q̂, X) 4β 2
= = 0. (3.35)
∂(p, q) qQ1 Q2
Solving (3.35) for qi = qi (p, X), substituting qi (p, X) into (3.34), and inverting
yields
X1 −X2
p= (2β q̂ − Q̂1 ); P1 = (t − Q̂1 );
2β 2β
(3.36a–c)
−X3
P2 = (2t + 2β q̂ − Q̂2 ).
2β
−1
= [X1 (2β q̂ − Q̂1 )2 + X2 (t − Q̂1 )2 + X3 (2t + 2β q̂ − Q̂2 )2 ]. (3.37)
4β 2
3
∂pj ∂F
p̂i = − qj − ; i = 1, 2, 3 (3.38a)
j=1
∂ q̂i ∂ q̂i
gives
1 1
p̂ = p − P2 , P̂1 = (P1 − p), P̂2 = P2 . (3.38b–d)
2β 2β
July 31, 2009 8:18 9.75in x 6.5in b684-ch03 FA
The transformations in (3.34) and (3.38) satisfy the Poisson brackets, and the first
square bracket term [•] in (3.33b) may be transformed to
1 P1
[p̂ + P̂1 + P̂2 ] = 1 − (p − P2 ) + . (3.39)
2β 2β
The primary Floquet resonance condition is, approximately, ωc − ωp ≈ O(ε) so that
a detuning parameter Ω may be defined by
ωp 1
1− = 1 − = εΩ; (3.40)
2ωc 2β
and the primary parametric Floquet resonance is ωp : ωc = 2:1. A generating
function is now given by
1 1 1
F (q̂, p, t) = p −q̂ + Q̂1 + P1 (t − Q̂1 ) + P2 (2t + 2β q̂ − Q̂2 ); (3.41)
2β 2β 2β
∂F
Ĥ − ⇒ H = H0 + H0 (t) + Hε + Hε (t)
∂t
+ Hγ + Hγ (t) + O(ε2 , εγ, γ 2 ), (3.42a)
where
1 βτ 1 1 βτ
H0 =− + (P1 − p) + − + − P2 , (3.42b)
8βξ 2ξλ6r τ12 β 8βξ 2ξλ6r τ12
√
ε(p − P2 ) P1 − p
Hε = εΩ(p − P2 ) + [(2βτ 2 − ξλ4r τλ (λ2r − βτ )) cos(2q)], (3.42c)
4 2bβξ 5 λ9r τ1 τλ
√
γ bf1 λr τ1 2β − 1
Hγ = P1 − p sin(2Q1 ) + γ
2ε 2ξβ 3 4βξ
variables (q, p) by
2 t
qorig = p − P2 sin q + Q1 + , (3.43a)
bξ 2β
t
porig = 2bξ p − P2 cos q + Q1 + , (3.43b)
2β
1 t
Q1orig = √ P1 − p sin 2Q1 +
Γ 2bβξ β
3t
+ P2 sin 2(q + Q1 + Q2 ) + , (3.43c)
β
bβξ t
P1orig = Γ P1 − p cos 2Q1 +
2 β
3t
+ P2 cos 2(q + Q1 + Q2 ) + , (3.43d)
β
1 t
Q2orig = √ − P1 − p sin 2Q1 +
Γ 2bβξ β
3t
+ P2 sin 2(q + Q1 + Q2 ) + , (3.43e)
β
bβξ t
P2orig = Γ − P1 − p cos 2Q1 +
2 β
3t
+ P2 cos 2(q + Q1 + Q2 ) + . (3.43f)
β
√
bβαΓ(2τ + λ4r ξτλ ) t
√ P − p cos 2Q +
τ τ λ 6 2ξ 1 1
β
1 λ r
√
bβαΓ(−2τ λ 4
ξτ
+
√ r λ ) 3t
+ P 2 cos 2(q + Q 1 + Q 2 ) +
τ τ λ
1 λ r
6 2ξ β
2Γεατ 2
t
D2 = − 9 2 2 (p − P2 ) sin 2(q + Q1 ) + ,
λr ξ τλ 2β
√
α 2
Γ bξ t
+ √ P − p sin 2Q +
λ2τ 1 1
β
r λ 2β
√
α 2
Γ bξ 3t
+ √ P 2 sin 2(q + Q 1 + Q2 ) +
λr τλ 2β
2 β
(3.44b)
t
√ − P − p cos 2Q +
1 1
β
αΓ
√ bβξ
2λ2τ
r 1
P Q Q
3t
+ 2 cos 2(q + 1 + 2 ) +
β
εαΓτ t
+ (p − P ) sin 2(q + Q ) +
D3 = λ 5 ξτ
r λ
2 1
2β ,
√
α2 Γ b(2τ − λ4r ξτλ ) t
+ √ P − p sin 2Q +
λ 6τ 2 2βξ
1 1
β
r λ
√
α 2
Γ b(2τ + λ 4
r ξτ λ ) 3t
+ √ P sin 2(q + Q + Q ) +
β
2 1 2
λr τλ 2βξ
6 2
(3.44c)
that are completely autonomous and will survive the KAM averaging operation.
Averaging H(q, p, t) over the dimensionless cross wave period 2π yields
H = H0 (p, P1 , P2 ) + Hε (q, p, P1 , P2 ) + Hγ (q, p, P1 , P2 , Q1 )
1 βτ 1 1 βτ
= − + (P 1 − p) + − + − P2
8βξ 2ξλ6r τ12 β 8βξ 2ξλ6r τ12
√
(p − P2 ) P1 − p
+ ε Ω(p − P2 ) + [(2βτ − ξλr τλ (λr − βτ )) cos(2q)]
2 4 2
4 2bβξ 5 λ9r τ1 τλ
√
bf1 λr τ1 2β − 1
+γ P1 − p sin(2Q1 ) + (P2 − p) sin 2(q + Q1 ) .
2ε 2ξβ 3 4βξ
(3.44d)
July 31, 2009 8:18 9.75in x 6.5in b684-ch03 FA
2P1 + P2 − 3p
q̇ = −a1 + εΩ + εa3 √ cos(2q), (3.46a)
P1 − p
July 31, 2009 8:18 9.75in x 6.5in b684-ch03 FA
ṗ = 4εa3 P1 − p(p − P2 ) sin(2q), (3.46b)
(p − P2 )
Q̇1 = a1 + εa3 √ cos(2q), (3.46e)
P1 − p
Q̇2 = a2 − εΩ − 2εa3 P1 − p cos(2q), (3.46f)
1 a1 − εΩ
q0 (P1 , P2 ) = cos−1 √ , (3.47c)
2 2εa3 P1 − P2
2
a1 − εΩ
P1 ≥ P2 + . (3.47d)
2εa3
The fixed point (q0 , p0 ) is a hyperbolic saddle point if the determinant satisfies
# #
# ∂ q̇ ∂ q̇ #
# #
# ∂q ∂p #
# # = (a1 − εΩ)2 − 4(P1 − P2 )ε2 a23 < 0. (3.48)
# ∂ ṗ ∂ ṗ #
# #
# ∂q ∂p #
(q=q0 ,p=p0 )
The symmetry properties of Hamiltonian systems imply that the stable and the
unstable manifolds of the hyperbolic saddle point (q0 , p0 ) have equal dimensions in
the full 6D phase space (q, p, P1 , P2 , Q1 , Q2 ) ∈ T1 × R1 × R2 × T2 . The unperturbed
system α = γ = 0 has a 4D (R2 × T2 ) normally hyperbolic invariant manifold given
by the union of the hyperbolic saddle points (q0 , p0 ) according to
The normally hyperbolic invariant manifold M has 5D (R1 ×R2 ×T2 ) stable manifold
W s (M) and unstable manifold W u (M) that coincide along the 5D heteroclinic
manifold H = W s (M) ∩ W u (M) − M, where W s (M) and W u (M) are the set of
July 31, 2009 8:18 9.75in x 6.5in b684-ch03 FA
initial conditions that approach the hyperbolic saddle points as t → ±∞ under the
action of the unperturbed flow.
Dynamics on M. The unperturbed vector field α = γ = 0 restricted to the
normally hyperbolic invariant manifold M may be determined from1,3
The unperturbed vector field α = γ = 0 restricted to M has the form of two degrees
of freedom, 4D phase space (T1 × R1 × T2 ), completely integrable Hamiltonian
system with the Hamiltonian given by the level energy surfaces
H(P1 , P2 ) =
H(q = q0 , p = p0 ; γ = 0) = a1 P1 + (a2 − a1 )P2 = E, (3.51)
where E is a constant energy that allows the phase space motion to be reduced from
four dimensions (R2 × T2 ) to three dimensions. Because P1 = P2 = 0, the constant
energy surfaces may be reduced to a 1D surface in the 3D constant energy space.
On this 1D surface, the angular motion is parameterized by the two frequencies
σ1 = a1 ,
(3.52a,b)
σ2 = a2 − a1 .
where each 2D torus is invariant. The level energy surface is a family of concentric
tori as shown in Fig. 3.5. Because the frequencies of motion σ1 and σ2 in (3.52) are
independent of P1 or P2 , they do not change from one concentric torus to another
concentric torus. The motion on the surface of the invariant torus Υ ⊂ M is quasi-
periodic. When the frequency ratio σ1 /σ2 = a1 /(a2 − a1 ) is an irrational number,
the motion on the surface of the 2D invariant nonresonant torus may no longer be
periodic; i.e., trajectories wind densely on the surface of the torus and never close
on themselves. This 2D torus has a 3D (R1 × T1 ) stable manifold W S and unstable
manifold W U that coincide along the 3D (R1 × T1 ) heteroclinic manifold H defined
for fixed values of P1 and P2 , as shown in Fig. 3.4. On a constant level energy surface,
neither the nonresonant invariant tori nor the stable and unstable manifolds may be
isolated. In addition, the 2D nonresonant invariant torus has a 2D center manifold
corresponding to nonexponentially expanding or contracting directions tangent to
the normally hyperbolic invariant manifold M.
July 31, 2009 8:18 9.75in x 6.5in b684-ch03 FA
Fig. 3.5. Motion of a phase space point for an integrable Hamiltonian system with two degrees
of freedom. (a) Invariant tori in a 3D constant energy space E; and (b) the flow on a 2D torus.1,3
July 31, 2009 8:18 9.75in x 6.5in b684-ch03 FA
Heteroclinic orbits. For fixed values of (P1∗ , P2∗ ) ∈ R2 , the (q, p) compo-
nents of the unperturbed vector field α = γ = 0 in (3.46) possess a 1D
heteroclinic orbit connecting the hyperbolic saddle points (q0 , p0 ). The hetero-
clinic orbit lies on each of the level energy surfaces defined by (3.51) and are
solutions to
H(γ = 0) − (a1 P1∗ + (a2 − a1 )P2∗ ) = 0. (3.55)
Values for q on the heteroclinic orbit are given by
1 % &
q = cos−1 (a1 − εΩ)/2εa3 P1∗ − p . (3.56)
2
In the full 6D phase space (q, p, P1 , P2 , Q1 , Q2 ) ∈ T1 ×R1 ×R2 ×T2 , the heteroclinic
manifold H may be determined by substituting
2
a1 − εΩ a1 − εΩ
cos[2q] = √ , sin[2q] = 1 − √ , (3.57a,b)
2εa3 P1 − p 2εa3 P1 − p
into the unperturbed vector field α = γ = 0 in (3.46), and integrating to
obtain1,3
A √
ph (t) = P2 + sech2 [ At] + p(0), (3.58a)
B
√
1 −1 A √
qh (t) = tan − tanh( At) + q(0), (3.58b)
2 a1 − εΩ
π
p(0) = 0, q = qn (0) = (2n + 1) ; n = 0, 1, 2, . . . . (3.58i,j)
2
The trajectories of the unperturbed system α = γ = 0 along the 5D phase space
(R1 × R2 × T2 ) heteroclinic manifold H may be expressed as
Ψ(P1 ,P2 ) = {qh (t), ph (t), P1 (0), P2 (0), Q1h (t), Q2h (t)}. (3.59)
The 6D phase space (q, p, P1 , P2 , Q1 , Q2 ) ∈ T × R × R × T is a direct product
1 1 2 2
coordinates (Q1 , Q2 ). Because P1 and P2 are constants, the motion in the 6D phase
space is reduced to the four dimensions (q, p, Q1 , Q2 ) shown in Fig. 3.6. Small per-
turbations are anticipated to break up the geometric structure of the unperturbed
system α = γ = 0 and to separate the manifolds. The behavior of the perturbed
systems γ > 0 and α > 0 near the unperturbed heteroclinic manifold H is required
in order to apply the GMM. The distance between the stable and unstable mani-
folds of any surviving invariant set in the perturbed system must be computed at
a point P on the unperturbed heteroclinic manifold H. Two perturbed systems are
evaluated: (1) γ > 0, α = 0 and (2) γ > 0 and α > 0.
Ṗ2 = 0, (3.60d)
(p − P2 ) sin[2Q1 ]
Q̇1 = a1 + εa3 √ cos[2q] + γc1 √ , (3.60e)
P1 − p ε P1 − p
Q̇2 = a2 − εΩ − 2εa3 P1 − p cos[2q] + γc2 sin[2(q + Q1 )], (3.60f)
H = H0 (p, P1 , P2 ) + Hε (q, p, P1 , P2 ) + Hγ (q, p, P1 , P2 , Q1 ). (3.61)
Because the perturbation is Hamiltonian, the 3D level energy surfaces are preserved.
In the 4D normally hyperbolic invariant manifold of the unperturbed space, the
locally stable and unstable manifolds and the flow describe the geometric structure
of the perturbed phase space given by the perturbed normally hyperbolic locally
invariant manifold, the locally stable and unstable manifolds, and the persistence
of the 2D nonresonant invariant tori Υγ (P1 , P2 ).
On Mγ there are locally stable and unstable manifolds that are of equal dimensions
and are close to the unperturbed locally stable and unstable manifolds. The per-
turbed normally hyperbolic locally invariant manifold Mγ intersects each of the 5D
level energy surfaces in a 3D set of which most of the two-parameter family of 2D
nonresonant invariant tori persist by the KAM theorem.1,3 The Melnikov integral
may be computed to determine if the stable and unstable manifolds of the KAM
tori intersect transversely.
KAM Theorem.1,3 The KAM theorem determines whether the recurrent motions
occur on the perturbed normally hyperbolic locally invariant manifold Mγ and
whether any of the two parameter families of 2D nonresonant invariant tori survive
the Hamiltonian perturbation. The unperturbed Floquet Hamiltonian
H(γ = 0) =
H0 (p, P1 , P2 ) + Hε (q, p, P1 , P2 ) satisfies the following nondegeneracy (or nonreso-
nance) condition:
# 2 #
# ∂
H ∂ 2
H #
# #
# ∂P 2 ∂P1 ∂P2 # (a1 − εΩ)2
# 1 #
# 2 # =− < 0. (3.63)
# ∂
H ∂ 2
H ## 4(P1 − P2 )2
#
# ∂P ∂P ∂P22 #(q=q0 ,p=p0 ;γ=0)
2 1
July 31, 2009 8:18 9.75in x 6.5in b684-ch03 FA
Most of the 2D nonresonant invariant tori Υ(P1 , P2 )) that persist are only
slightly deformed on the perturbed normally hyperbolic locally invariant manifold
Mγ and are KAM tori. In the phase space of the perturbed system γ > 0 and α = 0,
there are invariant tori that are densely filled with winding trajectories that are
conditionally periodic with two independent frequencies σ1 and σ2 . The resulting
conditionally-periodic motions of the perturbed system are smooth functions of the
perturbation γ. A generalization of the KAM theorem states that the KAM tori
have both stable and unstable manifolds by the invariance of manifolds.1,3 In order
to determine if chaos exists, two measurements are required in order to determine
whether or not W s (Υγ ) and W u (Υγ ) intersect transversely.
Melnikov integral. The distance between W s (Υγ ) and W u (Υγ ) at any point
P ∈ H may be computed from
βn
∂H γ (P1 ,P2 )
M(Q1 (0)) = − lim (Ψ )dt, (3.64a)
αn ,βn →∞ −αn ∂Q1
where
2c1
Hγ = P1 − p sin(2Q1 ) + c2 (P2 − p) sin[2(q + Q1 )], (3.64b)
ε
where Hγ = γH γ in (3.61), and where1,3
π
Q1 (0) = Q̄1n (0) = (2n + 1) − q(0); n = 0, 1, 2, . . . , (3.66a)
4
∂M(Q̄1n (0))
= −2 sin[2(Q̄1n (0) + q(0))][I1 + I2 + I3 ] = 0, n = 0, 1, 2, . . . .
∂Q1 (0)
(3.66b)
Consequently, the stable W s (Υγ (P1 , P2 )) and unstable W u (Υγ (P1 , P2 )) manifolds
of the KAM tori Υγ (P1 , P2 ) intersect transversely yielding Smale horseshoes1,3 on
the appropriate 5D level energy surfaces. This implies multiple transverse intersec-
tions and the corresponding existence of chaotic dynamics in the perturbed system
γ > 0 and α = 0.
The manifold Mγα has locally stable and unstable manifolds that are close to
the unperturbed locally stable and unstable manifolds; and if these manifolds
intersect transversely, then the Smale–Birkhoff theorem predicts the existence of
horseshoes and their chaotic dynamics in the perturbed dissipative system. A 2D
hyperbolic invariant torus Υγα (P1 , P2 ) may be located on Mγα by averaging the
perturbed dissipative vector field γ > 0 and α > 0 restricted to Mγα over the
angular variables Q1 and Q2 . The averaged equations have a unique stable hyper-
bolic fixed point (P1 , P2 ) = (0, 0) with two negative eigenvalues provided that the
determinant1,3
# #
# ∂
Ṗ1 ∂
Ṗ1 #
# #
# ∂P ∂P2 ## ν 2 γ 2 (λ4r ξτλ − τ )
# 1
# #= > 0. (3.68)
# ∂
Ṗ ∂
Ṗ # 2λ8r ξτ12 τλ
# 2 2 #
# #
∂P1 ∂P2
The Melnikov method fails to predict chaos for this dissipative system when
α > 0.1,3
∂ q̇ ∂ q̇ ∂ q̇ ∂ q̇ ∂ q̇ ∂ q̇
∂q ∂p ∂P1 ∂P2 ∂Q1 ∂Q2
∂ ṗ ∂ ṗ ∂ ṗ ∂ ṗ ∂ ṗ ∂ ṗ
δ q̇ ∂q ∂p ∂P1 ∂P2 ∂Q1 ∂Q2 δq
δ ṗ ∂ Ṗ1 ∂ Ṗ1 ∂ Ṗ1 ∂ Ṗ1 ∂ Ṗ1
∂ Ṗ1 δp
∂Q2
δ Ṗ1 ∂q ∂p ∂P1 ∂P2 ∂Q1 δP1
δ Ṗ = δP . (3.69)
2 ∂ Ṗ2 ∂ Ṗ2 ∂ Ṗ2 ∂ Ṗ2 ∂ Ṗ2 ∂ Ṗ2 2
δ Q̇ ∂q ∂p ∂P1 ∂P2 ∂Q1 ∂Q2 δQ
1
1
∂ Q̇1 ∂ Q̇1 ∂ Q̇1 ∂ Q̇1 ∂ Q̇1 ∂ Q̇1
δ Q̇2 δQ2
∂q ∂p ∂P1 ∂P2 ∂Q1 ∂Q2
∂ Q̇ ∂ Q̇2 ∂ Q̇2 ∂ Q̇2 ∂ Q̇2 ∂ Q̇
2 2
∂q ∂p ∂P1 ∂P2 ∂Q1 ∂Q2
For 20 different values of the dimensionless damping parameter and of the dimen-
sionless Floquet parametric forcing parameter, the largest LCEs were calculated and
a chaos diagram for the positive values of the largest Liapunov exponents identified
the parameter space in which chaotic motions exist.1,3
Appendix A
GHA:Type I
Equate N new variables to N old variables Ui = Ui (u, v, t) such that the Poisson
brackets satisfy [[Ui , Uj ]]uv = 0 and the determinant
# ∂U ∂U1 #
# 1 #
# · · · #
# ∂v1 ∂vN #
# #
# ∂Ui #
# · · · · # = 0. (3.A.1)
# ∂vj #
# #
# #
# ∂UN ∂UN #
# · · · #
∂v1 ∂vN
Equate N Herglotz auxiliary functions Xi (u, v) to the absolute value of the ratios
of the old variables in either of the following forms:
|ui /vi |
Xi = Xi (u, v) = ; i = 1, 2, . . . , N, (3.A.2a,b)
|vi /ui |
such that the Jacobian of the new variables U and the Herglotz auxiliary func-
tions X is nonzero ∂(U, X)/∂(u, v) = 0. Solve (3.A.2) for vi = vi (u, X); substitute
vi (u, X) into Ui (u, v, t) and invert to obtain ui = ui (U, X, t). Compute the gener-
ating function F (u, U, t) from
N
dF (u(U, X, t), U, t) = (vi dui − Vi dUi ), (3.A.3)
i=1
N
∂F ∂F
dXi + dUi
i=1
∂Xi ∂Ui
N N N
∂uj ∂uj
= vj dXi + vj dUi − Vi dUi . (3.A.4)
i=1 j=1
∂X i
j=1
∂U i
Equate the coefficients of like differentials and compute: (1) the generating function
N
N
Xi
∂uj
F (u, U, t) = vj dXi + C(U, t), (3.A.5)
∂Xi
i=1 j=1
N
∂uj ∂F
Vi (u, v, t) = vj − ; i = 1, 2, . . . , N. (3.A.6)
j=1
∂Ui ∂Ui
In order to compute the transformed Hamiltonian in terms of the new variables (U,
V), the inverse canonical transformation (u(U, V, t), v(U, V, t)) must be computed,
July 31, 2009 8:18 9.75in x 6.5in b684-ch03 FA
Appendix B
√
τ1 = 1+τ κτ = κ(1 + τ ) τλ = 1 + (τ /λ4r )
h
0 λ
r
2
1 − exp − ; for a full draft piston,
z λ2r
f1 = f (z) exp 2 ds =
−h λr
λ4r h h
− 1 + exp − 2 ; for a full draft hinge
h λ2r λr
√
1 βτ bf λ τ 1 2 τ − λ4r ξτλr
a1 = − + c1 = 1 r 1 d1 = 2 +
−8βξ 2ξλ6r τ12 4 2ξβ 3 4 τ1 βλ6r ξτλ2
1 1 βτ 2β − 1 1
a2 = − + − c2 = d2 =
β 8βξ 2ξλ6r τ12 4βξ 8 2bβξ 3 λ3r τ12
References
1. R. T. Hudspeth, Waves and Wave Forces on Coastal and Ocean Structures (World
Scientific, Singapore, 2006).
2. C. M. Bowline, R. T. Hudspeth and R. B. Guenther, Applicable Anal. 72, 287 (1999).
3. R. T. Hudspeth, R. B. Guenther and S. Fadel, Acta Mechanica 175, 139 (2005).
4. R. B. Guenther, H. Schwerdtfeger and G. Herglotz, Vorlesungen über die Mechanik der
Kontinua (1985).
5. S. Fadel, Application of the generalized Melnikov method to weakly damped para-
metrically excited cross waves with surface tension, PhD dissertation, Oregon State
University, USA (1998).
July 31, 2009 8:18 9.75in x 6.5in b684-ch04 FA
Chapter 4
Yoshimi Goda
ECOH CORPORATION, 2-6-4 Kita-Ueno
Taito-ku, Tokyo 110-0014, Japan
goda@ecoh.co.jp
4.1. Introduction
Breaking waves exert strong actions on maritime and coastal structures, while wave
dissipation through breaking plays a major role in the generation of nearshore cur-
rents. Without good understanding of wave-breaking process, we cannot pursue
any study for coastal engineering works. Nevertheless, wave breaking is an elusive
phenomenon. Not many people spend enough time to observe wonderful pictures
of wave deformation by breaking and regeneration after breaking. One needs some
kind of a pier at a beach to have a good look of waves that break and rush toward
the shore. Otherwise, one should work for some hours in a laboratory to measure
waves at various locations along a wave flume.
Our knowledge on wave breaking mostly comes from various literature based on
previous research works. Quite a number of people use the formulas, diagrams, and
other information listed there without examining the credibility of the information.
For example, many people regard the breaker index, or the ratio of wave height to
water depth at breaking, as a deterministic value without paying consideration to
the fact that the breaker height exhibits large fluctuations even for a given wave
87
July 31, 2009 8:18 9.75in x 6.5in b684-ch04 FA
88 Y. Goda
condition and that the breaker index has been obtained by drawing a mean curve
among scattered data.
Nowadays, many researchers are developing various numerical models of random
wave transformations. They have to adopt some kind of wave-breaking criteria and
energy dissipation mechanism so that the model can reproduce wave deformation
in the nearshore waters. However, modelers seem to pick up whatever is readily
available without deliberation on the physical features of the wave-breaking process
and the appropriateness of the breaking model.
The present chapter is a revised version of the author’s paper1 presented
at the 4th International Conference on Asian and Pacific Conference 2007, in
Nanjin, China. It aims at providing coastal engineers and researchers with the most
advanced knowledge on the statistics of wave breaking in the nearshore waters so
that they can make a correct approach to the problems related to wave-breaking
processes.
Waves are defined as breaking when the crest starts to contain foams or when
water particles jump out from the wave crest. It is a disruption of smooth water
surface, and breaking waves are often classified into three types: spilling, plunging,
and collapsing breakers. In the sea, there is no pure spilling breaker, because it is
always accompanied with a small portion of plunging water. For example, we can
observe from a window of an airplane such a scene of wave breaking in deep water
that a crest of large wave makes plunging, leaves a patch of white foam behind, and
moves forward with blue color. The following wave grows in height and breaks. It
is a manifestation of wave energy being transported by the group velocity, which is
one-half the phase velocity in deep water.
Theoretically, three criteria can be cited. The first is the condition that the
horizontal velocity of the water particle at the wave crest becomes equal to or greater
than the phase speed of wave profile. The second is the upward vertical acceleration
of the water particle at the wave crest to be equal to or to exceed the gravitational
acceleration. The third is the vertical gradient of the total pressure at the wave
crest to be zero or negative.
The first criterion has been employed by mathematicians to find out the lim-
iting waves of permanent form on a horizontal bed. Conventional perturbation
techniques are ineffective to derive the limiting waves, and the approach specific
to the limiting form needs to be employed. These limiting waves are characterized
with the angular crest having the angle of 120◦ . Yamada and Shiotani2 have pro-
duced the most reliable computation results so far, which are summarized by
Goda3 as reproduced in Table 4.1. The symbol H denotes the wave height, h is
the water depth, L is the wavelength, L0 is the deepwater wavelength, C is the
wave celerity, and ηc is the crest elevation. The subscripts “b” and “A” refer to
the quantities at breaking and those of the small amplitude waves (Airy’s wave),
respectively.
July 31, 2009 8:18 9.75in x 6.5in b684-ch04 FA
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 89
Table 4.1. Characteristics of breaking waves of permanent type (after Goda3 based on
Yamada and Shiotani2 ).
Because wave breaking attracts attention of many researchers, there have been pro-
posed a number of formulas to describe the ratio of wave height to water depth.
Kaminsky and Kraus8 called this ratio as the breaker height-to-depth index, but
the present chapter employs the term of breaker index for the ratio of wave height
to water depth at breaking for the sake of simplicity. Kamphuis9 has compared 11
formulas for testing of goodness-of-fit with his 225 sets of hydraulic model tests. Rat-
tanapitikon and Shibayama10 have collected 574 data points from 24 papers/reports
and calculated the root-mean-square errors of 24 breaking index formulas. Both
authors have proposed their own formulas by modifying some of the previous ones.
July 31, 2009 8:18 9.75in x 6.5in b684-ch04 FA
90 Y. Goda
Rattanapitikon et al.11 have further added 121 data points in large-scale flume tests
and proposed revised formulas.
Similar to the approach by Rattanapitikon et al.,11 the breaker index formulas
can be categorized into the following four functional forms:
where s denotes the bed slope. The wavelength L0 is the deepwater wavelength given
by L0 = g/(2π)T 2 , where T is the wave period. Because the relative water depth
hb /L0 is easily converted to hb /Lb through the dispersion relationship, the two
relative depths hb /L0 and hb /Lb are interchangeable. The formula of Hb /hb = 0.78
is a typical example of Eq. (4.1).
There are some other formulas using the parameter of deepwater wave steepness
H0 /L0 . They can be useful for predicting the breaker height of regular waves. Nev-
ertheless, they cannot be applied for breaking of random waves, because individual
zero-crossing waves in a train of random waves are unrelated to individual waves in
deepwaters. Thus, there is little room for the parameter H0 /L0 to function in the
breaker index for random waves.
Performance of a breaker index formula can be judged by the magnitude of the
bias of the predicted breaker height from the observed heights. It should also be
examined with either the root-mean-square error of predicted breaker heights or the
correlation coefficient between prediction and observation. The root-mean-square
error analysis by Rattanapitikon and Shibayama10 is not conclusive in differentiating
the merits of four functional forms, but they recommend a certain modification of
the slope effect in the function f4 (s, hb /L0 ), apparently indicating their preference
of this functional form.
Kamphuis9 calculated the correlation coefficients between 11 formulas and his
laboratory data. By assigning the best-fitting value to the proportionality coefficient
of each formula, he obtained the determination coefficient R2 = 0.69 to f1 (0),
R2 = 0.67 to f2 (hb /L0 ), R2 = 0.84 to f3 (s), and R2 = 0.88 to f4 (s, hb /L0 ). His
result clearly suggests the necessity to include both the parameters of bed slope and
relative water depth in the breaker index formula.
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 91
analyze this data set. Later, Goda12,13 approximated the breaker index curves with
the following empirical formula:
Hb A hb
= 1 − exp −1.5π 1 + 15s 4/3
: A = 0.17 . (4.5)
hb hb /L0 L0
Equation (4.5) was derived by a graphical curve-fitting technique without direct
comparison with the original laboratory data. Rattanapitikon and Shibayama10
have recommended a modification of the slope effect term of (1 + 15s4/3 )
into (1.033 + 4.71s − 10.46s2 ). Upon reexamination of the original laboratory
data, the author has also recognized a necessity of modifying the slope effect term.
The revised formula is as follows:
Hb A hb
= 1 − exp −1.5π 1 + 11 s4/3 : A = 0.17 . (4.6)
hb hb /L0 L0
Muttray and Oumeraci14 found the best-fitting coefficient of Eq. (4.5) being
0.167 instead of 0.170 for the slope of 1/30. When Eq. (4.6) is applied to their data,
the coefficient would have the value of 0.173.
Comparison of the laboratory data of breaker index with Eq. (4.6) is shown in
Fig. 4.1 for five groups of bed slopes, i.e., 1/7 to 1/12, 1/20, 1/30, 1/50, and 1/200
to 1/200. (See Goda3 for description of the laboratory data employed here.) Recent
data sets by Li et al.15 for s = 1/30 and s = 1/50, Li et al.16 for s = 1/200, and
Lara et al.17 for s = 1/20 are also added.
It is clear in Fig. 4.1 that the value of the breaker index increases as the bed
slope becomes steep. Thus, it is absolutely necessary to incorporate the slope effect
into the breaker index formula. Because the experimental data are scattered around
the index curves, the upper and lower bound curves with the range of 87%–115%
of the value by Eq. (4.6) are drawn in Fig. 4.1.
A quantitative evaluation of the degree of the scatter is made by means of
the relative error of the breaker index, i.e., E = (1 − γmeas/γest ), where γmeas is
the measured value of Hb /hb and γest is the predicted value by the breaker index
formula of Eq. (4.6). The mean and the standard deviation of the relative error are
calculated for each group of the bed slope. The mean Emean indicates a bias of the
breaker index and the standard deviation of E represents the degree of scatter of
the breaker index. Because E is defined as the relative error, the standard deviation
σ(E) is equivalent to the coefficient of variation (CoV). A positive bias indicates a
tendency of overestimate, while a negative bias shows an underestimate. Table 4.2
lists the bias and CoV of the breaker index of Eq. (4.6) for the data of various bed
slopes. The slope data of 1/9 and 1/12 are excluded from the analysis because of
their small sample sizes.
The bias varies from −2.9% to +6.2% depending on the bed slope, but the
formula of Eq. (4.6) can be regarded as yielding reasonable estimates of the breaker
heights. The scatter of the data as expressed by CoV is about 5% to 7% for the
bed slope of 1/200 to 1/50; it increases as the slope becomes steep, and it takes the
value of 14% for the bed slope of 1/10. Such scatter of data represents an inherent
stochastic nature of wave-breaking phenomenon. It resides in the data set itself,
being independent of the breaker index formula being applied.
July 31, 2009 8:18 9.75in x 6.5in b684-ch04 FA
92 Y. Goda
2.0
Slope = 1/9 to 1/12
Breaker Index, γ = Hb /hb
1.0
0.9
0.8
0.7
Kishi-Iohara 1/9
0.6 Iversen 1/10
Goda et al. 1/10
0.5 Galvin 1/10
Bowen et al. 1/12
Index curve 1/10
0.4 Upper 115%
Lower 87%
0.3
0.001 0.01 0.1
Relative water depth, hb /L0
1.5
Slope = 1/20
Breaker Index, γ = Hb /hb
1.0
0.9
0.8
0.7
0.6
Iversen 1/20
Galvin 1/20
0.5 Toyoshima 1/20
Lara 1/20
0.4 Index curve 1/20
Upper 115%
Lower 87%
0.3
0.001 0.01 0.1
Relative water depth, hb /L0
1.2
Slope = 1/30
1.0
Breaker Index, γ = Hb /hb
0.9
0.8
0.7
0.6
Iversen 1/30
0.5 Mitsuyasu 1/30
Toyoshima 1/30
Li Y.C. et al. 1/30
0.4 Index curve 1/30
Upper 115%
Lower 87%
0.3
0.001 0.01 0.1
Relative water depth, hb /L0
Fig. 4.1. Comparison of breaker index formula Eq. (4.6) with laboratory data of regular waves.
July 31, 2009 8:18 9.75in x 6.5in b684-ch04 FA
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 93
1.0
0.9
0.8
0.7
0.6
Slope = 1/100 to 1/200
Goda 1/100
0.5 Li Y.C. et al. 1/200
Index curve 1/100
0.4 Index curve 1/200
Upper 115%
Lower 87%
0.3
0.001 0.01 0.1
Relative water depth, hb /L0
(e) Breaker index data for the slope of 1/100–1/200.
Table 4.2. Bias and CoV of the breaker index formula of Eq. (4.6).
Even under a well-controlled laboratory test, the breaking point fluctuates over
some distance and the breaker height varies from wave to wave. Smith and Kraus18
reported on their regular wave tests that “despite care in conducting the tests and
use of the average value of the given quantity (i.e., over 10 waves), wide scatter
appeared in some quantities and must be considered inherent to the breaking process
of realistic waves.” One cause of the data scatter is the presence of small-amplitude,
long-period oscillations of water level in a laboratory flume, but the breaking
process itself is triggered by many small factors beyond the control of experi-
menters. We should regard the wave-breaking phenomenon as stochastic one and
accept a certain range of natural fluctuation. As listed in Table 4.2, the coefficient of
July 31, 2009 8:18 9.75in x 6.5in b684-ch04 FA
94 Y. Goda
variation is large for steep slope and it becomes small for gentle slope. In any research
work involving wave-breaking phenomena, due consideration should be given to
such stochastic nature of breaker heights.
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 95
waves or random waves, but such application does not yield correct answers. The
breaker index for regular waves may be utilized for the highest wave in an irregular
wave train, but it cannot be applied for the significant wave, the root-mean-square
wave, or any other characteristic wave. When random waves approach the shore,
breaking of individual waves occurs gradually with large waves first at the far dis-
tance, medium waves next at some distance from the shore, and small waves near
the shoreline. The variation of significant wave height from the offshore toward the
shore is so gradual that we cannot employ the concept of wave-breaking line, which
is so obvious in the case of regular waves.
Against such difficulty of defining the breaking point of significant wave,
Kamphuis9 introduced the definition of “incipient wave breaking.” He measured
cross-shore variations of significant wave height beyond and across the surf zone,
drew a curve of wave shoaling trend in the outside of the surf zone and a curve of
wave height decay within the surf zone, and called the condition at the cross-point
of the two curves as the incipient wave breaking. By using the data of the signif-
icant wave height at incipient breaking, he calibrated 11 breaker index formulas and
determined the best-fitting proportionality coefficient. For the formula of Eq. (4.5),
he obtained the proportionality coefficient of A = 0.12 for significant wave height.
Li et al.23 have also presented a data of the breaker index of (H1/3 /h)b on the
slope of 1/200, which is fitted to Eq. (4.5) with a modified constant value of 0.12
for the initial stage of breaking. Their breaking condition was some observation of
large individual breaking waves in an irregular wave train. Goda24 has prepared a
set of diagrams depicting variations of significant wave height across the surf zone
(reproduced as Figs. 3.29 to 3.32 in Ref. 25). The boundary lines of 2% decay in
these diagrams approximately correspond to the breaker index with A = 0.11, and
the water depth (h1/3 )peak at which the significant wave takes a peak value within
the surf zone (Fig. 3.34 in Ref. 25) corresponds to A = 0.11–0.13. Therefore, the
incipient breaker index of the significant wave can be expressed with the following
formula:
H1/3, b 0.12 (hb )incipient
= 1 − exp −1.5π 1 + 11 s 4/3
. (4.7)
hb incipient hb /L0 L0
Thus, the incipient breaker index of significant wave is about 30% lower than
that of regular waves. The incipient breaking of significant wave corresponds to the
condition that the high waves of upper several percent among individual waves have
begun to break.
96 Y. Goda
1.0
Breaker indices, H1/3 /h & Hrms /h Slope = 1/35
0.8
0.6
0.4
0.3
Broad spec. H1/3 /h
Narrow spec. H1/3 /h
0.2 Breaker envlp. H1/3 /h
Broad spec. Hrms /h
Narrow spec. Hrms /h
Breaker envlp. Hrms /h
0.10
0.01 0.03 0.1 0.3
Relative depth, h/L0
Fig. 4.2. Breaker indices for H1/3 and Hrms on s = 1/35 with the data by Ting.26,27
peak period of Tp = 2.0 s. He recorded wave profiles at an offshore station with the
depth of 0.457 m and at six stations on the slope with the depth of 0.27–0.0625 m.
Waves at the six stations on the slope had the percentage of breaking ranging from
5% to 94% (the case of broad-band spectrum).
Wave records were analyzed by the zero-downcrossing method, and calculated
results of characteristic wave heights and periods are presented in tabular forms.
From these results, the ratios of H1/3 and Hrms to the local depth (inclusive of
mean water level change) are calculated and plotted against the relative water depth
h/L0 , as shown in Fig. 4.2.
The curves denoted as breaker envelopes are calculated by Eq. (4.6) for s = 1/35
with the proportionality coefficient of A = 0.145 (85% of regular waves) for the
significant height H1/3 and to A = 0.111 (65%) for the root-mean-square height
Hrms . Because the percentage of breaking waves is high in these data, an A value
higher than that for incipient breaking fits to the data.
The breaker index data for Hrms by Tick is higher than the value proposed by
Sallenger and Holman,28 who gave an expression of Hrms /h = 3.2s + 0.32 without
inclusion of the relative depth (h/L0 ) term. They converted the orbital velocity
spectra to the surface wave spectra with the transfer function based on the linear
theory, and estimated the energy-based Hrms , which must have been smaller than
the statistical Hrms value based on direct measurement data of surface profiles.
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 97
Table 4.3. Summary of stationary coastal wave data employed in the present analysis.
observation data analyzed by Goda and Nagai29 and the data of long-traveled swell
recorded with an ultrasonic wave sensor reported by Goda30 were utilized. Table 4.3
lists the characteristics of these field data. Waves were recorded by means of either
step-resistance gauges or ultrasonic wave sensors so that they were reliable regis-
tration of surface wave profiles. The data at Tomakomai and Kanazawa as well as
Caldera were measured with ultrasonic wave sensors. They are not analyzed for
breaker limits but for wave nonlinearity effects to be described in Sec. 4.6.
Table 4.3 also lists the wave records at Sakata Port measured by means of
pressure gauges, which were utilized by Goda24 for calibration of his random wave-
breaking model. Although there remains a problem of pressure conversion to surface
profiles, the conversion error would have been small because of the relatively shallow
water depth at Sakata stations (10.5 and 14.5 m). They were included in the present
analysis to increase the size of database.
Other sources of nearshore waves are the photogrametric measurement data by
Hotta and Mizuguchi31,32 as well as by Ebersole and Hughes.33 Hotta and Mizuguchi
mobilized 11 motion-picture cameras set on top of a coastal observation pier at
Ajigaura Beach, Ibaragi, Japan. They took film pictures of instantaneous water
surfaces simultaneously at some 60 surveyor’s poles erected in the nearshore waters
on a line perpendicular to the shoreline stretched over a distance of about 120 m.
Films of surface wave records were taken every 0.2 s for an effective duration of 760 s.
The beach profile in September 1978 had a trough at the distance of 25 m from the
shoreline and the slope of about 1/60 offshore of the trough. The beach profile in
December 1978 was somewhat uniform without any bar or trough, and the slope was
about 1/70. The water depth inclusive of tides at the poles varied from 0.1 to 2.7 m.
Photogrametric measurements of nearshore waves were also executed by
Ebersole and Hughes33 during the DUCK85 campaign in Duck, North Carolina,
USA with the cooperation of Dr. Hotta who brought twelve cameras with him and
took charge of filming. They referred to this technique as “the photopole method”;
this terminology is employed in the present chapter. Over the distance of 64 m,
July 31, 2009 8:18 9.75in x 6.5in b684-ch04 FA
98 Y. Goda
12 poles were erected in the initial depth ranging from 0.4 to 1.9 m. Measurements
were taken for nine runs during the fourth and fifth days of September 1985. With
the variation of the tide level, the actual water depth varied from 0.5 to 2.4 m. The
effective duration of wave recording was about 650 s, judging from the number of
waves and average periods listed by Ebersole and Hughes.33
The beach profile was nearly flat for about 25 m from the shoreline with the
depth of about 0.5 m below the mean sea level, and it had the slope of about 1/30
beyond that. Another series of photopole measurements were carried out during the
SUPERDUCK campaign in 1986. Dr Hughes kindly supplied the author with the
data files of measured wave statistics. The number of poles was increased to 20 and
the water depth inclusive of tides varied from 0.4 to 3.7 m. The beach profile during
SUPERDUCK is not known, but it would have been nearly the same as DUCK85
because of the same season. All the photopole measurement data were analyzed
by the zero-downcrossing method, and various statistical wave characteristics were
calculated.
Table 4.4 lists the summary of the photopole wave measurement conditions.
The significant wave period T1/3 has been converted from the pole-averaged values
of either the mean period Tmean or the spectral period Tp by assumption of
T1/3 = 1.05Tmean or T1/3 = 0.95Tp, which would be appropriate for swell of very low
steepness. The offshore wave height H0 was converted from the significant height
H1/3 measured at the most offshore pole using the shoaling coefficient; no refraction
effect was taken into consideration as no information of wave direction was available.
All the waves were swell of very low steepness ranging from 0.0017 to 0.0081,
except one case of SUPERDUCK with H0 /L0 = 0.0254.
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 99
Step-resistance ( h=11-14m)
Presssure gage (h=10-14m)
Photopole Ajigaura (h=0.1-2.7 m)
Photopole DUCK85 (h=0.5-2.4m)
Photopole SUPERDUCK (h=0.5-3.2 m)
Breaker envelope (80% limit)
1.0
0.8
Wave height ratio, Hm0 /h
0.6
0.4
0.2
0.1
0.001 0.01 0.1 1
Relative depth, h/L0
Fig. 4.3. Breaker index for Hm0 based on the field wave data.
Coastal wave data, on the other hand, were not much affected by wave non-
linearity effects, and the zero-crossing significant height H1/3 was almost the same
as Hm0 .
Many data points in Fig. 4.3 belong to nonbreaking condition, but what interests
us is the upper envelope which provides an estimate of the upper limit of breaking
wave height. The curve of dash-dot line in Fig. 4.3 has been calculated by Eq. (4.6)
for the slope of s = 0.0143 (1/70) with the coefficient being reduced to A = 0.136
(80%). Similar with the laboratory data shown in Fig. 4.2, the wave height ratio
Hm0 /h is higher than the incipient breaker index of significant wave expressed by
Eq. (4.7). It is because the breaker index increases inside the surf zone as the
percentage of breaking waves increase.
It is seen that the energy-based significant wave height Hm0 on gentle slopes
does not exceed 0.7 times the local water depth except for the low-steepness swell
in very shallow water. For the range of h/L0 > 0.03, the upper limit of significant
wave height is about 0.6 times the local water depth. Some data points above the
dash-dot curve are those of DUCK85 and SUPERDUCK, which were conducted on
the beach steeper than the beach in Ajigaura.
100 Y. Goda
representation and analysis. Deviation from the linearity is measured with the values
of the skewness and kurtosis of the surface elevation with reference to the mean
water level during a wave record.
The skewness is zero when a distribution is symmetric with respect to the mean,
and takes a positive value when a distribution is asymmetric with a long tail toward
the right side (large value). The kurtosis takes a value of 3.0 for the Gaussian
distribution. When the mode of distribution has a sharp peak and the distribution
has long tails in both the left and right sides, the value of kurtosis becomes much
larger than 3.0. The degree of positive skewness and the deviation of kurtosis from
3.0 are the measure of the strength of wave nonlinearity. The skewness of ocean
waves is less than 0.5, and the kurtosis is below 4.0 for most cases, and thus the
nonlinearity of waves in deepwater is weak.
The variations of the skewness and kurtosis of field waves are examined with
coastal surface waves listed in Table 4.3 (excluding the pressure sensor data of
Sakata Port), and the photopole data in Ajigaura, DUCK85, and SUPERDUCK
listed in Table 4.4. The data of skewness and kurtosis of September 5th in Ajigaura
were kindly provided by Dr Hotta. The data of the photopole measurements of
SUPERDUCK provided by Dr Hughes had the data of skewness only.
As waves approach the shore, wave nonlinearity is enhanced and both the
skewness and kurtosis increase significantly. Figure 4.4(a) exhibits the increase of
the skewness with the nonlinearity parameter Π1/3 , which was introduced by Goda34
with the following definition:
H1/3 2πh
Π1/3 = coth3 , (4.9)
LA LA
where LA denotes the wavelength calculated by small amplitude wave theory or
Airy’s theory.
The data are grouped by the range of the offshore wave steepness H0 /L0 : the first
group for 0.001 < H0 /L0 < 0.0029, the second group for 0.003 < H0 /L0 < 0.0049,
the third group for 0.0050 < H0 /L0 < 0.0099, the fourth group for 0.010 < H0 /L0 <
0.029, and the fifth group for 0.030 < H0 /L0 < 0.049 (Legends are shown with
abbreviated figures).
The data shown in the left diagram of Fig. 4.4 are those outside the surf zone.
Because the boundary of surf zone is difficult to be set for random waves, an arbi-
trary boundary of h/H0 = 2.5 is employed here to separate the wave data outside
and inside the surf zone. As shown in the left diagram, the skewness outside the surf
zone shows a clear correlation with the wave nonlinearity parameter. The skewness
begins from the value of zero at Π1/3 = 0, increases almost linearly with Π1/3 ,
and attains the value of 2.0 around Π1/3 = 4. The dashed line represents a semi-
theoretical relationship, which is based on the analysis of finite amplitude regular
wave profiles by Goda34 with the consideration of the probability of individual wave
heights according to the Rayleigh distribution.
Variation of the skewness inside the surf zone (h/H0 < 2.5) is shown in
Fig. 4.4(b). The ordinate is the ratio of offshore wave height to water depth, H0 /h,
which increases rapidly as waves approach the shore. There is a clear trend of
skewness decreasing toward β1 = 0 with the increase of the height-to-depth ratio
July 31, 2009 8:18 9.75in x 6.5in b684-ch04 FA
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 101
2.5 2.5
0.001 <H0 /L0 <0.002 h/H0 > 2.5 h/H0 < 2.5 0.001<H0 /L0 <0.002
0.003 <H0 /L0 <0.004 0.003<H0 /L0 <0.004
2.0 0.005 <H0 /L0 <0.009 2.0 0.005<H0 /L0 <0.009
0.010 <H0 /L0 <0.029 0.010<H0 /L0 <0.029
0.030 <H0 /L0 <0.049 0.030<H0 /L0 <0.049
Semi-theoretical
1.5 1.5
Skewness, β 1
Skewness, β 1
1.0 1.0
0.5 0.5
0.0 0.0
-0.5 -0.5
0.01 0.1 1 10 0.4 0.6 0.8 1 1.2 1.4
Nonlinearity parameter, Π1/3 Height-to-depht ratio, H0 /h
(a) O utside the surf zone (h/H0 > 2.5) (b) Inside the surf zone (h/H0 < 2.5)
Fig. 4.4. Variation of the skewness of surface elevation outside and inside the surf zone.
10.5 10.5
0.001<H0 /L0 <0.002 h/H0 > 2.5 h/H0 < 2.5 0.001<H0 /L0 <0.002
0.003<H0 /L0 <0.004 0.005<H0 /L0 <0.009
9.0 0.005<H0 /L0 <0.009 9.0 0.010<H0 /L0 <0.029
0.010<H0 /L0 <0.029 0.030<H0 /L0 <0.049
0.030<H0 /L0 < 0.049
7.5 Empirical 7.5
Kurtosis, β 2
Kurtosis, β 2
6.0 6.0
4.5 4.5
3.0 3.0
1.5 1.5
0.01 0.1 1 10 0.4 0.6 0.8 1.0 1.2 1.4
Nonlinearity parameter, Π1/3 Height-to-depht ratio, H0 /h
(a) Outside the surf zone (h/H0 > 2.5) (b) Inside the surf zone (h/H0 < 2.5)
Fig. 4.5. Variation of kurtosis of surface elevation outside and inside the surf zone.
H0 /h. Use of the height-to-depth ratio in Fig. 4.4(b) is to provide a kind of contrast
of the increase and decrease of skewness in the outside and inside of the surf zone,
respectively. While the increase of skewness outside the surf zone seems indifferent
to the wave steepness, the value of skewness inside the surf zone is much affected by
the wave steepness; waves of low steepness maintain large values of skewness, while
waves of high steepness have small values only.
Variation of the kurtosis of surface elevation is shown in Fig. 4.5; Fig. 4.5(a)
shows kurtosis outside the surf zone and Fig. 4.5(b) shows kurtosis inside the surf
zone. The pattern of variation is the same as that of skewness, though the available
number of kurtosis data is smaller than the skewness data. The kurtosis starts from
the value of 3.0 at Π1/3 = 0, increases as the nonlinearity parameter increases, and
attains the value of 9 around Π1/3 = 4. In Fig. 4.5(b), the kurtosis inside the surf
July 31, 2009 8:18 9.75in x 6.5in b684-ch04 FA
102 Y. Goda
zone decreases toward β2 = 3.0 as waves approach the shoreline (i.e., increase of
H0 /h). Waves with the steepness larger than 0.01 exhibit the kurtosis value much
smaller than swell with very low steepness. It is because waves with large steepness
cannot experience the full process of nonlinear shoaling owing to the early start of
wave breaking.
Figures 4.6–4.8 show the evolutions of the wave height ratios H1/10 /Hm0 ,
H1/3 /Hm0 , and Hrms /Hm0 , respectively. The difference between H1/3 and Hm0 in
the nearshore waters has been pointed by several researchers such as Thompson and
Vincent36 and Ebersole and Hughes.33 In Figs. 4.6–4.8, the (a) parts are for those
outside the surf zone and the (b) parts are for those inside the surf zone, similar to
Figs. 4.4 and 4.5. The boundary of the surf zone is subjectively set at h/H0 = 2.5
in Figs. 4.6 and 4.7 and h/H0 = 1.5 in Fig. 4.8. In case of the root-mean-square
wave height Hrms , the surf zone should be defined in a shallower area than for the
significant wave, because Hrms is calculated with all individual waves. Thus, the
boundary of Hrms was set at h/H0 = 1.5.
The dashed lines in the left diagrams are semi-theoretical predictions based
on the potential energy calculation of finite amplitude waves and the Rayleigh
distribution of wave heights, by referring to the methodology employed by
Longuet-Higgins.37
July 31, 2009 8:18 9.75in x 6.5in b684-ch04 FA
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 103
2.2 2.2
0.001 <H0 /L0 <0.002 h/H0 > 2.5 h/H0 < 2.5 0.001 <H0 /L0 <0.002
0.003 <H0 /L0 <0.004 0.003 <H0 /L0 <0.004
2.0 0.005 <H0 /L0 <0.009 2.0 0.005 <H0 /L0 < 0.009
0.010 <H0 /L0 <0.029 0.010 <H0 /L0 <0.029
Wave height ratio, H1/10 /Hm0
1.6 1.6
1.4 1.4
1.2 1.2
1.0 1.0
0.01 0.1 1 10 0.4 1 4 10
Nonlinearity parameter, Π1/3 Height-to-depht ratio, H0 /h
(a) Outside the surf zone (h/H0 > 2.5) (b) Inside the surf zone (h/H0 < 2.5)
Fig. 4.6. Variation of wave height ratio H1/10 /Hm0 outside and inside the surf zone.
1.75 1.75
0.001 <H0 /L0 <0.002 h/H0 > 2.5 h/H0 < 2.5 0.001<H0 /L0 <0.002
0.003 <H0 /L0 <0.004 0.003<H0 /L0 <0.004
0.005 <H0 /L0 <0.009 0.005<H0 /L0 <0.009
0.010 <H0 /L0 <0.029 0.010<H0 /L0 <0.029
Wave height ratio, H1/3 /Hm0
1.25 1.25
1.00 1.00
0.75 0.75
0.01 0.1 1 10 0.4 1 4 10
Nonlinearity parameter, Π1/3 Height-to-depht ratio, H0 /h
(a) Outside the surf zone (h/H0 > 2.5) (b) Inside the surf zone (h/H0 < 2.5)
Fig. 4.7. Variation of wave height ratio H1/3 /Hm0 outside and inside the surf zone.
Outside the surf zone shown in the left diagrams of Figs. 4.6–4.8, the wave height
ratios H1/10 /Hm0 , H1/3 /Hm0 , and Hrms /Hm0 increase with the increase of the wave
nonlinearity parameter. Although there is much scatter of data, they follow the
trend of semi-theoretical curves. The one-tenth highest wave height H1/10 in Fig. 4.6
is given the initial value of 1.20Hm0 based on the trend of the data in consideration
of the spectral width effect.
The maximum values of the ratios of the statistical wave heights to the energy-
based significant wave height for swell of very low steepness are around 2.1 for
H1/10 /Hm0 , 1.6 for H1/3 /Hm0 , and 1.15 for Hrms /Hm0 . With reference to the
wave height ratios at the weak nonlinearity state, the statistical wave heights are
July 31, 2009 8:18 9.75in x 6.5in b684-ch04 FA
104 Y. Goda
1.2 1.2
h/H0 > 1.5 h/H0 < 1.5
0.001<H0 /L0 <0.002 0.001<H0 /L0 < 0.002
0.003<H0 /L0 <0.004 0.003<H0 /L0 < 0.004
1.1 0.005<H0 /L0 <0.009 1.1 0.005<H0 /L0 < 0.009
0.010<H0 /L0 <0.029 0.010<H0 /L0 < 0.029
Rayleigh Rayleigh
Semi-theoretical
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
0.01 0.1 1 10 0.4 1 4 10
Nonlinearity parameter, Π1/3 Height-to-depht ratio, H0 /h
(a) Outside the surf zone (h/H0 > 1.5) (b) I nside the surf zone (h/H0 < 1.5)
Fig. 4.8. Variation of wave height ratio Hrms /Hm0 outside and inside the surf zone.
enhanced by 1.75 times for H1/10 , 1.68 times for H1/3 , and 1.64 times for Hrms . Such
enhancement of statistical wave heights are apparent ones without real increase of
wave energy as discussed earlier.
Inside the surf zone shown in the right diagrams of Figs. 4.6–4.8, the data exhibit
large scatter but they indicate a clear tendency of decrease with the height-to-depth
ratio H0 /h; the wave height ratios H1/10 /Hm0 , H1/3 /Hm0 , and Hrms /Hm0 converge
to 1.27, 1.0, and 0.70, respectively, as H0 /h becomes larger than 2. As for the effect
of wave steepness, waves with low steepness in the range of 0.001 to 0.0049 have the
mean value of H1/3 /Hm0 = 1.4 around H0 /h = 0.4, while waves with the medium
steepness of 0.005 to 0.009 starts from the mean value of H1/3 /Hm0 = 1.2. Waves
with a high steepness of 0.010 to 0.049 have the H1/3 /Hm0 value less than 1.1.
Decrease of the wave height ratio with decreasing water depth has been pointed out
by Thompson and Vincent,36 who suggested earlier start of decrease for waves of
high steepness compared with waves of low steepness.
Such effects of wave steepness on the nonlinear features of waves inside the surf
zone are originated from the fact that waves of low steepness experience a high
degree of nonlinear wave shoaling before they are attenuated by breaking, while
waves of high steepness are attenuated much earlier before they experience strong
nonlinear shoaling. It is seen in the left diagrams of Figs. 4.6–4.8 that waves with the
steepness of 0.010–0.049 have the nonlinearity parameter up to Π1/3 = 0.3 only and
waves are transferred into the right diagrams which represent waves inside the surf
zone. Waves with the steepness of 0.005 to 0.0099 have the nonlinearity parameter
up to Π1/3 = 1.2 and moves into the group of those inside the surf zone. Waves with
the steepness below 0.0029 can have the nonlinearity parameter up to Π1/3 = 4
before they enter into the surf zone.
Thus, the evolution of wave nonlinearity expressed in terms of the skewness
β1 , kurtosis β2 , and wave height ratios H1/10 /Hm0 , H1/3 /Hm0 , and Hrms /Hm0
is summarized as follows. It is weak in the offshore, increases gradually as waves
propagate shoreward, becomes strongest in the outer half of the surf zone, begins
July 31, 2009 8:18 9.75in x 6.5in b684-ch04 FA
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 105
106 Y. Goda
1.0 1.0
H1/3/h = 0 - 0.399 H1/3/h = 0.400 - 0.499
52 records with 5111 waves 17 records with 1398 waves
0.8 0.8
Probability density, p (x)
0.4 0.4
0.2 0.2
0 0
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Normalized wave height, x = H /Hmean Normalized wave height, x = H /Hmean
1.0 1.0
Observation Observation
0.6 Rayleigh 0.6 Rayleigh
0.4 0.4
0.2 0.2
0 0
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Normalized wave height, x = H /Hmean Normalized wave height, x = H /Hmean
Fig. 4.9. Histograms of field wave data in four groups of relative water depth (after Goda24 ).
shore, the range of wave height begins to spread owing to regeneration of waves after
breaking, and the wave heights show a tendency to approach the Rayleigh again.
Some researchers such as Battjes and Groenendijk44 have tried to apply certain
statistical distributions to the wave heights in the surf zone and to empirically
formulate the parameter values. As demonstrated in Figs. 4.9 and 4.10, however, the
wave height distribution continues to vary throughout the surf zone, and any effort
to describe it with a combination of multiple distribution functions will be futile.
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 107
2.5
Ting's Lab. Data
0.99
Reduced variate, SQRT [ -ln (1 - P)]
2.0 Rayleigh
Nonexceedance Probability, P
h/H0'=1.90
h/H0'=1.60
h/H0'=1.28
0.95
h/H0'=0.97
1.5 0.90
h/H0'=0.69
h/H0'=0.48
0.75
1.0
0.50
0.5 0.25
0.10
0.05
0 0.01
0 0.5 1.0 1.5 2.0
Relative wave height, H/H1/3
Fig. 4.10. Evolution of cumulative distribution of wave heights in the surf zone based on the data
by Ting.42
Index for Spectral waves (PEGBIS). It has been demonstrated that the difference
among characteristic wave heights gradually decreases as the relative water depth
becomes shallow down to around h/H0 = 1, but the range of wave heights expands
again as the water becomes much shallower toward the shoreline.
Laboratory tests by Ting42,43 on the wave breaking on a uniform slope of
1/35 show a variation of the wave height ratio Hrms /H1/3, following the general
trend described above. Similar variations of the wave height ratios H1/10 /H1/3
and Hrms /H1/3 are found in the photopole data at Ajigaura, DUCK85, and
SUPERDUCK. Figure 4.11 presents a compilation of the laboratory and field data
concerning the wave height ratio changes. Also included in Fig. 4.11 is the pre-
diction of the wave height ratios across the surf zone by means of the PEGBIS model
applied for the Ajigaura beach condition. Although there are large scatters of data,
the ratio H1/10 /H1/3 takes the minimum value of about 1.15 around h/H0 = 1.4
for the field data, which indicate slightly larger values than the PEGBIS prediction,
probably owing to the wave nonlinearity effect. The ratio Hrms /H1/3 takes the
maximum value of about 0.81 around h/H0 = 1.0 for both the laboratory and field
data. The DUCK85 data exhibit the Hrms /H1/3 being closer to 1 than the Ajigaura
data. It might be due to lower wave steepness of the swell at DUCK85.
Such changes of the wave height ratios across the surf zone are the result of the
evolution of the pdf of wave heights discussed in Sec. 4.7.1. Narrowing of the wave
height distribution at the middle of the surf zone provides a larger safety margin
in the design of breakwaters and other structures than the case of applying the
Rayleigh distribution, because the design wave height is specified with either Hmax
(composite breakwaters) or H1/20 (mound breakwaters) for a prescribed offshore
July 31, 2009 8:18 9.75in x 6.5in b684-ch04 FA
108 Y. Goda
1.25
1.00
0.75
0.50
0.0 1.0 2.0 3.0 4.0 5.0 6.0
Relative depth, h/H0
Fig. 4.11. Variations of wave height ratios H1/10 /H1/3 and Hrms /H1/3 by photopole field data
and Hrms /H1/3 in laboratory tests by Ting together with prediction by PEGBIS model.
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 109
Table 4.5. Main characteristics of random wave-breaking models for comparative study.
Goda24,45 only. The energy dissipation is dealt directly with either the bore analogy
or the model by Dally et al.50 or indirectly through modification of the pdf of wave
heights.
Figure 4.12 is an example of differences in the variations of the root-mean-square
wave height in the nearshore waters predicted by the seven models on the slope of
s = 0.02 (Goda46 ). Differences are very large beyond expectation and Goda46 shows
other examples on the slope of s = 1/10, 1/30, and 1/100. Because the generation of
longshore currents is governed by the spatial gradient of the radiation stresses, these
models produce much different cross-shore profiles of longshore current velocities as
demonstrated by Goda.46
Smith51 has also demonstrated different performances of five wave-breaking
models with different breaker indices against field measurement of spectral waves.
Because of the large difference in the prediction of nearshore wave heights, one
should be careful in selecting a numerical model for engineering applications.
Development of numerical models for time-domain wave transformation has been
tried by many researchers, but random wave-breaking process has not been well
reproduced in these models. One of the exceptions is the Boussinesq-type model
developed by Hirayama et al.52 and Hirayama and Hiraishi,53 who employed the
breaking criterion of the vertical pressure gradient by Nadaoka et al.7 They raised
the threshold gradient from 0 to 0.5 to compensate the insufficiency in numerical
accuracy due to the features of weak nonlinearity inherent to the Boussinesq
equation. They have succeeded in reproducing the pdf variation across the surf zone.
July 31, 2009 8:18 9.75in x 6.5in b684-ch04 FA
110 Y. Goda
2.0
1.5
1.0
Battjes '72
Kuo & Kuo '74
Goda '75
0.5 Battjes & Janssen '78
Thornton & Guza: M1
Larson & Kraus '91
Goda '04 -spectr-
0
0 1.0 2.0 3.0 4.0 5.0 6.0
Depth, h (m)
Fig. 4.12. Variations of wave heights Hrms in the surf zone on planar beaches computed by seven
random wave-breaking models for (H1/3 )0 = 2.0 m, T = 8.004 s, and θ0 = 30◦ (Goda46 ).
4.9. Summary
(1) The breaker index formula by Goda13 for regular waves is revised by reducing
the bed slope effect for better agreement with laboratory data.
(2) Depth-limited breaker heights have inherent variability with the coefficient of
variation of about 6% for the slope of 1/100 and 14% for the slope of 1/10,
which increases as the bed slope becomes steep.
(3) Breaker index or the ratio of breaker height to water depth is governed by
both the relative water depth hb /L0 and the bed slope.
(4) Incipient breaker index for the significant height of random waves is smaller
than that for regular waves by about 30%, and the significant wave height has
an upper limit of about 0.6 times the water depth on gentle slopes.
(5) Wave nonlinearity expressed in terms of skewness, kurtosis, and the wave
ratios H1/10 /Hm0 , H1/3 /Hm0 , and Hrms /Hm0 increases with wave propagation
toward the shore and is most enhanced just outside the surf zone. For swell of
very low steepness, the skewness, kurtosis, and wave height ratio H1/3 /Hm0
may go up to 2.0, 9.0, and 1.6, respectively, at the wave nonlinearity parameter
around 4.0.
July 31, 2009 8:18 9.75in x 6.5in b684-ch04 FA
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 111
(6) For wind waves and swell of relatively large steepness, random wave breaking
begins at large depths compared with swell of small steepness. Thus, the former
waves do not exhibit conspicuous wave nonlinearity.
(7) Wave nonlinearity is weakened by wave breaking inside the surf zone and is
eventually lost near the shoreline with return to linear wave features.
(8) Probability density function (pdf ) of wave heights gradually deviates from
the Rayleigh as waves propagate toward the shore, with the largest deviation
taking place at the middle of the surf zone. Near the shoreline, the wave height
distribution approaches the Rayleigh again.
(9) Differences between characteristic wave heights such as H1/10 , H1/3 , and Hrms
decrease in the middle of the surf zone, but they return to those by the Rayleigh
near the shoreline. The PEGBIS model can simulate such changes of charac-
teristic wave heights.
(10) Various existing models of random wave breaking produce quite different
prediction of wave heights in the surf zone. Careful selection of a random
wave-breaking model should be made so that the model will yield reliable
information for engineering applications.
Acknowledgment
References
1. Y. Goda, How much do we know about wave breaking in the nearshore waters, Proc.
4th Int. Conf. Asian and Pacific Coast (APAC 2007), Nanjin, China, CD-RM (2007).
2. H. Yamada and T. Shiotani, On the highest water waves of permanent type, Bull.
Disaster Prevention Res. Inst., Kyoto Univ. 18–2(135), 1–22 (1968).
3. Y. Goda, A synthesis of breaker indices, Trans. Japan Soc. Civil Eng. 2(2), 39–49
(1970).
4. H. Yamada, G. Kimura and J. Okabe, Precise determination of the solitary waves of
extreme height on water of a uniform depth, Rep. Res. Inst. Appl. Mech., Kyushu
Univ. XVI(52), 15–32 (1968).
5. J. McCowan, On the highest waves in water, Phil. Mag. Ser. 5, 36, 351–358 (1894).
6. R. Miche, Mouvements ondulatoires de lamer en profondeur ou décroissante, Annales
de Ponts et Chaussées 19, 370–406 (1944).
7. K. Nadaoka, O. Ono and H. Kurihara, Near-crest pressure gradient of irregular water
waves approaching to break, Proc. Coastal Dynamics ’97 (1997), pp. 255–264.
8. G. M. Kaminsky and N. C. Kraus, Evaluation of depth-limited wave breaking criteria,
Second Int. Symp. Ocean Waves Measurement and Analysis, WAVES 1995 (1995),
pp. 180–193.
9. J. W. Kamphuis, Incipient wave breaking, Coastal Eng. 15, 185–203 (1991).
10. W. Rattanapitikon and T. Shibayama, Verification and modification of breaker height
formulas, Coastal Eng. J. 42(4), 389–406 (2000).
July 31, 2009 8:18 9.75in x 6.5in b684-ch04 FA
112 Y. Goda
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 113
35. Y. Goda and M. Kudaka, On the role of spectral width and shape parameters in
control of individual wave height distribution, Coastal Eng. J. 49(3), 311–335 (2007).
36. E. F. Thompson and C. L. Vincent, Significant wave height for shallow water design,
J. Waterway, Port, Coastal, Ocean Eng. ASCE 111(5), 828–841 (1985).
37. M. S. Longuet-Higgins, On the distribution of the heights of sea waves: Some effects
of nonlinearity and finite band effects, J. Geophys. Res. 85(C3), 1519–1523 (1980).
38. J. I. Collins, Probabilities of breaking wave characteristics, Proc. 12th Int. Coastal
Eng. Conf., Washington, DC, ASCE (1970) pp. 399–414.
39. J. A. Battjes, Setup due to irregular wave, Proc. 13th Int. Conf. Coastal Eng., Van-
couver, ASCE (1972) pp. 1993–2004.
40. C. T. Kuo and S. T. Kuo, Effect of wave breaking on statistical distribution of wave
heights, Proc. Civil Eng. Ocean, ASCE (1974) pp. 1211–1231.
41. T. E. Baldock, P. Holmes, S. Bunker and P. van Weert, Cross-shore hydrodynamics
with an unsaturated surf zone, Coastal Eng. 34(3–4), 173–196 (1998).
42. F. C. K. Ting, Laboratory study of wave and turbulence velocities in a broad-banded
irregular wave surf zone, Coastal Eng. 43, 183–208 (2001).
43. F. C. K. Ting, Laboratory study of wave and turbulence characteristics in a narrow-
banded irregular breaking waves, Coastal Eng. 46, 291–313 (2002).
44. J. A. Battjes and H. W. Groenendijk, Wave height distributions on shallow foreshores,
Coastal Eng. 40, 161–182 (2000).
45. Y. Goda, A 2-D random wave transformation model with gradational breaker index,
Coastal Eng. J. 46(1), 1–38 (2004).
46. Y. Goda, Examination of the influence of several factors on longshore current compu-
tation with random waves, Coastal Eng. 53, 156–170 (2006).
47. J. A. Battjes and J. P. F. M. Janssen, Energy loss and set-up due to breaking of
random waves, Proc. 16th Int. Conf. Coastal Eng., Hamburg, ASCE (1978), pp. 1–19.
48. E. B. Thornton and R. T. Guza, Transformation of wave height distribution,
J. Geophys. Res. 88(C10), 5925–5938 (1983).
49. M. Larson and N. C. Kraus, Numerical model of longshore current for bar and trough
beaches, J. Waterway, Port, Coastal, Ocean Eng., ASCE 117(4), 326–347 (1991).
50. W. R. Dally, R. G. Dean and R. A. Darlymple, Wave height variation across beaches
of arbitrary profile, J. Geophys. Res. 90(C6), 11917–11927 (1985).
51. J. M. Smith, Breaking in a spectral wave model, Ocean Wave Measurement and
Analysis (Proc. WAVES 2001), ASCE (2001) pp. 1022–1031.
52. K. Hirayama, N. Hara and T. Hiraishi, Application of bore model to non-linear wave
transformation, Proc. 13th Int. Offshore Polar Eng. Conf. (2003), pp. 796–801.
53. K. Hirayama and T. Hiraishi, A Boussinesq model for wave breaking and run-up in
a coastal zone, 1D, Fifth Int. Symp. Ocean Waves Measurement Analysis, WAVES
2005, CD-ROM, No. 151 (2005), pp. 1–10.
July 31, 2009 8:18 9.75in x 6.5in b684-ch04 FA
Chapter 5
Nobuhito Mori
Disaster Prevention Research Institute
Kyoto University, Gokasho, Uji, Kyoto 611-0011, Japan
mori@oceanwave.jp
Shohachi Kakuno
Osaka City University
3-3-138 Sugimoto, Sumiyoshi-ku, Osaka 558-8585, Japan
hachi@civil.eng.osaka-cu.ac.jp
Daniel T. Cox
Oregon State University, Corvallis, OR 97331-4501, USA
dan.cox@oregonstate.edu
This chapter presents a brief summary of aeration in the surf zone, beginning
with a review of air–water characteristics in surf zone waves. Second, measure-
ments techniques of the bulk of air and bubbles induced by breaking waves in
the surf zone are described, and third, the bulk of air and bubble characteristics
are summarized based on the in situ and visualization laboratory measurements.
Finally, the gas transfer in the surf zone is described and related to the wave
characteristics.
5.1. Introduction
Ocean surface gravity wave propagation from offshore to shoreline is often regarded
as a single phase flow using potential flow theories or solving the Navier–Stokes
equation. Generally, the single phase flow approach to ocean waves is successful for
simulating wave transformation in the coastal area. However, this assumption breaks
down because waves become steep and start to break owing to the bottom bathy-
metric effects in the near shore. The wave-breaking process creates dense bubble
plumes, dissipates energy and generates turbulence in the surf zone. Figure 5.1
shows an example of the crest of wave breaking on a gentle slope. Most of the waves
115
July 31, 2009 8:18 9.75in x 6.5in b684-ch05 FA
Fig. 5.1. Image of breaking wave crest at an initiation of breaking in the surf zone.
can be regarded as a single (liquid) phase flow, and only the near-crest region is
aerated by the jet attachment to the water surface. An accurate estimation of the
bulk of the void fraction, bubble size, and population distributions in the surf zone
are important for understanding the two-phase flow characteristics, solving engi-
neering problems, and elucidating environmental mechanisms of the coastal area.1
For example, Peregrine2 summarized that the entrapped air of breaking wave
reduces the wave impact pressure owing to its greater compressibility compared
with the single-phase approach. There is also a significant difference between salt
water and fresh water experiments on the wave impact pressure (bubbles generated
by turbulence in saltwater are smaller than fresh water), and the entrapped air
reduced the pressure by approximately 10%.3 The compressibility owing to air–
water mixture decreases the velocity of sound and is being used to estimate large-
scale prototype impacts, since the usual Froude scaling is unlikely to be correct
for engineering problems. Therefore, the connection between air-mixture, bubble
distribution, and wave-breaking-induced turbulence is essential to understand gas–
liquid interactions in the surf zone.
The air bubble plume is always present in the surf zone breaking waves. Exper-
imental data show that the entrained air is proportional to the energy dissipation.4
For these breaking waves, entrained air enhances the gas-exchange at the water
surface. For example, the dissolved oxygen is associated with the breaking waves in
the coastal waters,5,6 although the carbon dioxide gas transfer can be regarded as
a function of wind speed.
Understanding air entrainment and bubble distribution for the depth-limited
breaking of surf zone waves is limited in comparison with the wind-wave breaking.
The air bubble distributions under the wind-wave breaking in the upper layer of the
ocean surface has been summarized in detail by Thorpe.7 The wind-wave bubble
formation and its dependence on the turbulent dissipation rate of fluids have been
discussed,8 and it is proposed a −2 power-law scaling with bubble diameter. The
−10/3 power-law scaling is alternatively proposed based on the discussion of bubble
fragmentation due to strong turbulent shear flow. These bubble size scaling laws are
still under development.9 However, the knowledge of wind-wave breaking induced
July 31, 2009 8:18 9.75in x 6.5in b684-ch05 FA
bubbles and associated gas transfer has become well-known issue over the last
decade.10
Despite the fruitful knowledge of air entrainment of wind-wave breaking, few
studies of air entrainment for surf zone wave breaking exist. It has been found that
there is little difference in the bubble populations beneath the mechanically gen-
erated surface waves in saltwater and fresh water.11 On the other hand, it has been
reported that the void fraction depends on the turbulent intensity in the bore region
of surf zone waves.12 The large-scale bubbles entrained by breaking waves in the
surf zone are split by strong local turbulent shear at the bubble scale.13 However,
neither qualitative nor quantitative bubble characteristics in the surf zone and con-
nections between bubble characteristics and wave breaking are well known owing
to the lack of detailed observations. This detailed information of two-phase flow
characteristics is required for mathematical modeling to solve both basic scientific
interests and applied engineering problems.
There are several measurement methods for bubble population and size distribu-
tions, i.e., in situ measurements,14,15 video or photographic measurements,16 laser
measurements,17 and acoustic measurements.18 The bubble size measurements using
lasers probe such as a phase Doppler anemometer show high accuracy in small-scale
bubble measurements smaller than 500 µm but is of limited use in the presence
of high void fractions and large-scale bubble measurements due to fundamental
instrument limitations. Acoustic measurements of bubbles are useful in deepwater
but have limitation for very shallow water region owing to the multi-reflection of
sound beams. Therefore, conventional optical or resistivity type void probes are
useful for the surf zone breaking waves. The electrical conductivity probe (so-called
void probe) has been generally used for the high void fractions.19 This type of void
probe is robust and easy to handle for both wave flume experiments and field obser-
vations. Therefore, the void probe measurements have been widely used for the two-
phase flow measurement in the surf zone.12,14,20 However, the in situ measurement
has clear limitation for inhomogeneous surf zone breaking phenomena.
The photographic or imaging technique has been advanced to in situ measure-
ments which can cover the spatial and temporal variations of the two-phase flow
characteristics in the surf zone. Recent photographic studies have illustrated the dis-
integration of entrapped air cavities divided into bubbles.13,21 Several imaging tech-
niques have been proposed to measure the bubble characteristics in the surf zone.
One is an application of Particle Image Velocimetry (PIV) method with forward and
backward lighting.22,23 This method basically correlates the scattered light intensity
and bubble plume intensity, and it can be used for high void fraction region over 20%,
although it cannot measure the bubble shape itself. The other imaging technique is a
bubble shape tracking method with backlighting.24 This method is a so-called shadow
graph method because the bubble shape is enhanced by a backlight source toward a
camera. The shadow graph method can measure both bubble shape and velocity simul-
taneously. However, the density of the bubbles has to be within a certain limit to sep-
arate individual bubbles from the image by the shadow graph method.
July 31, 2009 8:18 9.75in x 6.5in b684-ch05 FA
In summary, there is no ideal method that can be used for universal purpose of
bubble measurements and one can choose appropriate method(s) for a particular
purpose.
Figure 5.2 shows successive images of breaking wave taken by high-speed video
camera in the surf zone. Figure 5.2(a) is just after attachment of jet on the water
surface which corresponds to the initial stage of bubble generation. Figures 5.2(b)
and 5.2(c) are 42 ms and 108 ms after image (a), respectively. The lifetime of the
large volume of entrained air can be categorized into two phases. The first phase
is the initial stage of breaking waves [Figs. 5.2(a) and 5.2(b)]. The newly injected
air column produces pluses of sound and, therefore, this phase is regarded as the
acoustically active phase.21 In the active phase, the trapped volume of air between
the overturning jet and water surface has fragmented and injected into the deeper
region, and the air pocket is fragmented into small size bubbles. This phase occurs
within 1 s of the initial stage of wave breaking. The second phase is the acoustically
silent phase. Most of the bubbles are advected by the surface roller of the carrier
waves with buoyancy, and some of them are trapped by oblique descending eddy
in this phase [Figs. 5.2(c)]. In the active phase, the void fraction ratio is relatively
high and there is no clear formation of bubbles in that region. Therefore, the only
possible approach to this phase is to account for the total amount of air, which is
the summation of the air pocket and bubbles. The silent phase has a lower void
fraction and is more dispersive, and it is possible to discriminate individual bubbles.
These injected air penetrate into a deeper region during the active phase. After
that, the volume of air and/or bubbles are advected by the wave motion due to their
buoyancy. Therefore, the gas phase characteristics in the surf zone, such as the void
fraction, are time–space variables and are inhomogeneous in both time and space.
The next part of this chapter discusses the total amount of air, bubble size
distributions, and gas-transfer in the surf zone, respectively.
(a)
(b)
(c)
Fig. 5.2. Formation of bubble plume under the breaking wave in the surf zone. (a) Attachment
of jet on the surface (initiation of bubble plume generation); (b) Active phase of bubble plume
generation [42 ms after (a)]; (c) Bubble plume in silent phase [108 ms after (a)].
July 31, 2009 8:18 9.75in x 6.5in b684-ch05 FA
10
(a)
<η> [cm]
5
0
5
1500
(b) z/Hb=0.033
1000
500
σu, σv, σw [cm /s ]
2 2
0
1500
2
(c) z/Hb=0.013
1000
2
500
2
0
1500
(d) z/Hb=0.0
1000
500
0
(e) z/Hb=0.033
0.2
<α >
z/Hb=0.013
z/Hb=0.0
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
t/T
Fig. 5.3. Example of phase-averaged surface elevation η and turbulence components 1.0 m from
B.P. in Case 1. (a) Free surface η (solid) and standard deviation η ± ση ; (b)–(d) temporal variation
of σu (thick solid), σv (thin solid), and σw (dotted) and (e) the phase-averaged void fraction α.20
The highest void fraction occurs at the wave front, and the void fraction is decreased
rapidly in time. The temporal variation of the void fraction depends on the air
supply from the surface, advection, and buoyancy, and these effects are difficult
to discriminate from each other. Although the process is complex, the normalized
temporal characteristics show a universal behavior. Similar laboratory experiments
were observed by several researchers.12,25,26 Furthermore, Cox and Shin12 found the
self-similar exponential decay of the temporal series of void fraction normalized by
the averaged value in different types of breaking waves. In addition, they reported
that the maximum void fraction in the time series was three to four times higher
than its average value.
C0 C0 C0 C0 C0
0 0.1 0.2 0 0.1 0.2 0 0.1 0.2 0 0.1 0.2 0 0.1 0.2
0 0 0 0 0
-2 -2 -2 -2 -2
-4 -4 -4 2 -4 -4
-6 -6 -6 -6 -6
-8 -8 -8 -8 -8
x-xb = 0.4 m -10 x-xb = 0.6 m -10 x-xb = 0.8 m -10 x-xb = 1.1 m -10 x-xb 1.35 m
2
×10
15
z/H =0.033
10 b
5
0 2
⋅ 10
15
z/H =0.013
10 b
5
α×104, k [cm2/s2]
0 2
⋅ 10
15
z/H =0.0
10 b
5
0
⋅ 102
15
z/Hb=-0.013
10
5
0
0 2 4 6 8 10 12 14
xb/Hb
Fig. 5.5. Horizontal relationship between time-averaged void fraction α (solid line) and time-
averaged TKE k (dotted line).20
Figure 5.5 shows experimental results of void fraction and total kinetic energy
(TKE) variations from the breaking point xb to the shore line.20 The spatial varia-
tions of the time-averaged void fraction α and time-averaged TKE, k show similar
tendency, although the peak of α is located approximately 0.2 × Hb ∼ 2 × Hb
seaward of the peak of k, as shown in Fig. 5.5. The short spatial lag between α and
k might be associated with the spatial difference of entrapped air and associated
turbulence. The spatial variations of the turbulence and void fraction are highly cor-
related near the mean water level. The entrained bulk of air and bubbles generate
turbulence, and turbulence splits the coarse bubbles into fine bubbles in this area.
Moreover, the TKE and void fraction increased by an order of magnitude from the
trough level (z/Hb = −0.013) to above the mean water level (z/Hb = 0.033); thus,
the two-phase flow characteristics are remarkably strong near the wave crest.
Neglecting the short spatial lag between the void fraction and TKE, the
direct comparison between the time-averaged void fraction α and normalized time-
averaged TKE k of all the measuring locations is shown in Fig. 5.6. For small
values of α or k, the relationship between the two shows a high correlation and
becomes scattered at high void fraction or TKE region. Overall, the relationship
between time-averaged void fraction and time-averaged TKE shows high corre-
lation12 as does the incident wave-dependence on the relationship between α and k.
July 31, 2009 8:18 9.75in x 6.5in b684-ch05 FA
0.14
0.12
0.1
0.08
α
0.06
0.04
0.02 Case1
Case2
Case3
0
0 0.1 0.2 0.3 0.4
1/2
( k/gh )
Fig. 5.6. Relationship between time-averaged TKE k and time-averaged void fraction α. (•: Case
1, +: Case 2, : Case 3).
The relation between the void fraction and TKE, the scale-dependence of void
fraction, and further theoretical considerations will be required to find the universal
relationship between the turbulence characteristics of depth-limited breaking waves
and air entrainment.
Fig. 5.7. Time series of bubble size distribution.24 (a) Spilling breaker; (b) Plunging breaker.
of the bubble number density distribution (Fig. 5.7) shows a similar tendency, but
the small-scale bubbles survive longer than large-scale bubbles.
Generally, entrapped air by breaking waves is split into small bubbles owing
to turbulence shear of surrounding flows. The early work of this problem for a
general condition is discussed by Kolmogorov32 and Hinze33 in the middle of the
last century, and a large bibliography has been generated.34 Longuet-Higgins35 pro-
posed a crushing air cavity model for the bubble distributions after the formation
of a bubble plume and shows good agreement of bubble aeration in general con-
dition, qualitatively. Figure 5.8 shows the instantaneous and time-averaged bubble
size spectra with the estimated power law measured by an imaging technique.24
The instantaneous bubble size spectra corresponds to the time of the maximum
void fraction observed. There is no significant difference of bubble size distribution
July 31, 2009 8:18 9.75in x 6.5in b684-ch05 FA
2
10
2
Number of bubbles /cm
0
10
−2
10
time-averaged
instantaneous
−1.0
d
−4 −3.4
10 d
−0.9
d
−1 0 1
10 10 10
diameter [mm]
between instantaneous and time-averaged spectra at small scale (diameter < 1 mm).
On the other hand, two scaling laws can be seen in the bubble size distribution. The
small and large size bubble power scaling laws are d−1 and d−3–−4 , respectively.
Garrett et al.9 proposed semi-empirical −10/3 power-law scaling for the bubble
which is larger than the Hinze scale, based on the discussion of bubble fragmentation
in the strong turbulence flow below the trough level. On the other hand, Deane and
Stokes21 empirically proposed a −3/2 power-law scaling, smaller than the Hinze
scale in the acoustically active phase near the crest. The node point between the
two power scaling corresponds to the Hinze scale of the bubble splitting theory.33
The measured power-law scalings of the small size bubble shown in Fig. 5.8 are close
to −3/2 and −10/3 power laws, respectively. These power-law scalings can be seen
in different conditions in mechanically generated waves in the flume.13,20,21
The Hinze scale dH 21 of bubble splitting can be given by
ρw
dH = 2−8/5 c ε−2/5 , (5.4)
γ
where ρw is the water density, γ is the surface tension coefficient, ε is turbulence
energy dissipation at the bubble diameter, and c = (W ec )3/5 , where W ec is the
critical Weber number. The Weber number W e is defined as
ρw 2
We = δu d, (5.5)
γ
where δu is turbulence velocity at the bubble scale d. The turbulence velocity δu at
the bubble scale in the inertial range can be given by δu = 2ε2/3 d2/3 . If the energy
dissipation rate ε can be estimated by wave energy dissipation, the Hinze scale can
be obtained. For example, if ε = 12 W/kg, the Hinze scale is 2 mm in diameter.
These theoretical predictions agree fairly well with the experimental data,20,21
although the Hinze scale is formulated under the assumption of isotropic and
homogeneous turbulence flow.
July 31, 2009 8:18 9.75in x 6.5in b684-ch05 FA
For moderately solvable gases such as carbon dioxide and oxygen, the gas transfer
across the air–sea interface is controlled by the aqueous condition, and is parame-
terized as a function of a gas transfer velocity kL
kL = αSc−1/2 u∗ , (5.6)
where Sc = νκ is the Schmidt number, α is the empirical coefficient, and u∗ is
the friction velocity, respectively.36 The role of bubbles in the exchange of gas in
the surf zone is poorly understood in comparison to the wind-wave breaking. An
analogy of wind-wave breaking to surf zone breaking waves is similar, but surface
renewal process and the bubble-mediated gas transfer are different from each other.
For example, the surface renewal by the wind stress is important for the wind-
wave scenario, but the turbulence generated by the wave overturning is dominant
in the surf zone. Moreover, the bubble-mediated gas transfer process depends on
the bubble population and size distribution in the surf zone. For the wind-wave
case, the large-scale bubbles with a diameter greater than approximately 1 mm play
an important role in air–sea gas exchange.37,38 However, there are a few confident
prediction models that exist for macro or microscopic bubble-mediated gas transfer
for surf zone breaking waves.
The oxygen gas transfer is important for fish and marine mammals near the
coastal region. Therefore, oxygen gas transfer, rather than carbon dioxide, plays a
vital role in aquatic environment in the surf zone.39 There are significant difference
between the wind-waves and surf zone waves generally. For example, the spatial
homogeneity can be assumed in wind-waves, but there is a strong spatial dependence
for surf zone waves. Therefore, a different approach to predict the gas transfer is
required. One may assume that the gas transfer is related to the total water mass
volume of surf zone (Fig. 5.9). The general expression of the gas transfer in an
aeration tank of constant volume can be formulated:
∂C A
= kL ∆C = k2 ∆C, (5.7)
∂t V
where C is the gas concentration, ∆C is the concentration difference between liquid
and gas phase, A is the aeration (surface) area, kL is the gas transfer velocity,
Vg
XTi
Xb
Fig. 5.10. Total bubble volume distribution in the surf zone (Xb : length of surf zone, Vg : bubble
volume per wave period).
July 31, 2009 8:18 9.75in x 6.5in b684-ch05 FA
Fig. 5.11. Gas transfer flow rate and bottom bathymetry in the surf zone (•: 1/10 slope, ◦: 1/20
slope, : 1/30 slope).
Research to estimate the gas transfer velocity kL from the wave characteristics
through the wave energy dissipation is ongoing.
In this chapter, we summarize our general review of gas–water two-phase flow char-
acteristics in the surf zone. There are several measurement methods for the aeration
in the surf zone, such as in situ measurements, video or photographic measure-
ments, laser measurements, and acoustic measurements. There is no one appropriate
method to measure the volume of air and bubbles in the surf zone. The spatial and
temporal characteristics of void fraction are summarized, and the bubble size dis-
tribution is described with simple theories. Furthermore, the gas transfer prediction
method is presented as a function of the bottom slope and wave characteristics for
the surf zone waves.
There are many other problems such as salinity effect, scale effects, and more
mathematical modeling should be investigated in the future.
References
26. D. Cox, N. Kobayashi and A. Okayasu, Vertical variations of fluid velocities and shear
stress in surf zones, Proc. 24th Int. Conf. on Coastal Eng., ASCE (1994), pp. 98–112.
27. A. Hoque and S. Aoki, Distributions of void fraction under breaking waves in the surf
zone, Ocean Eng. 32, 1829–1840 (2005).
28. M. Loewen, Inside whitecaps, Nature 418, 830 (2002).
29. R. Cipriano and D. Blanchard, Bubble and aerosol spectra produced by a laboratory
breaking waves, J. Geophys. Res. C86, 8085–8092 (1981).
30. C. Garrett, M. Li and D. Farmer, The connection between bubble size and energy
dissipation rates in the upper ocean, J. Phys. Oceanogr. 30, 2163–2171 (2000).
31. P. Bonmarin, Geometric properties of deep-water breaking waves, J. Fluid Mech. 209,
405–433 (1989).
32. A. Kolmogorov, On the breakage of drops in a turbulent flow, Dokl. Akad. Navk. SSSR
66, 825–828 (1949).
33. J. Hinze, Fundamentals of the hydrodynamic mechanism of splitting in dispersion
processes, Am. Inst. Chem. Eng. J. 1, 289–295 (1955).
34. C. Martı́nez-Bazán, J. Montanés and J. Lasheras, On the breakup of air bubble
injected into a fully developed turbulent flow. Part I: Breakup frequency, J. Fluid
Mech. 401, 157–182 (1999).
35. M. Longuet-Higgins, The crushing of air cavities in a liquid, Proc. Roy. Soc. London A
439, 611–626 (1992).
36. W. Melville, The role of surface-wave breaking in air-sea interaction, Ann. Rev. Fluid
Mechanics 28, 279–321 (1996).
37. L. Memery and L. Marlivat, Modeling of gas flux through bubbles at the air-water
interface, Tellus 37B, 272–285 (1985).
38. R. Keeling, On the role of large bubbles in air-sea gas exchange and supersaturation
in the ocean, J. Marine Res. 51, 237–271 (1993).
39. C. Moutzouris and E. Daniil, Water oxygenation in the vicinity of coastal struc-
tures due to wave breaking, Proc. 24th Int. Conf. Coastal Eng., ASCE (1994),
pp. 3167–3177.
40. W. Eckenfelder Jr., Absorption of oxygen from air bubbles in water, J. Sanitary Eng.
Div. Proc. ASCE, 89–99 (1959).
41. S. Kakuno, D. B. Moog, T. Tatekawa, K. Takemura and T. Yamagishi, The effect of
bubble on air-water oxygen transfer in the breaker zone, in Gas Transfer at Water
Surfaces, eds. M. Donelan, W. Drennan, E. Saltzman and R. Wanninkhof (AGU,
Washington, DC, 2002), pp. 265–277.
July 31, 2009 8:18 9.75in x 6.5in b684-ch06 FA
Chapter 6
Freak Wave
Nobuhito Mori
Disaster Prevention Research Institute
Kyoto University, Gokasho, Uji, Kyoto 611-0011, Japan
mori@oceanwave.jp
In this chapter, we describe mechanisms that lead to the occurrence of freak waves
in the ocean. It is now generally recognized that freak waves can be generated
by any one of four possible mechanisms. The traditional mechanism is a simple
linear superposition of waves, the theory of which is reviewed here. The newest
mechanism attempts to include the third-order nonlinearities that depart from
linear wave theory. Therefore, we present here a state-of-the-art review that is
based on nonlinear wave dynamics.
6.1. Introduction
The history contains many reports of encounters with monstrous waves.1 For
example, the captain of the cargo vessel Junior reported a wave estimated to be
100 ft high and the famous reliable report was that of the wave encountered by the
U.S.S. Ramapo in the North Pacific in 1993; that wave was estimated to be 112 ft
high. There are many more reports of encounter of similar waves in the history of
the seas. Over the last decade, more evidence has accumulated for such extreme
waves. This may be because of an increase in the number of measurement locations
together with the development of more accurate field measurements, for example,
several extreme wave events, the so-called freak waves, recorded in the North Sea.2
A sample wave profile is shown in Fig. 6.1, which was extracted from the data pro-
vided by wave radar shooting downward from the platform to the sea surface in the
North Sea. The data were collected over 12–15 hours at a depth of 40 m. Since the
early 1990s such an event has come to be called a freak wave or a rogue wave. More
evidence has been obtained by analyzing field data from the North Sea,3,4 the Sea
of Japan,5,6 and the Gulf of Mexico.7
The first clear definition of a freak wave was proposed by Klinting and Sand2 ;
in its original form, the definition required a wave to satisfy three conditions to be
131
July 31, 2009 8:18 9.75in x 6.5in b684-ch06 FA
132 N. Mori
Fig. 6.1. Freak wave recorded on the 24th November, 1981 at the Gorm Field in the Danish
Sector of the North Sea.8
It is now accepted that at least four mechanisms can explain the formation of freak
waves. A problem common to these four is whether a freak wave obeys an ordinary
July 31, 2009 8:18 9.75in x 6.5in b684-ch06 FA
statistical law or whether other physical mechanisms come into play. That is, if we
are to be able to predict freak waves, then any underlying mechanism must lead to
reliable statistical model for extreme wave events.
First Mechanism. The first mechanism is simply that a freak wave can be caused
by a linear superposition of waves. In this case, the probability distribution for
wave height in the limit of the narrow-band approximation obeys a Rayleigh dis-
tribution10 ; corrections due to finite spectral band width have been obtained.11–13
Wave crest statistics can be obtained by using a second-order theory.14 The proba-
bility distribution for wave crests has been found in a narrow-band approximation15
and for finite band width.16 This mechanism is theoretically well established and
relatively easy to follow; the basic theory will be reviewed in Sec. 6.3.
Second mechanism. The second mechanism is that freak wave can result when
waves interact with currents. When this situation occurs, ray theory can be used
to explain the formation of extreme waves. However, it is not yet known how the
statistical properties of surface are related to the properties of currents. From the
studies of freak waves generated over a random current, it has been found that wave
focusing with 1 m/s velocity fluctuations occurs in the uniform current.17
Third Mechanism. A third mechanism has its basis in crossing-sea states (i.e.,
two-sea systems). An example is a swell and a wind sea in which the two have
different directions coexisting over some region of the ocean, or diffraction of waves
behind an island. Usually, both the wave-current mechanism and the two-sea state
mechanism occur in specific locations, such as along the Angus current; therefore,
freak waves caused by either of these mechanisms are relatively predictable.
Fourth Mechanism. In the fourth mechanism, extreme wave events result from
the high-order nonlinear interactions; specifically, the underlying mechanism is four-
wave interactions.18,19 Numerical and experimental studies have demonstrated that
freak-like waves can be generated frequently in a two-dimensional wave flume,
even in the absence of a current, or wave refraction, or wave diffraction18,20–22 ;
see Fig. 6.2. Further, numerical studies clearly indicate that a freak wave having
a single, steep crest can be generated in deep-water by the third-order nonlinear
interactions.18
Freak wave generation is sometimes discussed in the context of the Benjamin–
Feir instability in deep-water waves because of the similarity in the steep wave
134 N. Mori
profile.18,22 Over the last two decades, the Benjamin–Feir instability of deep-
water gravity waves has been studied using the nonlinear Schrödinger-type
equations,23–25 mode-coupling equations,26 pseudo-spectral methods,27 and experi-
ments.28 However, the Benjamin–Feir instability vanishes when the wave spectrum
is sufficiently broad. This means, there is a discrepancy between the nonlinear
behaviors of periodic waves and random waves. Janssen19 investigated freak waves
that occur as a consequence of four-wave interactions, including the effects of non-
resonant four-wave interactions. Using Monte Carlo simulations of the Zakharov
equation, he found that homogeneous nonlinear interactions give rise to deviations
from the Gaussian distribution for the surface elevation.
The fourth mechanism for freak wave generation requires additional knowledge
before predictions can be made. Relationship between kurtosis and freak wave events
has been studied using laboratory data,29 numerical results,30 and field data.7 North
Sea data shows that the maximum crest, normalized by its wave heights, correlates
with skewness, while the wave heights, normalized by significant wave height, cor-
relate with the kurtosis.7 This is because the kurtosis is the fourth moment of the
probability density function; hence, it measures the relevance of tails in a distri-
bution.
These results explain the general characteristics of a freak wave. (a) A second-
order nonlinearity causes wave asymmetry over the short-time evolution, O(1/ε2 ),
of a wave train (Here ε is a small parameter that corresponds to the steepness of
deep-water waves). (b) But the kurtosis is associated with the third-order nonlinear
interactions.14 Therefore, it may be that third-order nonlinear interactions are the
fundamental factors contributing to the occurrence of freak waves. However, the
timescales for such interactions are longer than those for second-order interactions,
O(1/ε4 ), and they are two orders of magnitude longer than those for second-order
nonlinearities.
The formal relation between the kurtosis and the exceedance probability for wave
height have been discussed (Mori and Janssen31 ; see also Tayfun and Lo32 ; Mori
and Yasuda33 ). The kurtosis enters into the distribution function as a correction to
the Rayleigh distribution: when the kurtosis tends to three, the expected Gaussian
values, and then the distribution function tends to the Rayleigh distribution. In this
context, changes in the kurtosis can be evaluated once the evolution of the wave
spectrum is known.19 The wave height exceedance probability has been derived as a
function of the number of waves.31 This relation will be explained in a later part in
this chapter. In all these studies, an ultimate goal is the determination of the shape
of the exceedance probability function P (H). Once an analytical form for P (H) is
found for all sea conditions (for example, as a function of the wave spectrum), then
the probability of a wave exceeding a certain height can be estimated directly.
One difficulty in studying extreme wave events is the verification of theory.
Statistical theories are evaluated under the assumptions of stationarity; however,
the real seas change states both spatially and temporally. Therefore, statistical
theories are difficult to verify.34 In addition, freak wave prediction requires us to
predict the maximum wave height distribution. However, although study of the
maximum wave height distribution is necessary to verify a theory for freak wave
July 31, 2009 8:18 9.75in x 6.5in b684-ch06 FA
6
Surface Elevation (m)
−2
−4
45000 45100 45200 45300 45400 45500 45600
Time (s)
0.5
0.4
0.3
0.2
0.1
10 20 30 40 50 60 70 80 90 100
−5
79400 79500 79600 79700 79800 79900 80000
Time (s)
0.4
0.3
0.2
0.1
10 20 30 40 50 60 70 80
136 N. Mori
The surface gravity waves’ celerity depends on the wave length owing to dispersive
effects in deep-water. The nonlinear effects slightly modify the dispersion relation
in the order of wave steepness.35 The order of nonlinear effects, both dispersion
and wave profile, are square or cubic of the wave steepness if there is no resonant
interactions in the system. Therefore, in general, the ocean waves can be regarded
as a linear, ergodic, homogeneous, stationary, Gaussian processes. Because of these
assumptions, the probability density function for water surface elevation of unidi-
rectional waves can be expressed as the Gaussian distribution
1 η2
p(η) = √ exp − , (6.1)
2πm0 m0
∞
2 2
σ = ηrms = m0 = S(ω)dω , (6.2)
0
√
and η is the surface elevation while σ = m0 is its variance. Let η(t) be the sea
surface elevation at time t and let ζ(t) be an auxiliary variable such that η(t) and
ζ(t) are not correlated. Assuming both η(t) and ζ(t) are real zero-mean functions
with variance σ, we have
ζ(t)
φ(t) = tan−1 , (6.5)
η(t)
where A is the envelope of the wave train and φ is the phase. As there is no corre-
lation between η and ζ, the joint probability density function of η and ζ becomes
1 1
p(η, ζ) = exp − (η2 + ζ 2 ) . (6.6)
2π 2
√
All variables will be normalized by the variance of the surface elevation σ = m0 .
The probability density function (PDF) for the envelope A can now be obtained
by integrating the joint probability distribution
over φ; hence,
2π
p(A) = dφ p(A, φ). (6.8)
0
The results are the usual Rayleigh distributions for wave amplitude
2
p(A) = Ae−(1/2)A (6.9)
138 N. Mori
PDF of wave height p(H) and exceedance probability of the wave height P (H) are
known.37 Thus,
pm (Hmax )dHmax = N [1 − P (Hmax )]N −1 p(Hmax )dHmax , (6.12)
where N is the number of waves in the wave train. For sufficiently large N , we may
use the approximation
N N
lim [1 − P (Hmax )]N lim exp [−N P (Hmax )] . (6.13)
N →∞ N →∞
Substituting Eqs. (6.11) and (6.13) into Eq. (6.12), gives the PDF for the maximum
wave height pm ,
N 2
2
pm (Hmax )dHmax = Hmax e−Hmax /8 exp −N e−Hmax /8 dHmax , (6.14)
4
and the exceedance probability of maximum wave height Pm ,
2
Pm (Hmax ) = 1 − exp −N e−Hmax /8 . (6.15)
Note that Eqs. (6.14) and (6.15) are evaluated as functions of N number of waves
√
in the train. The freak wave condition in this study becomes Hmax / m0 > 8, and
we obtain from Eq. (6.15) the following simple formula for probability of a freak
wave occurring as a function of N ,
140 N. Mori
Second-order contributions, which are important for the distribution of wave crests,
can be included by using an approach like that used by Tayfun.15,45
The probability distribution function, pm , and the exceedance probability, Pm ,
of the maximum wave height can also be expressed as a function of the fourth
cumulant of the surface elevation κ40 and the number of waves N in the wave train,
N 2
pm (Hmax ) = Hmax e−Hmax /8 [1 + κ40 AH (Hmax )]
4
2
× exp −N e−Hmax /8 [1 + κ40 BH (Hmax )] , (6.30)
2
Pm (Hmax ) = 1 − exp −N e−Hmax /8 [1 + κ40 BH (Hmax )] , (6.31)
where Hmax is the maximum wave height normalized by ηrms . Qualitative agreement
has been found between the theoretical wave height distribution, given by Eq. (6.30),
and field data.31 Equations (6.30) and (6.31) are evaluated as functions of N and
κ40 (or µ4 ). When κ40 = 0, Eqs. (6.30) and (6.31) reduce to Eqs. (6.14) and (6.15),
which are provided by the Rayleigh distribution.
The occurrence probability of a freak wave as a function of N and κ40 , taking
into account nonlinear effects, now becomes
Pfreak = 1 − exp −e−8 N (1 + 8κ40 ) . (6.32)
Using Eq. (6.31) we see that the effect of kurtosis becomes of the same order as the
linear theory for κ40 = 1/8. This corresponds to µ4 = 3.125, which is not a strong
nonlinear condition. Hence, both the effects of finite kurtosis and the number of
waves N are important for determining the probability of maximum wave height in
the nonlinear wave train. It should be emphasized here that the above discussion
does not require any empirical, ad hoc parameters. A different approach, also based
on the dynamics of a cubic nonlinearity, has been used for finding the exceedance
probability for wave height.46 That approach involves an optimization problem,
requiring that at a specific space and time, the Fourier phases are all equal, under
the constraints that the conservation laws of the Zakharov equation are satisfied.
To summarize, quasi-resonant four-wave interactions introduce deviations from
linear statistics for surface elevation; in particular, for weakly nonlinear, narrow-
banded and long-crested wave trains, the kurtosis evolves according to Eq. (6.19).
In the narrow-band approximation, the kurtosis is related to the BFI [defined in
Eq. (6.25)]; the tail of the wave height distribution depends on the kurtosis/BFI and
it increases as the kurtosis increases. Finally, the maximum wave height distribution
depends on the number of waves in the wave train (record length) and on the
kurtosis of the nonlinear wave field, as shown in Eq. (6.30).
Figure 6.4 compares the linear (Rayleigh) theory with those from the present theory
for the occurrence probability of a freak wave, Pfreak ; the plot shows how Pfreak
July 31, 2009 8:18 9.75in x 6.5in b684-ch06 FA
142 N. Mori
0
10
Pfreak
−1
10
Rayleigh
µ4 = 3.1
µ4 = 3.2
µ4 = 3.3
µ4 = 3.4
µ4 = 3.5
−2
10 2 3 4
10 10 10
N
Fig. 6.4. Occurrence probability for freak waves as a function of number of waves N and
kurtosis µ4 .31
Pfreak
Rnonlinear = −1. (6.33)
Pfreak |κ40 =0
For a small number of waves N ≤ 250, the ratio Rfreak depends linearly on µ4 . If
µ4 = 3.1 and N ≤ 500, the occurrence probability of freak waves is 50% more from
the nonlinear theory than from the linear theory. However, the rate of increase in
Rfreak decreases as the number of waves increases. This is because, for very large
number of waves, the maximum wave height almost always exceeds 2H1/3 , even in
linear wave trains.
July 31, 2009 8:18 9.75in x 6.5in b684-ch06 FA
350 N=100
N=250
N=500
300 N=1000
N=5000
250
Rfreak [%]
200
150
100
50
0
3 3.1 3.2 3.3 3.4 3.5
µ4
Fig. 6.5. Ratio of freak wave occurrence predicted by Eq. (6.31) to the Rayleigh theory.31
The consequences of the maximum wave height distribution [Eq. (6.30)] are
generally hard to verify, not only because Eq. (6.30) depends on the number of
waves but also because large amount of data are required to characterize nonlinear
effects. In the following discussion, we compare the above theories with a set of long
time-series experimental data47 ; these data appear to be sufficiently long to make
useful comparisons.
Figure 6.6 shows the probability of occurrence of freak waves Hmax /ηrms ≥ 8
for short time records, as a function of kurtosis. We assumed 100 waves per record
and counted the number of freak waves. In the experimental data, the occurrence
probability for freak waves shows a clear dependence on the kurtosis (this effect of
course is not described by the Rayleigh distribution). But the experimental data
appear to depend quadratically on the kurtosis, while the nonlinear theory Eq. (6.31)
predicts a linear dependence. This is probably because the nonlinear theory includes
only the lowest-order correction for nonlinearity: it excludes high-order cumulants.
Figure 6.7 compares the maximum wave height distribution from the experi-
mental data with linear, Eq. (6.14), and nonlinear theories, Eq. (6.30), for N = 150.
In Fig. 6.7(a), the comparison between theory and experiment is shown for a nearby
linear wave condition (µ4 = 3.06). The peak in the observed maximum wave height
distribution is larger than the Rayleigh Hmax distribution and the MER Hmax dis-
tribution with µ4 = 3.06, but the observed distribution is not as wide (we ascribe
this difference to the effects of the finite width of the spectrum). The maximum
wave height distribution from the laboratory data departs from the Rayleigh dis-
tribution [Figs. 6.7(a)–6.7(c)] at some distance from the wavemaker, and the peak
of the observed distribution shifts to larger wave heights. While the Rayleigh Hmax
distribution, Eq. (6.14), is independent of the kurtosis, the MER Hmax distribution,
July 31, 2009 8:18 9.75in x 6.5in b684-ch06 FA
144 N. Mori
N=11900
0.45
Exp.
MER
0.4
Rayleigh
0.35
0.3
Probability of Freak Wave [%]
0.25
0.2
0.15
0.1
0.05
0
3 3.5 4 4.5
µ4
Fig. 6.6. Occurrence probability for freak waves in long recorded wave data: N = 11, 900
(•: Experimental data, solid line: MER distribution from Eq. (6.31), dashed line: Rayleigh
distribution).
Eq. (6.30), follows the behavior of the experimental data. The MER Hmax distri-
bution seems to reproduce the experimental data reasonably well; for very small
kurtosis, it overestimates the experimental data, and for large kurtosis it slightly
underestimates the data. This is consistent with what we have seen in Fig. 6.6 where
the occurrence probability of freak waves is plotted as a function of the kurtosis.
Similar results have been obtained in other studies.31
We also consider the general behavior of the probability density function of
maximum wave height in a nonlinear wave field. To do so, we plot in Fig. 6.8
the expected value of the maximum wave height, indicated by angle brackets,
as a function of µ4 and N . In the figure, we show the numerically integrated
value of Eq. (6.30) together with the ensemble average of the experimental data.
The quantity Hmax /H1/3 in Fig. 6.8 from the Rayleigh theory corresponds to
√
H/ m0 > 8. The dependence of Hmax /H1/3 on µ4 and N is clear in both the
experimental data and Eq. (6.30). Overall, Hmax /H1/3 from the experimental data
is smaller than that from the MER Hmax distribution; nevertheless, the theoretical
and experimental data show similar trends.
The above evidence shows that freak waves can be predicted for given strengths
of nonlinearities, as measured by the kurtosis. The remaining problems, therefore,
July 31, 2009 8:18 9.75in x 6.5in b684-ch06 FA
0.8
(a) µ4 =3.06 Exp.
MER
0.7
Rayleigh
0.6
)
0.5
rms
/η
max
0.4
P(H
0.3
0.2
0.1
0
4 5 6 7 8 9 10 11 12
H /η
max rms
0.8
(b) µ4 =3.43 Exp.
MER
0.7
Rayleigh
0.6
)
0.5
rms
/η
max
0.4
P(H
0.3
0.2
0.1
0
4 5 6 7 8 9 10 11 12
H /η
max rms
0.8
Exp.
(c) µ4 =4.10 MER
0.7 Rayleigh
0.6
P(Hmax/ηrms)
0.5
0.4
0.3
0.2
0.1
0
4 5 6 7 8 9 10 11 12
H max/ηrms
Fig. 6.7. Comparison of maximum wave height distribution Hmax /ηrms with N = 150 (•: exper-
imental data, solid line: MER Hmax distribution, dashed line: Rayleigh Hmax distribution).
July 31, 2009 8:18 9.75in x 6.5in b684-ch06 FA
146 N. Mori
2.2
2.1
2
<Hmax /H1/3 >
1.9
1.8
1.7
1.6
1.5
1.4
1000
800 4
600 3.8
3.6
400 3.4
200 3.2
N 3 µ4
2.2
2.1
2
<Hmax /H1/3 >
1.9
1.8
1.7
1.6
1.5
1.4
1000
800 4
600 3.8
3.6
400 3.4
200 3.2
N 3 µ4
are developing accurate methods for estimating the kurtosis and learning how to
take directional effects into account for the application in the ocean.
This chapter summarized generally accepted mechanisms that cause freak waves
and we have described linear and nonlinear mechanisms in more detail. Four-wave
July 31, 2009 8:18 9.75in x 6.5in b684-ch06 FA
References
148 N. Mori
36. A. Kimura, Probability of the freak wave appearance in a 3-dimensional sea condition,
Proc. 24th Int. Conf. Coastal Eng., Vol. 1, ASCE (1994), pp. 356–369.
37. Y. Goda, Random Seas and Design of Maritime Structures, 2nd Edn. (World Scientific,
2000).
38. I. Lavrenov, The wave energy concentration at the Agulhas current off South Africa,
Natural Hazard 17, 117–127 (1998).
39. A. Osborne, M. Onorato, M. Serio and S. Bertone, The nonlinear dynamics of rogue
waves and holes in deep water gravity wave trains, Phys. Lett. A 275(5–6), 386–393
(2000).
40. S. Haver, Evidences of the existence of freak waves, in Rogue Waves, eds. M. Olagnon
and G. Atanassoulis (IFREMER, 2001), pp. 129–140.
41. M. Olagnon and G. Athanassoulis (eds.), Rogue Waves 2000 (IFREMER, 2000).
42. K. Hasselmann, On the nonlinear energy transfer in gravity-wave spectrum. I. General
theory, J. Fluid Mechanics 12, 481–500 (1962).
43. V. Krasitskii, On reduced equations in the Hamiltonian theory of weakly nonlinear
surface waves, J. Fluid Mechanics 272, 1–20 (1994).
44. F. Edgeworth, On the representation of statistical ferquency by a series, J. Roy. Statist.
Soc. 70, 102–106 (1907).
45. M. A. Tayfun, Statistics of nonlinear wave crests and groups, Ocean Eng. 33(11–12),
1589–1622 (2006).
46. F. Fedele and M. A. Tayfun, Explaining freak waves by a stochastic theory of wave
groups, 25th Int. Conf. Offshore Mechanics and Arctic Engineering 1 (2006).
47. M. Onorato, A. R. Osborne, M. Serio, L. Cavaleri, C. Brandini and C. T. Stansberg,
Extreme waves, modulational instability and second order theory: Wave flume exper-
iments on irregular waves, Eur. J. Mech. B/Fluids 25(5), 586–601 (2006).
48. N. Mori, M. Onorato, P. A. E. M. Janssen, A. R. Osborne and M. Serio, On the extreme
statistics of long crested deep water waves: Theory and experiments, J. Geophys. Res.,
Ocean 112, C09011 (2007), doi: 10.1029/2006JC004024.
49. H. Socquet-Juglard, K. Dysthe, K. Trulsen, H. Krogstad and J. Liu, Probability dis-
tribution of surface gravity waves during spectral changes, J. Fluid Mechanics 542,
195–216 (2005).
50. O. Gramstad and K. Krulsen, Influence of crest and group length on the occurrence
of freak waves, J. Fluid Mechanics 582, 463–472 (2007).
July 31, 2009 8:18 9.75in x 6.5in b684-ch06 FA
Chapter 7
Akira Kimura
Department of Social Systems Engineering
Faculty of Engineering, Tottori University, 680-8552
Koyama Minami 4-101, Tottori, Japan
kimura@sse.tottori-u.ac.jp
This chapter deals with the short-term statistical properties of irregular sea waves.
It begins with a description of an early study by Rice [Bell Syst. Tech. J. 23,
282–332; 24, 46–156 (1944, 1945)] about the relation between wave spectra and
statistics of wave amplitudes. Next, an explanation is given of studies which deal
with the effect of spectrum width and wave nonlinearity on wave statistics as well
as with joint statistical properties between wave heights and periods. Finally, the
chapter closes with a look at the statistics of wave direction, length of wave crest,
and spatial maximum amplitude (which are the 3D wave’s properties) as well as
the time series of wave height.
7.1. Introduction
Waves in sea conditions have certain irregular properties. These comprise properties
such as wave heights and wave periods. For many years, engineers have known
that it is not sufficient to know only statistically representative values such as the
mean value. Larger wave heights can induce damage on coastal structures while
longer wave periods may induce resonance phenomena. To understand their overall
statistical properties, engineers and researchers have tried to clarify the statistical
characteristics of populations for these properties with the intention of finding out
their probability density function (pdf). In particular, their pdf have been studied
in relation to the wave spectrum.
The study of this development was initiated primarily by Rice.20 The appli-
cation of this idea in coastal engineering was started notably by Longuet-Higgins.15
He developed pdf for surface elevation and gradient, maxima and minima of water
surface, number of zero-crossings along a line, etc. Of course, sailors have carried out
visual observations of waves since ancient times. From the theoretical and empirical
knowledge of wave statistics, two types of statistical study have appeared: short-
term and long-term wave statistics. Since waves must be stationary stochastic to
151
July 31, 2009 8:18 9.75in x 6.5in b684-ch07 FA
152 A. Kimura
Rice20 proposed a model of the Fourier sum for irregular signals. The irregular signal
I(t) is expressed by the summation of trigonometric functions as
N
N
I(t) = (an cos σn t + bn sin σn t) = cn cos(σn t + εn ), (7.1)
n=1 n=1
in which σn is the angular frequency and εn (= tan−1 (−bn /an )) is the phase of the
nth component wave, respectively. an , bn , and cn are amplitudes of nth component
waves which are distributed normally with zero mean. Their variances are
2
cn
an = bn =
2 2
= S(σ)dσ , (7.2)
2
where means the ensemble mean, S(σ) is the energy spectrum of I(t). Rice
defined the envelope function for I(t) as
N
I(t) = cn cos(σn t + εn ) = Ic (t) cos σt − Is (t) sin σt , (7.3)
n=1
where σ is the mean frequency of S(σ), and Ic (t) and Is (t) are
N
N
Ic (t) = cn cos(σn t − σt + εn ), Is (t) = cn sin(σn t − σt + εn ) . (7.4)
n=1 n=1
since Ic (t) and Is (t) are independent (Ic (t) · Is (t) = 0). ϕ0 is the variance of Ic (t),
Is (t), and I(t), respectively:
Ic2 = Is2 = I 2 = ϕ0 . (7.7)
The variables R(t), I(t), Ic (t), and Is (t) are represented as R, I, Ic , and Is , respec-
tively. It is convenient to introduce the variable θ, where
Ic = R cos θ, Is = R sin θ . (7.8)
By substituting Eq. (7.8) into Eq. (7.6), the joint pdf of R and θ becomes
R R2
p(R, θ)dRdθ = exp − dRdθ . (7.9)
2πϕ0 2ϕ0
Since R and θ are independent and θ is distributed uniformly over the range 0 to
2π, the pdf of R is obtained as
R R2
p(R)dR = exp − dR . (7.10)
ϕ0 2ϕ0
Figure 7.1(a) shows the relation between I and R. The spectrum S(σ) for I is the
following normalized Wallops type (r = 10)8 :
−r −4
σ r σ
S(σ) = exp · 1− . (7.11)
σp 4 σp
154 A. Kimura
Fig. 7.1. (a) Wave profile I(t) (solid line) and its envelope R(t) (dotted line). (b) Wave profile
I(t) (solid line) and envelope R(t) (dotted line) (small discrepancies between I(t) and R(t) at
t = t∗1 , t∗2 , · · · ).
H = Il max − Il min ,
where Il max and Il min are the consecutive local maximum and minimum of the
wave profile, respectively. The pdf for H may not be the Rayleigh distribution if
the pdf for Il max and Il min are the Rayleigh distribution. Field measurements have
shown that there are small discrepancies between the Rayleigh distribution p(H)
and frequency distributions of measured data F (H): F (H) < p(H) for H < Hp
and F (H) > p(H) for H > Hp , where p(Hp ) is a peak of p(H). In particular, as
nonlinearity on the wave profile becomes prominent, this characteristic may bring
a considerable effect on the pdf. Forristall3 used the following Weibull distribution
to improve the goodness of fit:
α
−ς
P (ς) = exp , (7.14)
β
where P (ς) is the Weibull distribution function and α = 2.126, β = 8.42, and
H∗
ς= 1/2
, (7.15)
m0
July 31, 2009 8:18 9.75in x 6.5in b684-ch07 FA
in which H ∗ is the difference between the heights of consecutive wave crest and
trough within a zero-crossing interval of the wave profile. Longuet-Higgins17 pointed
out that the discrepancy should be improved using rms (root mean square) wave
amplitude.
The ith moment of a spectrum is determined as
∞
mi = σ i S(σ)dσ (i = 1, 2, . . .) . (7.16)
0
From the Khintchine theorem, the 0th moment is also defined in another way as
m0 = ηrms
2
, (7.17)
where η is the wave profile and ηrms is its rms. When the spectrum is narrow and
the nonlinearity on the wave profile is negligible,
(arms )2
m0 = , (7.18)
2
in which arms is the rms of wave amplitude a. The Rayleigh distribution function
given by Longuet-Higgins is
a2
P (a) = exp − 2 . (7.19)
arms
a2rms can be replaced by 2m0 if the nonlinearity is negligible. He discussed the non-
linearity effect on the wave profile as follows. Applying the third-order Stokes wave
theory, the 2nd moment of the periodic wave profile is derived from Eq. (7.17) as
1 2 1 19 3077
m0 = a 1 − (ak) − (ak) −
2 4
(ak) − · · · ,
6
(7.20)
2 2 12 720
in which k is the wave number. m0 for irregular waves is calculated by
∞
m0 = 2V (a)dp(a) , (7.21)
0
156 A. Kimura
in which an , σn , and εn are the amplitude, angular frequency, and phase of the
disturbance, respectively. Their amplitudes are b >> an . The spectrum of the dis-
turbances S (σ) is defined as
σn
+dσ
1 2
a = S (σn )dσ . (7.27)
σn
2 n
The rms of y is
∞
πσ
σ 2
yrms
2
= b2rms + sin2 + S (σ)dσ , (7.28)
0 2σ σ
and the 0th moment of the spectrum of y is
∞ ∞
b2rms
m0 = {S(σ) + S (σ)}dσ = + S (σ)dσ , (7.29)
0 2 0
where S(σ) is the spectrum of the periodic wave and brms is the rms of the periodic
wave profile. From Eqs. (7.28) and (7.29), yrms is
∞
2
σ 2 πσ
yrms
2
= 2m0 + − 1 − cos S (σ)dσ. (7.30)
0 σ2 2σ
Using the relations
2
σ
σ 2
σ
−1 =2 −1 + −1
σ2 σ σ
πσ 1 − cos π(σ/σ − 1)
π(σ/σ − 1)2
cos2
= = + ··· (7.31)
2σ 2 4
∞
i
µi = (σ − σ) S (σ)dσ ,
0
158 A. Kimura
1.2
1.0 r=5
r = 10
0.8
Rayleigh
p (ζ)
0.6
0.4
0.2
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
ζ
Zero-crossing wave amplitude in Eq. (7.19) extends the range from 0 to infinity if
the number of waves is infinity. When the population of waves is finite, such as data
from 20 min measurements every 2 h, maximum wave amplitude may be finite. If a
time series of wave amplitudes
a1 , a2 , . . . , aN , aN +1 , . . . , a2N , a2N +1 , . . . , a3N , a3N +1 . . . , a4N ,
Group No. 1 2 3 4
is divided into groups with N waves, the maximum values (amax ) from individual
groups show irregular properties. The pdf of the maximum wave amplitude is derived
by Longuet-Higgins.14
The probability that one wave amplitude a is greater than a0 is P (a0 )
[Eq. (7.19)]. When the size of the group is N , the probability that all the wave
amplitudes are a < a0 is {1 − P (a0 )}N . The probability that one amplitude, at
least, exceeds a0 is 1 − {1 − P (a0 )}N . Therefore, the probability that maximum
wave amplitude lies in the range amax ∼ amax + da is
1 − {1 − P (amax )}N − 1 + {1 − P (amax + da)}N
= −d[1 − {1 − P (amax )}N ]
dP (amax )
= −N {1 − P (amax )}N −1 da . (7.40)
da
The pdf for the maximum wave amplitude is
2 N −1 2
amax amax −amax
p(amax ) = 2N 1 − exp − exp . (7.41)
arms arms a2rms
with an approximation
2 N −θ
1 − e−(amax /arms ) ≈ e−e , (7.43)
in which
2
amax
θ = − ln N , (7.44)
arms
and γ is the Euler constant (γ = 0.5772 . . .). Maximum wave height E(Hmax )/Hrms
is usually compared also with Eq. (7.42).
The problem of deriving the pdf of zero-crossing wave periods is very difficult.
Rice derived a pdf for the zero-crossing interval. He treated the interval where I(t)
[Eq. (7.1)] crosses the 0 line upward at t = 0 and crosses the 0 line downward
between τ and τ + dτ .
1 τ 1 τ τ
p(τ ) = p0 (τ ) − p1 (r, τ )dr + p2 (r, s, τ )drds
1! 0 2! 0 0
1 τ τ τ
− p3 (r, s, u, τ )drdsdu + · · · , (7.45)
3! 0 0 0
with
2
2 −1/2
H = M23 M22 − M23 ,
where M22 and M23 are the cofactors of elements µ22 and µ23 of the matrix,
µ11 µ12 µ13 µ14 ψ0 0 ψτ ψτ
µ21 µ22 µ23 µ24 0
= −ψ0 −ψτ −ψτ
M = (7.47)
µ31 µ32 µ33 µ34 ψτ −ψτ −ψ0 0
µ41 µ42 µ43 µ44 ψτ −ψτ 0 ψ0
July 31, 2009 8:18 9.75in x 6.5in b684-ch07 FA
160 A. Kimura
and
∞
1
ψτ = S(σ) cos στ dσ ,
2π 0
∞
1
ψτ = − σS(σ) sin στ dσ ,
2π 0
(7.48)
∞
1
ψτ = − σ 2 S (σ) cos στ dσ ,
2π 0
ψ0 = ψτ |τ =0 , ψ0 = ψτ |τ =0 .
Equation (7.46) has a peak around the mean value of τ . The second and third peaks
appear for larger τ . This is because the higher terms in Eq. (7.45) are neglected.
No successful analytical solution for the zero-crossing wave period has been derived
at the present time. Instead of wave periods, Longuet-Higgins16 derived the pdf for
wave lengths. The wave profile of unidirectional irregular waves is expressed as
ς = R cos {χ} , (7.49)
where
χ = ux + φ, χx = u + φx , (7.50)
in which R is the amplitude of the wave envelope, φ is the phase, u is the mean
wave number in x-direction, and subscript x means a differentiation in terms of x.
Joint pdf for χ and χx was derived by Longuet-Higgins15 as
(m∗0 /µ∗2 )
1/2
p (χ, χx ) = % &3/2 , (7.51)
4π 1 + (χx − u) · m∗0 /µ∗2
2
with
∞
m∗n = E (u) un du , (n = 0, 1, . . .) (7.52)
0
∞
µ∗2 =
2
E (u) (u − u) du , (7.53)
0
with
m∗0 m∗2 − m∗2 µ∗2
ν ∗2 = 1
= . (7.57)
m∗21 u2 µ∗0
If ν ∗ is small,
π
≈ . (7.58)
u
The zero-up-crossing event of ς takes place when χ = (2r − 1/2)π with ∂χ/∂x > 0
or χ = (2r + 1/2)π with ∂χ/∂x < 0. The probability of the latter condition is
∞
p (χ, χx )χ=(2r+1/2)π |χx | dxdχx
−∞
1/2
1 m∗2 −1/2
= 1 − 1 + ν ∗2 dx . (7.59)
4 m∗0
This probability is negligible for small values of ν ∗ . Between any two successive
zero-crossing points x1 and x2 , χ must increase by π. Hence,
1
π = χ(x2 ) − χ(x1 ) = χx + 2 χxx + · · · .
2
The terms on the left of this equation, Taylor series expansion of χ(x2 ) around
x = x1 is used. Since χxx = O(ν 2 ), χxxx = O(ν 3 ), . . . and they are very small,
neglecting these terms,
π
≈ . (7.60)
χx
Applying the same method by Rice,20 a mean zero-crossing number of ς per unit
length, the conditional pdf of χx is
1/2
p(χ, χx )|χx | 1 m∗0 |χx |(m∗0 /µ∗2 )1/2
p(χx )χ = = , (7.61)
H(χ) 2 m∗2 {1 + (χx − u)2 (m∗0 /µ∗2 )}3/2
in which |χx |/H(χ) is the probability that χ falls between x and x + dx. Since
Eq. (7.61) does not involve χ, p(χx )χ directly becomes the pdf for χx . Substituting
Eqs. (7.57), (7.58), and (7.60) into Eq. (7.61),
1
p(ξ) = , (7.62)
2ν{1 + (ξ/ν ∗ )2 }3/2
in which ξ = ( − )/. If the frequency spectrum S(σ) is used instead of the wave
number spectrum E(u), the pdf for the zero-crossing interval (half-wave period) is
derived following the same method.
July 31, 2009 8:18 9.75in x 6.5in b684-ch07 FA
162 A. Kimura
Longuet-Higgins18 derived the joint pdf for wave amplitude and period. He used
the following expression for the wave profile:
ς = Re {A exp (iσt)} , (7.63)
where Re means a real part and
A = a exp{iφ} , (7.64)
in which a is the wave amplitude, φ is the phase of A, and σ is the mean angular
frequency given by
m1
σ= , (7.65)
m0
where m0 and m1 are calculated by Eq. (7.16). The phase of ς is χ = σt + φ.
Differentiation of the phase with t becomes
χt = φt + σ . (7.66)
When the spectrum is very narrow, the envelope of ς changes very slowly and
φt σ. The wave period can be approximated as
2π 2π
τ= = . (7.67)
χt σt + φt
Longuet-Higgins derived a joint pdf for a and φt :
2
a2 a (1/µ0 + φ2t /µ2 )
p(a, φt ) = exp − , (7.68)
(2πµ20 µ2 )1/2 2
where µ0 and µ2 are calculated with Eq. (7.31), and ν is calculated by Eq. (7.34).
With new variables
a τ
R= , T = ,
(2m0 )1/2 τ
Eq. (7.68) is expressed as
' '
' ∂(a, φt ) '
' = 2R L exp − R [1 + (1 − 1/T ) ] ,
2 2 2
p(R, T ) = p(a, φt ) '' (7.69)
∂(R, T ) ' π 1/2 νT 2 ν2
in which L is the correction factor to neglect the p(R, T ) in the region T < 0,
2
L= . (7.70)
1 + (1 + ν 2 )−1/2
Figure 7.3 shows the contour map of p(R, T ). Solid curves are contours for
p(R, T )/pmax = 0.99, 0.90, 0.70, 0.50, 0.30, 0.10 from inside to outside, respectively.
The marginal distribution for normalized wave amplitudes and periods are
R/ν
2L
p(R) = 1/2 R exp(−R2 ) exp(−β 2 )dβ, (7.71)
π −∞
−3/2
L (1 − 1/T )2
p(T ) = 1+ . (7.72)
2νT 2 ν2
July 31, 2009 8:18 9.75in x 6.5in b684-ch07 FA
Fig. 7.3. Joint pdf for wave amplitude and period Eq. (7.69) (after Longuet-Higgins18 ).
Fig. 7.4. Pdf for wave amplitude p(R) Eq. (7.71) (after Longuet-Higgins18 ).
Figures 7.4 and 7.5 show p(R) and p(T ), respectively. If ν = 0, p(R) is not the
Rayleigh distribution. Peak of the p(R) slightly shifts to the larger side of R with
increasing spectrum width ν. When ν = 0, p(T ) becomes a straight line (broken
line, T : constant).
Figure 7.6 shows the comparison of data and Eq. (7.69) by Goda.8 The number
of waves within a grid of ∆H = 0.2H and ∆T = 0.2T are shown. H and T are
the mean values of wave height and period, respectively. Many data sets, their
correlation coefficients between H and T distributed from 0.19 to 0.25, are used.
The spectrum width parameter in Eq. (7.69) is v = 0.26.
July 31, 2009 8:18 9.75in x 6.5in b684-ch07 FA
164 A. Kimura
Fig. 7.5. Pdf for wave period p(T ) Eq. (7.72) (after Longuet-Higgins18 ).
Fig. 7.6. Joint distribution of wave height and period (after Goda5 ).
July 31, 2009 8:18 9.75in x 6.5in b684-ch07 FA
Statistical properties of waves, such as length and direction of wave crest, are
important in the design of coastal structures. Wave height and period are deter-
mined from a record of point measurement in the ordinary method. In contrast, a
3D wave profile is necessary for the definition of wave crest length, direction, etc.
As a result, very few studies have examined their properties.
in which Ti is the zero-crossing wave period. The angle between the crest line and
y-axis (Fig. 7.7) is
−1 εi Li /2π
θi = sin , (i = 1, 2, . . .) , (7.74)
y0
where Li is the wave length corresponding to the zero-crossing wave period Ti and
θi is the direction of wave propagation measured from the x-axis.
166 A. Kimura
Assuming y0 u < 1 in which u is the mean wave number in the x direction, wave
profiles at these points are
∞
η1 (t) = ci cos(σi t + δi ) = ηc1 cos σt − ηs1 sin σt , (7.75)
i=1
∞
η2 (t) = ci cos(σi t + δi + vi y0 ) =ηc2 cos σt − ηs2 sin σt , (7.76)
i=1
in which ci , σi , δi , and vi are the amplitude, the angular frequency, the initial phase,
and the wave number in the y-direction of the ith component wave, respectively. σ
is the mean angular frequency. ηc1 , ηs1 , ηc2 , and ηs2 are derived in the same way as
Eq. (7.4). Their pdfs are the normal distribution with 0 mean, respectively. Their
joint pdf is
1
p(ηc1 , ηs1 , ηc2 , ηs2 ) =
(2π)2 (m20 − µ23 − µ24 )
1
× exp − (m0 ηc1
2
− 2µ3 ηc1 ηc2 − 2µ4 ηc1 ηs2
2(m20 − µ23 − µ24 )
+ m0 ηs1
2
+ 2µ4 ηs1 ηc2 − 2µ3 ηs1 ηs2 + m0 ηc2
2
+ m0 ηs2
2
) ,
(7.77)
July 31, 2009 8:18 9.75in x 6.5in b684-ch07 FA
with
m0 = ηc1
2
= ηs1
2
= ηc2
2
= ηs2
2
,
µ3 = ηc1 ηc2 = ηs1 ηs2 ,
(7.78)
µ4 = ηc1 ηs2 = −ηs1 ηc2 ,
ηc1 ηs1 = ηs2 ηc2 = 0 .
From the translations
ηc1 = R1 cos(σt + φ1 ) = R1 cos(χ1 ) ,
ηs1 = R1 sin(σt + φ1 ) = R1 sin(χ1 ) ,
(7.79)
ηc2 = R2 cos(σt + φ2 ) = R2 cos(χ2 ) ,
ηs2 = R2 sin(σt + φ2 ) = R2 sin(χ2 ) ,
the joint pdf for R1 , R2 , χ1 , and χ2 is
1
p(R1 , R2 , χ1 , χ2 ) =
(2π)2 (m20 − µ23 − µ24 )
1
× exp − (m0 (R12 + R22 )
2(m20 − µ23 − µ24 )
− 2R1 R2 [µ3 cos(χ1 − χ2 ) − µ4 sin(χ1 − χ2 )]) , (7.80)
R1 R2 1
= exp − [m0 (R12 + R22 )
2πm20 (1 − κ2 ) 2 m20 (1 − κ2 )
− 2R1 R2 (µ3 cos(2nπ − χ2 ) − µ4 sin(2nπ − χ2 ))] ,
(7.81)
in which
µ23 + µ24
κ2 = ≈ 1 − y02 u2 γ∗2 (7.82)
m20
and γ∗ is the long-crestedness parameter.15 Putting χ2 = 2nπ + ε, pdf for ε is
p(R1 , R2 , ε|χ1 = 2nπ)
p(ε|R1 , R2 ) =
p(R1 , R2 )
(
R1 R2 κ R1 R2 κ
= exp cos(ε − β) 2πI0 , (7.83)
m0 (1 − κ2 ) m0 (1 − κ2 )
July 31, 2009 8:18 9.75in x 6.5in b684-ch07 FA
168 A. Kimura
where I0 [ ] is the 0th order modified Bessel function and β = tan−1 (µ4 /µ3 ). β is an
angle of the dominant wave direction from x-axis. From Eq. (7.74), the pdf for the
wave direction from x-axis becomes
2π y0
p(θ|R1 , R2 ) = cos θ
L
(
R 1 R2 ) 2πy0 R1 R2 )
× exp 1 − y∗
2 cos sin θ − β 2πI0 1 − y∗
2 ,
m0 y∗2 L m0 y∗2
(7.84)
Figure 7.9 shows p(θ|R) for several R. R is the mean of R. The 3D spectrum is
the Bretschneider–Mitsuyasu type with Mitsuyasu-directional function modified by
Goda.8 The peakedness parameter for the directional function is Smax = 10. The
pdf for larger R shows a narrower width.
in which ηth is the threshold value (ηth = 0.1H1/3 in Fig. 7.10). The 3D wave
spectrum is the Bretschneider–Mitsuyasu type (H1/3 = 2.0 m, T1/3 = 8.01 s) with
Mitsuyasu-directional function modified by Goda, and Smax = 25.
The pattern recognition method is applied to determine the wave crest lengths.
Four threshold values ηth = 0.2 m, 0.5 m, 1.0 m, and 2.0 m were used. The calculated
July 31, 2009 8:18 9.75in x 6.5in b684-ch07 FA
crest lengths were arranged in descending order. The top 0.4%, 1%, 5%, and 10%
of total data for each ηth were used in the analysis, respectively. Figure 7.11 shows
the distribution of the mean length of wave crests. The four lines in the figure show
the results from the above four different data.
170 A. Kimura
given as
' '
' R (0) ' {R (0)}2
' '
' R (0) ' ≤ ∆L : ∆R = −
2R (0)
, (7.89)
' '
' R (0) '
' '
' R (0) ' > ∆L
R (0)(∆L)2
R (0) ≥ 0 : ∆R = R (0)∆L + , (7.90)
2
R (0)(∆L)2
R (0) < 0 : ∆R = −R (0)∆L + , (7.91)
2
July 31, 2009 8:18 9.75in x 6.5in b684-ch07 FA
in which R (0) and R (0) are the first and second derivatives of R(x ) at x = 0,
respectively. The probability that R∗ falls between c < R∗ < c + dR on condition
R(0) is
Prob(c < R∗ < c + dR)|R(0) = p[R (0), R (0) : R(0)]dS , (7.92)
S
in which p[R (0), R (0) : R(0)] is a conditional joint pdf of R (0) and R (0) on
condition R(0), and S is the region in which c < R∗ < c + dR in the R ∼ R plane.
The probability that R∗ falls between c < R∗ < c + dR regardless of R(0) is
c
Prob(c < R∗ < c + dR) = {Prob(c < R∗ < c + dR)|R(0) }p[R(0)]dR(0) , (7.93)
0
July 31, 2009 8:18 9.75in x 6.5in b684-ch07 FA
172 A. Kimura
Fig. 7.14. Pdf of local maximum wave amplitudes (dotted lines) within three finite width ∆L = 0
(Rayleigh dist.), 0.05L1/3 , and 0.10L1/3 .
in which p[R(0), R (0), R (0)] is the joint pdf of R(0), R (0), and R (0), and G is
the region in which the local maximum amplitude falls within R∗ and R∗ + dR∗ on
condition R(0). Rice20 formulated p[R(0), R (0), R (0)].
Figure 7.14 shows p(R∗ ) (dotted lines). The spectrum is the Bretschneider–
Mitsuyasu-type with H1/3 = 5.5 m and T1/3 = 10 s and Goda’s directional function
with Smax = 10, the water depth is h/L1/3 = 1.0, and the half-object width is ∆L =
0.050L1/3 and ∆L = 0.10L1/3, respectively. The Rayleigh distribution (solid line
∆L = 0) is also shown in the figure. Broken lines are the Weibull distribution with
m ≈ 2.5 (shape factor) shows good agreement when ∆L/L1/3 > 0.05 (h/L1/3 = 1.0,
Smax = 10).
Sawhney23 and Goda4 started to investigate the statistical properties of the time
series of irregular wave heights,
since this property may affect wave overtopping, slow drift oscillation of floating
bodies, etc. This property is analyzed by applying the concept of “run.”
Figure 7.15 shows a time series of wave heights. The run of high waves is
defined by waves, which are Hi > H∗ consecutively, in which H∗ is the threshold
wave height. In the case shown in Fig. 7.15, the length of the high wave run is 3.
July 31, 2009 8:18 9.75in x 6.5in b684-ch07 FA
Fig. 7.15. Run of wave heights (high waves, low waves, and total run).
174 A. Kimura
with
4H1 H2 H12 + H22 2κH1 H2
p(H1 , H2 ) = exp I0 , (7.107)
(1 − κ2 )Hrms
4 (κ2 − 1)Hrms
2 (1 − κ2 )Hrms
2
2H1 H2
p(H1 ) = 2 exp − 21 . (7.108)
Hrms Hrms
Equations (7.107) and (7.108) are the 2D Rayleigh and Rayleigh distributions,
respectively. k is the correlation parameter. The relation between κ and the corre-
lation coefficient γ(Hm Hm+1 ) is
1 1 π
γ(Hm Hm+1 ) = E(κ) − (1 − κ )K(κ) −
2
, (7.109)
1 − π/4 2 4
in which E( ) and K( ) are complete elliptic integrals of the first and second kind,
respectively.
Battjes et al.1 and Longuet-Higgins19 showed κ and the relation between κ and
γ(Hm Hm+1 ) as
2 ∞ 2
∞
1
κ =
2
S (σ) cos στ0 dσ + S (σ) sin στ0 dσ (7.110)
m0 0 0
July 31, 2009 8:18 9.75in x 6.5in b684-ch07 FA
Fig. 7.16. (a) Mean length of high wave runs (solid line: Eq. (7.103); circles: after Goda8 ).
(b) Mean length of total runs (solid line: Eq. (7.105); circles: Goda used j2 for the total run, after
Goda8 ).
July 31, 2009 8:18 9.75in x 6.5in b684-ch07 FA
176 A. Kimura
and
Figures 7.16(a) and 7.16(b) show the comparison between j1 and Eq. (7.103),
and jt and Eq. (7.105), respectively. The threshold wave height is H1/3 .
7.9. Remarks
Statistical studies have shown us the entire structures of wave irregularities in sea
conditions. A knowledge about their pdf has brought us an advance not only for
designing coastal structures but also for understanding nearshore hydraulics of the
wave and current system. However, we have to keep several things in mind when
we use these wave statistics.
The Rayleigh distribution for wave height is derived on the condition that the
wave spectrum is narrow. The problem still remains to clarify a good correspon-
dence between measured distributions and the Rayleigh distribution regardless of
the spectrum.
The assumption of the narrow band spectrum gives a considerable effect on the
pdf of wave periods. Although the theoretical models have the spectrum parameter
in the equation, we have to be careful about the stretch use for the spectrum with
finite band width.
The range of the pdf may be changed by wave dynamics such as wave breaking.
Extensive application of the theoretical pdf is sometimes risky.
The applications of the periodic wave theory for zero-crossing wave properties
raise problems. There have been very few studies about the physical properties of
zero-crossing waves, of wave period–wavelength/celerity, of wave height/period —
water particle velocities, etc.
Finally, we have to take into account the statistical properties for 3D irregular
waves. Common definitions for wave properties, first of all, become necessary: for
wave height, wave period, wavelength, celerity, wave direction, wave crest length, etc.
References
1. J. A. Battjes and G. Ph. van Vledder, Verification of Kimura’s theory for wave group
statistics, Proc. 19th ICCE (1984), pp. 642–648.
2. D. E. Cartwright and M. S. Longuet-Higgins, The statistical distribution of the
maxima of a random function, Proc. R. Soc. Lond. A 237, 212–232 (1956).
3. G. Z. Forristall, On the statistical distribution of wave heights in a storm, J. Geophys.
Res. 83(C5), 2353–2358 (1978).
4. Y. Goda, Numerical experiments on wave statistics with spectral simulation, Report
PHRI 9(3), 3–57 (1970).
July 31, 2009 8:18 9.75in x 6.5in b684-ch07 FA
5. Y. Goda, On the joint distribution of measured wave heights and periods, Tec. Note
of PHRI, No. 272, (1977) 19pp. (in Japanese).
6. Y. Goda, Statistics of wave crest lengths based on directional wave simulations, J. Off-
shore Mech. and Arctic Eng., Trans. ASME 116, 239–245 (1994).
7. Y. Goda, Numerical simulation of ocean waves for statistical analysis, J. Marine Res.
33, 5–14 (1999).
8. Y. Goda, Random Sea and Design of Maritime Structures, 2nd Edn. (World Scientific,
2000), p. 443.
9. A. Kimura, Statistical properties of random wave groups, Proc. 17th ICCE (1980),
pp. 2955–2973.
10. A. Kimura and T. Ohta, Statistical property of the irregular wave heights defined by
the zero-crossing method, Proc. Coastal Eng. 40, 146–150 (1993) (in Japanese).
11. A. Kimura and T. Ohta, Probability distribution of the maximum wave height along
a sea wall with finite width, Proc. 25th ICCE (1996), pp. 2272–2283.
12. A. Kimura, Probability distribution of 3-D irregular wave directions, Proc. Waves 97,
295–306 (1997).
13. B. M. Lake and H. C. Yuen, A new model for nonlinear wind waves, 1, Physical model
and experimental evidence, J. Fluid Mech. 88, 33–62 (1978).
14. M. S. Longuet-Higgins, On the statistical distribution of the heights of sea waves,
J. Marine Res. IX(3), 245–266 (1952).
15. M. S. Longuet-Higgins, The statistical analysis of a random, moving surface, Phil.
Trans. Roy. Soc. Ser. A 249(966), 321–387 (1957).
16. M. S. Longuet-Higgins, On the intervals between successive zeros of a random function,
Proc. Roy. Soc. Ser. A 246, 99–118 (1958).
17. M. S. Longuet-Higgins, On the distribution of the heights of sea waves: Some effects
of nonlinearity and finite band width, J. Geophys. Res. 85(C3), 1519–1523 (1980).
18. M. S. Longuet-Higgins, On the joint distribution of wave periods and amplitudes in a
random wave field, Proc. Roy. Soc. Ser. A 389, 241–258 (1983).
19. M. S. Longuet-Higgins, Statistical properties of wave groups in a random sea state,
Phil. Trans. Roy. Soc. Ser. A 312, 219–250 (1984).
20. S. O. Rice, The mathematical analysis of random noise, Bell System Tech. J. 23,
282–332, 24, 46–156 (1944).
21. S. O. Rice, Selected papers on Noise and Stochastic Processes, ed. N. Wax (Dover,
1945), pp. 133–294.
22. H. Rye, Wave group formation among storm waves, Proc. 14th ICCE (1974),
pp. 164–183.
23. M. D. Sawhney, A study of ocean wave amplitudes in terms of the theory of runs and
a Markov chain process, Tech. Rep. of New York University (1962), p. 29.
24. R. J. Sobey, Correlation between individual waves in a real sea states, Coastal Eng.
27, 223–242 (1996).
25. M. A. Tayfun, Effects of spectrum band width on the distribution of wave heights and
periods, Ocean Eng. 10(2), 107–118 (1983).
July 31, 2009 8:18 9.75in x 6.5in b684-ch07 FA
Chapter 8
M. P. C. de Jong
Formerly at Environmental Fluid Mechanics Section
Delft University of Technology; Presently at Deltares |Delft Hydraulics
P. O. Box 177, 2600 MH, Delft, The Netherlands
J. A. Battjes
Environmental Fluid Mechanics Section
Delft University of Technology, P. O. Box 5048, 2600 GA
Delft, The Netherlands
A resume is presented of recent work that has identified the mechanism through
which the seiches are generated that occasionally occur in Rotterdam Harbor
basins, and that has led to the development and implementation of an operational
warning system for the occurrence of significant seiche events. The generation
is due to moving systems of atmospheric convection cells that can arise over
the relatively warm water of the North Sea, behind a cold front. These cause
fluctuations in wind speed and atmospheric pressure that can generate long waves
at sea, as these cells move toward the coast. The average variance spectrum of
these long-wave surface elevations is found to have an approximately f −1.5 tail.
As these long waves at sea approach the harbor mouth, they can be resonantly
amplified inside certain semi-closed basins. The ratio of seiche amplitudes in two
different basins, derived through numerical simulations based on a 2DH mild-
slope model and an incident f −1.5 long-wave elevation spectrum, is in excellent
agreement with the mean ratio derived from observations for the same locations.
Finally, the principle of a possible system of prediction of the occurrence of seiche
events is described. Its predictive potential is proven through a series of hindcasts.
Based on these promising results, an operational forecasting system has been
developed and implemented by the authorities involved.
8.1. Introduction
179
July 31, 2009 8:18 9.75in x 6.5in b684-ch08 FA
Harbor system, the highest seiches occur in the Caland Canal, with measured crest
heights over 1 m at the closed end near Rozenburgse Sluis, ROZ (see Fig. 8.1).
This semi-closed basin has a length of approximately 20 km and a depth of
approximately 20 m. For almost all of the observed seiche events at ROZ, the
dominant frequency (determined through wavelet analysis) equals the lowest eigen-
frequency (quarter-wavelength mode), equivalent to a period of approximately
90 min. Smaller-amplitude seiches, though still quite significant, also occur in other
harbor basins.
The seiche oscillations in Rotterdam Harbor give rise to some practical concerns.
First, water levels used for the design of dikes (in particular their crest height) are
required by law to contain an allowance for seiches, superposed on the astronomical
high water level (spring tide) and wind setup. The appropriate measure of this
allowance is not well known.
Second, seiches affect the design and management of a movable storm surge
barrier (MSSB) in Rotterdam Waterway (Fig. 8.1), which protects the low areas
behind it against flooding. The MSSB must be closed during extreme storm condi-
tions. In that case a semi-closed basin is created on its seaside. Seiches are expected
to occur in this temporary basin, with a dominant eigenperiod of approximately
30 min. These contribute to the failure probability of the barrier, but their expected
heights were not known.
Third, the Rotterdam Port Management needs to guarantee a minimum depth
(“nautical depth”) in the approach channel and throughout the harbor for deep-
draught vessels. To this end, water levels are predicted operationally including tides
and wind-induced setup. Until recently, no method was available for inclusion of
seiche events in this short-range prediction, despite their practical importance: the
trough of a seiche could cause ships to run aground in case of a narrow predicted
tidal window.
The aforementioned practical applications in which seiches play a role have led to
a recent study on the origin of the seiche events in Rotterdam. The present chapter
summarizes the findings. Results are also presented of an analysis of the response
of Rotterdam Harbor basins to incoming wave energy in the seiche frequency band
using measured and theoretical ratios of seiche amplitudes at the closed ends of
two basins. Finally, a new system of routine prediction of seiche events is described,
July 31, 2009 8:18 9.75in x 6.5in b684-ch08 FA
8.2. Measurements
systems with cold fronts crossing the southern North Sea toward the Dutch coast.
However, the converse does not hold: not all such cold front passages give rise to
seiche events.
Analysis of measurements and numerical simulations showed that the seiches in
Rotterdam Harbor result from resonant amplification of oscillations at sea, which in
turn are generated by fluctuations in atmospheric pressure and wind speed following
a moving cold front. For the large majority of the seiche events, these fluctuations
are caused by atmospheric convection cells that can arise in the area behind a cold
front. These cells are a mesoscale atmospheric phenomenon, with typical horizontal
spatial scales of 30–100 km.7
Conditions conducive to the formation of convection cells occur when cold
(polar) air moves over relatively warm sea water, which is likely to occur in fall
and winter (also the seasons of significant seiche episodes). Especially in the area
behind the cold front, the relatively warm sea water causes heating of the air in the
lower layers, which results in an unstable lower atmosphere that causes air to rise
to higher altitudes in narrow plumes, (self-)organized into a more or less regular
pattern of convection cells.
If sufficient moisture is present in the air, condensation of water vapor can
locally cause a release of latent heat, further driving the circulation of air. Narrow
bands of clouds can then be formed surrounding clear skies, which can be seen
on satellite images (an example is shown in Fig. 8.2) and, in case of precipitation,
Fig. 8.2. Satellite image of infrared frequency band (10:12 GMT, 11 January 1995), taken during
a seiche event in the Port of Rotterdam. The image shows a pattern of convection cells that passes
the southern North Sea from the northwest. The location of the Port of Rotterdam is indicated
by the black and white dot. Source image: Dundee Satellite Receiving Station, Dundee University,
Scotland.
July 31, 2009 8:18 9.75in x 6.5in b684-ch08 FA
cold front
cloud formation
advection speed
warm air
2−4 km
cold air
30 −100 km
Fig. 8.3. Sketch of convection cells in the area behind a cold front together with a theoretical
surface elevation response.
also on radar precipitation images. These satellite and precipitation images showed
that the episodes of convection cells passing the southern North Sea more or less
coincided with the seiche episodes. The convection cells cause more or less spatially-
periodic fluctuations in atmospheric pressure and wind speed, which in a fixed point
will manifest themselves as temporal fluctuations as the cells are advected with the
weather system.
Because the system of convection cells moves, these fluctuations can generate
low-frequency waves at sea, which in turn can cause seiches in the harbor basins.
A sketch of the sea surface elevation response to the wind speed changes caused by
the convection cells is depicted in Fig. 8.3.
A 1D model based on the linearized shallow water equations was applied for a
transect across the North Sea along the main path of the convection cells, forced by
sinusoidal variations in atmospheric pressure and wind speed translating at a con-
stant speed, with values that are typical for those observed. Negligibly low responses
were found in the more remote, deeper parts of the North Sea. However, the model
showed that the fluctuations in wind speed and atmospheric pressure caused by
the passing convection cells are particularly efficient at generating long waves with
significant amplitudes in the relatively shallow part of the North Sea, the south-
ernmost 400 km near the Dutch coast. This is because in this region, with typical
depths of approximately 30 m, the phase speed of shallow-water waves is approx-
imately 17 m/s, i.e., near the advection velocity of the system of cold front and
convection cells (typically 15–20 m/s). The result is a near-resonant response of the
long waves at sea to the moving pattern of atmospheric convection cells, with ampli-
tudes up to 1–2 dm near the Dutch coast. These waves can in turn be amplified in
semi-closed harbor basins through (a quarter-wave) resonance, occasionally forming
seiches with observed amplitudes (crest heights) up to a meter.
184 variance density (cm /Hz) normalized by m0 (cm ) M. P. C. de Jong and J. A. Battjes
4 2
2
3
10
2
2
10
1
10 −1 0
10 10
Frequency (mHz)
Fig. 8.4. Three normalized variance density spectra of surface elevation measurements at the
EUR platform at sea. Nominal cut-off frequencies are indicated by the thick dashed vertical lines
at 0.1 mHz and 2.0 mHz. The thick line represents the best-fit power law of f −1.5 .
southern North Sea in the seiche frequency band, obtained during seiche events
at ROZ. Based on a relatively small number of seiche episodes he found that, on
average, the variance density spectrum in the seiche frequency band approximately
followed a f −1.2 trend. From June 2000, numerous processed and validated surface
elevation data have become available for a number of storm events. These have
now been used to determine the spectrum in the seiche frequency band at sea in
more detail (results described here are based on measurements from the platform
EUR). The variance density spectra have been estimated from filtered records (0.1–
2.0 mHz) of 4096 min (212 values) approximately centered around the seiche-event,
covering the complete storm duration of each event. Spectra of a number of events
normalized by the energy content of the seiche frequency band are shown in Fig. 8.4.
A least-squares fit of a power law to the averaged normalized spectral densities
results in a f −1.5 trend.
seiche activity with amplitudes (crest heights) exceeding 0.25 m. During most of
these events, the response at the lowest eigenperiod of the Caland Canal (90 min)
was dominant.
Surface elevation measurements at Europe Harbor (EH, see Fig. 8.1) have also
been studied for the same time interval as the observations at ROZ (1995–2001).
These time series were filtered for the frequency band 0.3–2.0 mHz in order to isolate
the main eigenfrequencies of this (shorter) harbor basin. Based on a wavelet analysis,
and using 0.1 m as an amplitude threshold, 46 seiche events at EH were identified,
all at the dominant eigenperiod of 50 min.
Measurement data at ROZ were available during 43 of the 46 events that were
identified at EH. Of these 43 events at EH, 86% coincided with a seiche-event at
ROZ. This percentage depends to some extent on the thresholds for the identifi-
cation of a seiche-event. These have been chosen subjectively: in case a threshold
of 0.15 m is chosen for the identification of seiche events at EH, 100% of the events
at EH coincide with a seiche-event at ROZ (with amplitude exceeding 0.25 m).
Conversely, of the 49 events at ROZ, 82% coincided with a seiche-event at EH.
In this case too, this result depends on the threshold of the definition of a seiche-
event. In case a threshold of 0.40 m is applied at ROZ, 100% of the events at ROZ
coincided with an event at EH (with a threshold of 0.10 m).
The fact that large-amplitude seiche events simultaneously occur in basins of
different dimensions indicates that the same mechanism generates seiches in basins
of different natural frequencies.
40
average seiche amplitude EH (cm)
30
20
10
y=0.5596x
0
0 10 20 30 40 50 60 70
average seiche amplitude ROZ (cm)
Fig. 8.5. Scatter plot of average seiche amplitudes (the average of the five highest crests and five
deepest troughs of an event) of coinciding seiche events at ROZ and Europe Harbor. Trend line
based on least-squares estimate.
where S(f ) is the (normalized) energy spectrum incident from the sea at the harbor
mouth and Ra (f ) and Rb (f ) are the amplitude amplification spectra for the con-
sidered locations a and b. These had been determined previously using a 2DH mild-
slope numerical wave model (PHAROS).9 Effects of bottom stress, inflow losses,
and seaward radiation were included. Applying these response functions and the
empirical seiche energy spectrum with a f −1.5 tail, the theoretical amplitude ratio
between EH and ROZ is found to be 0.57, nearly equal to the value of 0.56 that was
estimated from the measurements. This result, based on 30 seiche events, supports
the validity of the theoretical approach, and it confirms that the energy density at
sea in the seiche frequency band typically follows a f −1.5 trend, not just in the three
events from which it was estimated.
the uncertain range of conditional stability is excluded. A lower limit of the lapse
rate for unconditional instability of 10◦ C/km then applies. The following “predic-
tions” use this criterion and the estimated temperature gradient between sea level
and the observed level of 925 hPa pressure. For comparison, results based on the
average elevation of this pressure level (approximately 700 m) are shown as well.
The latter would be simpler operationally, albeit at the cost of enhanced errors in
the verification of the temperature difference criterion.
The results of the aforementioned approach are illustrated by presenting the
data of January till March 1999. This time interval was selected because the quality
of the data was relatively high during these months (only a small percentage of
missing data due to, e.g., atmospheric soundings that were not available on certain
days). Weather maps have been used to identify the days during which cold fronts
have passed the North Sea that resulted in a flow toward the Dutch coast (including
possible successive days during which this flow direction continued). This is obvi-
ously needed for seiche events to be generated in Rotterdam Harbor.
Figure 8.6 shows the results. The bars in the top panel show the observed dif-
ference between the air temperature at the (observed) altitude of the reference
pressure 925 hPa (at platform EKO) and the sea surface water temperature (at
platform AUK) for each day (not restricted to days of a front passage). Gaps are
present for those days for which no data were available. The lower panel shows the
15
temp. difference ( °C)
unstable
10
5
stable
0
5 10 15 20 25 30 5 10 15 20 25 5 10 15 20 25 30
Jan Feb Mar
15
temp. difference ( °C)
unstable
10
5
stable
0
5 10 15 20 25 30 5 10 15 20 25 5 10 15 20 25 30
Jan Feb Mar
Fig. 8.6. The bars indicate the observed temperature difference between the surface sea water
at AUK and the air at 925 hPa level at Ekofisk for January till March 1999. Top panel: all days;
bottom panel: days of occurrence of a cold front that moved from the North Sea toward the Dutch
coast. Straight dashed lines: critical temperature difference for an unstable atmosphere based on
average elevation of 925 hPa. Wiggly curves: values of the critical temperature difference for an
unstable atmosphere from day to day, based on an observed elevation of 925 hPa.
July 31, 2009 8:18 9.75in x 6.5in b684-ch08 FA
same data, now restricted to days during which cold fronts have passed the North
Sea toward the Dutch coast.
The straight dashed lines in both panels indicate the critical temperature dif-
ference for an unstable atmosphere based on the average altitude of the pressure
level of 925 hPa (∆T = 7◦ C). The wiggly curves in both panels indicate the value
of the critical temperature difference for an unstable atmosphere from day to day,
based on the temperature at the measured altitude of the reference pressure of
925 hPa for each day.
A seiche-event is predicted for those days in the lower panel for which the
observed temperature difference (the bar) exceeds the critical one (indicated by the
wiggly curve). The asterisks indicate the days for which seiche events have been
identified at ROZ. (Consecutive days with seiche events indicated represent a single
event that started during the first day and ended on the last day indicated.)
The bottom panel shows that an unstable atmosphere is found during a number
of potentially seiche-prone cold-front situations, correctly predicting a seiche-
episode. All the seiche events that occurred in this time interval are correctly pre-
dicted (no misses). A stable atmosphere is found during most other days indicated in
the bottom panel and no convection cells are expected during these days (regardless
of the availability of moisture in the air), which agrees with the fact that no seiche
events occurred.
Nevertheless, a limited number of false predictions is also found (17 January and
28 February). On these days, the temperature difference just slightly exceeded the
critical value. The resulting false predictions could therefore easily be due to the
criterion not being strict enough, or to the spatial variability of the temperature
of the sea surface water and the air at the reference levels: although the nominal
criterion is met locally, this does not necessarily have to be representative for the
complete southern North Sea region since only observations from sparse locations
are used. This will not be the case during the correctly predicted seiche events
in view of the fact that the prediction on these days is based on relatively large
temperature differences, which far exceed the critical values.
authorities and for the closure management of the movable storm surge barrier of
Rotterdam (which typically require prediction windows of, say, 6 h).
Based on the aforementioned findings, a pilot project was implemented, set up
by the Ministry of Public Works and Waterways in cooperation with the Royal
Dutch Meteorological Institute (KNMI) for operational prediction of the occurrence
of seiche events in the Port of Rotterdam, using parameter values derived from oper-
ational numerical weather prediction models. This pilot project has shown that the
system has sufficient predictive skill for operational use. Such a system has since
been implemented on an operational basis. Experience shows that all major seiche
events (above the seiche-amplitude threshold) have been successfully predicted.
However, in some cases a false prediction was made in the sense that a warning was
issued while no significant seiches occurred. This is a drawback that needs repair,
but even as it is, the system has merit because it does not miss significant events.
The false alarms mentioned above are related to an important limitation of the
above-described method, viz, the fact that so far it only predicts seiche-prone sit-
uations but not the expected energy level of the seiche-event. In case additional
parameters are taken into account, such as the (range in) sizes of the atmospheric
convection cells, the advection velocity, the excess of the temperature difference
above the critical value (not just the question whether it exceeds a critical value
or not), and the direction of movement of the weather systems, prediction of the
seiche energy level might become possible, for example, through the use of a neural
network that can be trained with a set of observed values of these quantifiable
parameters. This may open the possibility of operational forecasting of seiche-prone
conditions together with expected seiche crest heights. In that case, a seiche pre-
diction (warning) would only be issued when seiche-conducive atmospheric condi-
tions are expected that can lead to a seiche-event with amplitudes which justify a
warning.
8.6. Conclusions
the signal is too weak, and for the nearby stations the available prediction window
is too short to be of practical use.
A hindcast study has shown that the occurrence of significant seiche episodes
in Rotterdam Harbor can be accurately predicted on the basis of the prediction of
the occurrence of meteorological conditions conducive to the formation of mesoscale
atmospheric convection cells behind a cold front. This is presently realized using
output of air temperatures and moisture content from numerical atmospheric
models that are used in operational weather forecasting. The outcomes of this pre-
diction method are used by the port authorities for ship traffic control and, after
further development and testing, can eventually be incorporated into the closure-
management system of the storm surge barrier in Rotterdam Waterway.
Acknowledgments
The first author acknowledges the financial support from the Dr Ir Cornelis Lely
Foundation. A part of this study was financed by the Directorate-General for Public
Works and Water Management, South-Holland Directorate (Contract number: ZHA
23735). Furthermore, the authors thank the Directorate-General for Public Works
and Water Management and the Municipal Harbor Organization, Rotterdam for
providing the surface elevation measurements at sea and in the harbor, respectively.
The authors thank Dr A. J. van Delden of Utrecht University, The Netherlands,
for his extensive information on convection cells. Atmospheric sounding data
have been obtained from the Department of Atmospheric Science, University of
Wyoming, USA. Wavelet software was provided by Torrence and Compo,6 through
URL: http://paos.colorado.edu/research/wavelets/. The figures in this chapter are
reprinted from de Jong and Battjes,2 with permission from Elsevier.
References
8. C. Veraart, Events of Long Waves in the North Sea, Technical Report RIKZ-94.034,
Rijkswaterstaat/RIKZ and Delft University of Technology, Delft, The Netherlands
(1994).
9. P. Van den Bosch and J. J. Veldman, Europoortkering-Beerdam, Technical Report
H2438, WL | Delft Hydraulics, Delft, The Netherlands (1995).
10. J. R. Holton, An Introduction to Dynamic Meteorology, 3rd edn. (Academic Press,
London, 1992).
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
Chapter 9
Alexander B. Rabinovich
P.P. Shirshov Institute of Oceanology, Russian Academy of Sciences
36 Nakhimovsky Prosp., Moscow, 117997 Russia
Department of Fisheries and Oceans, Institute of Ocean Sciences
9860 West Saanich Road, Sidney, BC, Canada V8L 4B2
a.b.rabinovich@gmail.com
alexander.rabinovich@dfo-mpo.gc.ca
This chapter presents an overview of seiches and harbor oscillations. Seiches are
long-period standing oscillations in an enclosed basin or in a locally isolated part
of a basin. They have physical characteristics similar to the vibrations of a guitar
string or an elastic membrane. The resonant (eigen) periods of seiches are deter-
mined by basin geometry and depth and in natural basins may range from tens of
seconds to several hours. The set of seiche eigen frequencies (periods) and asso-
ciated modal structures are a fundamental property of a particular basin and
are independent of the external forcing mechanism. Harbor oscillations (coastal
seiches) are a specific type of seiche motion that occur in partially enclosed
basins (bays, fjords, inlets, and harbors) that are connected through one or more
openings to the sea. In contrast to seiches, which are generated by direct external
forcing (e.g., atmospheric pressure, wind, and seismic activity), harbor oscillations
are mainly generated by long waves entering through the open boundary (harbor
entrance) from the open sea. Energy losses of seiches in enclosed basins are mostly
due to dissipative processes, while the decay of harbor oscillations is primarily due
to radiation through the mouth of the harbor. An important property of harbor
oscillations is the Helmholtz mode (pumping mode), similar to the fundamental
tone of an acoustic resonator. This mode is absent in a closed basin.
Harbor oscillations can produce damaging surging (or range action) in some
ports and harbors yawing and swaying of ships at berth in a harbor. A property
of oscillations in harbors is that even relatively small vertical motions (sea level
oscillations) can be accompanied by large horizontal motions (harbor currents),
resulting in increased risk of damage of moored ships, breaking mooring lines
as well as affecting various harbor procedures. Tsunamis constitute another
important problem: catastrophic destruction may occur when the frequencies of
arriving tsunami waves match the resonant frequencies of the harbor or bay.
Seiches, as natural resonant oscillations, are generated by a wide variety of
mechanisms, including tsunamis, seismic ground waves, internal ocean waves, and
193
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
194 A. B. Rabinovich
jet-like currents. However, the two most common factors initiating seiches are
atmospheric processes and nonlinear interaction of wind waves or swell. At certain
places in the World Ocean, waves due to atmospheric forcing (atmospheric gravity
waves, pressure jumps, frontal passages, squalls) can be responsible for significant,
even devastating harbor oscillations, known as meteorological tsunamis. They have
the same temporal and spatial scales as typical tsunami waves and can affect
coasts in a similar damaging way.
9.1. Introduction
(2) Energy losses of seiches in closed basins are mostly associated with dissipation,
while the decay of harbor oscillations is mainly due to radiation through the
mouth of the harbor.
(3) Harbor oscillations have a specific fundamental mode, the Helmholtz mode,
similar to the fundamental tone of an acoustic resonator.54 This modes is absent
in closed basins.
Because harbor oscillations can produce damaging surging (or range action) —
yaw and swaying of ships at berth in a harbor — this problem has been extensively
examined in the scientific and engineering literature.4,11,12,36,41,46,47,62,67,68,70,78,79,
94,95
One of the essential properties of oscillations in harbors is that even rela-
tively small vertical motions (sea level oscillations) can be accompanied by large
horizontal water motions (harbor currents); when the period of these motions coin-
cides with the natural period of sway, or yaw of a moored ship, further resonance
occurs, which can result in considerable motion and possible damage of a moored
ship.82,94 Harbor oscillations can also break mooring lines, cause costly delays in
loading and unloading operations at port facilities, and seriously affect various
harbor procedures.79,80
Tsunamis constitute another important problem that have greatly stimulated
investigations of harbor oscillations. Professor Omori (Japan) was likely the first
to notice in 1902 that the dominant periods of observed tsunami waves are nor-
mally identical to those caused by ordinary long waves in the same coastal basin
(see Ref. 30). His explanation was that the bay or portion of the sea oscil-
lates like a fluid pendulum with its own period, i.e., the arriving tsunami waves
generate similar seiches as those generated by atmospheric processes and other
types of external forcing (see also Ref. 30). Numerous papers on the spectral
analysis of tsunami records for various regions of the world ocean have confirmed
this conclusion.13,48,69,74,75,84,90 Catastrophic destruction may occur when the fre-
quencies of arriving tsunami waves match the resonant frequencies of the harbor or
bay. One of the best examples of strong tsunami amplification due is the resonant
response of Port Alberni (located at the head of long Alberni Inlet on the Pacific
coast of Vancouver Island, Canada) to the 1964 Alaska tsunami.28,54
The basic theory of seiche oscillations is similar to the theory of free and forced
oscillations of mechanical, electrical, and acoustical systems. The systems respond
to an external forcing by developing a restoring force that re-establishes equilibrium
in the system. A pendulum is a typical example of such a system. Free oscillations
occur at the natural frequency of the system if the system disturbed beyond its
equilibrium. Without additional forcing, these free oscillations retain the same fre-
quencies but their amplitudes decay exponentially due to friction, until the system
eventually comes to rest. In the case of a periodic continuous forcing, forced oscilla-
tions are produced with amplitudes depending on friction and the proximity of the
forcing frequency to the natural frequency of the system.84
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
196 A. B. Rabinovich
n=2 n=1
n=3 n=2
n=4 n=3
Fig. 9.1. Surface profiles for the first four seiche modes in closed and open-ended rectangular
basins of uniform depth.
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
√
Table 9.1. Normalized periods, Tn∗ = Tn gH/(2L), for a
closed and open-mouth rectangular basin of uniform depth.
Mode
“tone”) have wavelengths equal to one-half, one-third, one-fourth, and so on, of the
wavelength of the fundamental mode (Fig. 9.1(a), Table 9.1).
The fundamental mode is antisymmetric: when one side of the water surface
is going up, the opposite side is going down. Maximum sea level oscillations are
observed near the basin borders (x = 0; x = L), while maximum currents occur at
the nodal lines, i.e., the lines where ζ = 0 for all time. Positions of the nodal lines
are determined by
(2m − 1)L
xm
n = , m, n = 1, 2, . . . ; m ≤ n. (9.4)
2n
Thus, for n = 1, there is one nodal line: x11 = L/2 located in the middle of the
basin; for n = 2, there are two lines: x12 = L/4 and x22 = 3L/4; for n = 3: x13 = L/6,
x23 = 3L/6 = L/2 and x33 = 5L/6 and so on. The number of nodal lines equals
the mode number n [Fig. 9.1(a)], which is why the first mode is called the uninodal
mode, the second mode is called binodal mode, the third mode the trinodal mode,
etc.31,95 The antinode positions are those for which ζ attains maximum values, and
are specified as
jL
xjn = , j = 0, 1, 2, . . . , n. (9.5)
n
For example, for n = 2 there are three antinodal lines: x02 = 0, x12 = L/2, and
x22= L. Maximum currents occur at the nodal lines, while minimum currents occur
at the antinodes. Water motions at the seiche nodes are entirely horizontal, while
at the antinodes they are entirely vertical.
The relationships (9.2) and (9.3) yield the well-known Merian’s formula for the
periods of (natural) in a rectangular basin of uniform depth68,78 :
2L
Tn = √ , (9.6)
n gH
where n = 1, 2, 3, . . .. Merian’s formula (9.6) shows that the longer the basin length
(L) or the shallower the basin depth (H), the longer the seiche period. The fun-
damental (n = 1) mode has the maximum period; other modes — the overtones
of the main fundamental — “tone” — have periods equal to one-half, one-third,
one-fourth, and so on, of the fundamental period (Fig. 9.1(a), Table 9.1). The fun-
damental mode and all other odd modes are antisymmetric, while even modes are
symmetric; an antinode line is located in the middle of the basin.
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
198 A. B. Rabinovich
The structures and parameters of open-mouth basins are quite different from
those of closed basins. Standing oscillations in a rectangular bay (harbor) with
uniform depth and open entrance also have the form (9.1) but with a nodal line
located near the entrance (bay mouth). In general, the approximate positions of
nodal lines are determined by the following expressions (Fig. 9.1(b), Table 9.1):
(2m + 1)L
xm
n = , m, n = 0, 1, 2, . . . ; m ≤ n, (9.7)
2n + 1
while antinodes are located at
2jL
xjn = , j, n = 0, 1, 2, . . . ; j ≤ n. (9.8)
2n + 1
In particular, for n = 1 there are two nodal lines: x01 = L/3 and x11 = L and two
antinodal lines: x01 = 0 and x11 = 2L/3; for n = 2 there are three nodal lines:
x02 = L/5, x12 = 3L/5 and x22 = L, and three antinodal: x02 = 0, x12 = 2L/5, and
x22 = 4L/5.
The most interesting and important mode is the lowest mode, for which n = 0.
This mode, known as the Helmholtz mode, has a single nodal line at the mouth of the
bay (x = L) and a single antinode on the opposite shore (x = L). The wavelength
of this mode is equal to four times the length of the bay; a basin oscillating in
this manner is known as a quarter-wave oscillator.34 The Helmholtz mode, which is
also called the zeroth mode,a the gravest mode and the pumping mode (because it is
related to periodic mass transport — pumping — through the open mouth,36,41 is of
particular importance for any given harbor. For narrow-mouthed bays and harbors,
as well as for narrow elongated inlets and fjords, this mode normally dominates.
The periods of the Helmholtz and other harbor modes can be approximately
estimated as84,95
4L
Tn = √ , for mode n = 0, 1, 2, 3, . . . . (9.9)
(2n + 1) gH
Using (9.9) and (9.6), the fundamental (Helmholtz) mode in a rectangular
√ open-
mouth basin of uniform depth H is found to have a period, T 0 = 4L/ gH, which √ is
double the period of the gravest mode in a similar but closed basin, T 1 = 2L/ gH.
Normalized periods of various modes (for n ≤ 4) are shown in Table 9.1.
Expressions (9.4)–(9.9), Table 9.1, and Fig. 9.1 are all related to the idealized
case of a simple rectangular basin of uniform depth. This model is useful for
some preliminary estimates of seiche parameters in closed and semi-closed natural
and artificial basins. Analytical solutions can be found for several other basins of
simple geometric form and nonuniform depth. Wilson95 summarizes results that
involve common basin shapes (Tables 9.2 and 9.3), which in many cases are quite
a In many papers and text books,8,94,95 this mode is considered the “first mode”. However, it is
more common to count nodal lines only inside the basin (not at the entrance) and to consider the
fundamental harbor mode as the “zeroth mode”.41,68,78,80,84 This approach is physically more
sound because this mode is quite specific and markedly different from the first mode in a closed
basin.
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
good approximations to rather irregular shapes of natural lakes, bays, inlets, and
harbors.
The main concern for port operations and ships and boats in harbors is not
from the sea level seiche variations but from the strong currents associated with the
seiche. As noted above, maximum horizontal current velocities occur at the nodal
lines. Therefore, its locations in the vicinity of the nodes that are potentially most
risky and unsafe. Maximum velocities, Vmax , can be roughly estimated as84 :
g
Vmax = An , (9.10)
H
where An is the amplitude of the sea level oscillation for the mode. For example, if
An = 0.5 m and H = 6 m, Vmax ≈ 0.64 m/s.
9.75in x 6.5in
Rectangular 0 x h(x) = h0 2L/(gh0 )1/2 1.000 0.500 0.333 0.250
h0 h(x)
A. B. Rabinovich
L0
Triangular h(x) x
h0 h(x) = h0 (1 − 2x/L) 1.305[2L/(gh0 )1/2 ] 1.000 0.628 0.436 0.343
(isosceles) ∠
2
L
0 x
h(x)
Parabolic h0 h(x) = h0 (1 − 4x2 /L2 ) 1.110[2L/(gh0 )1/2 ] 1.000 0.577 0.408 0.316
∠
2
L
0 x
h(x) = h0 (1 − 4x2 /L2 )2 1.242[2L/(gh0 )1/2 ]
h0 h(x)
Quartic 1.000 0.686 0.500 0.388
∠
b684-ch09
2
(Continued)
FA
July 31, 2009 8:18
Table 9.2. (Continued )
9.75in x 6.5in
L
Triangular 0
h1 h(x)
x h(x) = h1 x/L 1.640[2L/(gh1 )1/2 ] 1.000 0.546 0.377 0.288
(right-angled)
L
0 x h(x) = h0 + mx
Trapezoidal h0 1.000 0.546 0.377 0.288
h(x) h1 m = (h1 − h0 )/L
L1 L2
Coupled, 0 h(x) = h1 (x < 0)
h1 x L1 /L2 = 1/2 4L2 /(gh2 )1/2 1.000 0.500 0.250 0.125
rectangular h(x) h h(x) = h2 (x < 0)
2
h1 /h2 = 1/4 L1 /L2 = 1/3 3.13L2 /(gh2 )1/2 1.000 0.559 0.344 0.217
L1 /L2 = 1/4 2.73L2 /(gh2 )1/2 1.000 0.579 0.367 0.252
b684-ch09
L1 /L2 = 1/8 2.31L2 /(gh2 )1/2 1.000 0.525 0.371 0.279
201
FA
July 31, 2009 8:18
202
Table 9.3. Modes of free oscillations in semi-closed basins of simple geometric shape (modified after Ref. 95).
0 x
0 x h(x) h1
b(x) b1
h(x) = h1 2.000[2L/(gh1 )1/2 ] 1.000 0.333 0.200 0.143
L
9.75in x 6.5in
L
Rectangular Rectangular
L
A. B. Rabinovich
0 x
0 x h1
b(x) b1
h(x) h(x) = h1 x/L 2.618[2L/(gh1 )1/2 ] 1.000 0.435 0.278 0.203
L
Rectangular Triangular
L
x 0
x b(x) b1 0
h1 h(x) = h1 (1 − x2 /L2 ) 2.220[2L/(gh1 )1/2 ] 1.000 0.409 0.259 0.189
h(x)
L
Rectangular Semi-parabolic
b(x) L
b684-ch09
0 x 0 x b(x) = b1 x/L 1.308[2L/(gh1 )1/2 ] 1.000 0.435 0.278 0.230
b1 h(x) h1
h(x) = h1
L
Triangular Rectangular
(Continued)
FA
July 31, 2009 8:18
Table 9.3. (Continued )
9.75in x 6.5in
0 x 0 x b(x) = b1 x/L 1.653[2L/(gh1 )1/2 ] 1.000 0.541 0.374 0.283
b1 h1
h(x) h(x) = h1 x/L
L
Triangular Triangular
L
b1 /L = 2 2.220[2L/(gh1 )1/2 ] 1.000 0.707 0.578 0.378
0 x 0 x
b1 h1 = 4/3 0.554 0.493 0.323
=1 0.447 0.468 0.264
L
= 2/3 0.317 0.455 0.185
Semi-elliptic Semi-paraboloidal
L
L
0 b r 0
1
h(x) h1
b684-ch09
h(x) = h1 (1 − r 2 /L2 ) 2.220[2L/(gh1 )1/2 ] 1.000 0.707 0.578 0.500
r
Semi-circular Semi-paraboloidal
203
FA
July 31, 2009 8:18
204
Table 9.4. Mode parameters for free oscillations in uniform depth basins of rectangular and circular geometric shape.
9.75in x 6.5in
+
2 0 + 0.500 2 0 — — 3.054 0.603
A. B. Rabinovich
+
+
0 1 0.500 0 1 + 0.628 — 3.832 0.480
+ + +
1 1 0.447 1 1 0.719 — 5.331 0.345
+
+
+
+
2 1 0.354 2 1 + 0.766 — 6.706 0.275
+ + +
+ +
0 2 + 0.250 0 2 0.343 0.787 7.016 0.262
b684-ch09
+ + + +
1 2 0.243 1 2 + 0.449 0.820 8.536 0.216
+
FA
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
The roots of this equation determine the eigen values ksm (m, s = 0, 1, 2, 3, . . .), with
corresponding eigen modes described by Eq. (9.14) for various k = ksm . Table 9.4
presents the modal parameters and the free surface displacements of particular
modes.
As illustrated by Table 9.4, there are two classes of nodal lines, “rings” and
“spokes” (diameters). The corresponding mode numbers m and s give the respective
exact number of these lines. Due to mass conservation, the mode (0, 0) does not
exist in a completely closed basin.41 For the case s = 0, the modes are symmetrical
with respect to the origin and have annular crest and troughs.35 In particular, the
first symmetrical mode (s = 0, m = 1) has one nodal ring r = 0.628a (Table 9.4).
When the central part of the circular basin (located inside of this ring) is going up,
the marginal part (located between this ring and the basin border) is going down,
and vice versa. The second symmetrical mode (s = 0, m = 1) has two nodal rings:
r = 0.343a and r = 0.787a.
For s > 0, there are s equidistant nodal diameters located at an angle ∆θ = π/s
from each other; i.e., 180◦ for s = 1, 90◦ for s = 2, 60◦ for s = 3, etc. Posi-
tions of these diameters are indeterminate, since the origin of θ is arbitrary. The
indetermatability disappears if the boundary deviates even slightly from a circle.
Specifically, the first nonsymmetrical mode (s = 1, m = 0) has one nodal diameter
(θ = π/2), whose position is undefined; but if the basin is not circular but elliptical,
the nodal line would coincide with either the major or minor axis, and the corre-
sponding eigen periods would be unequal.35 The first unsymmetrical mode has the
lowest frequency and the largest eigen period (Table 9.4); in this case, the water
sways from one side to another relative to the nodal diameter. This mode is often
referred to as the “sloshing” mode.78
Most natural lakes or water reservoirs can support rather complex 2D seiches.
However, the two elementary examples of rectangular and circular basins help to
understand some general properties of the corresponding standing oscillations and
to provide rough estimates of the fundamental periods of the basins.
206 A. B. Rabinovich
mirror image relative to the mouth.b However, this condition is not strictly correct
because it does not take into account wave energy radiation through the mouth into
the open sea. The exact solutions may be obtained based on the Sommerfeld radi-
ation condition of free wave radiation through the open boundary.35,41 Following
application of the appropriate mouth correction (α), the nodal line is located close
to but outside the entrance. In other words, the effect of this correction is to increase
the effective length of the basin.95 The mouth correction depends on two param-
eters: the basin aspect ratio q = l/L, which relates the width of the basin (l) to
its length (L); and the aperture ratio ϑ = b/l, in which b is the actual width of
the mouth.
Mathematical determination of α is rather complicated but, as a rule, it increases
with increases of q and ϑ. For example, the fractional correction to Eq. (9.9) for
the fundamental mode in a rectangular basin of uniform depth and open mouth
(ϑ = 1.0) is determined as30,95
q 3 πq
α= − γ − ln , (9.16)
π 2 4
where γ = 0.5772 is Euler’s constant. Roughly speaking, radiation into the external
basin and the mouth correction are important when the semi-closed basin is broad
and has a large open entrance, and negligible when the basin is long and narrow
(i.e., when q is small); in the latter case, expressions (9.7)–(9.9), as well as those
presented in Table 9.3, are quite accurate.
The character of natural oscillations in a bay or harbor is strongly controlled
by the aperture ratio ϑ = b/l, which can vary from ϑ = 1.0 to ϑ = 0.0. These two
asymptotic cases represent a fully open harbor and a closed basin, respectively. It is
evident that the smaller is ϑ (i.e., the smaller the width of the entrance), the slower
water from the external basin (open sea) penetrates into the harbor. Thus, as ϑ
decreases, the periods of all harbor modes for n ≥ 1 in Table 9.1 increase, tending
to the periods of the corresponding eigen modes for a closed basin, while the period
of the fundamental (Helmholtz) harbor mode tends to infinity.c This is one of the
important properties of harbor oscillations.
Another important property is harbor resonance. The amplification factor for
long waves impinging on a harbor from the open sea is
1
H 2 (f ) = , (9.17)
(1 − f /f0 )2 + Q−2 (f /f0 )2
where f is the frequency of the long incoming waves, f0 is the resonant frequency of
the harbor, and Q is the quality factor (“Q-factor”), which is a measure of energy
damping in the system.47,95 Specifically,
dE/dt
Q−1 = = 2β, (9.18)
ωE
b This approach is used for numerical computation of eigen modes in natural 2D basins.70
c This is the reason for calling this the “zeroth mode.”
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
where E = E0 e−2βωt is the energy of the system as it decays from an initial value
E0 , β is a dimensionless damping coefficient, and ω = 2πf is the angular frequency.
The power amplification factor attains the value Q2 at resonance (f = f0 ), decreases
to unity at f = 0 and goes to zero as f goes to infinity. Therefore, Q for harbor
oscillations plays a double role: as a measure of the resonant increase of wave heights
for waves arriving from the open ocean and as an index of the time decay rate of wave
heights inside the harbor. The higher the Q, the stronger will be the amplification
of the incoming waves and the slower the energy decay, i.e., the longer the “ringing”
of seiche oscillations inside the harbor.
In closed basins, like lakes, bottom friction is the main factor controlling energy
decay. Normally, it is quite small, so in lakes with fairly regular topographic fea-
tures (low damping), a high Q-factor may be expected. Consequently, even a small
amount of forcing energy at the resonant frequency can produce significant seiche
oscillations that persist for several days.31,95 In contrast, the main factor of energy
decay in semi-closed water basins, such as gulfs, bays, fjords, inlets and harbors, is
wave radiation through the entrance. In their pioneering work, Miles and Munk47
concluded that narrowing the harbor entrance would increase the quality factor Q
and, consequently, the amplification of the arriving wave. This means that the con-
struction of dams, dikes, and walls to protect the harbor from wind waves and swell
could so constrict the entrance width that it leads to strong amplification of the
resonant seiche oscillations inside the harbor. Miles and Munk47 named this harbor
paradox.
As pointed out by Miles and Munk,47 there are two limitations to the previous
conclusions:
(1) A time of order Q/π cycles is necessary for the harbor oscillations to adjust
to the external forcing. This means that harbors with high Q would not respond to a
strong but short-lived incoming disturbance. In most cases, this limitation is not of
major concern because atmospheric disturbances (the major source of open-sea long
waves inducing harbor oscillations) are likely to last at least for several hours. Even
tsunami waves from distant locations “ring” for many hours, resonantly “feeding”
harbor seiches and producing maximum oscillations that have long (12–30 h) dura-
tions that persist well after the arrival of the first waves.74,75 This contrasts with the
case for near-field sites, where tsunamis normally arrive as short-duration impulsive
waves. Such tsunamis are much more dangerous at open coastal regions than in
bays or harbors, as was observed for the coast of Thailand after the 2004 Sumatra
tsunami.86
(2) As the harbor mouth becomes increasingly narrower, the internal harbor
dissipation eventually exceeds energy radiation through the mouth. At this stage,
further narrowing does not lead to a further increase in the Q-factor. However,
normally internal dissipation is small compared to the typical radiative energy losses
through the entrance.
Originally, Miles and Munk47 believed that their “harbor paradox” concept was
valid for every harbor mode provided the corresponding spectral peak was sharp
and well defined. Further thorough examination of this effect37,46,78 indicated that
the harbor paradox is only of major importance for the Helmholtz mode, while
for higher modes frictional and nonlinear factors, not accounted for in the theory,
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
208 A. B. Rabinovich
dampen this effect.95 However, the Helmholtz mode is the most important mode in
natural basins and is normally observed in bays, inlets, and harbors with narrow
entrance, i.e., in semi-closed basins with high Q-factor. Significant problems with
the mooring and docking of ships (and the loading and unloading of their cargo)
in ports and harbors are often associated with this fundamental mode and most
typically occur in ports with high Q.4,6,41,62,66–68,78–80
Rabinovich67 suggested reducing these negative effects in ports by artificially
increasing the internal dissipation. The idea is the same as that widely used in
rocket technology to damp eigen oscillations in fuel tanks.44,45 Radial piers in ports
and harbors play the same role as internal rings and ribs in rocket tanks, efficiently
transforming wave energy into vortical motions which reduce the wave energy and
therefore the intensity of the seiches and their associated horizontal currents. As
shown by Rabinovich,67 the logarithmic attenuation factor, δ0 = π/Q, for the
Helmholtz mode associated with the jth pier, is given by
j ∆E0j 1 j bj ζ̂0 ω0
δ0 = = Cx , (9.19)
2E0 6 r0 h0 σ0
where E0 is the energy of the mode inside the harbor, ∆E0j is the energy dissipated
at the pier over the mode period (T0 = 2π/ω0 ), bj is the length of the pier, r0
and h0 are the mean radius and depth of the harbor, ζ̂0 is the mean amplitude
of the Helmholtz mode in the harbor, Cxj is a dimensionless resistance coefficient,
and σ0 = (gh0 )1/2 /(πr0 ). Thus, the rate of damping of oscillations in a harbor
depends on the number of piers (N ) and a number of dimensionless parameters:
specifically, the relative amplitudes of the oscillations, ξ0 = ζ̂0 /h0 ; the normalized
harbor frequency, Ω0 = ω0 /σ0 ; the relative lengths of the piers, Bj = bj /r0 ; and the
coefficient Cxj . The parameter ξ0 depends on the intensity of the external forcing
while the two other parameters Ω0 and Bj do not depend on forcing but only on the
characteristics of the harbor. The coefficient Cxj strongly depends on the Keulegan–
Carpenter (KC) number which relates hydraulic resistance in oscillating flows to
those for stationary currents.32 For typical values Bj = 0.3, ξ0 = 0.1, Ω0 = 1.0,
N = 8, and Cxj = 10, we find δ0 ≈ 0.4 and Q ≈ 8.
Another important aspect of the harbor oscillation problem is that changes in
port geometry, and the construction of additional piers and dams can significantly
change the natural (eigen) periods of the port, thereby modifying considerably the
resonant characteristics of the basin.6 Helmholtz resonators in acoustics are used to
attenuate sound disturbances of long wavelengths, which are difficult to reduce using
ordinary methods of acoustical energy dissipation. Similarly, side channel resonators
are suggested as a method for attenuating incident wave energy in harbors.6,66,78
In general, estimation of the Q-factor is a crucial consideration for ports, harbors,
bay, and inlets. For a rectangular basin of uniform depth and entirely open mouth
(ϑ = b/l = 1.0), this factor is easily estimated as:
L
Q= , (9.20)
l
which is inversely proportional to the aspect ratio q = l/L. This means that
high Q-factors can be expected for long and narrow inlets, fjords, and waterways.
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
Honda et al.30 and Nakano and Unoki59 examined coastal seiches at more than
110 sites on the coast of Japan and found that strong and highly regular seiche
oscillations are most often observed in such elongated basins and that the periods
of these oscillations are in good agreement with the approximate period (9.9) for
the Helmholtz mode (n = 0):
4L
T0 = √ . (9.21)
gH
If the aperture ratio ϑ < 1.0, corresponding to a partly closed entrance, it is
more difficult to estimate the Q value and the resonant mode periods analytically. In
practice, special diagrams for a rectangular basin with various q and ϑ are used for
these purposes.80,84 For natural basins, these parameters can be estimated numer-
ically or from direct observations. If the respective spectral peak in observational
data is isolated, sharp and pronounced enough, then we can assume that Q 1.
In this case, it follows from (9.17) that the half-power frequency points (f1/2 ) are
given by the following expression47:
± 1
f1/2 = f0 1 ± , (9.22)
2Q
and the relative frequency bandwidth is simply
∆f
= Q−1 , (9.23)
f0
−
where ∆f = f1/2 +
− f1/2 and f0 = 1/T0 is the resonant frequency. This is a
useful practical method for estimating the Q-factor and amplification for coastal
basins based on results of spectral analysis of observational data. However, the
spatial structure of different modes, the distribution of currents, and sea levels
inside a natural basin, influence harbor reconstruction based on changes in these
characteristics, and many other aspects of harbor hydrodynamics, are difficult
to estimate without numerical computations. Numerical modeling has become a
common approach that is now widely used to examine harbor oscillations.4,14,70,92
210 A. B. Rabinovich
Sea of s
nd
Okhotsk sla
45° il I
K ur Iturup I. 4
Kunashir I.
Shikotan I.
Hokkaido I.
1 5
2 3
False Bay
Malokurilsk Bay
Pacific Ocean Shikotan I.
0.5 km
(a)
40°
140° 145° 150°
10 5
18.6
10 4 10 4 10 4
Spectra (cm 2 cpm -1)
1
2
10 3 18.6
10 2 10 3 95%
95% 95%
3
10 2
4
10 0
10 2
10 1
10 -3 10 -2 10 -1 10 -3 10 -2 10 -1 10 -2 10 -1
Frequency (cpm)
Fig. 9.2. (a) Location of cable bottom pressure stations near the northern coast of Shikotan
Island (Kuril Islands) and sea level spectra at (b) BPS-1, (c) BPS-2 (both in autumn 1986) and
(d) BPS-3 and BPS-4 (October–November 1990).
1300 m and a narrow neck of 350 m [Fig. 9.2(a)]. The oscillations recorded by these
instruments were significant, highly regular and almost monochromatic; the corre-
sponding spectra [Figs. 9.2(c) and 9.2(d)] have a prominent peak at a period of
18.6 min. An analogue tide gauge [#5 in Fig. 2(a)] situated on the coast of this
bay permanently measure oscillations with exactly the same period.70 It is clear
that this period is related to the fundamental mode of the bay. The Q-factor of the
bay, as estimated by expression (9.23) based on spectral analysis of the tide gauge
data for sites BPS-2 and BPS-3, was 12–14 and 9–10, respectively. The high Q-
factors are likely the main reason for the resonant amplification of tsunami waves
that arrive from the open ocean. Such tsunami oscillations are regularly observed in
this bay.13,69 In particular, the two recent Kuril Islands tsunamis of 15 November
2006 and 13 January 2007 generated significant resonant oscillations in Malokurilsk
Bay of 155 cm and 72 cm, respectively, at the same strongly dominant period
of 18.6 min.76
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
Fig. 9.3. Computed eigen modes and periods of the first six modes in Malokurilsk Bay (Shikotan
Island). Black triangles indicate positions of the BPS-2 and BPS-3 gauges (from Ref. 70).
Figure 9.3 shows the first six eigen modes for Malokurilsk Bay.70 The compu-
tations were based on numerical conformal mapping of the initial mirror reflected
domain on a circular annulus (for details, see Ref. 77) and the following application
of Ritz’s variational method to solve the eigenvalue problem. The computed period
of the fundamental (Helmholtz) mode (18.9 min) was close to the observed period of
18.6 min. The spectra at BPS-2 and BPS-2 indicate weak spectral peaks (three
orders of magnitude less than the main peak) with periods 4.1, 3.3, and 2.9 min
(the latter only at BPS-3), thought to be related to modes n = 2, 3, and 4. The
first mode (n = 1), with period of 6.5 min, was not observed at these sites appar-
ently because the nodal line for this mode passes through the positions of BPS-2
and BPS-3.
Thus, the computed periods of the bay eigen modes are in good agreement with
observation; plots in Fig. 9.3 give the spatial structure of the corresponding modes.
However, this approach does not permit direct estimation of the bay response to
the external forcing and the corresponding amplification of waves arriving from
the open ocean. In actuality, the main purpose of the simultaneous deployments
at sites BPS-3 and BPS-4 [Fig. 9.2(a)] in the fall of 1990 was to obtain observed
response parameters that could be compared with numerically evaluated values.14
The spectrum at BPS-4, the station located on the outer shelf of Shikotan Island
near the entrance to Malokurilsk Bay [Fig. 9.2(d)], contains a noticeable peak with
period of 18.6 min associated with energy radiation from the bay. This peak is about
1.5 orders of magnitude lower than a similar peak at BPS-3 inside the bay. The
amplification factor for the 18.6 min period oscillation at BPS-4 relative to that at
BPS-3 was found to be about 4.0. Numerical computations of the response charac-
teristics for Malokurilsk Bay using the HN-method14 gave resonant periods which
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
212 A. B. Rabinovich
were in close agreement with the empirical results of Rabinovich and Levyant70
(indicated in Fig. 9.3). Resonant amplification of tsunami waves impinging on the
bay was found to be 8–10.
Sec. 9.4); however, it is worth noting here that the coupling between the two inlets
can apparently amplify the destructive effects associated with each of the inlets
individually.38
9.3. Generation
Because they are natural resonant oscillations, seiches are generated by a wide
variety of mechanisms (Fig. 9.4), including tsunamis,14,28,54,69 seismic ground
waves,2,15,34,40 internal ocean waves,8,19,21 and jet-like currents.30,54,57 However, the
two most common factors initiating these oscillations in bays and harbors are atmo-
spheric processes and nonlinear interaction of wind waves or swell (Fig. 9.4).62,68,95
Seiches in lakes and other enclosed water bodies are normally generated by direct
external forcing on the sea surface, primarily by atmospheric pressure variations
and wind.31,95 In contrast, the generation of harbor oscillations is a two-step process
involving the generation of long waves in the open ocean followed by forcing of the
Period (sec)
10 5 10 4 10 3 10 2 101 sec
AT M O S P H E R E
Storm
surges Scattering
nonstationarity Nonlinearity
Long waves Swell Wind waves
Tsunami waves
OCEAN
Seismicity
BOTTOM
10 6 10 5 10 4 10 3 10 2 10 1 10 0
Wavelength (m)
Fig. 9.4. Sketch of the main forcing mechanisms generating long ocean waves.
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
214 A. B. Rabinovich
harbor oscillations as the long waves arrive at the harbor entrance where they lead
to resonant amplification in the basin.
Seiche oscillations produced by external periodic forcing can be both free and
forced. The free oscillations are true seiches (i.e., eigen oscillations of the cor-
responding basin). However, if the external frequency (σ) differs from the eigen
frequencies of the basin (σ = ω), the oscillations can be considered forced seiches.95
Open-ocean waves arriving at the entrance of a specific open-mouth water body
(such as a bay, gulf, inlet, fjord, or harbor) normally consist of a broad frequency
spectrum that spans the response characteristics of the water body from resonantly
generated eigenfree modes to nonresonantly forced oscillations at other frequencies.
Following cessation of the external forcing, forced seiches normally decay rapidly,
while free modes can persist for a considerable time.
Munk53 jokingly remarked that “the most conspicuous thing about long waves in
the open ocean is their absence.” This is partly true: the long-wave frequency band,
which is situated between the highly energetic tidal frequencies and swell/wind wave
frequencies, is relatively empty (Fig. 9.5). For both swell/wind waves and tides,
the energy is of order 104 cm2 , while the energy contained throughout the entire
intermediary range of frequencies is of order 1–10 cm2 . However, this particular fre-
quency range is of primary scientific interest and applied importance (Walter Munk
himself spent approximately 30 years of his life working on these “absent” waves!).
Long waves are responsible for formation and modification of the coastal zone and
shore morphology;5,68 they also can strongly affect docking and loading/unloading
of ships and construction in harbors, causing considerable damage.41,78,79,96 Finally,
Period
hr min sec
100 24 12 5 2 1 30 10 5 2 1 30 10 5 2 1
Tides
Swell, wind waves 10 3
10 2
-5
Variance (cm 2 )
10 1
-2
10 0
Storm surges Meteorological
waves
IG- waves 10 -1
10 -2
Tsunami
10 -3
10 -4 10 -3 10 -2 10 -1 10 0 10 1 10 2
Frequency (cpm)
Fig. 9.5. Spectrum of surface gravity waves in the ocean (modified from Ref. 68). Periods (upper
scale) are in hours (h), minutes (min), and seconds (s).
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
and probably the most important, are tsunamis and other marine hazardous long
waves, which are related to this specific frequency band. The recent 2004 Sumatra
tsunami in the Indian Ocean killed more than 226,000 people, triggering the largest
international relief effort in history and inducing unprecedented scientific and public
interest in this phenomenon and in long waves in general.86
Because of their resonant properties, significant harbor seiches can be produced
by even relatively weak open ocean waves. In harbors and bays with high Q-
factors, seiches are observed almost continuously. However, the most destructive
events occur when the incoming waves have considerable energy at the resonant
frequencies, especially at the frequency of the fundamental mode. Such a situation
took place in Port Alberni located in the head of long Alberni Inlet on Vancouver
Island (Canada) during the 1964 Alaska tsunami, when resonantly generated seiche
oscillations in the inlet had trough-to-crest wave heights of up to 8 m, creating total
economic losses of about $10 million (1964 dollars).28,54
(1) Direct generation of long waves by atmospheric forcing (pressure and wind) on
the sea surface.
(2) Generation of low-frequency motions (for example, storm surges) and subse-
quent transfer of energy into higher frequencies due to nonlinearity, topographic
scattering and nonstationarity of the resulting motions.
(3) Generation of high-frequency gravity waves (wind waves and swell) and sub-
sequent transfer of energy into larger scale, lower frequency motions due to
nonlinearity.
Long waves generated by the first two mechanisms are known as atmospher-
ically induced or meteorological waves.d Typical periods of these waves are from
a few minutes to several hours, typical scales are from one to a few hundreds of
kilometers. The first mechanism is the most important because it is this mech-
anism that is responsible for the generation of destructive seiche oscillations (mete-
orological tsunamis) in particular bays and inlets of the World Ocean (Sec. 9.4).
“Meteorological waves” can be produced by the passages of typhoons, hurricanes
or strong cyclones. They also have been linked to frontal zones, atmospheric
pressure jumps, squalls, gales, wind gusts and trains of atmospheric buoyancy
waves.10,59,68,71,87,95 The most frequent sources of seiches in lakes are baro-
metric fluctuations. However they can also be produced by heavy rain, snow,
d The Russian name for these waves is “anemobaric”68 because they are induced by atmospheric
pressure (“baric”) and wind (“anemos”) stress forcing on the ocean surface.
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
216 A. B. Rabinovich
or hail over a portion of the lake, or flood discharge from rivers at one end of
the lake.27,31,95
e Figure 9.6 does not include all possible types of IG waves and mechanisms of their generation; a
Fig. 9.6. Generation mechanisms for infragravity waves in the coastal zone.
harbor. Typical eigen periods of a harbor or a moored ship are the order of minutes.
Therefore, they cannot be excited directly by wind waves or swell, having typical
periods on the order of seconds.96 However, these periods exactly coincide with the
periods of wave groups and IG waves. So, it is conventional wisdom that surging
in harbors is the result of a triple resonance of external oscillations outside the
harbor, natural oscillations within the harbor, and natural oscillations of a ship.
The probability of such triple resonance is not very high, thus surging occurs only
in a limited number of ports. Ports and harbors having large dimensions and long
eigen periods (>10 min) are not affected by surging because these periods are much
higher than the predominant periods of the IG waves and the surging periods of the
vessels. On the other hand, relatively small vessels are not affected because their
natural (eigen) periods are too short.82 The reconstruction of harbors and the cre-
ation of new harbor elements, can significantly change the harbor resonant periods,
either enhancing or, conversely, reducing the surging.f Another important aspect of
the problem is that ship and mooring lines create an entirely separate oscillation
system.79 Changing the material and the length of the lines and their position,
f A famous example of this kind is the French port Le Havre. Before World War II it was known
for very common and strong surging motions that created severe problems for ships. During the
war a German submarine torpedoed by mistake a rip-rap breakwater, creating a second harbor
opening of 20–25 m width. After this, the surging in the port disappeared.68
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
218 A. B. Rabinovich
changes the resonant properties of the system (analogous to changing the material
and the length of a pendulum).
It is important to keep in mind that each oscillation mode has a specific
spatial distribution of sea level variability and associated current (as emphasized in
Sec. 9.2.1, maximum currents are observed near the nodal lines). The intensity of
the currents varies significantly from place to place. Moreover, topographic irregu-
larities within the harbor and the presence of structure elements (dams, dykes, piers
and breakwaters) can create intense local vortexes that may significantly affect the
ships.67 So, the effect of surging on a ship strongly depends on the exact location
of the ship, and even on its orientation, in the harbor.
In summary, harbor oscillations arise through co-oscillation of sea surface eleva-
tions and currents in the harbor with those at the entrance to the harbor. Seiche-
generating motions outside the harbor typically have periods of several minutes
and most commonly arise from bound and free long waves that are incident on the
harbor entrance.
9.3.3. Tsunami
Tsunami waves are the main factor creating destructive seiche oscillations in bays,
inlets and harbors.30,41,53,54,95 Tsunamis can produce “energies” of 103 –105 cm2 ,
although such events are relatively rare (depending on the region, from once every
1–2 years to once every 100–200 years). The main generation mechanisms for
tsunamis are major underwater earthquakes, submarine landslides and volcanic
explosions. Great catastrophic trans-oceanic tsunamis were generated by the 1946
Aleutian (magnitude Mw = 7.8), 1952 Kamchatka (Mw = 9.0), 1960 Chile
(Mw = 9.5), and 1964 Alaska (Mw = 9.2) earthquakes. The events induced strong
seiche oscillations in bays, inlets, and harbors throughout the Pacific Ocean.90
The magnitude Mw = 9.3 earthquake that occurred offshore of Sumatra in
the Indian Ocean on 26 December 2004 generated the most destructive tsunami
in recorded history. Waves from this event were recorded by tide gauges around
the world, including near-source areas of the Indian Ocean (Fig. 9.7), and remote
regions of the North Pacific and North Atlantic, revealing the unmatched global
reach of the 2004 tsunami.42,74,86,88 In general, the duration of tsunami “ringing”
increased with increasing off-source distance and lasted from 1.5 to 4 days.74,75 The
recorded oscillations were clearly polychromatic, with different periods for different
sites, but with clear dominance of 40–50 min waves at most sites. The analysis of
various geophysical data from this event indicates that the initial tsunami source
had a broad frequency spectrum, but with most of the energy within the 40–50 min
band. Therefore, although tsunami waves at different sites induced local eigen modes
with a variety of periods, the most intense oscillations were observed at sites having
fundamental periods close to 40–50 min.
Differences in spectral peaks among the various tide gauge records are indicative
of the influence of local topography. For example, for the Pacific coast of Van-
couver Island (British Columbia), the most prominent peaks in the tsunami spectra
were observed for Winter Harbor (period ∼30–46 min) and Tofino (∼50 min). In
fact, the frequencies of most peaks in the tsunami spectra invariably coincide with
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
E
Colombo (SL)
200 cm
Hanimaadhoo (Md)
Male (Md)
Sea level
Salalah (Om)
26 27 28
December 2004 (UTC)
Fig. 9.7. Tsunami records in the Indian Ocean for the 2004 Sumatra tsunami for six selected sites:
Colombo (Sri Lanka); Male and Gan (both Maldives); Salalah (Oman); Pointe La Rue (Seychelles);
and Port Louis (Mauritius). Solid vertical line labeled “E” denotes the time of the main earthquake
shock (from Ref. 74).
induce different resonant mode, specifically, large seismic sources generate low-frequency modes
and small seismic sources generate high-frequency modes.
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
220 A. B. Rabinovich
where W (ω) = H 2 (ω), H(ω) is the frequency admittance function describing the
linear topographic transformation of long waves approaching the coast, and E(ω) is
the source spectrum. It is assumed that the site-specific properties of the observed
spectrum Sj (ω) at the jth site are related to the topographic function Hj (ω) for that
site, while all mutual properties of the spectra at all sites are associated with the
source (assuming that the source is the same for all stations). For typical background
oscillations the source spectrum has the form, E(ω) = S0 (ω), where S0 (ω) = Aω −2 ,
and A = 10−3 −10−4 cm2 .68,69 During tsunami events, sea level oscillations observed
near the coast can be represented as
ζobs (t) = ζt (t) + ζb (t), (9.25)
where ζt are the tsunami waves generated by an underwater seismic source and ζb
are the background surface oscillations. If the spectra of both tsunami, St (ω), and
background oscillations, Ŝb (ω) and Sb (ω) (during and before the tsunami event,
respectively) have the form (9.24), and the admittance function, W (ω), is the same
for the observed tsunami and the background long waves, then the spectral ratio
R(ω), is estimated as
St (ω) + Ŝb (ω) E(ω) + Ŝ0 (ω)
R(ω) = = = A−1 ω 2 E(ω) + 1.0. (9.26)
Sb (ω) S0 (ω)
The function R(ω), which is independent of local topographic influence, is deter-
mined solely by the external forcing (i.e., by tsunami waves in the open ocean near
the source area) and gives the amplification of the longwave spectrum during the
tsunami event relative to the background conditions. The close similarity of Rj (ω)
for various sites confirms the validity of this approach.69
The topographic admittance function Hj (ω), which is characteristic of the res-
onant properties of individual sites, can be estimated as
1/2 j
1/2
1/2
Sj (ω) Sb (ω) Sbj (ω)
Hj (ω) = = =ω . (9.27)
E(ω) S0 (ω) A
these events.2,15,40 The mechanism for seiche generation by seismic waves from
distant earthquakes is not clear, especially considering that seismic waves normally
have much higher frequencies than seiches in natural basins. McGarr40 concludes
that there are two major factors promoting efficient conversion of the energy from
distant large-magnitude earthquakes into seiches:
(1) A very thick layer of soft sediments that amplify the horizontal seismic ground
motions.
(2) Deeper depths of natural basins, increasing the frequencies of eigen periods for
the respective water oscillations.
It should be noted, however, that seismic origins for seiches must be considered
as very rare in comparison, for example, with seiches generated by meteorological
disturbances.95
222 A. B. Rabinovich
the Inland Sea. Here, the semi-diurnal tidal currents move large volumes of water
back and forth between the Pacific and the Inland Sea twice per day with typical
speed of 13–15 km/h. This region is one of the greatest attractions in Japan because
of the famous “Naruto whirlpool”, occurring twice a month during spring tides,
when the speed of tidal flow reaches 20 km/h. Honda et al.30 noticed that flood tidal
currents generate near both coasts significant seiche oscillations with a period of
2.5 min, which begin soon after low tide and cease near high tide; the entire picture
repeats with a new tidal cycle. No seiches are observed during ebb tidal currents (i.e.,
between high and low water) when the water is moving in the opposite direction.
Nakano57 explained this phenomenon by assuming that a strong current passing
the mouth of a bay could be the source of bay seiches, similar to the way that a
jet of air passing the mouth piece of an organ pipe produces a standing oscillation
within the air column in the pipe. Special laboratory experiments by Nakano and
Abe58 demonstrated that jet-like flow with a speed exceeding a specific critical
number generates a chain of antisymmetric, counter-rotating von Karman vortexes
on both sides of the channel. The checker-board pattern of vortexes induce standing
oscillations in nearby bays and harbors if their fundamental periods match the
typical vortex periods,
l
Tvor = , (9.28)
u
where l is the distance between vortexes, and u is the speed of the vortexes (u = 0.4V
to 0.6V , where V is the speed of the tidal currents). For the parameters of the
Naruto tidal currents, the laboratory study revealed that values of Tvor agreed with
the observed seiche period of 2.5 min. Apparently, the same mechanism of seiche
generation can also work in other regions of strong jet currents.
As discussed in Sec. 9.3.3, tsunamis are the main source of destructive seiches
observed in various regions of the World Ocean. However, atmospheric disturbances
(atmospheric gravity waves, pressure jumps, frontal passages, squalls) can also
be responsible for significant, even devastating, long waves, which have the same
temporal and spatial scales as typical tsunami waves. These waves are similar to
ordinary tsunami waves and can affect coasts in a similar damaging way, although
the catastrophic effects are normally observed only in specific bays and inlets.
Nomitsu,61 Defant,10 and Rabinovich and Monserrat 71,72 suggested to use the term
“meteorological tsunamis” (“meteotsunami”) for this type of waves.
At certain places in the World Ocean, these hazardous atmospherically induced
waves occur regularly and have specific local names: “rissaga” in the Balearic
Islands, “šćiga” on the Croatian coast of the Adriatic Sea, “marubbio” (“marrobio”)
in Sicily, “milghuba” in Malta, “abiki” and “yota” in Japan, “Seebär” in the Baltic
Sea, “death waves” in Western Ireland, “inchas” and “lavadiads” in the Azores and
Madeira islands. These waves are also documented in the Yellow and Aegean seas;
the Great Lakes; northwestern Atlantic; Argentina and New Zealand coastal areas;
and Port Rotterdam.7,9–12,16,27,29,30,34,43,51,64,68,71,72,91,92 Table 9.5 gives a list of
destructive harbor oscillations, which apparently have the same atmospheric origin
and similar resonances due to similarities in the characteristics of the atmospheric
disturbances and local geometry and topography of the corresponding basins.
Because of the strong likeness between “meteotsunamis” and seismically generated
tsunamis,51,88 it is quite difficult sometimes to recognize one from another. Cata-
logues of tsunamis normally contain references to numerous “tsunami-like” events
of “unknown origin” that are, in fact, atmospherically generated ocean waves.
“Rissaga” (a local Catalan word that means “drying”, similar to a Spanish
word “resaca”) is probably the best known example of meteorological tsunamis.h
These significant short-period sea level oscillations regularly occur in many bays
and harbors of the Catalan and Valencian coasts of the Iberian Peninsula, and on
the coast of the Balearic Islands. The waves in Ciutadella Harbor, Menorca Island
[Fig. 9.8(a)] are particularly high and occur more frequently than in any other
location.18,22,49–51,71–73,81,85
Ciutadella Inlet is about 1 km long, 100 m wide, and 5 m deep; the harbor is
located at the head of the inlet [Fig. 9.8(a)]. The fundamental period of the inlet
(Helmholtz mode) is approximately 10.5 min [Figs. 9.8(b,c)]. Due to the particular
geometry of Ciutadella Inlet, it has a large Q-factor, which results in significant
resonant amplification of long-wave oscillations arriving from the open sea. Seiche
h For this reason Derek Goring, a wave specialist from New Zealand, suggested to apply the term
“rissaga” to all rissaga-like meteorological seiches in other areas of the World Ocean.23 However,
if we were to adopt this term, then we would loose information on the cause of the oscillations
and the fact that they are part of a family of events that include seismically generated tsunamis,
landslide tsunamis, volcanic tsunamis, and meteotsunamis.
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
224 A. B. Rabinovich
Table 9.5. Extreme coastal seiches in various regions of the World Ocean.
Maximum
Typical observed
Region Local name peroid height References
Comment: Exact references can be found in: Wiegel (1964), Korgen (1995), Rabinovich and
Monserrat (1996), de Jong et al. (2003) and Monserrat et al. (2006).
oscillations of duration ranging from a few hours to several days and wave heights
exceeding 0.5 m recur in Ciutadella every summer. However, rissaga events (large-
amplitude seiches) having wave heights more than 3–4 m, with dramatic conse-
quences for the harbor, usually take place once in 5–6 years. During the rissaga of
21 June 1984 (Fig. 9.9), about 300 boats were destroyed or strongly damaged.71
More recently, on 15 June 2006, Ciutadella Harbor was affected by the most dra-
matic rissaga event of the last 20 years, when almost 6-m waves were observed in
the harbor and the total economic loss was of several tens millions of euros.51
Fontseré,17 in the first scientific paper on extreme seiches for the Catalan
coast, showed that these seiches always occur from June to September and first
suggested their atmospheric origin. This origin of rissaga waves was supported by
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
Fig. 9.8. A map of the Balearic Islands and positions of four tide gauges (M0, M1, M2 and MW3)
deployed in Ciutadella and Platja Gran inlets and on the shelf of Menorca Island during the LAST-
97 experiment.50 The arrow shows the predominant direction of propagation of atmospheric waves
during “rissaga” events. (b) The strong “rissaga” event recorded in Ciutadella Inlet on 31 July
1998 by a tide gauge located at position M0. (c) Spectra for “rissaga” of 24 July 1997 (solid line)
and background oscillations (dashed line) for four tide gauges indicated in (a). The actual four-day
records during this event are shown in the insets.
Ramis and Jansà81 based on observed oscillations on the Balearic Islands. These
authors also defined a number of typical synoptic atmospheric conditions nor-
mally associated with rissaga events. The atmospheric source of rissaga is now well
established.18,22,49,50,85 During late spring and summer, meteorological conditions in
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
226 A. B. Rabinovich
Fig. 9.9. Ciutadella Harbor during the rissaga of 21 June 1984. Photo by Josep Gornes (from
Ref. 71).
the western Mediterranean are favorable for the formation of high-frequency atmo-
spheric pressure disturbances with parameters promoting the generation of rissaga
waves. These conditions include the entrance of warm air from the Sahara at near-
surface levels, and relatively strong middle level winds from the southwest. When
this synoptic meteorological situation exists, trains of atmospheric pressure gravity
waves (with periods of minutes) are reported traveling from SW to NE.49 If these
atmospheric pressure disturbances propagate from SW to NE with a phase speed
of about 22–30 m/s, resonant conditions are set up for the southeastern shelf of
Mallorca Island (“Proudman resonance”) and dynamic energy associated with the
atmospheric waves is efficiently transferred into the ocean waves. When these waves
reach the coast of Menorca Island, they can generate significant (and sometimes
even hazardous) seiche oscillations inside Ciutadella and other inlets due to harbor
resonance.
The Q-factor for the fundamental Helmholtz mode in Ciutadella Inlet (10.5
min), roughly estimated by Eq. (9.20), is about 9. Spectral estimates based on
Eq. (9.23) give a similar value, Q ≈ 10.73 As shown in Fig. 9.8(b), rissaga oscillations
in Ciutadella Inlet have a very regular monochromatic character. Maximum wave
heights occur during the fourth to the sixth oscillations, in good agreement with the
criterion by Miles and Munk47 that time of the order of Q/π cycles is necessary for
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
the harbor oscillations to adjust themselves to external forcing. The peak period of
10.5 min for the Helmholtz mode strongly dominates the spectra for the M0 and
M2 gauges located in Ciutadella Inlet [Fig. 9.8(c)] both for rissaga and background
spectra, while in the adjacent inlet Platja Gran (M1), where rissaga waves are also
observed but weaker than in Ciutadella, the dominant peak associated with the
Helmholtz mode is 5.5 min. In contrast, on the shelf (MW3) both peaks are absent
and oscillations are significantly weaker.
Spectral analysis results [Fig. 9.8(c)] reveal that harbor resonance is a crucial
factor in the formation of rissaga waves, as well as “meteorological tsunamis” in
other bays, inlets, and harbors of the World Ocean. Barometric data from the
Balearic Islands49–51,81 as well as from Japan,29 and Eastern Adriatic Sea64,91,92
demonstrate that generation of these destructive waves is associated with strong
atmospheric disturbances, e.g., trains of atmospheric gravity waves, or isolated
pressure jumps. These atmospheric disturbances may have different origin: dynamic
instability, orographic influence, frontal passages, gales, squalls, storms, tornados,
etc.24 However, even during the strongest events, the atmospheric pressure oscilla-
tions at the meteotsunami scales (from a few minutes to a few hours) reach only
2–6 hPa, corresponding to only a 2–6 cm change in sea level. Consequently, these
atmospheric fluctuations may produce significant sea level response only when reso-
nance occurs between the ocean and the atmosphere. During the resonance process,
the atmospheric disturbance propagating above the ocean surface generates signif-
icant long ocean waves by continuously pumping additional energy into these waves.
Possible resonances that are responsible for the formation of meteorological
tsunamis are68 :
These resonant effects may significantly amplify ocean waves approaching the
coast. Nevertheless, even strong resonant amplification of atmospherically generated
ocean waves normally cannot produce waves with sufficient energy to extensively
affect the open coast (for example, a 3–4 hPa pressure jump and a factor of 10
resonant amplification, will only produce ocean wave heights of 30–40 cm). It is
when energetic ocean waves arrive at the entrance of a semi-closed coastal basin
(bay, inlet, fjord or harbor) that they can induce hazardous oscillations in the basin
due to harbor resonance.
On the other hand, intense oscillations inside a harbor (bay or inlet) can only
be formed if the external forcing (i.e., the waves arriving from the open sea)
are energetic enough. Seismically generated tsunami waves in the open ocean can
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
228 A. B. Rabinovich
• A harbor (bay, inlet or fjord) with definite resonant properties and high Q-factor.
• The occurrence of strong small-scale atmospheric disturbance (a pressure jump or
a train of internal atmospheric waves).
• A propagation direction that is head-on toward the entrance of the harbor.
• The occurrence of an external resonance (Proudman, Greenspan or shelf) between
the atmospheric disturbance and ocean waves.
• The occurrence of internal resonance between the dominant frequency of the
incoming open-ocean waves and the fundamental harbor mode frequency.
are: (1) well-defined resonant characteristics of the harbor (bay, inlet, etc.); and
(2) specific properties of the shelf favorable for external resonance (between atmo-
spheric and open-ocean waves) and internal resonance (between arriving open-ocean
waves and harbor oscillations). The combination of these factors for some particular
sites is like a “time-bomb”: sooner or later it will explode (when the atmospheric
disturbance is strong enough and the parameters of disturbance coincide with the
resonant parameters of the corresponding topography/geometry). Locations with
known regular extreme seiches are just the places for these “time-bombs”.51
The catastrophic abiki wave event of 31 March 1979 best illustrates the physical
mechanisms responsible for the generation of meteotsunamis [Fig. 9.10(a)]. Hibiya
and Kajiura29 (HK in the following text) examined this event in detail and con-
structed an effective numerical model that agrees well with observational data.
Nagasaki Bay is a narrow, elongated bay located on the western coast of Kyushu
Island, Japan [Fig. 9.10(b)]; the length of the bay is about 6 km, the width is 1 km,
and the mean depth is 20 m. The fundamental period of the bay (Helmholtz mode)
is 35 min, and this period prevails in seiche oscillations inside the bay (95% of all
observed events) and it was specifically this period that was observed on 31 March
1979.1 HK noticed that almost all known cases of significant abiki waves are asso-
ciated with pressure jumps. In the case of the 1979 event, there was an abrupt
pressure jump (∆Pa ) of 2–6 hPa (according to the observations at several sites)
that propagated eastward (more precisely, 5.6◦ north of east) over the East China
Sea with an approximate mean speed U = 31 m/s (Fig. 9.5). HK approximated
this jump as ∆Pa = 3 hPa over a linear increase distance L1 = 28 km and a linear
decrease distance L2 = 169 km. So, the corresponding static inverted barometer
response of sea level was ∆ζ̄ ≈ −3 cm [Fig. 9.10(a)]. Moreover, the depth of the
East China Sea between mainland China and Kyushu Island is between 50 and
150 m, and the corresponding long-wave speed c ≈ 22–39 m/s. Thus, it was a clas-
sical example of Proudman resonance. HK presented a simple expression describing
resonant amplification of forced open-ocean long waves as:
∆ζ̄ xf
∆ζ = , (9.29)
L1 2
where xf = U t is the distance traveled by the pressure jump during time t. If
L1 = 28 km and xf = 300 km [from the source area to the Goto Islands — see
Fig. 9.10(b)], then ∆ζ ≈ 16 cm. More precise numerical computation with real-
istic two-dimensional bathymetry gives the resonant factor ε = ∆ζ/∆ζ̄ = 4.3 and
∆ζ ≈ 12.9 cm in good agreement with observation. Therefore, due to the reso-
nance, the initial disturbance of 3 cm increased in the open sea by four to five times
[Fig. 9.10(a)]. It is interesting to note that the resonant amplification is inversely
proportional to L1 [see Eq. (9.29)], so the faster the change in atmospheric pressure
(the more abrupt is the pressure jump), the stronger is the amplification of the
generated waves (HK).
According to the HK computations, the outer shelf region between the Goto
Islands and the mainland of Kyushu (“Goto Nada”) has resonant periods of 64, 36,
and 24 min. The second period (36 min) almost coincides with the fundamental
period of Nagasaki Bay (35 min). The Goto Nada shelf did not significantly amplify
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
230 A. B. Rabinovich
(a) U = 31 m/s
3 hPa
4.8 m
measured at air-pressure disturbance
tide gauges U c
2.8 m
Proudman resonance
1.3 m
45 cm 16 cm
3 2 1 3 cm
inlet
shelf wave amplification
through Proudman resonance
Nagasaki Bay
a
East China Se
harbor
resonance shelf
amplification
Kyushu 0
Pa Goto Is
Sea level (cm)
I. -50
31 m/s Nagasaki
-100
-150
China
Kyushu I. Nezumi (1)
3 50
East China Nagasaki 1 2
2 0
Sea
0
Bay
50
-50
0
0 5
10
km 10 12 14 16 18hr
31 March 1979
Fig. 9.10. (a) A sketch illustrating the physical mechanism for formation of the catastrophic
meteotsunami at Nagasaki Bay (Japan) on 31 March 1979. Numbers “1”, “2”, and “3” correspond
to locations shown in (b). (b) Map of Nagasaki Bay and the initial atmospheric pressure disturbance
(shaded rectangular region). (c) Tide records of the catastrophic “abiki waves” of 31 March 1979
at Nezumi (9.1) and Nagasaki (2); positions of the tide gauges are shown in the inset in panel (b).
the incoming wave (the first crest height was 16 cm at the shelf depth of 60 m)
but it selected and amplified waves with specific periods, in particular those with
a period of 36 min. Between the outer sea (depth 60 m) and the head of Nagasaki
Bay, the arriving waves were amplified by a factor of 2.4 due to the combined effects
of topographic convergence, partial reflection, and shoaling inside the bay. Finally,
resonant amplification in Nagasaki of incoming wave train with a period of about
35 min formed catastrophic oscillations within the bay with a maximum recorded
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
wave height of 278 cm [as measured by a tide gauge located in the middle of the
bay — see Fig. 9.10(c)] and an estimated wave height in the head of the bay of
478 cm.1
Thus, for this extreme event, we observe the full combination of “hazardous”
conditions (factors) responsible for the formation catastrophic oscillations inside
Nagasaki Bay: (1) A pronounced atmospheric disturbance (pressure jump of
2–6 hPa), (2) propagating toward the bay with (3) near-resonant phase speed of
31 m/s; this disturbance resonantly generated open-sea long waves with selected
(over the shelf) 36 min period that matched (4) the fundamental 35-min period of
the bay that has (5) high Q-factor and well-defined resonant properties. As a result,
3 cm ocean waves in the source area resulted in 478 cm waves at the head of the bay
(Fig. 9.4).
Analysis of destructive meteotsunami events in the Mediterranean18,22,50,51,64,
71,72,91,92
indicate that the physical mechanisms of these events were similar to
those for Nagasaki Bay event. Tides in the Mediterranean are small; consequently,
harbors are not designed to accommodate large amplitude sea level changes asso-
ciated with occasional meteotsunamis. Consequently, it is atmospherically generated
phenomena (not ordinary tsunamis) that are normally responsible for significant
flooding and damage in this region. However, the main reason for the damaging
nature of meteotsunamis is likely due to the strong currents in the harbor that
accompany the sea level oscillations. Seiches with a 10 min period give raise to cur-
rents that are 70 times stronger than semi-diurnal tides having the same amplitude.
Acknowledgments
This work was initiated by Professor Fred Raichlen (CalTech, Pasadena, CA); the
author sincerely appreciates his help and friendly support. He is also very grateful
to Professor Young Kim (California State University, Los Angeles), the Editor of
this Handbook, for his patience and useful comments and to Drs. Sebastian Mon-
serrat (Universitat de les Illes Balears, Palma de Mallorca, Spain) and Ivica Vilibić
(Institute of Oceanography and Fisheries, Split, Croatia) for their assistance and
providing various observational data. Dr. Richard Thomson (Institute of Ocean Sci-
ences, Sidney, BC, Canada) did tremendous work editing this chapter and encour-
aging the author, Patricia Kimber (Sidney, BC) helped to draft the figures. Partial
financial support was provided by the Russian Foundation on Basic Research, grants
06-0565210a, 08-05-13582, 09-05-00599 and 09-05-01125a and by NATO Science for
Peace project SfP-981382.
References
232 A. B. Rabinovich
3. J. A. Battjes, Surf-zone dynamics, Ann. Rev. Fluid Mech. 20, 257–293 (1988).
4. W. A. M. Botes, K. S. Russel and P. Huizinga, Modeling harbor seiching compared
to prototype data, Proc. 19th Int. Coastal Eng. Conf., Houston (1984), pp. 846–857.
5. A. J. Bowen and D. A. Huntley, Waves, long waves and nearshore morphology, Marine
Geol. 60, 1–13 (1984).
6. E. C. Bowers, The modelling of waves and their effects in harbors, in Hydraulic Mod-
elling in Maritime Engineering (Thomas Telford, London, 1982), pp. 121–127.
7. J. Candela, S. Mazzola, C. Sammari, R. Limeburner, C. J. Lozano, B. Patti and
A. Bonnano, The “Mad Sea” phenomenon in the Strait of Sicily, J. Phys. Oceanogr.
29, 2210–2231 (1999).
8. D. C. Chapman and G. S. Giese, Seiches, Encyclopedia of Ocean Sciences, eds
J. H. Steele, K. K. Turekian and S. A. Thorpe, Vol. 5 (Academic Press, London, 2001)
pp. 2724–2731.
9. P. Colucci and A. Michelato, An approach to study of the ‘Marubbio’ phenomenon,
Boll. Geofis. Theor. Appl. 13, 3–10 (1976).
10. A. Defant, Physical Oceanography, Vol. 2 (Pergamon Press, Oxford, UK, 1961), 598 pp.
11. M. P. C. de Jong, L. H. Holthuijsen and J. A. Battjes, Generation of seiches
by cold fronts over the southern North Sea, J. Geophys. Res. 108, 3117 (2003),
doi:10.1029/2002JC001422.
12. M. P. C. de Jong and J. A. Battjes, Low-frequency sea waves generated by atmospheric
convection cells, J. Geophys. Res. 109, C01011 (2004), doi:10.1029/2003JC001931.
13. V. A. Djumagaliev, E. A. Kulikov and S. L. Soloviev, Analysis of ocean level oscilla-
tions in Malokurilsk Bay caused by tsunami of 16 February 1991, Science Tsunami
Hazards, 11, 47–58 (1993).
14. V. A. Djumagaliev, A. B. Rabinovich and I. V. Fine, Theoretical and experimental esti-
mation of transfer peculiarities of the Malokurilsk Bay coast, the Island of Shikotan,
Atmosph. Oceanic Phys. 30, 680–686 (1994).
15. W. L. Donn, Alaskan earthquake of 27 March 1964: Remote seiche simulation, Science
145, 261–262 (1964).
16. A. F. Drago, A study on the sea level variations and the ‘Milghuba’ phenomenon in the
coastal waters of the Maltese Islands, Ph.D. thesis, University of Southampton (1999).
17. E. Fontseré, Les ‘seixes’ de la costa catalana. Servei Meteorològic de Catalunya, Notes
d’Estudi, 58 (1934) (in Catalan).
18. M. Garcies, D. Gomis and S. Monserrat, Pressure-forced seiches of large amplitude
in inlets of the Balearic Islands. Part II: Observational study, J. Geophys. Res. 101,
6453–6467 (1996).
19. G. S. Giese and D. C. Chapman, Coastal seiches, Oceanus 36, 38–46 (1993).
20. G. S. Giese, D. C. Chapman, P. G. Black and J. A. Fornshell, Causations of large-
amplitude harbor seiches on the Caribbean coast of Puerto Rico, J. Phys. Oceanogr.
20, 1449–1458 (1990).
21. G. S. Giese and R. B. Hollander, The relation between coastal seiches at Palawan
Island and tide-generated internal waves in the Sulu Sea, J. Geophys. 92, 5151–5156
(1987).
22. D. Gomis, S. Monserrat and J. Tintoré, Pressure-forced seiches of large amplitude in
inlets of the Balearic Islands, J. Geophys. Res. 98, 14437–14445 (1993).
23. D. G. Goring, Rissaga (long-wave events) on New Zealand’s eastern seaboard: A
hazard for navigation, Proc. 17th Australasian Coastal Ocean Eng. Conf., 20–23
September 2005, Adelaide, Australia (2005), pp. 137–141.
24. E. E. Gossard and B. H. Hooke, Waves in Atmosphere (Elsevier, New York, 1975),
456 pp.
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
234 A. B. Rabinovich
49. S. Monserrat, A. Ibberson and A. J. Thorpe, Atmospheric gravity waves and the
“rissaga” phenomenon, Q. J. R. Meteorol. Soc. 117, 553–570 (1991).
50. S. Monserrat, A. B. Rabinovich and B. Casas, On the reconstruction of the transfer
function for atmospherically generated seiches, Geophys. Res. Lett. 25, 2197–2200
(1998).
51. S. Monserrat, I. Vilibić and A. B. Rabinovich, Meteotsunamis: Atmospherically
induced destructive ocean waves in the tsunami frequency band, Nat. Hazards Earth
Syst. Sci. 6, 1035–1051 (2006).
52. W. H. Munk, Surf beats, Eos, Trans. Amer. Geophys. Un. 30, 849–854 (1949).
53. W. H. Munk, Long waves, The Sea (J. Wiley, New York, 1962), pp. 647–663.
54. T. S. Murty, Seismic Sea Waves — Tsunamis (Bull. Fish. Res. Board Canada 198,
Ottawa, 1977), 337 pp.
55. T. S. Murty, Modification of hydrographic characteristics, tides, and normal modes
by ice cover, Marine Geodesy 9, 451–468 (1985).
56. M. Nakano, Secondary undulations in bays forming a coupled system, Proc. Phys.
Math. Soc. Japan 3, 372–380 (1932).
57. M. Nakano, Possibility of excitation of secondary undulations in bays by tidal or
oceanic currents, Proc. Imp. Acad. Japan 9, 152–155 (1933).
58. M. Nakano and T. Abe, Standing oscillation of bay water induced by currents, Records
Oceanogr. Works in Japan, Spec. No. 3, 75–96 (1959).
59. M. Nakano and S. Unoki, On the seiches (secondary undulations of tides) along the
coast of Japan, Records Oceanogr. Works Japan, Spec. No. 6, 169–214 (1962).
60. M. Nakano and N. Fujimoto, Seiches in bays forming a coupled system, J. Oceanogr.
Soc. Japan 43, 124–134 (1987).
61. T. Nomitsu, A theory of tsunamis and seiches produced by wind and barometric
gradient, Mem. Coll. Sci. Imp. Univ. Kyoto A 18, 201–214 (1935).
62. M. Okihiro, R. T. Guza and R. J. Seymour, Excitation of seiche observed in a small
harbor, J. Geophys. Res. 98, 18201–18211 (1993).
63. J. Oltman-Shay and R. T. Guza, Infragravity edge wave observations on two California
beaches, J. Phys. Oceanor. 17, 644–663 (1987).
64. M. Orlić, About a possible occurrence of the Proudman resonance in the Adriatic,
Thalassia Jugoslavica 16, 79–88 (1980).
65. J. Proudman, The effects on the sea of changes in atmospheric pressure, Geophys.
Suppl. Mon. Notices R. Astr. Soc. 2, 197–209 (1929).
66. D. Prandle, The use of wave resonators for harbor protection, Dock Harbour Author
(1974), pp. 279–280.
67. A. B. Rabinovich, Possible vorticity effect in longwave motions (surging) in harbors,
Trans. (Doklady) Russian Acad Scie, Earth Sci. Sec. 325, 224–228 (1992).
68. A. B. Rabinovich, Long Ocean Gravity Waves: Trapping, Resonance, and Leaking
(Gidrometeoizdat, St. Petersburg, 1993), 325 pp. (in Russian).
69. A. B. Rabinovich, Spectral analysis of tsunami waves: Separation of source and topog-
raphy effects, J. Geophys. Res. 102, 12663–12676 (1997).
70. A. B. Rabinovich and A. S. Levyant, Influence of seiche oscillations on the formation
of the long-wave spectrum near the coast of the Southern Kuriles, Oceanology 32,
17–23 (1992).
71. A. B. Rabinovich and S. Monserrat, Meteorological tsunamis near the Balearic and
Kuril Islands: Descriptive and statistical analysis, Nat. Hazards 13, 55–90 (1996).
72. A. B. Rabinovich and S. Monserrat, Generation of meteorological tsunamis (large
amplitude seiches) near the Balearic and Kuril Islands, Nat. Hazards 18, 27–55 (1998).
July 31, 2009 8:18 9.75in x 6.5in b684-ch09 FA
236 A. B. Rabinovich
Chapter 10
10.1. Introduction
Tsunamis are the ocean water surface waves generated by an undersea earth-
quake, a landslide, a volcanic eruption, or even a meteoric impact on water surface.
As crust movement of the earth became active, the region along the rim of the
Pacific Ocean experienced large-scale tsunamis including the 1946 Alaskan tsunami,
the 1960 Chilean tsunami, and the 1964 Alaskan tsunami, for instance. These
large-scale tsunamis propagated across the Pacific Ocean, and attacked the coastal
area along the Pacific rim. The damage including loss of human lives caused
by these catastrophic tsunamis was well documented by Lander and Lockridge.3
More recent tsunamis in this area have been investigated by many scientists, and
reported by Satake et al.4 for the 1992 Nicaragua tsunami, and Hokkaido Tsunami
Survey Group5 for the 1993 Hokkaido tsunami. On the other hand, large-scale
tsunamis including the 1992 and 1994 Indonesia tsunamis, 1995 Philippine tsunami,
1996 Indonesia tsunami, and 1998 Papua New Guinea tsunami were also generated
along the rim of the Indian Ocean. In particular, the tsunami of magnitude 9.3
237
July 31, 2009 8:18 9.75in x 6.5in b684-ch10 FA
238 S. B. Yoon
was generated off the west coast of Sumatra in the Indian Ocean on 26 December
2004. The west coast of northern Sumatra was most severely damaged by this
tsunami. The coastal areas along the rim of the Indian Ocean including Malaysia,
Thailand, Myanmar, Bangladesh, India, Sri Lanka, Arabian, and African countries
were attacked by this tsunami. More than 290,000 people were killed due to this
tsunami event. And the property damages and injuries cannot be countable. In
order to reduce such damages caused by tsunamis, a proper monitoring, forecasting,
and warning system should be established. For this we need to fully understand
the mechanism such as generation, propagation, and inundation of tsunamis, and
predict the possible quantities such as arrival times and run-up heights of a tsunami
event. Thus, the development of an accurate and efficient numerical model for com-
puting all the aspects of tsunami propagation is indispensable.
When the tsunamis are generated by an earthquake, they have a shape of
solitary wave with a long crest. After they propagate a long distance from the
source area over a deep ocean, the tsunamis evolve into a train of waves due to
the dispersion effect of waves. When the source area is narrow, the dispersion
effect plays an important role in the deformation of tsunamis also in the rela-
tively shallow water area. Boussinesq equations are one of the best choices for gov-
erning equations to construct a numerical model for the far-field tsunamis, because
they can take into account the dispersion effects. The numerical models such as
FUNWAVE6 and COULWAVE7,8 were developed based on a fully nonlinear and
dispersive Boussinesq equations, and were applied to simulate the 26 December
2004 Indian Ocean tsunami.9,10 However, these Boussinesq models require a small
mesh size to suppress the numerical dispersion error. The small mesh system con-
sumes huge computer resources due to the implicit nature of the solution technique
to deal with dispersion terms. Thus, the Boussinesq model is not preferred for the
simulation of the far-field tsunamis, and the shallow-water equations are generally
employed instead.
Considerable efforts to construct a numerical model for the simulation of
tsunamis have been made. For example, finite difference models were developed by
Hwang and Divoky,11 Goto and Shuto,12 Kowalik and Murty,13 and Mader and
Curtis.14 Most of the existing numerical models are based on the shallow-water
equations and have been successfully applied to the near-field tsunamis which have
little dispersion effect of waves. However, distant tsunamis generated far from the
region of interest are mainly transformed due to the accumulation of dispersion
effects during the propagation. Thus, the numerical models based on the shallow-
water equations will suffer from the lack of accuracy for the simulation of distant
tsunamis.
Imamura et al.15 presented a finite difference model for the simulation of trans-
oceanic propagation of tsunamis. The model solves the shallow-water equations
using a leap-frog scheme. The physical dispersion is compensated by the numerical
dispersion introduced by the truncation error of the numerical scheme. This can be
done only if the grid size is appropriately selected for the given water depth and
the time step satisfying the criterion proposed by Imamura et al.15 Cho16 improved
the numerical model of Imamura et al.15 for tsunamis obliquely propagating to
July 31, 2009 8:18 9.75in x 6.5in b684-ch10 FA
the principal axes of computational grids. However, strictly speaking, those models
have no capability to consider the dispersion effect of waves in the case of practical
application. The use of numerical models developed by Imamura et al.15 and Cho16
is limited to the case of constant water depth. Yoon17 developed a finite-difference
scheme that uses a uniform grid, but the actual computations are made on a hidden
grid of variable size determined from the condition proposed by Imamura et al.15
This model satisfies local dispersion relationships of waves for a slowly varying
topography. However, the accuracy of the model becomes degraded when applied to
short waves due to numerical errors associated with the interpolation of quantities
at hidden grid points.
Yoon et al.1 developed an efficient and relatively accurate finite-difference model
to simulate the propagation of distant tsunamis over varying topography. This
finite-difference model solves a linear Boussinesq-type wave equation employing a
uniform finite-difference grid. A considerable accuracy is achieved by eliminating the
numerical dispersion error and by providing the physical dispersion. The dispersion-
correction is performed by controlling the parameters determined from the given
time step, grid size, and local water depth. As a result, an accurate dispersion-
correction is possible although the grid size does not meet Imamura’s mesh
condition.
As the tsunamis approach the coastal area, they experience a shoaling due to
decreasing water depth over a continental shelf. The wavelength becomes short, and
the grid resolution becomes low near the coastline. Thus, the accuracy of the far-
field model proposed by Yoon et al.1 is degraded. This difficulty can be solved by
nesting a finer grid system to the coarse grid trans-oceanic model. As the tsunamis
approach the coastline, nonlinearity and bottom friction play a significant role in
the transformation of tsunami waves. In this case, a near-field model such as full
Boussinesq or shallow-water equation model can be nested to the trans-oceanic
model as described in Yoon17 and Lim et al.2 Thus, various nesting schemes between
systems with different grid sizes or governing equations are indispensable to extend
the applicability of the far-field numerical model.
The objective of this chapter is to review a recent development of a prac-
tical finite-difference model to simulate tsunami propagation and runup for con-
structing inundation maps. The contents in this chapter are not an original work
but a remake of research works presented by Yoon et al.1 and Lim et al.2,18
Section 10.2 describes governing equations and numerical scheme employed in the
dispersion-correction finite-difference model developed by Yoon et al.1 and Lim
et al.2,18 Section 10.3 presents the test of numerical model for uniform and variable
depth cases. Section 10.4 describes the near-field model. Section 10.5 introduces
the grid nesting scheme developed by Lim et al.2,18 to extend the applicability of
the numerical model. In order to evaluate the applicability of the present model to
real tsunami events, a historical tsunami is simulated and the results are compared
with the observations at two harbors along the east coast of Korea in Sec. 10.6.
Then, the effect of underwater topography on the propagation of tsunamis in this
sea is analyzed based on the simulated results. Finally, this study is summarized
in Sec. 10.7.
July 31, 2009 8:18 9.75in x 6.5in b684-ch10 FA
240 S. B. Yoon
Tsunamis are generally regarded as long waves in comparison with the wind-
generated waves. However, tsunamis can be short when the tsunami source is
narrow. In this case the dispersion effect can play an important role in the trans-
formation of tsunami waves. The dispersion effect, rather than the nonlinear or
the bottom friction effects, dominates the transformation of the far-field tsunamis.
For the far-field tsunami, the linear Bousinessq equations, which neglect the non-
linear effects but include the dispersion effects, can be used as a governing equation.
On the other hand, the nonlinear effect and the bottom friction effect become
important for the transformation of tsunamis in the near-field where the water is
shallow near the coastal area. For the near-field tsunami, the nonlinear shallow-
water equations are generally employed as a governing equation. In the present
numerical model, the linear Boussinesq-type wave equation proposed by Yoon et al.1
is employed for the far-field tsunami propagation, and the nonlinear shallow-water
equations with the bottom friction term are used for the near-field tsunamis.
For trans-oceanic propagation of tsunamis, the nonlinearity of waves can be
neglected because the free surface displacement is much smaller in comparison with
the water depth. However, the dispersion effect of tsunami waves should be con-
sidered properly for the far-field tsunamis. Thus, the following linear Boussinesq
equations can be used as governing equations.
∂ζ ∂P ∂Q
+ + = 0, (10.1)
∂t ∂x ∂y
∂P ∂ζ h2 ∂ ∂ ∂P ∂ ∂Q
+ gh = +
∂t ∂x 2 ∂x ∂x ∂t ∂y ∂t
2
h3 ∂ ∂ P ∂2 Q
− + , (10.2)
6 ∂x ∂t∂x h ∂t∂y h
∂Q ∂ζ h2 ∂ ∂ ∂P ∂ ∂Q
+ gh = +
∂t ∂y 2 ∂y ∂x ∂t ∂y ∂t
2
h3 ∂ ∂ P ∂2 Q
− + , (10.3)
6 ∂y ∂t∂x h ∂t∂y h
where ζ represents the free surface elevation from still water level (m), and P and
Q are the depth-integrated volume fluxes (m2 /s) in the x- and y-directions, respec-
tively, g is the acceleration of gravity (m/s2 ), and h is the still water depth (m).
The equations of motion (10.2) and (10.3) include the dispersion terms in the
right-hand side. These dispersion terms cause numerical difficulties in practice
because of the mixed form of differentiations with respect to both time and space.
Consequently, it calls for the use of implicit scheme, which solves a matrix system.
The implicit scheme requires very fine grids to reduce the numerical dispersion
errors inherent in the numerical scheme such as finite-difference method. A fine grid
July 31, 2009 8:18 9.75in x 6.5in b684-ch10 FA
needs the tremendous computational effort with a huge computer memory space
and excessive execution time.
To develop an efficient and relatively accurate numerical model for the propa-
gation of dispersive tsunamis over slowly varying topography, Yoon et al.1 derived
the following linear Boussinesq-type wave equation (LBTWE) from the linear
Boussinesq equations (10.1)–(10.3) as
∂2ζ ∂ ∂ζ ∂ ∂ζ h2 ∂ 2 ∂ ∂ζ ∂ ∂ζ
− gh + gh − gh + gh
∂t2 ∂x ∂x ∂y ∂y 3 ∂x2 ∂x ∂x ∂y ∂y
∂2 ∂ ∂ζ ∂ ∂ζ
+ gh + gh = 0, (10.4)
∂y 2 ∂x ∂x ∂y ∂y
where the water depth is assumed to be slowly varying in the computational domain.
This equation includes the dispersion terms in the last part. Equation (10.4) can
be split into two first-order partial differential equations in time as
2
∂v ∂h ∂ζ ∂h ∂ζ ∂ ζ ∂ 2ζ
=g + + gh + 2 , (10.5)
∂t ∂x ∂x ∂y ∂y ∂x2 ∂y
2
∂ζ ∂ v ∂ 2v
= v − γ∆x2 + , (10.6)
∂t ∂x2 ∂y 2
where v denotes an auxiliary variable introduced for computational convenience,
and γ is the dispersion-correction parameter. The solution of (10.5) and (10.6) is
identical with that of (10.4) if γ is set to be −h2 /3∆x2 . Since the linear Boussinesq-
type wave equations (10.5) and (10.6) can be solved by a pure explicit numerical
scheme, the dispersive tsunami propagation over a large domain can be simulated
with low computational cost.
Yoon et al.1 developed a dispersion-correction finite-difference model which
solves the equations (10.5) and (10.6) using a relatively simple explicit numerical
scheme given by
n+1/2 n−1/2 n n
vi,j − vi,j hi+1,j − hni−1,j n
ζi+1,j − ζi−1,j
+g
∆t 2∆x 2∆x
n n
hi,j+1 − hni,j−1 n
ζi,j+1 − ζi,j−1
+
2∆y 2∆y
n n n n n n
ζi+1,j − 2ζi,j + ζi−1,j ζi,j+1 − 2ζi,j + ζi,j−1
− gh (1 − α) +
∆x2 ∆y 2
n n n n n n
α ζi+1,j−1 − 2ζi,j−1 + ζi−1,j−1 ζi+1,j+1 − 2ζi,j+1 + ζi−1,j+1
+ +
2 ∆x2 ∆x2
n n n n n n
α ζi−1,j+1 − 2ζi−1,j + ζi−1,j−1 ζi+1,j+1 − 2ζi+1,j + ζi+1,j−1
+ + = 0,
2 ∆y 2 ∆y 2
(10.7)
July 31, 2009 8:18 9.75in x 6.5in b684-ch10 FA
242 S. B. Yoon
and
n+1 n
ζi,j − ζi,j n+1/2
− vi,j + γ∆x2
∆t
n+1/2 n+1/2 n+1/2 n+1/2 n+1/2 n+1/2
vi+1,j − 2vi,j + vi−1,j vi,j+1 − 2vi,j + vi,j−1
× + = 0, (10.8)
∆x2 ∆y 2
where indices i (and j) and n represent a spatial grid point and a time level, respec-
tively, as shown in Fig. 10.1. ∆x(= ∆y) is the spatial grid size, and ∆t denotes the
time step. The weighting factor α to get orientation-free solution is given as 1/6,
and the dispersion-correction parameter γ is given by
∆x2 − (4h2 + gh∆t2 )
γ= . (10.9)
12∆x2
If the dispersion-correction parameter γ calculated by (10.9) for given local water
depth, spatial grid size, and time step are used, a considerably accurate solution
can be achieved for the linear Boussinesq wave equation (10.4). Thus, the selection
of grid size ∆x is free from the condition proposed by Imamura et al.15 This means
that the present dispersion-correction scheme is much more flexible in the selection
of grid size than the conventional scheme.
The stability analysis of the dispersion-correction numerical scheme shows that
the stability criterion is Cr ≤ 0.67 and −0.125 ≤ γ ≤ 0.083, where Cr is the
Courant number. The lower limit of γ imposes a limitation on the ratio of grid
size to water depth as 1.27. This means that the uniform grid size greater than
approximately 1.27 times of a local water depth must be employed for the stability
of the present dispersion-correction scheme. To satisfy the stability criterion of the
dispersion-correction finite-difference model, the range of applicable water depth is
limited if a uniform grid size is used for varying water depth. For practical purposes,
this stability criterion can be solved by imposing intentionally a limitation on the
Fig. 10.1. Sketch of the arrangement of variables and grid points; (a) grid system in space, (b) grid
system in time.
July 31, 2009 8:18 9.75in x 6.5in b684-ch10 FA
γ value within the range of −0.125 ≤ γ ≤ 0.083. This means that the γ value is set
to be −0.125 if the γ calculated by (10.9) is less than −0.125. The γ value is always
less than 0.083. Thus, the upper limit of γ is automatically satisfied for very large
grid size.
where J0 is the 0th order Bessel function of the first kind and g = 9.81 m/s2 . The
parameters used for the solution are a = 7500 m, ∆x = 2086 m, and ∆t = 6 s. The
numerical simulations using the present scheme are performed for various constant
water depths of h = 500 m, 1000 m, and 1500 m.
Figures 10.3 and 10.4 present the comparisons of analytical solutions and
numerical results calculated considering or neglecting the dispersion-correction
Fig. 10.2. Coordinate system and initial free surface profile to test the accuracy of the present
numerical scheme.
July 31, 2009 8:18 9.75in x 6.5in b684-ch10 FA
244 S. B. Yoon
Fig. 10.3. Comparison of free surface calculated using FDM and analytical solution for the case
of h = 500 m (∆xIm = 1085 m, ∆x = 2086 m): (a) γ = 0.0, α = 0.015 ; (b) γ = 0.0, α = 1/616 ;
(c) γ = 0.061, α = 1/6 (present).1
parameters γ and the weighting factor α. We remark here that the numerical scheme
of Imamura et al.15 is recovered if we set intentionally γ = α = 0.0.
The orientation-free scheme of Cho16 is obtained if γ = 0.0 and α = 1/6.
These figures show a time history of free surface displacements at the location of
150∆x(r = 312, 900 m) far from the center of the initial Gaussian hump.
The free surface profiles calculated by the present finite-difference model consid-
ering or neglecting dispersion-correction are compared with the analytical solution
of Carrier19 for the case of 500 m water depth in Fig. 10.3.
July 31, 2009 8:18 9.75in x 6.5in b684-ch10 FA
Fig. 10.4. Comparison of free surface calculated using FDM and analytical solution for the case
of h = 1500 m (∆xIm = 3087 m, ∆x = 2086 m): (a) γ = 0.0, α = 0.015 ; (b) γ = 0.0, α = 1/616 ;
(c) γ = −0.099, α = 1/6 (present).1
Since the grid size ∆x(= 2086 m) employed in the computationis larger than
∆xIm (= 1085 m) evaluated by Imamura condition, i.e., ∆xIm = 4h2 + gh∆t2 ,
the free surface profiles calculated without dispersion-correction are more dispersive
than those of the analytical solutions as shown in Fig. 10.3(a). Especially, the
numerical model gives a less dispersive profile along the diagonal direction (θ = 45◦ )
than along principal axes (θ = 0◦ , 90◦ ), while the numerical solution with α = 1/6
shows no directional dependency as shown in Fig. 10.3(b). The present model with
the dispersion-correction parameter γ = 0.061 and the weighting factor α = 1/6
July 31, 2009 8:18 9.75in x 6.5in b684-ch10 FA
246 S. B. Yoon
gives a good agreement with the analytical solution in the surface profiles in all
directions as shown in Fig. 10.3(c). On the other hand, for the case of 1500 m water
depth, ∆x(= 2086 m) is smaller than ∆xIm (= 3087 m). Thus, the numerical solu-
tions calculated with γ = 0.0 show less dispersive nature in the surface profiles
than the analytical solutions based on the linear Boussinesq equations as shown
in Figs. 10.4(a) and 10.4(b). The present numerical model using γ = −0.099 and
α = 1/6, however, still gives a correct dispersion effect in each direction as shown
in Fig. 10.4(c).
From the results discussed above, it is concluded that the present numerical
model is less sensitive to the choice of grid size than the models of Imamura et al.15
and Cho.16 As a result, the present model gives more flexibility in selecting grid size.
Fig. 10.5. Schematic diagram of a submerged circular shoal, wave gages, and initial condition.
July 31, 2009 8:18 9.75in x 6.5in b684-ch10 FA
the water depth is 1000 m. Gage ➁ is located at the center of the shoal where the
water depth is 500 m. Gage ➃ is placed behind the shoal. Sponge layers are placed
along both ends of the computational domain to absorb the energy of outgoing
waves.
Since no analytical solution is known for this case, numerical solutions using
the present dispersion-correction finite-difference model are compared with those
of the linearized Nwogu’s Boussinesq equations20 implemented in the FUNWAVE
code.6,21 FUNWAVE is a fully nonlinear Boussinesq wave model with improved dis-
persion relationships for short waves. The accuracy of FUNWAVE has been verified
for various coastal problems such as shoaling, refraction, diffraction, and breaking
of waves. The numerical simulation using FUNWAVE is performed with a uniform
grid size of 500 m, the finest grid allowable on the personal computer employed in
this study, to minimize the numerical dispersion. An additional computation using
FUNWAVE is also conducted with 2000 m grid size to test the sensitivity of the
grid resolution. On the other hand, the numerical simulation employing the present
dispersion-correction model is made with a uniform grid size of 2000 m. The time
step ∆t is determined from the stability criteria for each wave model. The algorithms
to compute various physical processes, such as nonlinear advection, nonlinear dis-
persion, and wave dissipation due to bottom friction and breaking, are eliminated
from the source code of the FUNWAVE model. Thus, the computational time for
FUNWAVE model to calculate only the propagation step of small-amplitude waves
can be measured for fair comparison with that of the present model.
The numerical simulation is conducted for 9000 s after the initial water surface
displacement imposed along x = 0 is released. The computational time elapsed
for different models is presented in Table 10.1. The FUNWAVE model employing
a predictor-corrector scheme consumes a long computational time, while the
present fully explicit model takes only 1/10 of the computational time required for
FUNWAVE in the case of using the same grid size, i.e., ∆x = 2000 m. The compu-
tational efficiency of the present model can be realized even more dramatically if
the computational time is compared with that of the FUNWAVE using finer grid
of ∆x = 500 m. The present model is approximately 2200 times faster than the
FUNWAVE model. If the accuracy of the present model is comparable to that of
FUNWAVE, it can be concluded that the present model is highly efficient for prac-
tical problems.
Figure 10.6 presents the comparison of the free surface time history at various
gage locations computed by FUNWAVE and the present FDM. Figure 10.6(a) shows
248 S. B. Yoon
Fig. 10.6. Comparison of time history computed by FUNWAVE and the present FDM:
(a) location ➀ (h = 1000 m); (b) location ➁ (h = 500 m); (c) location ➂ (h = 1000 m); (d) location
➃ (h = 1500 m).1
July 31, 2009 8:18 9.75in x 6.5in b684-ch10 FA
the free surface time history recorded at the wave gage ➀ located on the front slope
of the shoal. This figure shows that the initial solitary wave evolves into a group of
waves due to dispersion of waves. Figure 10.6(b) presents the time history at the
location ➁ on the top of the shoal. The free surface calculated by the present model
shows a good agreement with that of the FUNWAVE with fine grid of ∆x = 500 m.
The results calculated using FUNWAVE with 2000 m grid size, however, show a
considerable phase lag in the tail of wave train. This phase lag is caused by the
accumulation of numerical dispersion errors due to poor resolution of computational
grid near the top of the shoal.
Although the same grid size of 2000 m is employed, the present model still gives
more accurate results. This implies that the present model is capable of minimizing
the numerical dispersion efficiently through the dispersion-correction algorithm.
Figure 10.6(c) shows the time history at location ➂ on the back slope of the shoal.
Two trains of waves arrive here. The first train of waves propagating around the
shoulders of the shoal meets at this location at t = 6000 s. The second train of waves
traveling over the top of the shoal arrives here at t = 6300 s. A good agreement
between the results calculated using the present model and the FUNWAVE with
fine grids is achieved. The results calculated using the FUNWAVE with coarse
grids, however, suffers from the numerical dispersion. Figure 10.6(d) shows the time
history at location ➃ behind the shoal where the water depth h is 1500 m. The
agreements between the numerical solutions are reasonable.
In summary, the present FDM is proven to be sufficiently accurate in comparison
with the FUNWAVE model which can deal with a full coverage of dispersion effects
for the varying water depth region. On top of this, the present model is shown to
be highly efficient. Thus, the present model can be used as a practical numerical
model to simulate the propagation of trans-oceanic tsunamis over slowly varying
topography.
As the tsunamis propagate over a continental shelf and approach a coastal area, the
dispersion effect of waves becomes weak, and the nonlinearity and bottom friction
of waves dominantly influence the transformation of the tsunamis. However, the
linear Boussinesq-type wave equation does not include the nonlinear and bottom
friction terms. Thus, the nonlinear shallow-water equations (NSWE) are employed
for the near-field transformation of tsunamis.
∂ζ ∂P ∂Q
+ + = 0, (10.15)
∂t ∂x ∂y
∂P ∂ P2 ∂ PQ ∂ζ gn2
+ + + gD + 7/3 P P 2 + Q2 = 0, (10.16)
∂t ∂x D ∂y D ∂x D
2
∂Q ∂ PQ ∂ Q ∂ζ gn2
+ + + gD + 7/3 Q P 2 + Q2 = 0, (10.17)
∂t ∂x D ∂y D ∂y D
July 31, 2009 8:18 9.75in x 6.5in b684-ch10 FA
250 S. B. Yoon
where n denotes the Manning’s roughness coefficient (sec/m1/3 ) for the bottom
friction, and D(= h + ζ) is the total water depth. A conventional leap-frog finite
difference scheme with wet–dry technique17 can be employed.
In the central area of the East Sea of Korea, there exists a huge submerged shoal
called Yamato Rise. The depth difference between shallow and deep areas is signif-
icant. To achieve the best performance from the dispersion-correction model, the
grid size for deep area of 4 km water depth should be approximately 6 km. This grid
size causes a problem of poor resolution for the shallow water area near the top of
submerged shoal, because the wavelength becomes shorter due to shoaling. Thus,
the accuracy of the simulation will be seriously degraded. To overcome this diffi-
culty, a nesting between coarser grid for deep area and finer grid for shallow area is
indispensable.
Yoon17 developed a dynamic grid nesting scheme for the finite-difference model
to solve shallow-water equations. This grid nesting scheme uses a second-order
interpolation in space and a first-order interpolation in time. Numerical tests of
the nesting scheme developed by Yoon17 show some reflection for short waves
along the nesting line. To improve the performance of the nesting scheme, a new
dynamic grid nesting scheme has been developed by Lim et al.2,18 for the present
finite-difference model based on the linear Boussinesq-type wave equations. They
employed a third-order interpolation in space and a second-order interpolation
in time. The higher order interpolation helps short waves propagate with weak
reflection from the nesting line. Figures 10.7(a) and 10.7(b) show the layout of vari-
ables in space and time, respectively, to nest the coarse and fine grid regions. The
Fig. 10.7. Layout of variables for grid nesting; (a) grid nesting system in space, (b) grid nesting
system in time.
July 31, 2009 8:18 9.75in x 6.5in b684-ch10 FA
grid size and the time step for the fine grid region are 1/3 of the values of the coarse
grid region. Along the boundary between two regions, the information calculated
in one region should be transferred to another region for each time step. When the
information is transferred from the coarse grid region to the fine grid region, both
spatial and temporal interpolations of the variables of the coarse grid region are
required.
Lim et al.2 tested their grid nesting scheme for the propagation of a linear
solitary wave over a submerged circular shoal with a significant depth difference
between the top and the base of the shoal. Numerical solutions calculated using
their nested model were compared with those of the linearized Nwogu’s Boussinesq
equations20 implemented in the FUNWAVE code.6,21 They showed that the present
model system is approximately 4000 times faster than the FUNWAVE model, while
the accuracy of the present model is comparable to that of FUNWAVE. Thus, it can
be concluded that the present model system is highly efficient for practical problems.
To link the near-field model using nonlinear shallow-water equations to the far-
field model employing linear Boussinesq-type wave equations, special care should
be taken because the governing equations are different between two regions. If the
dispersion-correction is neglected along the nesting line, i.e., γ = 0.0, (10.6) can be
simplified to give
∂ζ
= v. (10.18)
∂t
Furthermore, the continuity equation (10.15) of the nonlinear shallow-water
equations can be rewritten as
∂ζ ∂P ∂Q
=− + . (10.19)
∂t ∂x ∂y
The right-hand side of (10.18) and (10.19) should be identical. Thus, the aux-
iliary variable v can be approximated as
∂P ∂Q
v=− + . (10.20)
∂x ∂y
The auxiliary variable v calculated in the far-field propagation model using LBTWE
is transferred to the near-field model in the form of the divergence of flow rate
as a boundary condition to solve the continuity equation of NSWE. The auxiliary
variable v required for the far-field model can be computed by (10.20) using the
flow rates P and Q calculated in the near-field model.
A historical tsunami event, i.e., the 1993 Hokkaido Tsunami, is selected to verify
the nested dispersion-correction finite-difference model (NDCFDM). The calculated
free surface displacements are compared with the tide gage records obtained at two
harbors: Mukho and Ulsan harbors, along the east coast of Korea.
July 31, 2009 8:18 9.75in x 6.5in b684-ch10 FA
252 S. B. Yoon
where φ and ψ represent the latitude and longitude of grid point, respectively. ∆φ
and ∆ψ are the radian grid size in latitudinal and longitudinal directions, respec-
tively. Since ∆ψ is chosen as 1 min, ∆φ is obtained by multiplying ∆ψ by cos φ,
and ∆x is identical with ∆y everywhere. Thus, the actual grid size is given by
where R(= 6.378 × 106 m) is the radius of the earth. Subsequently, the finite-
difference mesh system taken into account the curvature of the earth can be regarded
Fig. 10.8. Computational domain and bathymetry of the East Sea (depth unit: m).
July 31, 2009 8:18 9.75in x 6.5in b684-ch10 FA
as locally uniform, and the actual grid size slowly changes in association with the
latitude of grid point. It is approximately 1.4 km around the central part of compu-
tational domain.
The time step ∆t is chosen as 3 s, and the actual grid size ∆x slowly varies in the
range of 1312–1556 m. The dispersion-correction parameters γ calculated by given
conditions such as time step ∆t, actual spatial grid ∆x, and local depth h, give the
values between −2.93 and 0.08 in the whole computational domain. As mentioned
above, the minimum value of γ is restricted to −0.125 to maintain the stability for
the deep areas of the East Sea. Due to this limitation, the dispersion effects are
underestimated in the numerical solutions. However, the present scheme gives more
accurate solutions than the conventional models which use a uniform γ value, i.e.,
γ = 0.0, everywhere.
As they approach the shallow-water region, the wavelength of tsunamis becomes
short due to shoaling, and the resolution of grids becomes lower. Thus, a set of finer
grid system, B, C, D, and E, is nested dynamically to the coarse grid region A.
As shown in Fig. 10.8, two subregions, Mukho (C1) and Ulsan (C2) harbors, are
selected for computations with finer grid system, because tide gage records are
available at these harbors. The computational information such as grid size and time
step employed for the nesting is listed in Table 10.2. The subregions are coupled by
the nesting scheme as proposed above. The linear Boussinesq-type wave equation
(LBTWE) is employed to compute the regions, A, B, C, and D. For the finest
grid regions, E1 and E2, the nonlinear shallow-water equations (NSWE) are solved
because the nonlinearity increases as the water depth decreases. Manning-type
energy dissipation for bottom friction is implemented in the nonlinear model. The
Manning’s roughness coefficient is set to be equal to 0.05 s/m1/3 .
To verify the applicability of the present model to the real topography, the
calculated tsunami-free surface elevation is compared with the elevation measured
by tide gage records at two harbors, Mukho and Ulsan harbors. The bathymetry
of computational subregions and the locations of tide gage at Ulsan harbor are
presented in Fig. 10.9.
Two kinds of boundary conditions, no-flux boundary and absorbing boundary,
are mainly used in the tsunami propagation model. No-flux boundary condition is
employed to express the rigid impermeable wall, and an absorbing boundary con-
dition proposed by Larsen and Darcy22 is placed in front of the open-sea boundary.
The fault parameters of the historical tsunami event, such as epicenter location
◦ ◦
in latitude ( N) and longitude ( E), depth of upper rim of fault H(km), orientation
254 S. B. Yoon
Fig. 10.9. Computational subregions and detailed bathymetry near Ulsan harbor (depth unit: m).
Fig. 10.10. Definition sketch for fault parameters of three historical tsunami events.
◦ ◦ ◦
of fault with strike angle θ( ), dip angle δ( ), and slip angle λ( ) from the north,
length of fault L(km), width of fault W (km), and dislocation of fault D(m), are
defined in Fig. 10.10.
It is assumed that the water surface displacement created by the earthquake
is identical to the vertical displacement of sea bed induced by the ground motion.
July 31, 2009 8:18 9.75in x 6.5in b684-ch10 FA
The displacement field can be determined by using the fault model proposed by
Mansinha and Smylie.23
Fig. 10.11. Initial free surface profile of the 1993 Tsunami (unit: m).2
July 31, 2009 8:18 9.75in x 6.5in b684-ch10 FA
256 S. B. Yoon
Fig. 10.13. Distribution of calculated highest water levels due to the 1993 Tsunami (unit: m).2
July 31, 2009 8:18 9.75in x 6.5in b684-ch10 FA
Fig. 10.14. Comparison of calculated and measured time history of free surface displacements of
the 1993 Tsunami at (a) Mukho harbor and (b) Ulsan harbor.2
of the 1993 Tsunami. The Shimane Peninsula (JF) of Japan and the Imwon area
(KA) of the east coast of Korea are shown to be vulnerable to the tsunami attacks
generated in the East Sea because of the large wave heights due to the wave trapping
over the Yamato Rise (YM) and the submerged ridges.
The calculated time histories of free surfaces at two harbors are compared with
the tide gage records of the 1993 Tsunami in Fig. 10.14. The calculated results
are generally in good agreement with measured data in many respects even though
there are slight phase and height differences.
258 S. B. Yoon
Fig. 10.15. Schematic diagram of energy stream path for the 1993 Tsunami.2
effect, and then trapped again by the underwater ridge connecting the Yamato
Rise and the Shimane Peninsula of Japan. As a result, the main stream of tsunami
energy radiated toward the open sea turns back to the Shimane Peninsula (JF) of
Japan. On the other hand, the K-shaped submerged ridges catch the tsunami energy
propagating toward the east coast of Korea and carry it along the two branch ridges
toward the Imwon area (KA) where two ridges meet. Thus, the tsunami energy is
focused at this spot. Even though these regions, i.e., Shimane Peninsula (JF) of
Japan and Imwon (KA) of Korea, are located far from the tsunami source area,
these two regions are vulnerable to the attack of various tsunamis generated in the
East Sea of Korea.
10.7. Conclusions
that the results agree well with the observations. Based on the simulated results
of the historical tsunami that occurred in the East Sea of Korea, it is found that
the underwater topography, such as submerged rises and ridges, plays an important
role in the propagation of tsunamis in this region. The Yamato Rise located in the
central part of the sea behaves as a topographic lens to focus the tsunami energy,
and the submerged ridge connecting the rise to the Shimane Peninsula of Japan
behaves like a waveguide along which the focused energy is trapped and carried to
the Shimane Peninsula. The K-shaped submerged ridges exert a similar influence
on the tsunami incident on the Imwon area in the east coast of Korea.
Acknowledgment
This research was supported by a grant (No. NEMA-06-NH-06) from the Natural
Hazard Mitigation Research Group, National Emergency Management Agency of
Korea.
References
260 S. B. Yoon
11. L. S. Hwang and D. Divoky, Tsunami generation, J. Geophys. Res. 75, 6802–6817
(1970).
12. C. Goto and N. Shuto, Numerical simulation of tsunami propagations and run-up,
Tsunami — Their Science and Engineering, eds. K. Iida and T. Iwasaki (Terra Science
Publishing Company, Tokyo, 1983), pp. 439–451.
13. Z. Kowalik and T. S. Murty, Computation of tsunami amplitudes resulting from a
predicted major earthquake in the Shumagin seismic gap, Geophys. Res. Lett. 11,
1243–1246 (1984).
14. C. L. Mader and G. D. Curtis, Numerical modeling of tsunami inundation of Hilo
harbor, JIMAR Contribution No. 91–251, University of Hawaii, Honolulu (1991).
15. F. Imamura, N. Shuto and C. Goto, Numerical simulation of the transoceanic prop-
agation of tsunamis, Proc. 6th Congress Asian and Pacific Regional Division, IAHR,
Japan (1988), pp. 265–271.
16. Y. S. Cho, Numerical simulations of tsunami propagation and run-up, PhD thesis,
School of Civil and Environmental Engineering, Cornell University, Ithaca, NY (1995).
17. S. B. Yoon, Propagation of distant tsunamis over slowly varying topography,
J. Geophys. Res. 107(C10), 3140 (2002), doi:10.1029/2001JC000791.
18. C. H. Lim, J. S. Bae, Y. J. Jeon and S. B. Yoon, Grid nesting for dispersion-correction
finite difference model for tsunami simulation, Proc. 32nd Congress of the Interna-
tional Association of Hydraulic Engineering and Research, Venice, Italy, Paper No.
D2.b-050-O (2007).
19. G. F. Carrier, Tsunami propagation from a finite source, Proc. 2nd UJNR Tsunami
Workshop, NGDC, Hawaii (1991), pp. 101–115.
20. O. Nwogu, Alternative form of Boussinesq equations for nearshore wave propagation,
J. Waterway, Port, Coastal Ocean Eng. 119(6), 618–638 (1993).
21. G. Wei and J. T. Kirby, A time-dependent numerical code for the extended Boussinesq
equations, J. Waterway, Port, Coastal Ocean Eng. 121(5), 251–261 (1995).
22. J. Larsen and H. Dancy, Open boundaries in short wave simulations — A new
approach, Coastal Eng. 7, 285–297 (1983).
23. L. Mansinha and D. E. Smylie, The displacement fields of inclined faults, Bull. Seismol.
Soc. Am. 61, 1433–1440 (1971).
24. T. Takahash, N. Shuto, F. Iammura and M. Ortis, Fault model to describe Hokkaido
Nansei offshore earthquake for tsunami, J. Coastal Eng. 41, 251–255 (1994) (in
Japanese).
July 31, 2009 8:18 9.75in x 6.5in b684-ch11 FA
Chapter 11
11.1. Introduction
Tsunami waves represent extreme, often catastrophic events, which significantly and
adversely impact coastal areas. In spite of the lower frequency of occurrence com-
paring to storms and storm-induced surges, tsunami-induced coastal flooding often
leads to massive casualties and tremendous economic losses.1–3 Hence, tsunamis are
rare events, high-impact natural disasters.
The devastating effects of the 26 December 2004 Tsunami on many coun-
tries bordering the Indian Ocean raised public concern and revealed existing defi-
ciencies within the current warning and defense systems against tsunamis. One of
the important elements that needs significant improvement is the estimation of
forces generated by tsunami-induced bores, as well as water-borne debris. Before
the 2004 Indian Ocean Tsunami, the design of structures against tsunami-induced
forces was considered of minor importance when compared to the attention given
to tsunami warning systems. This was due to the assumption that tsunamis are
261
July 31, 2009 8:18 9.75in x 6.5in b684-ch11 FA
(a) (b)
(c) (d)
Fig. 11.1. Tsunami damage in Thailand and Indonesia (December 2004 Indian Ocean Tsunami):
(a) severe structural damage, Khao Lak, Thailand; (b) column failure of a reinforced con-
crete frame, Phuket, Thailand; (c) column failure due to debris impact, Banda Aceh, Indonesia;
(d) punching failure of infill walls, Banda Aceh, Indonesia.4
rare events, with significantly high return periods (sometimes more than 500 years).
Reconnaissance missions of the December 2004 Indian Ocean Tsunami disaster
revealed that tsunami-induced forces can lead to severe damage or collapse of struc-
tures as shown in Fig. 11.1.3–11 Therefore, these forces should be properly accounted
for in the design of infrastructure built within a certain distance from the shoreline
in tsunami-prone areas.
The design of coastal structures such as breakwaters, jetties, and groins against
waves is typically based on considering the effect of breaking waves and their asso-
ciated forces, and is well established. Unlike coastal structures, the evaluation of
tsunami-induced hydrodynamic forces on structures used for habitation and/or eco-
nomic activity, received little attention by researchers and engineers.
Results of field surveys conducted in the aftermath of the December 2004 Indian
Ocean Tsunami in Indonesia and Thailand showed that poorly detailed concrete
structures experienced severe damage.3,4 This highlighted the fact that the current
structural design codes do not account for tsunami-induced forces and the impact
of associated debris. Reinforced concrete structures have been observed to with-
stand tsunamis with acceptable low levels of damage.12 However, as shown in
July 31, 2009 8:18 9.75in x 6.5in b684-ch11 FA
Damage
Partial Damage
Withstand
0 1 2 3 4 5 6 7 20
Lack of data
Inundation Depth (m)
Fig. 11.2. Relation between the inundation depth and degree of damage to reinforced concrete
buildings.13
Fig. 11.2, inundation depths of more than 5 m can induce partial damage to concrete
structures.
Currently, there are no clearly established procedures to address the aforemen-
tioned forces. Moreover, significant disagreement on existing empirical formulae fos-
tered new research interest in an effort to properly address the inclusion of both
tsunami-induced forces and the impact of debris into design codes. Aspects related
to these forces are discussed in this chapter. Some of the shortcomings and incon-
sistencies of existing codes are also highlighted.
(a) (b)
(c) (d)
Fig. 11.3. Tsunami wave in Khao Lak, Thailand (December 2004 Indian Ocean Tsunami):
(a) water recedes; (b) waves approach the shoreline; (c) tsunami waves break close to the shoreline;
(d) tsunami waves inundate the shoreline.14
3 2
F H 1 1 H H
= 1.325 + 0.347 ++ , (11.1)
Fl h 7160 h
58.5 h
2 3
M H 1 H 1 H
= 1.923 + 0.454 + + , (11.2)
Ml h 8.21 h 808 h
July 31, 2009 8:18 9.75in x 6.5in b684-ch11 FA
Fig. 11.4. Comparison of: (a) wave profile; (b) runup; (c) pressure head; (d) force due to a strong
turbulent bore and a dry-bed surge.31
where F is the force on the wall; Fl is the force on the wall due to a runup equal to
twice the wave height, assuming hydrostatic pressure; H is the wave height at the
wall; h is still water depth; M is the moment on the wall; and Ml is the moment
corresponding to Fl .
Okada et al.32 conducted a survey of previous studies on tsunami wave forces
and pressures, and identified five empirical formulae for tsunami-induced forces or
pressures. It was found that calculation of tsunami load on structures using these
formulae would result in approximately the same magnitude of load. These formulae
are as follows:
Fig. 11.5. Measured and nondimensional force history for a square column.37
350
250
Force (N)
150
h0 = 1.00 m
h0 = 0.85 m
50
-50
10 12 14 16 18 20 22
Time (s)
Fig. 11.6. Time history of exerted forces on a circular structure38 ; h0 is the impoundment depth.
60 60
Height of the structure (cm)
Height of the structure (cm)
50 50
40 t=0.000 s 40
t=0.160 s
t=0.006 s t=0.170 s
30 t=0.006 s 30 t=0.180 s
t=0.009 s t=0.190 s
20 20
10 10
0 0
-2 3 8 13 18 -2 3 8 13 18
P (kPa) P (kPa)
Fig. 11.7. Variation of pressure distribution on the front face — circular structure38 ; t = 0.0 s is
the instant when the bore impacts the structure.
Fm/γ D2L
σf / γL = 2000
1500
120
1000
500
80
200
100
40
20
0
0.5
0.0 1.0 2.0 3.0 VA0/(gD)
Fig. 11.8. Impact forces of wood logs for bores and surges.39
where γw is the specific weight of wood, D is the diameter of the log, L is the
length of the log, CM is a coefficient which depends on the flow passing around the
receiving wall (≈ 1.7 for bore or surge, and 1.9 for steady flow), u is the velocity of
the log at impact, and σf is the yield stress of the log. Figure 11.8 shows the design
chart based on Eq. (11.3).
Currently, three basic models are proposed for estimating the forces due to the
impact of debris on structures, which are used by a few design codes. In these
models, the impact force is calculated based on the mass and velocity of debris,
while ignoring the mass and rigidity of the structure. However, other than the mass
and velocity of debris, each model needs an additional parameter. The three models
and their corresponding additional parameters are
a large basin where water was stationary and logs were placed on a movable car-
riage. The effect of parameters such as added mass of the water and the eccentricity
and obliqueness of the collision were also considered. Figure 11.9 shows the effect of
impact orientation on the measured force. It was found that the maximum impact
force, Fi,max , can be calculated using Eq. (11.4).
Fi,max = Maxk̂x = u k̂m1 , (11.4)
where u is the impact velocity of the log, k̂ is the constant effective stiffness between
the log and the structure, and m1 is the mass of the log. Based on experiments,
Haehnel et al.40 found the value of k̂ = 2.4 MN/m to be the representative for the
upper envelope of the collected data.
At present, only four design codes and guidelines specifically account for tsunami-
induced loads as listed below:
• FEMA 55: The code is adopted by the Federal Emergency Management Agency,
the United States, and recommends formulae for tsunami-induced flood and wave
loads.43
• The City and County of Honolulu Building Code (CCH): The code, developed
by the Department of Planning and Permitting of Honolulu, Hawaii, United
States, makes provisions for regulations that apply to districts located in flood
and tsunami-risk areas.44
• Structural Design Method of Buildings for Tsunami Resistance (SMBTR): The
code is proposed by the Building Center of Japan32 and outlines the structural
design for tsunami refuge buildings.
• Development of Guidelines for Structures that Serve as Tsunami Vertical Evacu-
ation Sites: The guidelines were prepared by Yeh et al.45 for estimating tsunami-
induced forces on structures for the Washington State Department of Natural
Resources.
There are several other design codes (sometimes country-specific) which contain
prescriptions and design guidelines for flood-induced loads. Examples of widely used
codes are indicated below:
However, none of the above codes address directly the tsunami-induced forces,
which represent the focus of this chapter. The reader is advised to refer to these
codes when seeking guidance for the design of structures subjected to flood-induced
loads other than tsunamis: coastal flooding due to storm surges, flooding of river
banks above bank-full conditions, etc.
(drag) force, (3) buoyant force, (4) surge force, and (5) impact of debris. A brief
description of these forces is further presented.
2
1 u2p
FHS = ρg dS + . (11.5)
2 2g
FB = ρgV, (11.6)
16
CCH: u=h
FEMA 55: u=2(gh)^0.5
Iizuka: u=1.1(gh)^0.5
12 Kirkoz: u=(2gh)^0.5
Murty: u=1.83(gh)^0.5
Bryant: u=1.67h^(0.7)
V (m/s)
0
0 2 4 6
d S (m)
building
qx
3h
Design inundation depth
z
h
z
3ρgh
Fig. 11.11. Tsunami wave pressure for structural design recommended by SMBTR.32
where qx is the tsunami wave pressure for structural design, z is the height of the
relevant portion from ground level (0 ≤ z ≤ 3h), ρ is the mass per unit volume of
water, and g is the gravitational acceleration.
Integration of the wave pressure formula for walls with heights equal to or greater
than 3h results in the same equation as the surge force formula recommended by
CCH [Eq. (11.9)]. The magnitude of the surge force calculated using Eqs. (11.9)
and (11.10) will generate a value equal to nine times the magnitude of the hydro-
static force for the same flow depth. However, a number of experiments31,37 did not
capture such differences in magnitude. Yeh et al.45 commented on the validity of
Eq. (11.9) and indicated that this equation gives “excessively overestimated values.”
On the other hand, Nakano and Paku46 conducted extensive field surveys in order
to examine the validity of the proposed tsunami wave pressure formula [Eq. (11.10)].
The coefficient 3.0 in Eq. (11.10) was taken as a variable, α, and was calculated
such that it could represent the boundary between damage and no damage in the
surveyed data. A value of α equal to 3.0 and 2.0 was found for walls and columns,
respectively. The former is in agreement with the proposed formulae by both CCH
and SMBTR [Eqs. (11.9) and (11.10)].
The tsunami wave force may be composed of drag, inertia, impulse, and
hydraulic gradient components. However, SMBTR does not specify different compo-
nents for the tsunami-induced force, and the proposed formula presumably accounts
for other components.
dub ui
Fi = mb =m , (11.11)
dt ∆t
where Fi is the impact force, mb is the mass of the body impacting the structure,
ub is the velocity of the impacting body (assumed equal to the flow velocity), ui is
approach velocity of the impacting body (assumed equal to the flow velocity), and
∆t is the impact duration taken equal to the time between the initial contact of the
floating body with the building and the instant of maximum impact force.
The only difference between CCH and FEMA 55 resides in the recommended
values for the impact duration which has a noticeable effect on the magnitude of
the force. For example, CCH recommends the use of impact duration of 0.1 s for
concrete structures, while FEMA 55 provides different values for walls and piles for
various construction types as shown in Table 11.1.
According to FEMA 55, the impact force (a single concentrated load) acts hor-
izontally at the flow surface or at any point below it. Its magnitude is equal to
July 31, 2009 8:18 9.75in x 6.5in b684-ch11 FA
the force generated by 455 kg (1000-pound) of debris traveling with the bore and
acting on a 0.092 m2 (1 ft2 ) surface of the structural element. The impact force is
to be applied to the structural element at its most critical location, as determined
by the structural designer. It is assumed that the velocity of the floating body goes
from ub to zero over some small finite time interval (∆t). Finding the most critical
location of impact is a trial and error procedure that depends, to a large extent, on
the experience and intuition of the engineer.
(i) FEMA 55 does not provide loading combinations specifically for calculation
of tsunami force. However, flood load combinations can be used as guidance.
Flood load combinations for piles or open foundations, as well as solid walls
(foundation) in flood hazard zones and coastal high hazard zones are presented
July 31, 2009 8:18 9.75in x 6.5in b684-ch11 FA
as follows:
Pile or open foundation:
Fbrkp (on all piles) + Fi (on one corner or critical pile only), or
Fbrkp (on front row of piles only) + Fdyn (on all piles but front row) +
Fi (on one corner or critical pile only).
Solid (wall) foundation:
Fbrkw (on walls facing shoreline, including hydrostatic component) +
Fdyn (assumes one corner is destroyed by debris),
where Fbrkp , Fi , Fdyn , and Fbrkw refer to breaking force on piles, impact force,
hydrodynamic force, and breaking force on walls, respectively. The reader can
refer to FEMA 5543 for more details.
(ii) Yeh et al.45 modified flood load combinations provided by FEMA 55 and
adapted them for tsunami forces as follows:
Pile or open foundation:
Fbrkp (on column) + Fi (on column), or
Fd (on column) + Fi (on column),
where Fd is the drag force.
Solid (wall ) foundation (perpendicular to flow direction):
Fbrkw (on walls facing shoreline) + Fi (on one corner), or
Fs (on walls facing shoreline) + Fi (on one corner), or
Fd (on walls facing shoreline) + Fi (on one corner),
where Fs is the surge force on walls.
(iii) Dias et al.54 proposed two load combinations called “point of impact” and
“post-submergence/submerged” (Fig. 11.12). These load combinations are
based on two conditions: (i) the instant that tsunami bore impacts the
structure, and (ii) when the whole structure is inundated.
Point of impact:
Fd (on walls facing shoreline) + Fs (on walls facing shoreline),
where Fs is defined as the hydrostatic force by Dias et al.54
Post-submergence/submerged:
Fd (on walls facing shoreline) + Fb (on submerged section of the structure).
The net hydrostatic force is zero and Fb (γV) is the buoyant force.
(a) (b)
Fd Fd
Fs Fs Fs
W W- V
Fig. 11.12. Loading combinations: (a) point of impact/not submerged; and (b) post-submergence/
submerged.54
July 31, 2009 8:18 9.75in x 6.5in b684-ch11 FA
(a)
Fi
h FS
(b)
Fi
dS Fd
FHS FHS
Fig. 11.13. Proposed loading conditions: (a) point of impact; and (b) post-impact.42
(iv) Nouri et al.42 proposed two new load combinations based on the two condi-
tions considered by Dias et al.,54 as shown in Fig. 11.13. The proposed load
combinations by Nouri et al.42 are adapted to follow a consistent format as the
above combinations:
Columns:
Fs (on front row of piles only) + Fi (on one corner or critical column in the
front row only), or
Fd (on all piles) + Fi (on one corner or critical column only),
where Fs is the surge force on walls.
Solid (wall) foundation:
Fs (on walls facing shoreline) + Fi (on walls facing shoreline), or
Fd (on walls facing shoreline) + Fi (on one critical wall facing shoreline) +
Fb (on submerged section of the structure).
Building codes provide guidance for the design of lateral force resisting systems
subjected to wind and seismic excitations. Tsunami-induced loading is normally
not considered. The objective of this example is to demonstrate the levels of
lateral loading associated with tsunamis for a prototype reinforced concrete building
located in a tsunami-prone area. Specifically, the loads generated by a tsunami
bore are addressed. Other researchers have provided comparisons between tsunami
loading and other lateral loads (Okada et al.,32 Pacheco and Robertson,55 Nouri
et al.,42 and Palermo et al.56 ).
July 31, 2009 8:18 9.75in x 6.5in b684-ch11 FA
6.0
B
Tsunami
6.0
C
6.0
D
1 2 3 4 5 6
Fig. 11.14. Plan view of structural layout of reinforced concrete moment-resisting frame.56
the entire surface of the building. In this situation, the drag coefficients are taken
as 1.5 and 1.25 for CCH and FEMA 55, respectively.
Drag Force:
ρCD Au2
FD =
2
m m
u = C gds = 2 9.81 2 (5 m) = 14 ; FEMA 55
s s
m
u = ds = 5 ; CCH
s
2
1030 kg/m3 (2)(56 m2 ) (14m/s)
FD = = 11,317 × 103 N = 11,317 kN; FEMA 55
2
2
1030 kg/m3 (2)(56 m2 ) (5 m/s)
FD = = 1442 × 103 N = 1442 kN. CCH
2
Debris Impact Force:
ui 14m/s
Fi = m = 455 kg = 21.2 × 103 N = 21 kN; FEMA 55
∆t 0.3 s
5 m/s
Fi = 455 kg = 22.8 × 103 N = 23 kN. CCH
0.1 s
m kg
g = 9.81 ; ρ = 1030 ; ds = 5 m
s2 m3
Surge Force:
N
FS = 1.14 × 106 (18 m + 0.45 m)
m
Drag Force:
ρCD Au2
FD =
2
CD = 1.25 Walls FEMA 55
A = (5 m)(18 m + 0.45 m) = 92 m2 ;
1030 kg/m3 (1.25)(92 m2 ) (14 m/s)2
FD =
2
= 11,652 × 103 N = 11,652 kN; FEMA 55
2
1030 kg/m3 (1.5) 92 m2 (5 m/s)
FD = = 1782 × 103 N = 1782 kN. CCH
2
Debris Impact Force:
ui 14 m/s
Fi = m = 455 kg = 21.2 × 103 N = 21 kN; FEMA 55
∆t 0.3 s
5 m/s
Fi = 455 kg = 22.8 × 103 N = 23 kN. CCH
0.1 s
11.4.6. Results
Tables 11.2 through 11.7 provide the results for the calculation of the individual
force components for the structure considered using CCH, FEMA 55, and SMBTR,
respectively.
Given the force components, a loading combination must be specified in order
to evaluate the maximum tsunami load that would be used for either design or
analysis purposes. Yeh et al.45 suggested loading combinations that are applicable
for tsunami loading. CCH does not specifically provide guidance to evaluate the
maximum tsunami load. Nouri et al.42 proposed a two-part loading combination:
Initial impact and Post-impact flow. For this example, these loading combinations
are similar to those of Nouri et al.42 Table 11.8 provides the results of the tsunami
CCH 1 1 82 12 5
2 2 327 92 9
3 3 737 311 14
4 4 1310 738 18
5 5 2046 1442 23
July 31, 2009 8:18 9.75in x 6.5in b684-ch11 FA
CCH 1 1 839 14 5
2 2 3356 114 9
3 3 7550 385 14
4 4 13,423 912 18
5 5 20,973 1782 23
FEMA 55 1 6 453 10
2 9 1811 13
3 11 4074 16
4 13 7243 19
5 14 11,317 21
FEMA 55 1 6 466 10
2 9 1864 13
3 11 4195 16
4 13 7457 19
5 14 11,652 21
load calculation for CCH, FEMA 55 and SMBTR for an inundation depth of 5 m
based on loading combinations of Nouri et al.42
For the prototype moment-resisting frame structure with the short side per-
pendicular to the advancing bore, it is apparent that nonbreakaway walls or rigid
exterior nonstructural components can lead to large design base shears. CCH and
SMBTR estimate significantly larger base shears relative to FEMA 55 due to the
omission of a surge component in FEMA 55. It is evident that the width of exposed
surfaces affects the magnitude of total forces exerted on a structure. Therefore, it
would be prudent to orient buildings such that the short side is placed parallel to the
shoreline. Furthermore, using breakaway or flexible walls at the lower level would
reduce the lateral force that is transmitted to the lateral force resisting system. Note
that although the debris impact force is a negligible component in the base shear,
July 31, 2009 8:18 9.75in x 6.5in b684-ch11 FA
SMBTR 1 36
2 146
3 327
4 582
5 909
SMBTR 1 839
2 3356
3 7550
4 13,423
5 20,973
Table 11.8. Tsunami-induced load based on CCH, FEMA 55, and SMBTR for
5 m inundation depth.
Acknowledgment
References
1. D. C. Cox and J. F. Mink, Bull. Seism. Soc. Am. 53, 1191–1209 (1963).
2. H. Yeh, Natural Hazards 4, 209–220 (1991).
July 31, 2009 8:18 9.75in x 6.5in b684-ch11 FA
31. J. D. Ramsden, J. Waterways, Port, Coastal and Ocean Eng. 122(3), 134–141 (1996).
32. T. Okada, T. Sugano, T. Ishikawa, S. Takai and T. Tateno, The Building Centre of
Japan (2005).
33. R. Asakura, K. Iwase and T. Iketani, Proc. Coastal Eng. JSCE 47, 911–915 (2000).
34. M. Ikeno, N. Mori and H. Tanaka, Proc. Coastal Eng. JSCE 48, 846–850 (2001).
35. H. Iizuka and H. Matsutomi, Proc. Conf. Coastal Eng., JSCE 47 (2000) (in Japanese).
36. M. Omori, N. Fujii and O. Kyotani, Proc. Coastal Eng. JSCE 47, 376–380 (2000).
37. H. Arnason, PhD thesis, University of Washington, Seattle (2005), p. 172.
38. Y. Nouri, I. Nistor, D. Palermo and A. Cornett, Coastal Structures 2007, Venice, Italy
(2007).
39. H. Matsutomi, J. Hyd. Coastal Environ. Eng. JSCE, No. 621/II-47, 111–127 (1999)
(in Japanese, with English abs.).
40. R. B. Haehnel and S. F. Daly, Technical Report: ERDC/CRREL TR-02-2, US Army
Corps of Engineers (2002), p. 40.
41. R. B. Haehnel and S. F. Daly, J. Hydraul. Eng. 130(2), 112–120 (2004).
42. Y. Nouri, I. Nistor, D. Palermo and M. Saatcioglu, 9th Canad. Conf. Earthquake
Engineering, Ottawa, Canada, June (2007).
43. Federal Emergency Management Agency, Coastal Construction Manual (3 Vols.),
3rd edn. (FEMA 55) (Jessup, MD, 2003).
44. Department of Planning and Permitting of Honolulu Hawai, Chapter 16, City and
County of Honolulu Building Code, Article 11 (2000).
45. H. Yeh, I. Robertson and J. Preuss, Report No 2005-4, Washington Dept. of Natural
Resources (2005).
46. Y. Nakano and C. Paku, Summaries of technical papers of Annual Meeting Architec-
tural Institute of Japan (Kinki) (2005).
47. UBC, Int. Conf. Building Officials, 1997 Uniform Building Code, California (1997).
48. ASCE Standard, Minimum design loads for buildings and other structures, SEI/ASCE
7-05, 424 (2006).
49. International Code Council (INC), 2006 International Building Code 2006, Country
Club Hills, IL (2006), p. 675.
50. M. S. Kirkoz, 10th IUGG Int. Tsunami Symposium, Sendai-shi/Miyagi-ken, Japan
(Terra Scientific Publishing, Tokyo, Japan, 1983).
51. T. S. Murty, Bull. Fisheries Res. Board of Canada, No. 198, Department of Fisheries
and the Environment, Fisheries and Marine Service, Scientific Information and Pub-
lishing Branch, Ottawa, Canada (1977).
52. E. A. Bryant, Tsunami: The Underrated Hazard (Cambridge University Press, London,
UK, 2001), p. 320.
53. F. Camfield, Tsunami Engineering (Coastal Engineering Research Center, US Army
Corps of Engineers, 1980), p. 222, Special Report (SR-6).
54. P. Dias, L. Fernando, S. Wathurapatha and Y. De Silva, Int. Conf. Disaster Reduction
on Coasts, Melbourne (2005).
55. K. H. Pacheco and I. N. Robertson, Evaluation of Tsunami Loads and Their Effect
on Reinforced Concrete Buildings. University of Hawaii Research Report (2005).
56. D. Palermo, I. Nistor, Y. Nouri and A. Cornett, PROTECT 2007, Whistler, Canada
(2007).
July 31, 2009 8:18 9.75in x 6.5in b684-ch12 FA
Chapter 12
Hocine Oumeraci
Leichtweiss-Institute for Hydraulic Engineering and Water Resources
Technical University Braunschweig
Beethovenstr. 51a, 38106 Braunschweig, Germany
h.oumeraci@tu-bs.de
After a brief discussion of the necessity (i) to consider in addition to the reduction
of wave heights in the sheltered area further aspects of the hydraulic performance
when developing and designing new coastal structures for the protection against
wave action and (ii) to develop a systematic “roadmap” of the existing concepts
and types of structures together with their hydraulic performance characteristics as
an important tool for practicing engineers and decision-makers, five examples from
selected research studies performed in the last years by the Leichtweiss-Institute
are presented in order to illustrate the development of nonconventional structures
with substantially improved performance as compared to their conventional coun-
terparts. These examples include (i) a multi-chamber structure to overcome the
drawbacks of the perforated JARLAN-type breakwater concept, (ii) an artificial reef
made of successive submerged permeable screens to increase the wave damping per-
formance and to better control both wave reflection and wave transmission, (iii) a
“High Mound Composite Breakwater” (HMCB) concept to decrease breaking wave
loads, wave reflection, overtopping, and spray generation at the structure, (iv) an
onshore wave damping barrier made of staggered walls, and (v) a rubble mound
breakwater with a core made of geotextile sand containers.
12.1. Introduction
In the past, the development of wave damping structures used in coastal protection
and harbor engineering was rather directed toward achieving a maximum reduction
of wave height in the sheltered area, while less attention was generally paid to
further performance aspects such as
(i) the reduction of wave loads, wave reflection, and wave overtopping,
(ii) the reduction of environmental impacts (degradation of the marine landscape
and of the water quality in the sheltered area, erosion of the foreshore, down
coast erosion, etc.),
287
July 31, 2009 8:18 9.75in x 6.5in b684-ch12 FA
288 H. Oumeraci
However, before starting with these example studies, a first attempt is made to
discuss, why and how a parameterization of the existing concepts and structures for
wave damping might help to develop a systematic “roadmap” of these concepts as
an important decision tool for practicing engineers and coastal managers.
A large variety of concepts and structures for wave damping is available. However,
the designer and coastal manager faced with the problem of finding the concept
mostly suited for the intended use have no proper tool to quickly make the right
choice before embarking into detailed investigations. For this purpose, a compact
overview of the most important concepts/types of structures is needed, including the
basic information on their hydraulic functioning (“roadmap”). The roadmap should
not only provide the required performance characteristics. Moreover, the unwanted
effects and impacts may also be directly provided or easily derived for each type
of structure. The latter issue becomes increasingly important, because the choice
will not, as in the past, mainly or exclusively be dictated by the wave damping
efficiency, but also by further criteria associated with the sustainable protection of
coastal zones.18
Unfortunately, such a practical decision tool, which is expected to substantially
help both engineers and coastal managers, is still missing. Due to the considerable
July 31, 2009 8:18 9.75in x 6.5in b684-ch12 FA
Ei Er Et
SWL
H
h1
L
ε hs h
h2
hs > h hs < h hs = 0
h1 = 0 h1 > 0 h1 = 0 h1 > 0
h1 + h2 = h
h2 = 0 h2 = 0 h2 > 0 h2 > 0
1.1 1.2 1.3 1.4 1.5
ε=0
impermeable no structure
b=0
thin 2.1 2.2 2.3 2.4 2.5
walls
ε≠ 0
permeable no structure
ε=0
impermeable
b>0
wide 4.1 4.2 4.3 4.4 4.5
structures
ε≠ 0
permeable
Fig. 12.2. Tentative compact overview of existing wave damping structures in parametrized form.
July 31, 2009 8:18 9.75in x 6.5in b684-ch12 FA
290 H. Oumeraci
It is seen from Figs. 12.1 and 12.2 that only five parameters (structure width B,
porosity ε, structure height dB , submergence depth h1 , and clearance h2 ) together
with the water depth h are sufficient to achieve a parametric description of most
types of existing wave damping structures. But this represents only a first step. The
next important step is to provide the characteristic curves describing the hydraulic
performance for each of the parametrized structure types in Fig. 12.2. For this
purpose, the simplest procedure is to use the fact that the incident wave energy
Ei is converted into a reflected (ER ), a dissipated (ED ), and a transmitted (ET )
energy component:
Ei = ER +E D +E T . (12.1)
Equation (12.1) written in terms of the reflection (KR ), dissipation (KD ), and
transmission (KT ) coefficients yields
2 2
1 = KR + KD + KT2 . (12.2)
PW = 1 − KT2 (12.3)
Pw = K 2
R +K 2
D (
= 1 +KT 2
)
KR = Reflection Coeff.; KD = Dissipation Coeff.; KT = Transmission Coeff.
1.0
Horizontal Breakwater
on Piles (3.5)
Curtain-Wall on
Piles (1.3)
Submerged Breakwater
(3.2)
0.0
0 Wave Period T [s] 16
Fig. 12.4. Hydraulic functioning of a single perforated wall under nonbreaking and nonover-
topping wave conditions (adapted from Ref. 4).
for longer periods. Widening the wall to a caisson (Type 3.3) will considerably
improve the wave damping performance over a wider range of wave periods. Very
often, however, the wave damping performance alone is not sufficient and a “full”
description of the hydraulic functioning as a function of the most relevant parameter
governing wave reflection, wave transmission, and energy dissipation [see Eq. (12.1)]
is required. For instance, in the case of a single surface piercing perforated screen
(Type 2.1) and for nonbreaking and nonovertopping wave conditions, a kind of
“dynamic porosity” (here reflection number RN ) which combines the influence of
incident wave height Hi , water depth h, and wall porosity ε, is found to be the
most appropriate parameter to describe the reflected, transmitted, and dissipated
components of the total incident wave energy (Fig. 12.4).
For most of the types of structure shown in Fig. 12.2, curves describing the
hydraulic performance in terms of wave reflection, transmission, and damping
already exist which are derived empirically, analytically, or numerically.2,11,19,21,22
292 H. Oumeraci
20%
0%
0 0.1 0.2 0.3 0.4 0.5
relative chamber width B/L [-]
Fig. 12.5. Wave damping friction and destructive interference (Adapted from Ref. 5).
with zero-porosity) exhibits a very similar behavior to the reflection coefficient with
respect to the B/L ratio (compare Figs. 12.6 and 12.7).
This first example study has illustrated how a detailed insight into the physical
processes responsible for the hydraulic functioning of an existing concept may lead
to a clear identification of the drawbacks of this concept by indicating how to
overcome them and how to achieve substantial improvements through the intro-
duction of additional structure members. Given its potential to substantially reduce
and better control wave reflection (less risk to navigation and less sea bed scour),
wave loads, wave runup and overtopping, spray generation, etc., the new Multi-
Chamber System represents an ideal alternative as a breakwater, jetty, and quay
wall as well as a seawall for the protection of reclaimed sea fronts and artificial
islands. Due to the flexibility of caisson structures to allow any shape and size,
the seawalls can be adapted to incorporate promenades and any further facility for
recreation activities, etc.
294 H. Oumeraci
Fig. 12.6. Reflection coefficient versus relative chamber width for a traditional JARLAN-type
caisson and for the new multi-chamber system (Adapted from Ref. 4).
Fig. 12.7. Resulting horizontal wave force versus relative chamber width for a traditional
JARLAN-type caisson and for the new multi-chamber system (Adapted from Ref. 4).
July 31, 2009 8:18 9.75in x 6.5in b684-ch12 FA
However, the existing artificial reef concepts have serious drawbacks: the wave
damping performance is limited and sometimes not sufficient. Moreover, the overall
hydraulic performance is difficult to control, due to the limitations associated with
the variation of the structure parameters, etc. Therefore, a reef concept made
of submerged permeable screens with predetermined porosity and spacing has
been experimentally tested in the Large Wave Flume of Hannover (GWK). As
schematically shown in Fig. 12.8 for a three-filter system, this new reef concept
is particularly appropriate for the protection of such coastal areas which are
frequently used for recreation activities.12−14
Before starting with the systematic study on the hydraulic performance of this
new reef concept, it was important to demonstrate first its efficiency with respect
to the protection against beach erosion. For this purpose, a submerged two-filter
system with porosities ε = 11% (front screen) and ε = 5% (back screen), spacing
B = 10.3 m and height dB = 4 m was installed in front of a beach profile. The same
beach profile was previously tested in the Large Wave Flume of Hannover17 without
any protection under the same storm surge conditions (storm surge water level
h = 5.0 m, TMA-wave spectrum with significant wave height Hs = 1.20 m and peak
Fig. 12.8. Submerged three-filter system as an artificial reef for coastal protection.23
July 31, 2009 8:18 9.75in x 6.5in b684-ch12 FA
296 H. Oumeraci
Fig. 12.9. Beach profile development under storm wave attack with and without reef protection.
period TP = 6.6 s, storm duration t = 10 h). The comparison of the results related
to the development of the beach profile at the different stages of the tests with and
without the reef structure as shown in Fig. 12.9 provides a good demonstration of
the efficiency of the new reef concept as a soft protection alternative. In both cases
(i.e., with and without protection), one can indeed observe a beach erosion, a net
seaward transport, and a sand bar formation. However, the net seaward transport
rate is about twice larger for the unprotected beach than for the protected beach.
In addition, the reef causes the seaward transport to occur only within a limited
narrow zone, so that this sand bar does not extend further seaward as in the case
of an unprotected beach. Moreover, the eroded volume above the storm water level
(h = 5.0 m) for the protected beach is only half as much as the eroded volume
for the beach without any protection. As a result, the recession of the shoreline of
the protected beach is only half as much as the recession of the unprotected beach
(Fig. 12.10).
Regarding the hydraulic performance of submerged wave absorbers, a brief illus-
tration is given in Fig. 12.11, showing how a two-filter system damps the waves
over the full range of B/L ratios. The contribution of the seaward screen to the
total wave damping of the system varies between 30% for B/L ≈ 0.5 and 85% for
B/L ≈ 0.3. The maximum wave damping performance of the system also occurs at
B/L = 0.3, while the minimum value is at B/L = 0.5. These and further results
with a three-filter system clearly show that for a given submergence depth (Rc /Hi )
the relative spacing B/L represents the most decisive parameter to describe the
hydraulic performance of wave absorbers.23
July 31, 2009 8:18 9.75in x 6.5in b684-ch12 FA
Fig. 12.10. Beach profile after storm with and without reef (Adapted from Ref. 11).
2.5
dB/h = 0.98
2 wave
Hi/ Ht damping at
filter 1
relative wave height
1.5
H i
1
wave damping at
Ht1/ Ht H H H
filter 2 r t1 t
0.5 h
dB =
Hi/Ht (measurement) B
3 .9 4 m 10m
Ht1/Ht (measurement) 11% 5%
0 F i l ter 1 F i l ter 2
rel. damping [%]
100
50
Fig. 12.11. Wave damping of a submerged two-filter system (Adapted from Ref. 11).
July 31, 2009 8:18 9.75in x 6.5in b684-ch12 FA
298 H. Oumeraci
Fig. 12.12. Hydraulic performance of submerged single screens and filter systems.
From the comparative analysis of the hydraulic performance shown in Fig. 12.12
for a submerged single screen with different porosities (ε = 0%, 5%, and 11%) and
submerged two- or three-filter systems it is seen that
(i) using a filter system instead of a single screen substantially increases the
amount of dissipated energy,
(ii) unlike a single screen, a filter system can substantially reduce and simultane-
ously control both wave reflection and wave transmission,
(iii) using an optimized three-filter system, more than 80% of the incident wave
energy can be dissipated,
(iv) the relative submergence depth Rc /Hi is an important parameter for the wave
damping performance of both single screens and filter systems, and
(v) for the range of practical submergence depths (Rc /Hi ≈ −1), the highest
improvement of the wave damping performance is achieved when using a two-
filter system instead of a single filter. A further increase of the number of
filters would lead to comparatively less improvement of the wave damping
performance.
Further results have indeed shown that the relative spacing B/L is much more
important than the number and porosity of the filters to control the reflected,
July 31, 2009 8:18 9.75in x 6.5in b684-ch12 FA
transmitted, and dissipated wave energy. A more detailed discussion on these issues
is provided by Oumeraci and Koether23 and Koether.11
As mentioned in the introduction, there are many cases where the reduction
of wave periods in the sheltered area may also be very important. Such situations
occur, for instance, when the reef is principally used to control sediment transport
or to reduce wave loads and overtopping at a seawall located behind the reef. An
example result to illustrate the efficiency of the new reef concept by reducing both
transmitted wave periods and heights is shown in Fig. 12.13 for two relative heights
of a three-filter system.
This second example study has shown again that substantial improvements of the
existing concepts can only be achieved through a better insight into the mechanisms
and processes governing the hydraulic performance. In fact, using a single submerged
screen, only a very limited amount of incident wave energy can be dissipated, i.e.,
a decrease of wave transmission can only be achieved at the cost of the increase of
wave reflection (see Ref. 23 for more results).
Using a conventional reef made of rubble material would require a very wide
structure and a progressive decrease of the porosity in wave direction in order
to achieve a satisfactory wave damping performance. This is not only costly and
difficult to construct and to maintain, but it is also very difficult to control the
hydraulic performance by means of a variation of the structure parameters as it is
the case for this new artificial reef concept.
300 H. Oumeraci
The understanding of the underlying mechanisms has shown that a new reef
concept made of two or three submerged thin filters is an elegant and cost-
effective alternative to overcome most of the drawbacks of the existing reef con-
cepts, including a substantial reduction and a better control of the reflected and
transmitted components of the incident wave energy by means of the variation of
the structure parameters (submergence depth, porosity, number and spacing of sub-
merged slit walls.) A theoretical model to optimize submerged wave absorbers has
been developed by Koether.11,23 The model has been successfully validated by large-
scale experimental data for regular and irregular waves as well as for submerged
single-slit wall and wave absorbers with two and three filters.
Preliminary tests in the Large Wave Flume of Hannover using comparatively a
single-wall as well as a two-wall and a three-wall submerged wave absorber subject
to 1 m high solitary waves have shown that this reef concept might also be applied
for the protection against tsunami. In fact, the part of wave energy of the incident
solitary waves dissipated is more than 75% and 85% for a two-wall and a three-
wall system, respectively. The experimental results showing the incident, reflected,
and transmitted waves for a single-, two-, and three-wall systems are plotted and
discussed by Oumeraci and Koether.23
The new HMCB-concept was intended to be used mainly as a seawall for the
protection of artificial islands (offshore airport) and roads with heavy traffic along
the coast, but also as a breakwater [Fig. 12.14(c), 12.14(d)] in harbors. This concept
has recently been implemented at MORI Port in Hokkaido, Japan.14 The governing
characteristic of the HMCB-concept is to cause the highest waves in the spectrum to
break before reaching the crest structure by means of a relatively flat slope (about
1:3). This concept has the following advantages:
(i) the required amount of rubble material is much less than for a conventional
rubble mound breakwater,
(ii) the required armor units are smaller since they are all placed below still water
level, and
(iii) the required crest structure is much smaller than a conventional caisson
breakwater.
302 H. Oumeraci
x̄new HMCB
CoV = x̄ σx (Coeff. of vacation); τ̄x = x̄old HMCB
(mean
mean
reduction factor)
July 31, 2009 8:18 9.75in x 6.5in b684-ch12 FA
(ii) longer-term impacts such as salt corrosion of buildings and other facilities,
damage to agriculture and inland vegetation, etc.
Therefore, one of the most challenging tasks to cope with salt spray consists of
the development of innovative shapes of the structure crest to substantially reduce
the splash/spray height induced by the breaking waves at the structure. For this
purpose, special tests were performed to analyze the effectiveness of various alter-
natives to reduce splash/spray height.8,25,28
As shown in Fig. 12.16, the splash/spray height observed in the Large Wave
Flume tests for different structure alternatives is related to the corresponding value
observed for the new HMCB without any splash reducer as a reference structure
(Type 1). Even without any splash reducer, the new HMCB-concept is capable of
reducing the splash/spray height by half as compared to a conventional impermeable
wall (Type 0). Adding a conical splash reducer on each pile of the front slit wall
and a splash reducer on the back wall will further reduce the splash/spray height by
about half (Type 3). The same improvement was achieved by using a continuous slab
on the crest of the front wall (Type 4), but this has the drawback to substantially
increase the uplift force.28 Since the efficiency of Type 2, both with respect to the
reduction of splash heights and overtopping, was found very low, no further analysis
was performed.
This third example study has highlighted how a very old concept can basically
be improved to provide a new solution with a considerable improvement of the wave
Fig. 12.16. Efficiency of various alternatives for the reduction of splash and spray height.
July 31, 2009 8:18 9.75in x 6.5in b684-ch12 FA
304 H. Oumeraci
Fig. 12.17. Onshore wave damping barrier for the North Sea Island Norderney (Adapted from
Ref. 29).
July 31, 2009 8:18 9.75in x 6.5in b684-ch12 FA
306 H. Oumeraci
On the other hand, the use of sand as a quasi-impervious core instead of quarry
stone would result in an increase of
308 H. Oumeraci
Fig. 12.21. Class of geocore structures in comparison to conventional rubble mound structures.
tan α
Kr = 0.43 · √ , (12.4)
ko d
July 31, 2009 8:18 9.75in x 6.5in b684-ch12 FA
In fact, the best-fit in analyzing the experimental results for the geocore and con-
ventional breakwater was found by using the modified freeboard R∗ according to
Eq. (12.5). As a result, the following formula was obtained for the transmission
coefficient (Fig. 12.23):
where ar and br are constants, which depend on the permeability of the structure.
The structure parameters ar and br are obtained for the conventional and the
geocore breakwater (Fig. 12.23).
July 31, 2009 8:18 9.75in x 6.5in b684-ch12 FA
310 H. Oumeraci
Fig. 12.23. Wave transmission coefficient for conventional breakwater and geocore breakwater.
Fig. 12.24. Runup for conventional breakwater versus runup for geocore breakwater.
where
q
Q∗ = 3 )0.5
,
(g · Hm0
312 H. Oumeraci
Fig. 12.25. Average wave overtopping for conventional and geocore breakwater.
not available in sufficient quantities and at affordable costs or when the reduction
of wave transmission and sediment infiltration through the rubble mound structure
is an important issue. However, one should keep in mind that using a GSC-core
instead of a conventional core will result in larger runup heights and higher wave
overtopping rates. Moreover, larger armor units are required to ensure hydraulic
stability. More results are given by Oumeraci et al.24
The advantages of the nonconventional “geocore” concept are expected to
be particularly revealed when applied to a seawall protecting reclaimed land
(Fig. 12.17(d)), due to the difficulties to protect the latter from being washed out
through a conventional core made of quarry run.
This chapter has attempted to show that one of the challenges of a sustainable pro-
tection against wave attack is the development of innovative structures/concepts
which are not only capable to further reduce the wave heights in the sheltered area,
but also to achieve substantial improvements with respect to the reduction of wave
loads, wave reflection, and overtopping. Moreover, the novel types of structures
should also be able to cope with new emerging requirements such as (i) a better
control of the wave periods in the sheltered area, (ii) a reduction of spray gener-
ation at the structure, and (iii) a better adaptability to multi-purpose use. Since
the importance of these new emerging design criteria is expected to considerably
increase in the context of the sustainable development of coastal zones, they might
July 31, 2009 8:18 9.75in x 6.5in b684-ch12 FA
represent challenging candidate topics for future research. A further important can-
didate issue for applied research will certainly consist of the elaboration of a kind
of “roadmap” of the existing wave damping concepts (possibly in a parametrized
form!), together with the associated characteristic curves which not only summarize
the associated hydraulic performance, but also the possible negative impacts and
further information required for a fast decision-making.
Acknowledgments
Most of the ideas, material, and illustrative examples used herein are based
on inputs taken from completed research projects supported by the German
Research Council (DFG, Bonn), the Federal Ministry for Education and Research
(BMBF, Bonn), the Science and Technology Agency of Japan (STA, Tokyo),
and the Fifth Port Construction Bureau of the Ministry of Transport (MOT,
Tokyo), the European Community within the EU-MAST research program, NLWK
Norden and BBG Bauberatung GmbH & Co. KG, Germany. The author grate-
fully acknowledges the financial support. Since most of these projects were
joint projects with Port and Harbor Research Institute (PHRI, Yokosuka), Civil
Engineering Research Institute (CERI, Sapporo), the Institute of Naval Archi-
tecture and Ocean Engineering, TU Berlin, and many further European Insti-
tutes, the author also gratefully acknowledges the fruitful cooperation with these
partners.
References
314 H. Oumeraci
Chapter 13
Kyung-Duck Suh
Department of Civil and Environmental Engineering
Seoul National University, San 56-1, Shillim-dong, Gwanak-gu
Seoul 151-744, Republic of Korea
kdsuh@snu.ac.kr
During the past several decades, breakwaters including perforated walls have been
introduced to resolve various problems associated with gravity-type breakwaters.
In this chapter, first, the mathematical models are described that predict various
hydrodynamic characteristics of single- or multiple-row curtain-wall-pile break-
waters, the upper part of which is a curtain wall and the lower part consisting of an
array of vertical piles. Their extension to irregular waves is also described. These
models can be used for curtain-wall breakwaters by just removing the piles. They
can also be used for pile breakwaters by removing the curtain-wall and extending
the piles to the water surface. Second, the mathematical model to predict wave
reflection from a fully-perforated-wall caisson mounted on a rubble foundation is
described, and its applicability to a partially-perforated-wall caisson and irregular
waves is described. Third, a discussion is given for the calculation of the so-called
permeability parameter, which represents the energy dissipation and phase-shift
of flows passing across a perforated wall.
13.1. Introduction
Gravity-type breakwaters using rubble mound or vertical caissons have been widely
used to provide a calm basin for ships and to protect harbor facilities from rough
seas. In general, the width of these breakwaters increases with water depth, leaving
a large footprint and requiring a great amount of construction material, especially
when built in deeper water. Often, they block littoral drift and cause severe erosion
or accretion in neighboring beaches. In addition, they prevent the circulation of
water, thus deteriorating the water quality within the harbor. In some places, they
obstruct the passage of fishes and bottom-dwelling organisms. A solid soil foun-
dation is also needed to support such heavy structures.
317
July 31, 2009 8:18 9.75in x 6.5in b684-ch13 FA
Fig. 13.1. A definition sketch of a curtain-wall-pile breakwater: (a) side view; and (b) front view.
b = thickness of the wall. A Cartesian coordinate system (x, z) is defined with the
positive x directing downwave from the crest line of the breakwater and the vertical
coordinate z being measured vertically upward from the still water line. The distance
between the centers of two neighboring piles is denoted as 2A, and the width of an
opening is 2a so that the porosity of the lower part of the breakwater at x = 0 is
defined as r0 = a/A. A regular wave train with wave height Hi is incident on the
positive x-direction. The fluid domain is divided into region 1 (x ≤ 0) and region 2
(x ≥ 0).
Assuming incompressible fluid and irrotational flow motion, the velocity
potential exists, which satisfies the Laplace equation. Linearizing the free-surface
boundary conditions, the following boundary value problem for the velocity
potential Φ(x, z, t) is obtained:
∂ 2Φ ∂ 2Φ
+ = 0, (13.1)
∂x2 ∂z 2
∂Φ ω 2
− Φ = 0 at z = 0 , (13.2)
∂z g
∂Φ
= 0 at z = −h , (13.3)
∂z
where ω = wave angular frequency, and g = gravitational acceleration. Assuming
periodic motion in time t, we can assume the solution to the above problem as
igHi 1
Φ(x, z, t) = Re − φ(x, z) exp(−iωt) , (13.4)
2ω cosh(kh)
√
where i = −1; and the symbol Re represents the real part of a complex value.
The wave number k must satisfy the dispersion relationship, ω 2 = gk tanh(kh).
The spatial variation of the velocity potential φ(x, z) should be determined in each
region.
July 31, 2009 8:18 9.75in x 6.5in b684-ch13 FA
We assume that the length scale of the flow near the breakwater is of the order of
the wall thickness, which is much smaller than the far-field length scale of O(k−1 ),
so that the wall has no thickness mathematically and the 3D feature near the
breakwater does not significantly affect the 2D far-field solutions. Then, φ1 (x, z)
and φ2 (x, z) must satisfy the following matching conditions at x = 0:
∂φ1 ∂φ2
= = 0 for − d ≤ z ≤ 0, x = 0 , (13.5)
∂x ∂x
∂φ1 ∂φ2
= = iG(φ1 − φ2 ) for − h ≤ z ≤ −d, x = 0 , (13.6)
∂x ∂x
where the subscripts indicate the regions of the fluid domain. The first matching
condition describes that the horizontal velocities vanish on both sides of the curtain
wall of the breakwater. The second one for the lower part of the breakwater describes
that the horizontal mass fluxes (or indirectly horizontal velocities) in the two regions
must be the same at the breakwater and that the horizontal velocity at the opening is
proportional to the difference of velocity potentials, or the pressure difference, across
the breakwater. The proportional constant G, often called permeability parameter,
is in general complex, which will be discussed in detail later.
To solve the boundary value problem [Eqs. (13.1) to (13.3)], we use the eigen-
function expansion method of Isaacson et al.4 The velocity potential is expressed in
a series of infinite number of solutions:
∞
φ1 = φi − Am cos[µm (h + z)] exp(µm x) , (13.7)
m=0
∞
φ2 = φi + Am cos[µm (h + z)] exp(−µm x) , (13.8)
m=0
where φi = cosh[k(h+z)] exp(ikx) is the incident wave potential. The wave numbers
µm are the solution of the dispersion relation, ω 2 = −gµm tan(µm h), which has an
infinite discrete set of real roots ±µm (m ≥ 1) for nonpropagating evanescent waves
and a pair of imaginary roots µ0 = ±ik for propagating waves. We take µ0 = −ik
so that the propagating waves in Eqs. (13.7) and (13.8) correspond to the reflected
and transmitted waves, respectively. We also take the positive roots for m ≥ 1 so
that the nonpropagating waves die out exponentially with the distance from the
breakwater.
Now, Eqs. (13.7) and (13.8) satisfy the free-surface boundary condition in
Eq. (13.2) and the bottom boundary condition in Eq. (13.3). Also, they automat-
ically satisfy the requirement that the horizontal velocities must be matched at
the breakwater. In order to solve for the unknown coefficients, Am s, we use the
matching conditions at the breakwater. First, Eqs. (13.7) and (13.8) are substi-
tuted into Eqs. (13.5) and (13.6), respectively. Multiplying each resulting equation
by cos[µn (h + z)], integrating with respect to z over the appropriate domain of z
(i.e., z = −d to 0, or z = −h to −d), and finally adding them, we obtain a matrix
July 31, 2009 8:18 9.75in x 6.5in b684-ch13 FA
equation for Am :
∞
Cmn Am = bn for n = 0, 1, 2, . . . , ∞ , (13.9)
m=0
where
Cmn = µm fmn (−d, 0) + (µm − 2iG)fmn (−h, −d) , (13.10)
bn = −µ0 [f0n (−d, 0) + f0n (−h, −d)] , (13.11)
q
fmn (p, q) = cos[µm (h + z)] cos[µn (h + z)]dz
p
q
1 sin[(µm + µn )(h + z)] sin[(µm − µn )(h + z)]
2 + for m = n , (13.12)
µm + µn µm − µn p
=
1
[2µm (h + z) + sin[2µm (h + z)]]qp for m = n .
4µm
where ρ = density of fluid. The second term on the right-hand side represents
the second-order force contribution of the wave crest regions on the upwave and
downwave sides of the breakwater.26 Without this term, in the limiting case of a
full-depth impermeable vertical wall (d → h or r0 → 0), Am = 0 for all m ≥ 1 and
A0 = −1 so that the preceding equation becomes
s 1 sinh(kh)
Fmax = ρgHi , (13.17)
k cosh(kh)
where the superscript s stands for the standing wave in front of an impermeable
vertical wall. In another limiting case of no breakwater (d → 0 and r0 → 1), Fmax
becomes zero, as expected.
The spatial variation of the velocity potential φj (x, z) should be determined in each
region.
The solutions in the regions of Ω0 and ΩJ are given by
z
Hi
Ω0 Ω1 Ω2 Ωj-1 Ωj ΩJ-1 ΩJ
x
bj dj
d1 d2 dj dJ
h
b1 b2 bj bJ
r1 r2 rj rJ
x1 x2 xj xJ
2aj
2Aj
(a) (b)
Fig. 13.2. A definition sketch of a multiple-row curtain-wall-pile breakwater: (a) side view; and
(b) bird’s-eye view of jth row.
∞
φJ = AJm cos[µm (h + z)] exp[−µm (x − xJ )] , (13.20)
m=0
respectively. Here, Ajm and Bjm are the coefficients of the component waves prop-
agating forward and backward, respectively. The first subscript (j) indicates the
row of the wall, while the second one (m) indicates the wave component. The
wave numbers µm are the same as those used for a single-row breakwater. Taking
A00 = 1, the first term on the right-hand side of Eq. (13.19) represents the incident
wave potential, and |B00 | and |AJ0 | are the reflection and transmission coefficients,
respectively. On the other hand, the solutions in the regions from Ω1 to ΩJ−1 are
given by
∞
φj = {Ajm exp[−µm (x − xj )] + Bjm exp[µm (x − xj+1 )]} cos[µm (h + z)] ,
m=0
j = 1, 2, . . . , J − 1. (13.21)
The matching conditions at each row of the breakwater are the same as
Eqs. (13.5) and (13.6), i.e.,
∂φj−1 ∂φj
= = 0 for − dj ≤ z ≤ 0, x = xj , (13.22)
∂x ∂x
∂φj−1 ∂φj
= = iGj (φj−1 − φj ) for − h ≤ z ≤ −dj , x = xj . (13.23)
∂x ∂x
These matching conditions are used to solve for the unknown coefficients Ajm s and
Bjm s. First, for x = x1 , Eqs. (13.19) and (13.21) are substituted into Eqs. (13.22)
and (13.23). Multiplying each resulting equation by cos[µn (h + z)], integrating with
July 31, 2009 8:18 9.75in x 6.5in b684-ch13 FA
+ iG1 fmn (−h, −d1 )A1m + iG1 exp(−µm ∆x1 )fmn (−h, −d1 )}B1m ]
= µ0 A00 {f0n (−d1 , 0) + f0n (−h, −d1 )} + iG1 A00 f0n (−h, −d1 )
for n = 0, 1, 2, . . . , ∞ (13.24)
and
∞
[µm {fmn (−d1 , 0) + fmn (−h, −d1 )}B0m
m=0
where
and fmn (p, q) are the same as given in Eq. (13.12). For x = xj (j = 2 to J − 1),
Eq. (13.21) is substituted into Eqs. (13.22) and (13.23). Again, multiplying each
resulting equation by cos[µn (h + z)], integrating with respect to z over the appro-
priate domain of z, and finally adding them, we obtain
∞
[exp(−µm ∆xj−1 ){−µm fmn (−dj , 0)+µm fmn (−h, −d1 )
m=0
+ iGj fmn (−h, −dj )}Aj−1,m + {µm fmn (−dj , 0) − µm fmn (−h, −dj )
+ iGj fmn (−h, −dj )}Bj−1,m − iGj fmn (−h, −dj )Ajm
− iGj exp(−µm ∆xj )fmn (−h, −dj )Bjm ] = 0
for n = 0, 1, 2, . . . , ∞ (13.27)
and
∞
[µm exp(−µm ∆xj−1 ){fmn(−dj , 0)+fmn(−h, −dj )}Aj−1,m
m=0
Finally, for x = xJ , Eqs. (13.20) and (13.21) are substituted into Eqs. (13.22) and
(13.23). Using the similar procedure as above, we obtain
∞
[exp(−µm ∆xJ−1 ){−µm fmn(−dJ , 0) + µm fmn (−h, −dJ )
m=0
and
∞
[µm exp(−µm ∆xJ−1 ){fmn (−dJ , 0)+fmn (−h, −dJ )}AJ−1,m
m=0
If we take M wave modes (i.e., one progressive and M − 1 evanescent modes), the
number of unknown coefficients is 2JM . Equations (13.24), (13.25), and (13.27)–
(13.30) give 2JM equations, which can be solved for the unknown coefficients.
For a multiple-row breakwater, the reflection and transmission coefficients are
given by
Cr = |B00 | , (13.31)
and
Ct = |AJ0 | , (13.32)
respectively. The wave runup on the upwave side of the first row of the breakwater
is given by
Hi ∞
1
Ru = A00 cos(µ0 h) + B0m cos(µm h) . (13.33)
2 cosh(kh) m=0
The wave force on each wall can be calculated by integrating the wave pressure
acting on upwave and downwave sides of the wall. The maximum horizontal wave
force Fmax per unit width of the first wall of breakwater is given by
ρgHi 1 A00
Fmax = sin(µ0 h)
2 cosh(kh) µ0
∞
1
+ {B0m − A1m − B1m exp[µm (x1 − x2 )]} sin(µm h)
m=0 m
µ
A00
−r1 sin[µ0 (h − d1 )]
µ0
July 31, 2009 8:18 9.75in x 6.5in b684-ch13 FA
∞
1
+ {B0m − A1m − B1m exp[µm (x1 − x2 )]} sin[µm (h − d1 )]
µ
m=0 m
2
ρgHi2 1 ∞
+ A00 cos(µ0 h) + B0m cos(µm h)
8 cosh2 (kh) m=0
2
∞
− {A1m + B1m exp[µm (x1 − x2 )]} cos(µm h) .
m=0
(13.34)
Again, the second term on the right-hand side represents the second-order force
contribution of the wave crest regions on the upwave and downwave sides of the
wall. Without this term, in the limiting case of a full-depth impermeable vertical
wall, B0m = 0 for all m ≥ 1, B00 = A00 = 1, and A1m = B1m = 0 for all m ≥ 0, so
that the preceding equation becomes Eq. (13.17).
where m0,i , m0,r , and m0,t = zeroth moments of the incident, reflected, and trans-
mitted wave spectra, respectively. Note that the reflection and transmission coef-
ficients calculated by the conventional method, as the ratio of the reflected and
July 31, 2009 8:18 9.75in x 6.5in b684-ch13 FA
transmitted rms wave heights to the incident one, are equivalent to the frequency-
√
averaged coefficients, because the rms wave height is proportional to m0 .
The force spectrum, SF (f ), is calculated by
2
SF (f ) = |TF (f )| Sη,i (f ) , (13.39)
where TF (f ) is the frequency-dependent transfer function between wave amplitude
and force amplitude, which can be computed by using the linear term in
Eq. (13.16) as
∞
1 Am
TF = 2ρg {sin(µm h) − r0 sin[µm (h − d)]} . (13.40)
cosh(kh) m=0 µm
Note that a factor of 2 instead of 4 was used in order to represent the force amplitude.
Fig. 13.3. A schematic diagram and coordinate system for calculation of wave reflection form a
perforated-wall caisson breakwater.
July 31, 2009 8:18 9.75in x 6.5in b684-ch13 FA
coordinate z is measured vertically upward from the still water line. In region 2
(−b ≤ x ≤ 0), the water depth h(x) is a varying function of x. For x ≤ −b (region 1)
and 0 ≤ x ≤ B (region 3), the water depth is constant and equal to h1 and h3 ,
respectively. The rubble mound foundation is assumed to be impervious.
Assuming incompressible fluid and irrotational flow motion, the velocity
potential Φ(x, y, z, t) for the monochromatic wave propagating over the water depth
h(x) with the angular frequency ω and wave height H can be expressed as
igH
Φ(x, y, z, t) = Re − φ(x, y, z) exp(−iωt) . (13.42)
2ω
∂ 2φ
∇2 φ + = 0, (13.43)
∂z 2
∂φ
− λφ = 0 at z = 0 , (13.44)
∂z
∂φ ∂φ dh
+ = 0 at z = −h(x) , (13.45)
∂z ∂x dx
M
φ(x, y, z) = ϕm (x, y)Zm (x, z) . (13.46)
m=0
cos[µm (z + h)]
Zm (x, z) = , (13.47)
cos(µm h)
where
νj,m = µ2j,m + χ2 , χ = kj sin θj = k1 sin θ1 = constant (j = 1, 2, 3) . (13.51)
If m = 0 (propagating mode), the following relationships are obtained:
λ = kj tanh(kj hj ), νj,0 = ±ikj cos θj . (13.52)
For νj,0 , we take the negative sign for the reflected wave in region 1, while we need both
positive and negative signs for the waves inside the wave chamber (i.e., region 3).
The potential φj (x, y, z) must satisfy the matching conditions which provide
continuity of pressure and horizontal velocity, normal to the vertical planes sepa-
rating the fluid regions and no-flux condition at the wall on the downwave side of
the wave chamber, i.e.,
∂φ1 ∂φ2
φ1 = φ2 , = (x = −b, −h1 ≤ z ≤ 0) , (13.53)
∂x ∂x
i ∂φ2 ∂φ2 ∂φ3
φ3 = φ2 + , = (x = 0, −h3 ≤ z ≤ 0) , (13.54)
G ∂x ∂x ∂x
∂φ3
= 0 (x = B, −h3 ≤ z ≤ 0) . (13.55)
∂x
As shown in Eq. (13.46), the Galerkin solution consists of a free propagating
wave mode and nonpropagating evanescent modes. The evanescent wave modes
would be of importance in the region near the breakwater. However, since we are
interested in the reflected wave far from the breakwater, we focus on the solution
for the propagating mode. Massel28 showed that in region 2, the function ϕ̃0 (x)
satisfies the following ordinary differential equation:
d2 ϕ̃0 dϕ̃0
+ D(x) + E(x)ϕ̃0 = 0 , (13.56)
dx2 dx
where
k 2τ dh
D(x) = 1 − 3τ +
2
, (13.57)
τ + kh(1 − τ 2 ) τ + kh(1 − τ 2 ) dx
2
u1 dh u 2 d2 h
E(x) = k 1 + 2
2
+ 2 − χ2 , (13.58)
k u0 dx k u0 dx2
July 31, 2009 8:18 9.75in x 6.5in b684-ch13 FA
1 K
u0 = tanh(kh) 1 + , (13.59)
2k sinh K
sech2 (kh)
u1 = (sinh K − K cosh K) , (13.60)
4(K + sinh K)
ksech2 (kh)
u2 = [K 4 + 4K 3 sinh K − 9 sinh K sinh(2K)
12(K + sinh K)3
+ 3K(K + 2 sinh K)(cosh2 K − 2 cosh K + 3)] , (13.61)
where the abbreviation K = 2kh was used.
In order to solve Eq. (13.56) in region 2, we need the boundary conditions at
x = −b and x = 0. Suh and Park23 derived these boundary conditions as
dϕ̃0 (−b)
= i[2 − ϕ̃0 (−b)]k1 cos θ1 , (13.62)
dx
1.0
0.9
0.8
Mound Slope
0.7
0.6
0.5
0.4
0.3
Fig. 13.4. Contours of the ratio of the reflection coefficient of the breakwater with a mound to
that of the breakwater without a mound as a function of mound height and slope.
the height and slope of the mound as expected. Within the range of the practical
design situations, however, the rubble mound effect on wave reflection is not so large.
However, it should be noted that the effect of permeability of the rubble mound was
not take into account because the rubble mound was assumed to be impervious.
Suh et al.30 also examined the effect of angle of incidence upon wave reflection.
The reflection coefficients were calculated by changing the wave angle from 0◦ to
80◦ and B cos θ/L from 0.1 to 0.5. The caisson was situated in a constant depth
without a mound so that h1 = h2 = h3 = 30 cm. Wave height H = 3 cm, wave
period T = 1.554 s, porosity of the wall r = 0.25 were used, while the perforated
wall thickness was taken to be zero (i.e., the inertia effect of the perforated wall was
neglected). The calculated results are shown in Fig. 13.5. For normal incidence of
waves, the wave reflection is minimum at B cos θ/L = 0.25 and increases as it gets
away from this value. The same trend is kept for oblique incidence of waves, but
the solution exhibits singular behavior (i.e., the reflection coefficient shows a sharp
peak) at B/L = 0.5, 1.0, 1.5, and so on, as shown in Fig. 13.5. In the practical design
of a perforated-wall caisson, the value of B/L is usually less than 0.5. Thus, when
the waves are obliquely incident to the breakwater, to minimize the wave reflection,
the value of B cos θ/L should be around 0.25, but the condition B/L ∼ = 0.5 should
be avoided.
0.5
B/L = 0.5 1.0 1.5
0.4
Bcos θ / L
0.3
0.2
0.1
0 20 40 60 80
Wave Angle (degree)
Fig. 13.5. Contours of the reflection coefficient as a function of wave incident angle and B cos θ/L;
the lines of B/L = 0.5, 1.0, and 1.5 are also included.
weight is concentrated on the rear side of the caisson. Therefore, difficulties are
sometimes met in the design of a perforated-wall caisson to satisfy the design criteria
against sliding and overturning. In addition, particularly in the case where the
bearing capacity of the seabed is not large enough, the excessive weight on the rear
side of the caisson may have an adverse effect. In order to solve these problems,
a partially-perforated-wall caisson as shown in Fig. 13.6(b) is often used, which
provides an additional weight to the front side of the caisson. In this case, however,
other hydraulic characteristics of the caisson such as wave reflection and overtopping
may become worse than a fully-perforated-wall caisson.
The mathematical model described in Sec. 13.3.1 assumes that the water depth
inside the wave chamber is the same as that on the mound berm as in a fully-
perforated-wall caisson breakwater shown in Fig. 13.6(a). However, for a partially-
perforated-wall caisson [see Fig. 13.6(b)], these water depths are different from each
other, having depth discontinuity at the location of the perforated wall. In order
to apply the model to the case of a partially-perforated-wall caisson, it can be
assumed that the lower part of the front face of the caisson (below the perforated
wall) is not vertical but has a very steep slope. As can be seen in Eq. (13.58), the
model includes the terms proportional to the square of the bottom slope and to the
bottom curvature, so that it can be applied over a bed having substantial variation
of water depth. In order to examine the effect of the slope of the lower part of the
caisson (which is infinity in reality), Suh et al.24 calculated the reflection coefficient
by changing the slope from 0.1 to 10. They found that the reflection coefficient
July 31, 2009 8:18 9.75in x 6.5in b684-ch13 FA
(a)
(b)
Fig. 13.6. Bird’s-eye views of (a) a fully-perforated-wall caisson breakwater and (b) a partially-
perforated-wall caisson breakwater.
virtually does not change for slopes greater than 2.0, and they used the slope of 4.0
in their comparison between model prediction and experimental results.
Equations (13.6) and (13.23) state that the fluid velocity normal to the perfo-
rated wall is proportional to the pressure difference across the wall with a complex
July 31, 2009 8:18 9.75in x 6.5in b684-ch13 FA
= 2C . (13.68)
For rectangular piles, Flagg and Newman35 proposed the blockage coefficient as
b 1 2A 1 2 281 4
C= −1 + 1 − log(4r0 ) + r0 + r . (13.69)
2 r0 π 3 180 0
July 31, 2009 8:18 9.75in x 6.5in b684-ch13 FA
where Cc = empirical contraction coefficient, for which Mei et al.31 suggested using
the formula:
and r0 cos θ denotes the effective ratio of the opening of the perforated wall taking
into account the oblique incidence of the waves to the wall. Very recently, on the
other hand, Yoon et al.37 proposed the head-loss coefficient taking into account the
effect of wall thickness as
2
−2.5(b/2a) 1
α = (0.6 + 0.4e ) −1 , (13.72)
r0 Cc
for normally incident waves.
For circular piles, Kakuno and Oda38 proposed the blockage coefficient as
π 1 π2
C= A(1 − r0 )2 ; ξ= (1 − r0 )2 . (13.73)
4 1−ξ 12
Park et al.11 proposed a formula for the head-loss coefficient for circular piles as
2
1
α= −1 , (13.74)
rCc
where Cc is the contraction coefficient given by Eq. (13.71), and the ad hoc porosity
r is given by
D/2
1 1 dx
= , (13.75)
r2 D −D/2 r(x)2
The governing equation (13.56) and the upwave boundary condition in Eq. (13.62)
do not change. After solving this problem, the transmission coefficient Ct is given by
Ct = Re{ϕ̃(0)}, from which Hw is calculated as Ct times the incident wave height
on the flat bottom.
The above method to compute the permeability parameter is advantageous com-
pared with other methods in that all the related parameters are known, i.e., incident
wave height and period and the geometrical parameters of the barrier. However,
Suh et al.12 showed that this method gives a wrong result for very long waves. As
can be seen in Eq. (13.66), as kh goes to zero, β goes to infinity, indicating complete
dissipation of the long waves. This leads to zero transmission for long waves, which
is definitely wrong because in the limit of long waves the breakwater is invisible to
the waves so that complete transmission of the waves should occur.
This method can be used for irregular waves as well. Using the regular wave
model, the reflection and transmission coefficients are calculated differently for each
frequency component. In the computation of the energy dissipation coefficient β in
Eq. (13.66), however, the rms wave height should be used instead of the height of
the component wave, because the flow separation due to irregular waves and the
resulting energy dissipation are induced not by the individual component waves but
by the superposition of the component waves.
1 − r0
s = 1 + Cm , (13.79)
r0
Acknowledgments
This chapter was prepared while the author was supported by the Korea Sea Grant
Program and the Project for Development of Reliability-Based Design Method for
Port and Harbor Structures of Korea Ministry of Marine Affairs and Fisheries.
References
1. R. L. Wiegel, J. Waterways Harbors Coastal Eng. Div., Am. Soc. Civ. Eng. 86, 1
(1960).
2. D. L. Kriebel and C. A. Bollmann, Proc. 25th Int. Conf. Coastal Eng., ASCE (1997),
p. 2470.
3. D. L. Kriebel, C. Sollitt and W. Gerken, Proc. 26th Int. Conf. Coastal Eng., ASCE
(1999), p. 2069.
4. M. Isaacson, S. Premasiri and G. Yang, J. Waterways, Port, Coastal, Ocean Eng. 124,
118 (1998).
5. T. Hayashi, M. Hattori, T. Kano and M. Shirai, Proc. 10th Int. Conf. Coastal Eng.,
ASCE (1966), p. 873.
6. H. Kojima, M. Utsunomiya, T. Ijima, A. Yoshida and T. Kihara, Proc. 35th Jpn Conf.
Coastal Eng., ASCE (1988), p. 542.
July 31, 2009 8:18 9.75in x 6.5in b684-ch13 FA
7. T. Uda, A. Omata and T. Kawamura, Rep. Public Works Res. Inst., 2891, Japan
Ministry of Construction (1990) (in Japanese).
8. S. Kakuno and P. L.-F. Liu, J. Waterways, Port, Coastal, Ocean Eng. 119, 302 (1993).
9. K. Hagiwara, Proc. 19th Int. Conf. Coastal Eng., ASCE (1985), p. 2810.
10. G. S. Bennett, P. McIver and J. V. Smallman, Coastal Eng. 18, 231 (1992).
11. W. S. Park, B. H. Kim, K. D. Suh and K. S. Lee, Coastal Eng. J. 42, 253 (2000).
12. K.-D. Suh, S. Shin and D. T. Cox, J. Waterways, Port, Coastal, Ocean Eng. 132, 83
(2006).
13. K.-D. Suh and C.-H. Ji, Proc. 30th Int. Conf. Coastal Eng., ASCE (2006), p. 4303.
14. S. Takahashi and K. Shimosako, Proc. Hydro-Port ’94, Port and Harbour Res. Inst.,
Yokosuka, Japan (1994), p. 747.
15. S. Takahashi, K. Tanimoto and K. Shimosako, Proc. Hydro-Port ’94, Port and Harbour
Res. Inst., Yokosuka, Japan (1994), p. 489.
16. G. E. Jarlan, The Dock and Harbour Authority, Vol. XII(486) (1961), p. 394.
17. M. Marks and G. E. Jarlan, Proc. 11th Int. Conf. Coastal Eng., ASCE (1968), p. 1121.
18. F. L. Terret, J. D. C. Osorio and G. H. Lean, Proc. 11th Int. Conf. Coastal Eng.,
ASCE (1968), p. 1104.
19. K. Tanimoto, S. Haranaka, S. Takahashi, K. Komatsu, M. Todoroki and M. Osato,
Technical Report, Port and Harbour Res. Inst., Min. of Transport, Japan, No. 246
(1976) (in Japanese).
20. H. Kondo, Proc. Coastal Structures ’79, ASCE (1979), p. 962.
21. S. Kakuno, K. Oda and P. L.-F. Liu, Proc. 23rd Int. Conf. Coastal Eng., ASCE (1992),
p. 1258.
22. M. Fugazza and L. Natale, J. Waterways, Port, Coastal, Ocean Eng. 118, 1 (1992).
23. K. D. Suh and W. S. Park, Coastal Eng. 26, 117 (1995).
24. K.-D. Suh, J. K. Park and W. S. Park, Ocean Eng. 33, 264 (2006).
25. D. L. Kriebel, Proc. Coastal Structures ’99, Balkema, Rotterdam (2000), p. 525.
26. R. G. Dean and R. A. Dalrymple, Water Wave Mechanics for Engineers and Scientists
(World Scientific, River Edge, New Jersey, 1991).
27. Y. Goda, Random Seas and Design of Maritime Structures (World Scientific,
Singapore, 2000).
28. S. R. Massel, Coastal Eng. 19, 97 (1993).
29. W. H. Press, S. A. Teukolsky, W. T. Vetterling and B. P. Flannery, Numerical Recipes
in FORTRAN: The Art of Scientific Computing, 2nd edn. (Cambridge University
Press, Cambridge, 1992).
30. K. D. Suh, W. S. Park and K. D. Yum, Proc. Int. Conf. Coastal and Port Eng.
Developing Countries (1995), p. 2311.
31. C. C. Mei, P. L.-F. Liu and A. T. Ippen, J. Waterways, Harbors Coastal Eng. Div.,
Vol. WW3, ASCE (1974), p. 217.
32. C. K. Sollitt and R. H. Cross, Proc. 13th Int. Conf. Coastal Eng., ASCE (1972),
p. 1827.
33. B. H. Kim, Interactions of waves, seabed and structures, PhD. dissertation, Seoul
National University, Seoul, Korea (1998).
34. K. D. Suh, S. Y. Son, J. I. Lee and T. H. Lee, Proc. 28th Int. Conf. Coastal Eng.,
ASCE (2002), p. 1709.
35. C. N. Flagg and J. N. Newman, J. Ship Res. 15, 257 (1971).
36. C. C. Mei, The Applied Dynamics of Ocean Surface Waves (Wiley, New York, 1983).
37. S. B. Yoon, J. I. Lee, D. H. Nam and S. H. Kim, J. Korean Soc. Coastal Ocean Engrs.
18, 321 (2006) (in Korean).
38. S. Kakuno and K. Oda, J. Jpn. Soc. Civil Engrs. 369, 213 (1986) (in Japanese).
July 31, 2009 8:18 9.75in x 6.5in b684-ch13 FA
39. I. J. Losada, M. A. Losada and A. Baquerize, Appl. Ocean Res. 15, 207 (1993).
40. X. Yu, J. Waterway, Port, Coastal, Ocean Eng. 121, 275 (1995).
41. M. Isaacson, J. Baldwin, S. Premasiri and G. Yang, Appl. Ocean Res. 21, 81 (1999).
42. S. Zhu and A. T. Chwang, J. Eng. Mech. 127, 326 (2001).
43. A. Hossain, W. Kioka and T. Kitano, Coastal Eng. J. 43, 83 (2001).
44. Y. Li, Y. Liu and B. Teng, Coastal Eng. J. 48, 309 (2006).
July 31, 2009 8:18 9.75in x 6.5in b684-ch13 FA
Chapter 14
Prediction of Overtopping
Tim Pullen
HR Wallingford
Howbery Park, Wallingford, Oxon, OX10 8BA, UK
tap@hrwallingford.co.uk
William Allsop
HR Wallingford
Howbery Park, Wallingford, Oxon, OX10 8BA, UK
w.allsop@hrwallingford.co.uk
Tom Bruce
School of Engineering, University of Edinburgh
King’s Buildings, Edinburgh, EH9 3JL, UK
tom.bruce@ed.ac.uk
Holger Schüttrumpf
RWTH-Aachen University
Institute of Hydraulic Engineering and Water Resources Management
Mies-van-der-Rohe-Str. 1, 52065 Aachen, Germany
schuettrumpf@iww.rwth-aachen.de
Andreas Kortenhaus
Leichtweiss-Institute for Hydraulics, Technical University of Braunschweig
Beethovenstr, 51a, 38106 Braunschweig, Germany
kortenhaus@tu-bs.de
341
August 13, 2009 14:43 9.75in x 6.5in b684-ch14 FA
This chapter describes the processes of wave overtopping at sea defense and
related coastal or shoreline structures. It introduces a range of methods to
calculate mean overtopping discharges, individual and maximum overtopping
volumes, and the proportion of waves overtopping a seawall. It describes the
principal hazards from wave overtopping and will help engineers by suggesting
limiting tolerable discharges for frequent, design, and extreme wave conditions.
This chapter is supported by more detailed material in Chaps. 15 and 16 which
focus on the methods to predict overtopping for rubble mound structures (with
partly sloping embankments), and on vertical structures and battered walls. All
of these three chapters have been based closely on the new EurOtop Overtopping
Manual.5
14.1. Introduction
formed from relict sand dunes or shingle ridges. Rubble mound structures tend to
be more common in areas where harder rock is available.
Along urban frontages, especially close to ports, erosion or flooding defense
structures may include vertical (or battered/steep) walls (Fig. 14.3). Such walls
August 13, 2009 14:43 9.75in x 6.5in b684-ch14 FA
may be composed of stone or concrete blocks, mass concrete, or sheet steel piles.
Typical vertical seawall structures may also act as retaining walls to material behind.
Shaped and recurved wave return walls may be formed as walls in their own right,
or smaller versions may be included in sloping structures. Some coastal structures
are relatively impermeable to wave action. These include seawalls formed from
blockwork or mass concrete, with vertical, near-vertical, or steeply sloping faces.
Such structures may be liable to intense local wave impact pressures, may overtop
suddenly and severely, and will reflect much of the incident wave energy. Reflected
waves cause additional wave disturbance and/or may initiate or accelerate local bed
scour.
It is worth noting that developments along waterfronts are highly valued with
purchase or rental prices substantially above those for properties not on the water-
front. Yet, direct (or indirect) effects of wave overtopping have the potential to
generate significant hazards to such developments and their users. Residential and
commercial properties along a waterfront will often be used by people who may
be unaware of the possibility, of the severity, or of the effects of wave overtopping
in storm conditions. Regulatory authorities may therefore wish to impose onerous
flood defense requirements on new developments. For instance, protection against
flooding (including wave overtopping) for any new developments in the United
Kingdom is now required to be 0.5% annual probability, equivalent to 1:200 year
return. Exposure to overtopping of many coastal sites will, however, be influenced
by climate change, probably increasing wave heights and periods as well as sea
level rise.
August 13, 2009 14:43 9.75in x 6.5in b684-ch14 FA
enjoying themselves, designers and owners of these defenses must, however, also
deal with potential direct hazards from overtopping. This requires that the level of
hazard and its probability of occurrence be assessed, allowing appropriate action
plans to be devised to ameliorate risks arising from overtopping. Section 14.6 deals
with tolerable wave overtopping.
into how wave overtopping behaves for different structures. Those structures con-
sidered here are: smooth sloping structures (dikes, seawalls); rubble mound struc-
tures (breakwaters, rock armored slopes); and vertical structures (caissons, sheet
pile walls).
The principal prediction formula for many types of wave overtopping is
q
= a exp(−bRc /Hm0 ). (14.1)
3
gHm0
It is an exponential function with the dimensionless overtopping discharge
3
q/(gHm0 )1/2 and the relative crest freeboard Rc /Hm0 . This type of equation shows
a straight line on a log-linear graph, which makes it easy to compare formulae for
different structures. Specific equations are given in Chaps. 15 and 16.
Two equations are considered for pulsating waves on a vertical structure. Allsop
et al.1 consider relatively shallow water and Franco et al.6 more deep water
(caissons). Vertical structures in shallow water, and often with a sloping foreshore in
front, may become subject to impulsive forces, i.e., high impacts and water splashing
high up into the air. Specific formulae have been developed for these kinds of
situations.
For easy comparison of different structures, like smooth and rubble mound
sloping structures and vertical structures for pulsating and impulsive waves, some
simplifications will be assumed.
In order to simplify the smooth structure, no berm is considered (γb = 1), only a
normal wave attack is considered (γβ = 1), and the sloping seawall does not feature
any wavewall on top (γv = 1). As the slope is smooth and impermeable, γf = 1.
This limits the structure to a smooth and straight slope with normal wave attack.
The slope angles considered for smooth slopes are cotα = 1−8, which means from
very steep to very gentle. If relevant, a wave steepness of sm−1,0 = 0.04 (steep storm
waves) and 0.01 (long waves due to swell or wave breaking) will be considered.
Detailed definitions are given in Chap. 15.
The same equation as for smooth slopes is applicable for rubble slopes, but now
with a roughness factor of γf = 0.5, simulating a rock armoured structure. Rubble
mound structures are often steep, but rock armoured slopes may also be gentle.
Therefore, slope angles with cotα = 1.5 and 4.0 are considered.
For vertical structures under pulsating waves, both formulae of Allsop et al.1
and Franco et al.6 will be compared, together with the formula for impulsive waves.
Impulsive waves can only be reached with a relatively steep foreshore in front of
the vertical wall. For comparison, values of the breaker ratio (wave height/water
depth) of Hm0 /hs = 0.5, 0.7, and 0.9 will be used. These will be discussed further
in Chap. 16.
Overtopping on smooth slopes can be compared with rubble mound slopes and
with vertical structures under pulsating or impulsive conditions. First, the tradi-
tional graph is given in Fig. 14.4 with the relative freeboard Rc /Hm0 versus the
3 1/2
logarithmic dimensionless overtopping q/(gHm0 ) .
In most cases the steep smooth slope gives the largest overtopping. Steep means
cotα < 2, but also a little gentler if long waves (less steepness) are considered. Under
these conditions, waves surge up the steep slope. For gentler slopes, waves break as
August 13, 2009 14:43 9.75in x 6.5in b684-ch14 FA
1.E+00
steep smooth slopes, cota<2
3 0.5
1.E-03
1.E-04
1.E-05
1.E-06
0 0.5 1 1.5 2 2.5 3 3.5 4
Relative freeboard Rc/Hm0
Fig. 14.4. Comparison of wave overtopping formulae for various structure types.
plunging waves and this reduces wave overtopping. The gentle slope with cotα = 4
gives much lower overtopping than the steep smooth slopes. Both slope angle and
wave period have influence on overtopping for gentle slopes.
The high roughness and permeability of a rubble mound can reduce over-
topping substantially (Fig. 14.4). A roughness factor of γf = 0.5 was used here
although γf = 0.4 (two layers of rock on a permeable under layer) would reduce
the overtopping further. A gentle rubble mound slope with cotα = 4 gives very low
overtopping.
Vertical structures under pulsating waves1,6 give lower overtopping than steep
smooth slopes, but more than a rough rubble mound slope. The impulsive conditions
give a different trend. First of all, the influence of the relative water depth is fairly
small as all curves with different Hm0 /hs are quite close. For low vertical structures
(Rc /Hm0 < 1.5), there is hardly any difference between pulsating and impulsive
conditions. The large difference is present for higher vertical structures and certainly
for the very high structures. With impulsive conditions, water can be thrown high
into the air, which means that overtopping occurs even for very high structures.
The vertical distance that the discharge travels is more or less independent of the
actual height of the structure. For Rc /Hm0 > 3 the curves are almost horizontal.
Another way of comparing the effectiveness of structure types is to show the
influence of slope angle on wave overtopping, as in Fig. 14.5. A vertical structure
means cotα = 0. Battered walls have 0 < cotα < 1. Steep slopes are generally
described by 1 ≤ cotα ≤ 3. Gentle slopes have roughly cotα ≥ 2 or 3. Overtopping
August 13, 2009 14:43 9.75in x 6.5in b684-ch14 FA
1.E+00
vertical, Allsop (1995)
vertical, Franco et al. (1994)
Dimensionless overtopping q/(gHm03)0.5
1.E-01
impulsive vertical, Hm0/hs=0.9, so=0.04
smooth slope, so=0.04
1.E-02 smooth slope, so=0.01
Rc/Hm0=1.5
1.E-03
Rc/Hm0=3.0
1.E-04
1.E-05
1.E-06
1.E-07
0 1 2 3 4 5 6 7 8
Slope angle cot α
prediction curves for two relative freeboards: Rc /Hm0 = 1.5 and 3.0 are shown in
Fig. 14.5.
Of course, similar conclusions can be drawn as for the previous comparison.
Steep slopes give the largest overtopping, which reduces for gentler slopes, for a
given wave condition and water level. Vertical slopes give less overtopping than steep
smooth slopes, except for a high vertical structure under impulsive conditions. This
graph gives also the method to calculate for a battered wall: interpolate between a
vertical wall and a slope 1:1 with cotα as the parameter to interpolate.
Details of all equations used here are described in more detail in Chaps. 15 and
16 (sloping smooth structures, rubble mound structures, and vertical structures).
16
b=3
overtopping volume (m /m)
14 b=2
3
b=1
12 b=0.75
b=0.65
10 b=0.85
4
2
0
100 90 70 50 30 20 10 5 2 1 0.5 0.1
probability of exceedance (%)
Fig. 14.6. Various distributions on a Rayleigh scale graph. A straight line (b = 2) is a Rayleigh
distribution.
exceedance and has been plotted according to the Rayleigh distribution. It is known
that wave heights in deep water generally conform to a Rayleigh distribution; so,
responses governed by deep water wave conditions will plot on or close to a straight
line, whilst shallow water effects will show deviations from the Rayleigh distribution.
When waves enter shallow water and the highest waves break, wave heights more
closely match a Weibull distribution with b > 2. An example with b = 3 is shown
in Fig. 14.6, and this indicates that there are more large waves of similar height.
The exponential distribution (often found for extreme wave climates) has b = 1,
and shows that extremes become larger compared to most of the data. Such an
exponential distribution would give a straight line in a log-linear graph.
The distribution of overtopping volumes for all kinds of structures has average
values even smaller than b = 1. Such a distribution is even steeper than an expo-
nential distribution. It means that the wave overtopping process can be described
by many fairly small overtopping volumes and a few very large volumes. The EA-
manual 2 gives values for b and a, based on limited data sets. The b-values are mostly
within the range 0.6 < b < 0.9. For comparison, curves with b = 0.65 and b = 0.85
are given in Fig. 14.6. The curves are very similar, except that the extremes differ
a little. It is for this reason that an average value of b = 0.75 was chosen for smooth
slopes and not different values for various subsets. The same average value has been
used for rubble mound structures, which makes smooth and rubble mound struc-
tures easily comparable. The exceedance probability, PV , of an overtopping volume
per wave is then similar to
0.75
V
PV = P (V ≤ V ) = 1 − exp − , (14.2)
a
with
q
a = 0.84 · Tm · = 0.84 · Tm · q · Nw /Now = 0.84 · q · t/Now . (14.3)
Pov
August 13, 2009 14:43 9.75in x 6.5in b684-ch14 FA
Equation (14.3) shows that the scale parameter a depends not only on the over-
topping discharge, q, but also on the mean period, Tm , and the probability of over-
topping, Now /Nw , or (similarly) on storm duration, t, and the actual number of
overtopping waves Nw .
The maximum overtopping during a certain event is fairly uncertain as most
maxima, but depends on the duration of the event. In a 6-h period, individual peak
volume is likely to be larger than that in 15 min. The maximum overtopping volume
by a single wave during an event depends on the actual number of overtopping
waves, Now , and can be calculated by
A comparison can be made between the mean overtopping discharge, q, and the
maximum overtopping volume in the largest wave. Note that the mean overtopping
is given in l/s per m width and that the maximum overtopping volume is given in
liters per m width.
Again, three example structures are considered: a smooth 1:4 slope; a rubble
slope at 1:1.5; and a vertical wall, and again specific equations are given in Chaps. 15
and 16. The storm duration has been assumed as 2 h (the peak of the tide), and
a fixed wave steepness of sm−1,0 = 0.04 has been considered. Figure 14.7 gives the
q − Vmax lines for the three structures and for relatively small waves of Hm0 = 1 m
and for fairly large waves of Hm0 = 2.5 m.
100000
Smooth slope Hs=2.5m
Maximum volume in overtopping wave (l per m)
1000
100
0.1 1 10 100 1000
Fig. 14.7. Relationship between mean discharge and maximum overtopping volume in one wave
for smooth, rubble mound, and vertical structures for wave heights of 1 m and 2.5 m.
August 13, 2009 14:43 9.75in x 6.5in b684-ch14 FA
A few conclusions can be drawn from Fig. 14.7. First of all, the ratio q/Vmax
is about 1000 for small q (roughly around 1 l/s per m), and about 100 for large q
(roughly around 100 l/s per m). So, the maximum volume (in liters per m width) in
the largest wave is about 100–1000 times larger than the mean overtopping discharge
(in l/s per m).
Secondly, the lines for a 1 m wave height are lower than for a larger 2.5 m wave
height, which means that for lower wave heights, but similar mean discharge, q, the
maximum overtopping volume is also smaller. For example, a vertical structure with
a mean discharge of 10 l/s per m gives a maximum volume of 1000 l/m for a 1 m
wave height and a volume of 4000 l/m for a 2.5 m wave height.
Finally, the three different structures give different relationships, depending on
the equations to calculate q and the equations to calculate the number of over-
topping waves.
with a minimum Kt = 0.075 and maximum Kt = 0.8; and limitations 1 < ξop < 3;
0◦ ≤ β ≤ 70, and 1 < B/Hm0,i < 4, where β is the angle of wave attack and B is
the crest width (and not berm width).
The transmission coefficient Kt is shown in Fig. 14.8 as a function of the relative
freeboard Rc /Hm0 and for a smooth structure with slope angle cotα = 4 (a gentle
smooth low-crested structure). Three wave steepnesses have been used: sop = 0.01
(long waves), 0.03 and 0.05 (short wind waves). Again, normal wave attack has been
assumed. Wave transmission decreases for increasing crest height, but longer waves
give more transmission. Wave overtopping can be calculated for the same structure
and wave conditions (see Chap. 15 and Fig. 14.9).
Wave overtopping and transmission can be related if Figs. 14.8 and 14.9 are
combined, and Fig. 14.10 shows this relationship. For convenience, the graphs are
not dimensionless, but for a wave height of Hm0,i = 3 m.
A small transmitted wave height of Hm0,t = 0.1 m is found if overtopping exceeds
q = 30−50 l/s per m. In order to reach a transmitted wave height of Hm0,t = 1 m
(Kt ≈ 0.33), the overtopping discharge should exceed q = 500−2500 l/s per m or
August 13, 2009 14:43 9.75in x 6.5in b684-ch14 FA
0.6
so=0.01
0.5
so=0.03
0.4 so=0.05
Kt
0.3
0.2
0.1
0.0
0 0.5 1 1.5 2
Rc /Hm0
Fig. 14.8. Wave transmission for a gentle smooth structure of 1:4 and for different wave steepneses.
1.000
so=0.01
so=0.03
0.100 so=0.05
3 0.5
q/(gHm0 )
0.010
0.001
0 0.5 1 1.5 2
Rc /Hm0
Fig. 14.9. Wave overtopping for a gentle smooth structure of 1:4 and for different wave
steepnesses.
0.5–2.5 m3 /s per m. One may conclude that any significant wave transmission is
always associated with (very) large wave overtopping.
Wave transmission for rubble mound structures has also been investigated in
the European DELOS project,4 and the following prediction formulae were derived
for wave transmission:
3500
so=0.01
3000
so=0.03
2500 so=0.05
q (l/s per m)
2000
1500
1000
500
0
0 0.5 1 1.5 2
Hm0 transmitted (m)
Fig. 14.10. Wave transmission versus wave overtopping for a smooth 1:4 slope and a wave height
of Hm0 = 3 m.
3500
3000 so=0.01
so=0.03
2500
so=0.05
q (l/s per m)
2000
1500
1000
500
0
0 0.5 1 1.5 2
Hm0 transmitted (m)
Fig. 14.11. Wave transmission versus wave overtopping discharge for a rubble mound structure,
cotα = 1.5, 6–10 ton rock, B = 4.5 m, and Hm0 = 3 m.
example, transmitted wave heights between Hm0,t = 0.5 m and 1 m are found for
overtopping discharges smaller than q = 100−200 l/s per m.
A simple equation for wave transmission at vertical structures has been given
by Goda,7 although again more complete methods are given in Chap. 16:
For vertical walls, transmission is primarily governed by the relative crest height
with little or no effect of wave period or steepness. A simple vertical structure has
been used as an example with Hm0,i = 3 m. Overtopping and wave transmission are
compared in Fig. 14.12, where in the calculations the crest height has been changed
to calculate wave transmission as well as wave overtopping.
For comparison, the same rubble mound structure has been used as the example
in Fig. 14.11, with cotα = 1.5, 6–10 ton rock (Dn50 = 1.5 m) as armour, a crest
width of 4.5 m (3Dn50 ), and a wave steepness sop = 0.03. The curve for a smooth
structure (Fig. 14.10) and for sop = 0.03 has been given too in Fig. 14.12.
A rubble mound structure gives more wave transmission than a smooth structure
for similar overtopping discharge, but a vertical structure can give even more trans-
mission. The reason may be that overtopping water over the crest of a vertical
breakwater always falls in a mass from a distance into the water, rather than flowing
relatively smoothly over or through a sloping structure.
One may conclude that even without considerable wave overtopping discharge at
the crest of a vertical structure, there still might be significant wave transmission.
In this example of a vertical structure, transmitted wave heights between 0.5 m and
1 m are found for overtopping discharges smaller than 100–200 l/s per m.
Finally, an example is shown of both wave overtopping and wave transmission
on a rubble mound breakwater in Fig. 14.13.
3500
vertical
3000 rubble mound (so=0.03)
2000
1500
1000
500
0
0 0.5 1 1.5 2
Hm0 transmitted (m)
Fig. 14.12. Comparison of wave overtopping and transmission for a vertical, rubble mound, and
smooth structure.
August 13, 2009 14:43 9.75in x 6.5in b684-ch14 FA
Fig. 14.13. Wave overtopping and transmission at breakwater IJmuiden, the Netherlands.
14.3. PC-OVERTOPPING
The computer program pc-overtopping was created using the results of the TAW8
Report “Wave runup and wave overtopping at dikes” and is used for the five-yearly
safety assessment of all water defenses in the Netherlands. The TAW8 Report has
now been replaced by Chap. 5 (dikes and embankments) in the EurOtop Over-
topping Manual5 and extended for rubble mound and vertical structures in Chaps. 6
and 7 of that manual. pc-overtopping has been translated into English, and is
available from the EurOtop Overtopping Manual5 web site.
The program was based on dike-type structures. The structure should be sloping,
although a small vertical wall on top of the dike may be included. Some effects of
roughness and/or permeability can be included, but not a crest with permeable and
rough rock or armour units. In such a case, the structure should be modeled up to
the transition to the crest, and other formulae should be used to take into account
the effect of the crest.
pc-overtopping was set up so that almost every sloping structure can be modeled
by an unlimited number of sections. Each section is given by x–y coordinates, and each
section can have its own roughness factor. The program calculates most relevant over-
topping parameters (except flow velocities and flow depths), such as:
• 2% runup level;
• mean overtopping discharge;
• percentage of overtopping waves;
• overtopping volumes per wave (maximum and for every percentage defined by
the user);
• required crest height for given mean overtopping discharges (defined by the user).
August 13, 2009 14:43 9.75in x 6.5in b684-ch14 FA
Fig. 14.15. Input of geometry by x–y coordinates and choice of top material.
August 13, 2009 14:43 9.75in x 6.5in b684-ch14 FA
The output is given in three columns (Fig. 14.17). The left column gives the
2% runup level, the mean overtopping discharge, and the percentage of overtopping
waves. If the 2% runup level is higher than the actual dike crest, this level is cal-
culated by extending the highest section in the cross section. The middle column
gives the required dike height for the given mean overtopping discharges. Again,
the highest section is extended if required. Finally, the right-hand column gives
the number of overtopping waves in the given storm duration, together with the
maximum overtopping volume, and volumes for specified overtopping percentages,
given as a percentage of the total number of overtopping waves.
pc-overtopping also provides a check on whether the results for the 2% runup
level and mean overtopping discharge fall within the measured ranges. Results on
which the formulae were based are shown in the runup or overtopping graphs (see
Figs. 14.18 and 14.19). These show the measured runup or overtopping, including
effects of reductions due to roughness, berms, etc. The curve gives the maximum
for smooth straight slopes with normal wave attack. The program then plots the
calculated point in these graphs (the point within the circle).
Artificial neural networks are tools that allow meaning to be extracted from very
large quantities of data. Neural networks (NN) are organized in the form of layers,
within which there are one or more processing elements called “neurons.” The first
layer is the input layer, and the number of neurons in this layer is equal to the
number of input parameters. The last layer is the output layer and the number of
neurons in this layer is equal to the number of output parameters to be predicted.
The layers between the input and output layers are the hidden layers, consisting of a
number of neurons to be defined in configuring the NN. Neurons in each layer receive
August 13, 2009 14:43 9.75in x 6.5in b684-ch14 FA
information from the preceding layer, carry out a standard operation, and produce
an output. Each connection has a weight factor assigned from the calibration of the
NN. The input of a neuron consists of a weighted sum of the outputs of the preceding
layer. This procedure is followed for each neuron; the output neuron generates the
final prediction of the NN.
Artificial NNs have applications in many fields including coastal engineering
where examples have been applied to predicting armour stability, forces on walls,
wave transmission, and wave overtopping. The development of an artificial NN is
useful, where:
INPUT
LAYER
HIDDEN
LAYER
OUTPUT
LAYER
q (m3/s/m)
Fig. 14.21. Overall view of possible structure configurations for the NN.
August 13, 2009 14:43 9.75in x 6.5in b684-ch14 FA
Gc = 5 m
Rc = 4 m
Ac = 5 m
Hs = 5 m; Tp = 10 s β = 0
Xbloc 1:1.5
h = 12 m ht = 9 m
Bt = 4 m
Fig. 14.22. Example cross section with parameters for application of NN.
Very often one is not only interested in one calculation, but in more. As the
input file has no limitations in the number of rows (= number of calculations), it
is easy to incrementally increase one or more parameters and to find a trend for a
certain (design) measure. As an example for the calculation of a trend with the NN
tool, an example cross section of a rubble mound embankment with a wave wall has
been chosen (Fig. 14.22).
If, for example, the cross section in Fig. 14.22 experiences too much overtopping,
the following measures could be considered:
An example input file with the first six calculations is shown in Table 14.1,
where incremental increase of the crest will show the effect of raising the crest on
the amount of wave overtopping. Calculations will give an output file with the mean
overtopping discharge q (m3 /s per m width) and with confidence limits. Table 14.2
shows as an example the output related to the input in Table 14.1.
Assembling the input file for this example took 1 h and resulted in 1400 rows or
calculations. The calculation of the NN took less than 10 s. The results were copied
into the Excel input file, and the resulting graph was plotted within Excel, which
took another hour. Figure 14.23 gives the final result, where the four trends are
Table 14.1. Example input file for NN with first six calculations.
q
(m3 /s per m) 2.50% 5.00% 25.00% 50.00% 75.00% 95.00% 97.50%
shown. The base situation had an overtopping discharge of 59 l/s per m. The graph
clearly shows what measures are required to reduce the overtopping by, for example,
a factor 10 (to 5.9 l/s per m) or to only 1 l/s per m. It also shows that increasing
the structure height is most effective, followed by increasing only the crest wall.
Within the CLASH research project, two different NNs exist. One is the official
NN developed by Delft Hydraulics in the main part of the CLASH project. It runs
as executable and can be downloaded from the CLASH web site or the EurOtop
Overtopping Manual web site.5
The other NN has also been developed within CLASH, but as part of a PhD
thesis at Gent University.9 The network was developed in MatLab R
and so it can
R
only be run if the user has a licence for MatLab . The advantage of this NN is that
it first decides whether there will be overtopping or not (classifier). If there is no
overtopping, it gives q = 0. If there is overtopping, it will quantify the overtopping
with a similar network as the CLASH network (quantifier). This use of both “qual-
ifier” and “quantifier” NNs is certainly an advantage over the single-stage CLASH
August 13, 2009 14:43 9.75in x 6.5in b684-ch14 FA
NN. The CLASH network was only trained with overtopping data (tests with “no
overtopping” were not considered) and, therefore, this network always gives a pre-
diction of overtopping, even in the range where no overtopping should be expected.
1.E+00
series 508
1.E-01 series 914
series 113
1.E-02 series 033
3 0.5
Franco et al.,(1994)
q/(gHm0 toe )
1.E-03
1.E-04
1.E-05
1.E-06
1.E-07
0 0.5 1 1.5 2 2.5 3 3.5 4
Rc /Hm0 toe
Fig. 14.24. Overtopping for large wave return walls — first selection.
1.E+00
series 914
1.E-01 series 113
Franco et al.,(1994)
1.E-02
3 0.5
q/(gHm0 toe )
1.E-03
1.E-04
1.E-05
1.E-06
1.E-07
0 0.5 1 1.5 2 2.5 3 3.5 4
Rc/H m 0 toe
Fig. 14.25. Overtopping for large wave return walls — second selection with more criteria.
wall, given by the dashed line. But for Rc /Hm0 toe > 1, there are quite some tests
without any overtopping.
As still quite some data are remaining in Fig. 14.25, it is possible to narrow the
search area even further. With a wave steepness of sm−1,0 = 0.04 in the considered
case, the wave steepness range can be limited to 0.03 < sm−1,0 < 0.05. The width
August 13, 2009 14:43 9.75in x 6.5in b684-ch14 FA
1.E+00
series 113
1.E-01
Franco et al.,(1994)
1.E-02
3 0.5
q/(gH m0 toe )
1.E-03
1.E-04
1.E-05
1.E-06
1.E-07
0 0.5 1 1.5 2 2.5 3 3.5 4
R c/Hm 0 toe
Fig. 14.26. Overtopping for wave return wall with so = 0.04, seaward angle of 45◦ , a width of
2 m and a crest height of Rc = 3 m. For Hm0 toe = 3 m, the overtopping can be estimated from
Rc /Hm0 toe = 1.
of the wave return wall of 2 m, with the wave height of Hm0 = 3 m, gives a
relative width of 0.67. The range can be limited to 0.5 < −cotαu∗ (Ac + hb )/Hm0 <
0.75. Finally, the transition from vertical to wave return wall is 1 m above the
design water level, giving hb /Hm0 toe = −0.33. The range can be set at −0.5 < hb /
Hm0 toe < −0.2.
The final selection obtained after filtering is given in Fig. 14.26. Only four tests
remain from one test series, one of which gave no measurable overtopping. The data
now give a clear picture. For a relative freeboard lower than about Rc /Hm0 toe = 0.7,
the overtopping will not be much different from the overtopping at a vertical wall.
The wave return wall, however, becomes very efficient for large freeboards and
effectively prevents any measurable overtopping for Rc /Hm0 toe > 1.2. For the
structure considered with Rc /Hm0 toe = 1, the wave overtopping will be 20–40 times
less than that for a vertical wall and will probably amount to q = 0.5 − 2 l/s per m
width. In this particular case, it was possible to find four tests in the database with
very close similarities to the considered structure and wave conditions.
of drainage ditches, damage versus inundation curves, and return period. Guidance
on modeling inundation flows is being developed within the Floodsite project (see:
http://www.floodsite.net/html/project overview.htm), but flood volumes, per se,
are not distinguished further in this chapter. Instead, advice here focuses on direct
hazards from wave overtopping.
For sea defenses that protect people living, working, or enjoying themselves,
designers and owners of these defenses must deal with potential direct hazards from
overtopping. This requires that the level of hazard and its probability of occurrence
be assessed, allowing appropriate action plans to be devised to ameliorate risks
arising from overtopping.
The main hazards on or close to sea defense structures are of death, injury,
property damage, or disruption from direct wave impact or by drowning. On
average, approximately 2–5 people are killed each year in each of the United
Kingdom and Italy through wave action, chiefly on seawalls and similar structures
(although this rose to 11 in the United Kingdom during 2005). It is often helpful
to analyze direct wave and overtopping effects, and their consequences under three
general categories:
The character of overtopping flows or jets, and the hazards they cause, also
depend upon the geometries of the structure, the hinterland behind the seawall,
and the form of overtopping. Rising ground behind the seawall may allow people
at potential risk to see incoming waves, and the slope will slow overtopping flows.
Conversely, a defense that is elevated significantly above the land defended will
obscure visibility of incoming waves (Fig. 14.27), and post-overtopping flows may
increase in speed rather than decreasing. Hazards caused by overtopping therefore
depend upon both the local topography and structures as well as on the direct
overtopping characteristics.
It is not possible to give unambiguous or precise limits to tolerable overtopping
for all conditions. Some guidance is, however, offered here on tolerable mean dis-
charges and maximum overtopping volumes for a range of circumstances or uses,
and on inundation flows and depths. These limits may be adopted or modified
depending upon the circumstances and uses of the site.
Fig. 14.27. Defended area below seawall and foreshore (saltmarsh) level.
• overtopping velocities over the crest, horizontally and vertically, vxc and vzc or
vxp and vzp ;
• overtopping flow depth, again measured on crest or promenade, dxc or dxp .
Less direct responses (or similar responses, but farther back from the defense)
may be used to assess the effects of overtopping, perhaps categorized by:
The main response to direct overtopping hazards has most commonly been the
construction of new defenses, but should now always consider three options, in
increasing order of intervention:
• move human activities away from the area subject to overtopping or flooding
hazard, thus modifying the land-use category and/or habitat status;
August 13, 2009 14:43 9.75in x 6.5in b684-ch14 FA
For any structure expected to reduce overtopping, the crest level and/or the
front-face configuration will be dimensioned to give acceptable levels of overtopping
under specified extreme conditions or combined conditions (e.g., water level and
waves). Setting acceptable levels of overtopping depends on:
Under most forms of wave attack, waves tend to break before or onto sloping
embankments with the overtopping process being relatively gentle (see Fig. 14.1).
Relatively few water levels and wave conditions may cause “impulsive” breaking
where the overtopping flows are sudden and violent. Conversely, steeper, vertical, or
compound structures are more likely to experience intense local impulsive breaking,
and may overtop violently and with greater velocities (see Fig. 14.3). The form of
breaking will therefore influence the distribution of overtopping volumes and their
velocities, both of which will have impact on the hazards that they cause.
Additional hazards that are not dealt with here are those that arise from wave
reflections, often associated with steep-faced defenses. Reflected waves increase
wave disturbance, which may cause hazards to navigating or moored vessels; may
increase waves along neighboring frontages, and/or may initiate or accelerate local
bed erosion thus increasing depth-limited wave heights.
intermittent) can cause sudden loss of visibility in turn leading drivers to veer
suddenly.
Effects of wind and generation of spray have not often been modeled. Some
research studies have suggested that effects of onshore winds on large green water
overtopping are small, but that overtopping under q = 1 l/s per m might increase
up to four times under strong winds, especially where much of the overtopping is as
spray. Discharges between q = 1 and 0.1 l/s per m are, however, already greater than
some discharge limits suggested for pedestrians or vehicles, suggesting that wind
effects may influence overtopping at and near acceptable limits for these hazards.
This is discussed further in Sec. 14.7.
The design life for flood defenses like dikes which are fairly easy to upgrade,
is taken in the Netherlands as 50 years. In urban areas, where it is more difficult
to upgrade a flood defense, the design life is taken as 100 years. This design life
increases for very special structures with high capital costs, like the Eastern Scheldt
storm surge barrier, Thames barrier, or the Maeslandtkering at the entrance to
Rotterdam. A design life of around 200 years is then usual.
Variations from simple “acceptable risk” approach may be required for publicly
funded defenses based on benefit — cost assessments, or where public aversion
to hazards causing death requires greater efforts to ameliorate the risk, either by
reducing the probability of the hazard or by reducing its consequence.
Mean Max
discharge volumea
Hazard type and reason q (l/s per m) Vmax (l/m)
a These limits relate to overtopping velocities well below vc ≤10 m/s. Lower
a These
limits probably relate to overtopping defined at highway.
b These
limits relate to overtopping defined at the defense, but assumes the highway to
be immediately behind the defense.
speed and with some suddenness. These lower limits are however based on few site
data or tests, and may therefore be relatively pessimistic.
Rather fewer data are available on the effects of overtopping on structures,
buildings, and property. Site-specific studies suggest that pressures on buildings by
overtopping flows will vary significantly with the form of wave overtopping, and
with the use of sea defense elements intended to disrupt overtopping momentum
(not necessarily reducing discharges). Guidance derived from the CLASH research
project and previous work suggests limits in Table 14.6 for damage to buildings,
equipment, or vessels behind defenses.
A set of limits for structures in Table 14.7 has been derived from early work
by Goda7 and others in Japan. These give a first indication of the need for specific
protection to resist heavy overtopping flows. It is assumed that any structure close
to the sea will already be detailed to resist the erosive power of heavy rainfall and/or
spray. Two situations are considered:
• embankment seawall or dike, elevated above the defended area, and so overtopping
flows pass over the crest and down the rear face;
• promenade defense in which overtopping flows remain on or behind the seawall
crest before returning seaward.
Table 14.6. Limits for overtopping for property behind the defense.
Table 14.7. Limits for overtopping for damage to the defense crest or rear slope.
Mean discharge
Hazard type and reason q (l/s per m)
The limits for the latter category cannot be applied where the overtopping flows
can fall from the defense crest where the nature of the flow may be more impulsive.
The limits in Table 14.7 are precautionary and are generally based on old data.
In order to clarify the erosion resistance of grass protection under wave over-
topping, tests were performed using the overtopping simulator (Figs. 14.29 and
14.30) in 2007 on a sea dike in the Netherlands. The grass dike had a 1:3 inner slope
of fairly good clay, sand content smaller than 30%. The overtopping simulator was
used to test the erosion resistance of this inner slope for a simulated 6 h storm for
each overtopping condition. These started with overtopping equivalent to a mean
discharge of 0.1 l/s per m and increased to 1, 10, 20, 30, and finally even 50 l/s
per m. After all these simulated storms, the slope was still in good condition and
showed little erosion. The erosion resistance of this dike was very high.
Another test was performed on bare clay by removing the grass sod over the full
inner slope to a depth of 0.2 m. Overtopping conditions of 0.1 l/s per m, 1, 5, and
finally, 10 l/s per m were performed, again for 6 h each. Erosion damage started
for the first condition (two erosion holes) and increased during the other overtopping
conditions. After 6 h at a mean discharge of 10 l/s per m (see Figs. 14.30 and
14.31), there were two large erosion holes, about 1 m deep, 1 m wide, and 4 m long.
This situation was considered as “not too far from initial breaching.” The overall
conclusion of this first overtopping test on a real dike is that clay with grass can
be highly erosion-resistant. Even without grass the good quality clay also survived
extensive overtopping. The conclusions may not yet be generalized to all dikes as
August 13, 2009 14:43 9.75in x 6.5in b684-ch14 FA
Fig. 14.30. Overtopping simulator discharging a large volume on the inner slope of a dike.
Fig. 14.31. Wave overtopping test on bare clay; result after 6 h with 10 l/s per m width.
clay quality and type of grass cover still may play a role and, therefore, more testing
is required to come to general conclusions.
One remark, however, should be made on the strength of the inner slopes of dikes
by wave overtopping. Direct erosion of the slope is one possible failure mechanism.
A major failure mechanism, especially in the past, was slip failure of the (rear)
slope. Slip failures may directly lead to a breach, and such slip failures often occur
August 13, 2009 14:43 9.75in x 6.5in b684-ch14 FA
mainly for steep inner slopes like 1:1.5 or 1:2. For this reason, most dike designs in
the Netherlands in the past 50 years have used a 1:3 inner slope, where it is unlikely
that slip failures will occur due to overtopping. This mechanism might, however,
occur for steep inner slopes, and so it should be taken into account in safety analysis.
strongly site-specific, these results suggest that quite low discharges (0.1–1.0 l/s
per m) may lead to loadings up to 5 kPa.
in Chaps. 15 and 16 for dikes, for rubble mound slopes, and for vertical walls,
respectively.
None of these methods give the universally “best” results, and indeed there may
still be a need for site-specific model tests for some defenses. The most reliable
method to be used will depend on the type and complexity of the structures, and
the closeness with which it conforms to simplifying assumptions used in the previous
model testing (on which all of the methods above are inherently based).
In selecting which method to use, or which set of results to prefer when using
more than one method, the user will need to take account of the origins of each
method. It may also be important in some circumstances to use an alternative
method to give a check on a particular set of calculations. To assist these judgments,
a set of simple rules of thumb are given here, but as ever, these should not be treated
as universal truths.
In almost all instances, the use of any of these methods will involve some degree
of simplification of the true situation. The further the structure or design (analysis)
conditions depart from the idealized configurations tested to generate the methods/
tools discussed, the wider will be the uncertainties. Where the importance of the
assets being defended is high, and/or the uncertainties in using these methods are
large, then the design solution may require use of site-specific physical model tests.
It is clear that increased attention to flood risk reduction, and to wave overtopping
in particular, have increased interest and research in this area. The EurOtop Over-
topping Manual5 is therefore not expected to be the “last word” on the subject;
indeed even whilst preparing the first version of the manual, the author team
expected that there will be later revisions. The reader of this handbook is therefore
advised to check whether an improved version of the EurOtop Overtopping Manual
has been released. Beyond that manual, it is probable that there will be signif-
icant improvements in numerical modeling, although it should be acknowledged that
improved numerical models will require substantial measurement data to validate
them before their results can be relied upon in detailed analysis or design.
Acknowledgments
This chapter, and also Chaps. 15 and 16, are based on the EurOtop Overtopping
Manual,5 which was funded in the the United Kingdom by the Environmental
Agency, in Germany by the German Coastal Engineering Research Council (KFKI),
in the Netherlands by Rijkswaterstaat and Netherlands Expertise Network (ENW)
on Flood Protection.
The Project Team for the creation, editing, and support of the manual was
formed by
Prof William Allsop (HRW)
Dr Tom Bruce (UoE)
Dr Andreas Kortenhaus (LWI)
Dr Jentsje van der Meer (Infram)
Dr Tim Pullen (HRW)
Dr Holger Schüttrumpf (BAW)
August 13, 2009 14:43 9.75in x 6.5in b684-ch14 FA
References
References in this chapter have been kept to a real minimum. An extensive list of
relevant references, however, can be found in the EurOtop Overtopping Manual.5
Chapter 15
Holger Schüttrumpf
Institute of Hydraulic Engineering and Water Resources Management
RWTH-Aachen University, 52065 Aachen, Germany
schuettrumpf@iww.rwth-aachen.de
Andreas Kortenhaus
Leichtweiss-Institute for Hydraulics
Technical University of Braunschweig
Beethovenstr 51a, 38106, Braunschweig, Germany
kortenhaus@tu-bs.de
Tom Bruce
School of Engineering, University of Edinburgh
King’s Buildings, Edinburgh, EH9 3JL, UK
tom.bruce@ed.ac.uk
Leopoldo Franco
Department of Civil Engineering, University of Roma Tre
Via V. Volterra 62, 00146 Roma, Italy
leopoldo.franco@uniroma3.it
383
August 14, 2009 9:48 9.75in x 6.5in b684-ch15 FA
Wave overtopping and to a lesser extent wave run-up for armored rubble slopes
and mounds have been subject to a number of investigations in the past. The
objective of the present chapter is to summarize existing information to be present
as a closed guidance on the use of wave run-up and wave overtopping formulae
for a wide range of possible applications in practice. Therefore, guidance is given
first on the use of wave run-up and wave overtopping formulae for simple slopes,
excluding the effects of composite slopes, direction of wave attack, roughness,
wave walls, etc. Then, formulae are presented to include these parameters in the
calculation procedure. Guidance is also given on wave overtopping volumes, over-
topping velocities, and the spatial distribution as well as for wave overtopping for
shingle beaches. Finally, the effect of model and scale effects on the calculation of
average overtopping rates are discussed. This chapter has mainly been composed
from Chap. 6 of the EurOtop Overtopping Manual (2007), with some additions
from Chap. 5. The present chapter is related to the previous Chap. 14 and the
next Chap. 16 of this manual.
15.1. Introduction
Armored rubble slopes and mounds are characterized by a mound with some
porosity or permeability, covered by a sloping porous armor layer consisting of large
rock or concrete units. In contrast to dikes and embankment seawalls, the porosity
of the structure and armor layer plays a role in wave run-up and overtopping. The
cross section of a rubble mound slope, however, may have great similarities with an
embankment seawall and may consist of various slopes.
As an example for armored slopes and mounds, a rock-armored embankment is
given in Fig. 15.1.
Wave Run-Up and Wave Overtopping at Armored Rubble Slopes and Mounds 385
RC
Ru2%
SWL Hm0
α
h
Fig. 15.2. Definition of the wave run-up height Ru2% on a smooth impermeable slope.
August 14, 2009 9:48 9.75in x 6.5in b684-ch15 FA
Ru2%
= 1.65 · γb · γf · γβ · ξm−1,0 (15.1)
Hm0
with a maximum of RHu2% = 1.00 · γb · γf · γβ 4.0 − √ 1.5 ,
m0 ξm−1,0
where
Ru2% =wave run-up height exceeded by 2% of the incoming waves [m];
Hm0 =significant wave at the toe of the structure [m];
γb =influence factor for a berm [see 15.3.4(b)] [–];
γf =influence factor for roughness on the slope [–];
γβ =influence factor for oblique wave attack (see
15.3.3) [–];
√ 2
ξm−1,0 = breaker parameter = tan α/ sm−1,0 = tan α 2πHm0 /(gTm−1,0 ) [–];
tanα = average slope angle (see Fig. 15.2) [–];
Tm−1,0 = spectral moment at the toe of the structure, based on m−1 and m0 [s].
The breaker or surf similarity parameter ξm−1,0 relates the slope steepness tan α
2
(or 1/n) to the fictitious wave steepness sm−1,0 = 2πHm0 /(gTm−1,0 ) and is often
used to distinguish different breaker types. The combination of structure slope and
wave steepness gives a certain type of wave breaking (Fig. 15.3).
For ξm−1,0 > 2–3 waves are considered not to be breaking (surging waves),
although there may still be some breaking, and for ξm−1,0 < 2–3 waves are breaking.
Waves on a gentle foreshore break as spilling waves and more than one breaker
line can be found on such a foreshore. Plunging waves break with steep and over-
hanging fronts and the wave tongue will hit the structure or back washing water.
The transition between plunging waves and surging waves is known as collapsing.
The wavefront becomes almost vertical, and the water excursion on the slope (wave
run-up + run-down) is often largest for this kind of breaking. Values are given for
the majority of the larger waves in a sea state. Individual waves may still surge for
generally plunging conditions or plunge for generally surging conditions.
The relative wave run-up height increases linearly with increasing ξm−1,0 in
the range of breaking waves and small breaker parameters less than about 2.
For nonbreaking waves and higher breaker parameters, the increase is less and
Wave Run-Up and Wave Overtopping at Armored Rubble Slopes and Mounds 387
becomes more or less horizontal. The relative wave run-up height Ru,2% /Hm0 is also
influenced by the geometry of the structure, the effect of wind, and the properties
of the incoming waves.
4.0
3.5
3.0
2.5
Ru2%/Hm0
2.0
1.5
0.0
0 1 2 3 4 5 6 7 8 9 10
Fig. 15.4. Relative run-up on straight rock slopes with permeable and impermeable core, com-
pared to smooth impermeable slopes.
August 14, 2009 9:48 9.75in x 6.5in b684-ch15 FA
The prediction for the 2% mean wave run-up value for rock or rough slopes can
be described by
Ru2%
= 1.65 · γb · γf · γβ · ξm−1,0 , (15.2)
Hm0
Ru2%
with a maximum of Hm0
= 1.00 · γb · γf surging · γβ 4.0 − √ 1.5
.
ξm−1,0
Ru2%
= 1.75 · γb · γf · γβ · ξm−1,0 , (15.3)
Hm0
Ru2%
with a maximum of Hm0
= 1.00 · γb · γf surging · γβ 4.3 − √ 1.6
.
ξm−1,0
Wave Run-Up and Wave Overtopping at Armored Rubble Slopes and Mounds 389
where
Pov = probability of overtopping [–];
Now = number of overtopping waves in a sea state [–];
Nw = number of waves in a sea state [–];
Rc = crest freeboard [m].
Equation (15.4) can be used to calculate the probability of overtopping, given a
crest freeboard Rc or to calculate the required crest freeboard, given an allowable
probability or percentage of overtopping waves.
One warning should be given in applying Eqs. (15.2)–(15.4). The equations give
the run-up level in percentage or height on a straight (rock-armored) slope. This
is not the same as the number of overtopping waves or overtopping percentage.
Figure 15.5 gives the difference. The run-up is always a point on a straight slope,
where for a rock-armored slope or mound the overtopping is measured some distance
away from the seaward slope and on the crest, often behind a crown wall. This means
that Eqs. (15.2)–(15.4) always give an overestimation of the number of overtopping
waves.
Figure 15.6 shows measured data for rubble mound breakwaters armored with
Tetrapods, AccropodeTM , or a single layer of cubes. All tests were performed at
Delft Hydraulics.
The test setup was more or less similar to Fig. 15.4 with a crown wall height
Rc a little lower than the armor freeboard Ac . CLASH-data on specific over-
topping tests for various rock and concrete armored slopes were added to Fig. 15.6.
This figure gives only the percentage of overtopping waves passing the crown wall.
August 14, 2009 9:48 9.75in x 6.5in b684-ch15 FA
overtopping
run-up level (eq. 2 and 3) measured
calculated here CREST behind wall
Gc
1.5 Hm0toe Ac Rc
swl
1.5 Hm0toe
100
Percentage of overtopping waves (%)
90
tetrapod DH
80 accropode DH
1 layer cube DH
70
CLASH
60
50
40
30
20
10
0
-0.2 0 0.2 0.4 0.6 0.8 1
2
Dimensionless crest height Ac*Dn/Hm0
Fig. 15.6. Percentage of overtopping waves for rubble mound breakwaters as a function of relative
(armor) crest height and armor size (Rc ≤ Ac ).
Analysis showed that the size of the armor unit relative to the wave height had
2
influence, which gave a combined parameter Ac · Dn /Hm0 , where Dn is the nominal
diameter of the armor unit.
The figure covers the whole range of overtopping percentages, from com-
plete overtopping with the crest at or lower than SWL to no overtopping at all.
The CLASH-data give maximum overtopping percentages of about 30%. Larger
August 14, 2009 9:48 9.75in x 6.5in b684-ch15 FA
Wave Run-Up and Wave Overtopping at Armored Rubble Slopes and Mounds 391
percentages mean that overtopping is so large that it can hardly be measured and
that wave transmission starts to play a role.
Taking 100% overtopping for zero freeboard (the actual data are only a little
lower), a Weibull curve can be fitted through the data. Equation (15.5) can be used
to predict the number or percentage of overtopping waves or to establish the armor
crest level for an allowable percentage of overtopping waves.
1.4
Ac Dn
Pov = Now /Nw = exp − 2 . (15.5)
0.19Hm0
It is clear that Eqs. (15.2)–(15.4) will come to more overtopping waves than
Eq. (15.5). But both estimations together give a designer enough information to
establish the required crest height of a structure given an allowable overtopping
percentage.
0.5
1.5
2
imp; cota=2 imp; cota=3 imp; cota=4
perm; cota=1.5 perm; cota=2 perm; cota=3
hom; cota=2 imp; Deltaflume perm; Deltaflume
2.5
Fig. 15.7. Relative 2% run-down on straight rock slopes with impermeable core (imp), permeable
core (perm), and homogeneous structure (hom).
August 14, 2009 9:48 9.75in x 6.5in b684-ch15 FA
The graph shows clearly the influence of the permeability of the structure as the
solid data points (impermeable core) generally show larger run-down than the open
data symbols of the permeable core. Furthermore, the breaker parameter ξm−1,0
gives a fairly clear trend of run-down for various slope angles and wave periods.
Figure 15.7 can be used directly for design purposes, as it also gives a good idea of
the scatter.
Wave Run-Up and Wave Overtopping at Armored Rubble Slopes and Mounds 393
The dimensionless overtopping discharge Q∗ = q/ gHm0 3
is a function of the
wave height, originally derived from the Weir formula.
q 0.067
=√ γb · ξm−1,0
3
g · Hm0 tan α
Rc
· exp −4.3 (15.7)
ξm−1,0 · Hm0 · γb · γf · γβ · γv
q
with a maximum of √ 3
= 0.2 · exp −2.3 Hm0R·γcf ·γβ ,
g·Hm0
where γv = the influence of a small wall on top of the embankment.
q 0.067
= √ γb · ξm−1,0
3 tan α
g · Hm0
Rc
· exp −4.75 (15.8)
ξm−1,0 · Hm0 · γb · γf · γβ · γv
q
with a maximum of √ 3
= 0.2 · exp −2.6 Hm0R·γcf ·γβ .
g·Hm0
The reliability of Eq. (15.8) is described by taking the coefficients 4.75 and
2.6 as normally distributed stochastic parameters with means of 4.75 and 2.6 and
standard deviations σ = 0.5 and 0.35, respectively. For probabilistic calculations,
Eq. (15.8) should be taken together with these stochastic coefficients. For predictions
of measurements or comparison with measurements also Eq. (15.8) should be taken
with, for instance, 5% upper and lower exceedance curves.
It has to be mentioned that the first part of Eqs. (15.7) and (15.8) is valid
for mostly breaking waves. Considering the steep slopes of armored rubble slope
and mounds this part has less importance in practice than the second part of the
equation, describing the maximum of overtopping. In that case, the relative free-
board does not depend on the breaker parameter ξm−1,0 for nonbreaking waves
(Fig. 15.8), as the line is horizontal.
This means that a composite slope and even a, not too long, berm leads to the
same overtopping discharge as for a simple straight rubble mound slope. Only when
the average slope becomes so gentle that the maximum part in Eqs. (15.7) and
(15.8) do not apply anymore, then a berm and a composite slope will have effect on
the overtopping discharge. Generally, average slopes around 1:2 or steeper do not
show influence of the slope angle, or only to a limited extent, and the maximum
part in Eqs. (15.7) and (15.8) are valid.
August 14, 2009 9:48 9.75in x 6.5in b684-ch15 FA
0,016
[-]
3 1/2
1:4 slope
0,012 1:6 slope
.
0,01
0,008
0,006
0,004
0,002
0
0 0,01 0,02 0,03 0,04 0,05 0,06
2
wave steepness 2πHm0/(gTm-1,0 ) [-]
Fig. 15.8. Wave overtopping as a function of the fictitious wave steepness sm−1,0 = 2πHm0 /
(gTm−1,0 2 ) and a smooth slope.
Wave Run-Up and Wave Overtopping at Armored Rubble Slopes and Mounds 395
1.E+00
Smooth Rock
Cube rough Cube flat
1.E-01 Anitfer Haro
3 0.5
Accropde Coreloc
Xbloc smooth gf=1.0
1.E-02 rough gf=0.45
1.E-03
Eq. 8 f = 1.0
1.E-04
1.E-05
Eq. 8 f = 0.45
1.E-06
1.E-07
0 0.5 1 1.5 2 2.5 3 3.5 4
Fig. 15.9. Mean overtopping discharge for 1:1.5 smooth and rubble mound slopes.
be applied:
Cr = 3.06 exp(−1.5Gc /Hm0 ), with maximum Cr = 1. (15.9)
Equation (15.9) gives no reduction for a crest width smaller than about
0.75 Hm0 . This is fairly close to about 3Dn50 and is, therefore, consistent. A crest
width of 1Hm0 reduces the overtopping discharge to 68%, a crest width of 2Hm0
gives a reduction to 15%, and for a wide crest of 3Hm0 , the overtopping reduces
to only 3.4%. In all cases, the crest wall has the same height as the armor crest:
Rc = Ac .
Equation (15.9) was determined for a rock-armored slope and can be considered
as conservative, as for a slope with Accropode, more reduction was found.
Wave Run-Up and Wave Overtopping at Armored Rubble Slopes and Mounds 397
of slope from the breaking point to the maximum wave run-up height. Therefore,
often it has been recommended to calculate the characteristic slope from the point
of wave breaking to the maximum wave run-up height. This approach needs some
calculation effort, because of the iterative solution since the wave run-up height
Ru2% is unknown. For the breaking limit, a point on the slope can be chosen which
is 1.5Hm0 below the still water line.
It is recommended to use also a point on the slope 1.5Hm0 above water as a first
estimate to calculate the characteristic slope and to exclude a berm (Fig. 15.10).
3 · Hm0
First estimate: tan α = . (15.11)
LSlope − B
As a second estimate, the wave run-up height from the first estimate is used to
calculate the average slope [LSlope has to be adapted (see Fig. 15.11)]:
1.5 · Hm0 + Ru2%(from 1st estimate)
Second estimate: tan α = . (15.12)
LSlope − B
1st estimate:
3H m0
tan α =
LSlope − B
B
Average slope
1.5 HS
SWL dB
1.5 HS B
LSlope
2nd estimate:
Ru2% (from 1st estimate)
1.5 H m 0 + Ru 2% ( from 1st estimate)
tan α =
LSlope − B
Average slope
SWL dB
1.5 HS B
LSlope
If the run-up height or 1.5Hm0 comes above the crest level, then the crest level
must be taken as the characteristic point above SWL.
Also, the influence of a berm in a sloping profile has been adapted from smooth
sloping sea dikes for rubble mound structures. There is, however, often a difference
in the effect of composite slopes or berms for rubble mound and smooth gentle
slopes. On gentle slopes, the breaker parameter ξm−1,0 has large influence on wave
overtopping (see Eqs. (15.2) and (15.3) as the breaker parameter will be quite
small). Rubble mound structures often have a steep slope, leading to the formula for
“nonbreaking” waves, the maximum in Eqs. (15.7) and (15.8). In these equations,
no slope angle or breaker parameter is present, and the effect of a small berm will
be very small and probably negligible.
Bhorizontal
SWL db
Bangled
1.0 H S
SWL 1.0 HS db
LBerm
Wave Run-Up and Wave Overtopping at Armored Rubble Slopes and Mounds 399
A berm reduces wave run-up or wave overtopping. The influence factor γb for a
berm consists of two parts.
The first part (rB ) stands for the width of the berm LBerm and becomes 0 if no
berm is present.
B
rB = . (15.14)
LBerm
The second part (rdb ) stands for the vertical difference dB between the SWL and
the middle of the berm and becomes 0 if the berm lies on the still water line. The
reduction of wave run-up or wave overtopping is maximum for a berm on the still
water line and decreases with increasing dB . Thus, a berm lying on the still water
line is most effective. A berm lying below 2 · Hm0 or above Ru2% has no influence
on wave run-up and wave overtopping.
Different expressions are used for rdB in Europe. Here, an expression using
a cosine-function for rdb (Fig. 15.13) is recommended which is also used in pc-
overtopping (see Chap. 14).
db
rdb = 0.5 − 0.5 cos π for a berm above SWL ,
Ru2%
db
rdb = 0.5 − 0.5 cos π for a berm below SWL , (15.15)
2 · Hm0
rdb = 1 for berms lying outside the area of influence .
1
0.9
0.8
0.7
0.6
rdb [-]
0.5
0.4
0.3
0.2
0.1
0
-1 -0.5 0 0.5 1
.
db/Ru2% [-] db/(2 Hm0) [-]
Wave Run-Up and Wave Overtopping at Armored Rubble Slopes and Mounds 401
Fig. 15.16. Nonreshaping Icelandic berm breakwater with various classes of big rocks.
reshaping berm breakwaters and γf = 0.35 for nonreshaping Icelandic berm break-
waters. The method of composite slopes and berms should be applied as described
above.
The second method is to use the CLASH neural network. As overtopping
research at that time on berm breakwaters was limited, this method also gives quite
some scatter, but a little less than the first method described above.
Recent information on berm breakwaters has been described by Andersen.5 Only
part of his research was included in the CLASH database and consequently in the
Neural Network prediction method. He performed about 600 tests on reshaping
berm breakwaters and some 60 on nonreshaping berm breakwaters (fixing the steep
slopes by a steel net). The true nonreshaping Icelandic type of berm breakwaters
with large rock classes, has not been tested and, therefore, his results might lead to
an overestimation.
One comment should be made on the application of the results of Andersen.5
3 0.5
The maximum overtopping discharge measured was only q/(gHm0 ) = 10−3. In
practical situations with wave heights around 5 m, the overtopping discharge will
then be limited to only a few l/s per m width. For berm breakwaters and also for
conventional rubble slopes and mounds, allowable overtopping may be much higher
than this value.
August 14, 2009 9:48 9.75in x 6.5in b684-ch15 FA
The final result of the work of Andersen5 is a quite complicated formula, based
on multi-parameter fitting. The advantage of such a fitting is that by using a large
number of parameters, the data set used will be quite well described by the formula.
The disadvantage is that physical understanding of the working of the formula,
certainly outside the ranges tested, is limited. But, due to the fact that so many
structures were tested, this effect may be neglected.
The formula is valid for berm breakwaters with no superstructure and gives the
overtopping discharge at the back of the crest (Ac = Rc ). In order to overcome
the problem that one encounters when calculating the reshaped profile before any
overtopping calculation can be done, the formula is based on the “as built” profile,
before reshaping. Instead of calculating the profile, a part of the formula predicts the
influence of waves on recession of the berm. The parameter used is called fHo , which
is an indicative measure of the reshaping, and can be defined as a “factor accounting
for the influence of stability numbers.” Note that fHo is a dimensionless factor
and not the direct measure of recession and that Ho and To are also dimensionless
parameters (see below).
fHo = 19.8s−0.5 ∗
om exp(−7.08/Ho ) for To ≥ To ,
(15.17)
fHo = 0.05Ho To + 10.5 for To < To∗ ,
where
Ho = Hm0 /∆Dn50 , To = (g/Dn50 )0.5 Tm0,1 ,
and
To∗ = {19.8s−0.5
om exp(−7.08/Ho) − 10.5}/(0.05Ho).
The berm level dh is also taken into account as an influence factor, d∗h . Note
that the berm depth is positive if the berm level is below SWL, and therefore, for
berm breakwaters often negative. Note also that this influence factor is different
from that for a bermed slope. This influence factor is described by
d∗h = (3Hm0 − dh )/(3Hm0 + Rc ) for dh < 3Hm0 ,
(15.18)
d∗h = 0 for dh ≥ 3Hm0 .
The final overtopping formula then takes into account the influence factor on
recession, fHo , the influence factor of the berm level, d∗h , the geometrical param-
eters Rc , B, and Gc , and the wave conditions Hm0 and the mean period Tm0,1 . It
means that the wave overtopping is described by a spectral mean period, and not
by Tm−1,0 .
3
q/(gHm0 )0.5 = 1.79 · 10−5 (fHo
1.34
+ 9.22)s−2.52
op ∗,
Equation (15.19) is only valid for a down slope of 1:1.25 and an upper slope of
1:1.25. For other slopes, one has to reshape the slope to a slope of 1:1.25, keeping
the volume of material the same and adjusting the berm width B and for the upper
slope also the crest width Gc . Note also that in Eq. (15.19), the peak wave period
August 14, 2009 9:48 9.75in x 6.5in b684-ch15 FA
Wave Run-Up and Wave Overtopping at Armored Rubble Slopes and Mounds 403
Tp has to be used to calculate sop , where the mean period Tm0,1 has to be used in
Eq. (15.17).
Although no tests were performed on the nonreshaping Icelandic berm break-
waters (see Fig. 15.16), a number of tests were performed on nonreshaping struc-
tures by keeping the material in place with a steel net. The difference may be that
Icelandic berm breakwaters show a little less overtopping, due to the presence of
larger rocks and, therefore, more permeability. The tests showed that Eq. (15.19) is
also valid for nonreshaping berm breakwaters, if the reshaping factor fHo = 0.
Zeebrugge (1:1.4)
Ostia (1:4)
30,0
Eq. (20), f_q,max = 7.9
Eq. (20), f_q,max = 30
20,0
fq [-]
10,0
0,0
1,E-06 1,E-05 1,E-04 1,E-03 1,E-02
3
qss [m /s/m]
Fig. 15.17. Proposed adjustment factor applied to data from two field sites (Zeebrugge 1:1.4
rubble mound breakwater, and Ostia 1:4 rubble slope).
lower than 1.0 l/s/m but may include significant adjustment factors below these
rates. Due to the inherent uncertainties, the proposed approach tries to be conser-
vative. It has, however, been applied to pilot cases in CLASH and has proved good
corrections with these model data.
The adjustment factor fq for the model and scale effects can be determined as
follows (Fig. 15.17):
1.0 for qss ≥ 10−3m3 /s/m ,
fq = (15.20)
min{(− log qss − 2)3 ; fq,max } for qss ≥ 10−3m3 /s/m ,
where fq,max is an upper bound to the adjustment factor fq and can be calculated
as follows:
fq,r
for γf ≤ 0.7 ,
fq,max = 5 · γf · (1 − fq,r ) + 4.5 · (fq,r − 1) + 1 for 0.7 < γf ≤ 0.9 , (15.21)
1.0 for γf > 0.9 .
In Eq. (15.21), fq,r is the adjustment factor for rough slopes which is mainly
dependent on the slope of the structure and whether wind needs to be included or
not:
1.0 for m ≤ 0.6 ,
fq,r = fw · (8.5 · m − 4.0) for 0.6 < m ≤ 4.0 , (15.22)
fw · 30 for m > 4.0 ,
in which m = cot α (slope of structure); fw accounts for the presence of wind and
is set to fw = 1.0 if there is wind and fw = 0.67 if there is no wind.
This set of equations include the case of smooth dikes which will, due to γf = 0.9
in this case, always lead to an adjustment factor of fq = 1.0. In case of a very rough
August 14, 2009 9:48 9.75in x 6.5in b684-ch15 FA
Wave Run-Up and Wave Overtopping at Armored Rubble Slopes and Mounds 405
1:4 slope with wind fq,max = fqr = 30.0, which is the maximum, the factor can get
to (but only if the mean overtopping rates get below qss = 10−5 m3 /s per m). The
latter case and a steep rough slope is illustrated in Fig. 15.14.
Wave Run-Up and Wave Overtopping at Armored Rubble Slopes and Mounds 407
described by
Equation (15.27) can be rewritten to calculate directly the travel distance x (at
a certain level y) by rewriting the above equation:
Shingle beaches differ from the armored slopes principally in the size of the beach
material, and hence its mobility. The typical stone size is sufficiently small to permit
significant changes of beach profile, even under relatively low levels of wave attack.
A shingle beach may be expected to adjust its profile to the incident wave conditions,
provided that sufficient beach material is available. Run-up or overtopping levels on
a shingle beach are therefore calculated without reference to any initial slope.
The equilibrium profile of shingle beaches under (temporary constant) wave
conditions is described by van der Meer.11 The most important profile parameter
for run-up and overtopping is the crest height above SWL, hc . For shingle with
Dn50 < 0.1 m, this crest height is only a function of the wave height and wave
August 14, 2009 9:48 9.75in x 6.5in b684-ch15 FA
steepness. Note that the mean wave period is used, not the spectral wave period
Tm−1,0 .
hc /Hm0 = 0.3s−0.5
om . (15.29)
Only the highest waves will overtop the beach crest and most of this water will
percolate through the material behind the beach crest. Equation (15.29) gives a
run-up or overtopping level which is more or less close to Ru2% .
Acknowledgments
This chapter was based on the EurOtop Overtopping Manual,4 which was funded
in the United Kingdom by the Environmental Agency, in Germany by the German
Coastal Engineering Research Council (KFKI), in the Netherlands by Rijkswater-
staat and Netherlands Expertise Network (ENW) on Flood Protection.
The Project Team for the creation, editing, and support of the manual; the
Project Steering Group for guidance and supervision; and a number of individual
persons, have been listed and acknowledged in Chap. 14.
References
References in this chapter have been kept to a real minimum. An extensive list
of relevant references, however, can be found in the EurOtop Overtopping Manual
(2007).
Wave Run-Up and Wave Overtopping at Armored Rubble Slopes and Mounds 409
Chapter 16
Tom Bruce
School of Engineering, University of Edinburgh
King’s Buildings, Edinburgh, EH9 3JL, UK
tom.bruce@ed.ac.uk
Tim Pullen
HR Wallingford
Howbery Park, Wallingford, Oxon, OX10 8BA, UK
tap@hrwallingford.co.uk
William Allsop
HR Wallingford
Howbery Park, Wallingford, Oxon, OX10 8BA, UK
w.allsop@hrwallingford.co.uk
411
August 14, 2009 10:4 9.75in x 6.5in b684-ch16 FA
16.1. Introduction
This chapter presents guidance for the assessment of overtopping and post-
overtopping processes at vertical and steep-fronted coastal structures such as caisson
and blockwork breakwaters and vertical seawalls (see Figs. 16.1 and 16.2). Also
included are composite vertical wall structures (where the emergent part of the
structure is vertical, fronted by a modest berm) and vertical structures which include
a recurve/bull-nose/parapet/wave return wall as the upper part of the defense.
Large vertical breakwaters (Fig. 16.1) are almost universally formed of sand-filled
concrete caissons usually resting on a small rock mound. Such caisson breakwaters
may reach depths greater than 100 m, under which conditions with no wave breaking
at all at the wall would be expected. Conversely, older breakwaters may, out of
necessity, have been constructed in shallower water or indeed, built directly on
natural rock “skerries.” As such, these structures may find themselves exposed to
breaking wave, or “impulsive” conditions when the water depth in front of them is
Fig. 16.1. Examples of vertical breakwaters: (left) modern concrete caisson and (right) older
structure constructed from concrete blocks.
Fig. 16.2. Examples of vertical seawalls: (left) modern concrete wall and (right) older stone
blockwork wall.
August 14, 2009 10:4 9.75in x 6.5in b684-ch16 FA
sufficiently low. Urban seawalls (e.g., Fig. 16.2) are almost universally fronted by
shallow water, and are likely to be exposed to breaking or broken wave conditions,
especially in areas of significant tidal range.
There are three principal sources of guidance on this topic preceding this chapter;
in the United Kingdom, the Environment Agency’s Overtopping of Seawalls: Design
and Assessment Manual 2 ; in the United States, the US Army Corps of Engineers’
Coastal Engineering Manual 6 ; in Japan, Goda’s design charts.11 The guidance
presented in this chapter builds upon that of Besley,2 with adjustments to many
formulae based upon further testing since 1999.
For those familiar with Besley,2 the principal changes/additions are:
• minor adjustments to recommended approach for distinguishing impulsive/
non-impulsive conditions (Sec. 16.2);
• all formulae are now given in terms of wave period Tm−1,0 resulting in an adjusted
definition of the h∗ , d∗ , and Vbar parameters (Secs. 16.2.2, 16.2.3, and 16.4.2,
respectively) in order to maintain comparability with the earlier work;
• in line with convergence on the Tm−1,0 measure, formulae using fictitious wave
steepness sop have been adjusted to use the new preferred measure sm−1,0
(Secs. 16.3.1 and 16.4.2);
• new guidance on mean overtopping under conditions when all waves break before
reaching the wall (part of Sec. 16.3.1);
• new guidance on overtopping at vertical walls at zero freeboard (Sec. 16.3.1);
• minor adjustments to formulae for mean overtopping under impulsive condi-
tions due to the availability of additional data, from e.g., the CLASH database7
(Sec. 16.3.1);
• extension of method for mean overtopping to account for steep (but not vertical)
“battered” walls (Sec. 16.3.2);
• new guidance on prediction of mean and wave-by-wave overtopping for oblique
wave attack under impulsive conditions (Sec. 16.3.4);
• new guidance on reduction in mean overtopping discharge due to wave return
walls/parapets/recurves (Sec. 16.3.5);
• new guidance on the effect of wind on mean overtopping discharge (Sec. 16.3.6);
• inclusion of summary of new evidence on scale effects for laboratory study of
overtopping at vertical and steep walls (Sec. 16.3.7);
• new guidance on “post-overtopping” processes, specifically; velocity of “throw”;
landward spatial extent of overtopping (Sec. 16.5);
• all formulae are now given explicitly in terms of basic wave and structural param-
eters without recourse to intermediate definitions of dimensionless overtopping
discharge and freeboard parameters specific to impulsive conditions.
The qualitative form of the physical processes occurring when the waves
reach the wall are described in Sec. 16.2. Distinctions drawn between different
wave/structure “regimes” are reflected in the guidance for assessment of mean over-
topping discharges given in Sec. 16.3. The basic assessment tools are presented for
plain vertical walls (Sec. 16.3.1), followed by subsections giving advice on how these
basic tools should be adjusted to account for other commonly-occurring configura-
tions, such as battered walls (Sec. 16.3.2); vertically composite walls (Sec. 16.3.3);
August 14, 2009 10:4 9.75in x 6.5in b684-ch16 FA
the effect of oblique wave attack (Sec. 16.3.4); the effect of recurve/wave-return
walls (Sec. 16.3.5). The effect of wind is described in Sec. 16.3.6, with scale and
model effects reviewed in Sec. 16.3.7. Methods to assess individual “wave-by-wave”
overtopping volumes are presented in Sec. 16.4. The current knowledge and advice
on post-overtopping processes including velocities, spatial distributions, and post-
overtopping loadings are reviewed in Sec. 16.5.
Fig. 16.3. A non-impulsive (pulsating) wave condition at a vertical wall, resulting in a non-
impulsive (or “green water”) overtopping.
Fig. 16.4. An impulsive (breaking) wave at a vertical wall, resulting in an impulsive (violent)
overtopping condition.
August 14, 2009 10:4 9.75in x 6.5in b684-ch16 FA
Fig. 16.5. Broken wave at a vertical wall, resulting in a broken wave overtopping condition.
Fig. 16.6. Definition sketch for assessment of overtopping at plain vertical walls.
inshore:
hs 2π hs hs 2π hs
h∗ = 1.3 2
or h∗ = 2
, (16.1)
Hm0 g Tm−1,0 Hm0 g Tm
where Hm0 = the significant wave height at the toe of the structure and Tm−1,0 =
the spectral wave period based on spectral moments m−1 and m0 , at the same
location.
Non-impulsive (pulsating) conditions dominate at the wall when h∗ > 0.3, and
impulsive conditions occur when h∗ < 0.2. The transition between conditions for
which the overtopping response is dominated by breaking and nonbreaking waves
lies over 0.2 < h∗ < 0.3. In this region, overtopping should be predicted for both
non-impulsive and impulsive conditions, and the larger value assumed.
It should be noted that, for certain long-period waves, h∗ < 0.2 can be found
under conditions for which no breaking at the vertical wall would be expected.
Although the term “impulsive overtopping” may not be accurate in these conditions,
the overtopping responses nevertheless follow those described by the “impulsive”
formulae.
Fig. 16.7. Definition sketch for assessment of overtopping at composite vertical walls.
d 2π hs d 2π hs
d∗ = 1.3 2 or d∗ = , (16.2)
Hm0 g Tm−1,0 Hm0 g Tm
2
be one standard deviation. Each equation will be given as a mean prediction with
uncertainty, followed by the application in a deterministic way.
For simple vertical breakwaters, the following equations should be used.
1.E+00
CLASH database set 028
CLASH database set 106
0.5
5%
1.E-04
5%
1.E-05
1.E-06
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
dimensionless freeboard Rc/Hm0
Fig. 16.8. Mean overtopping at a plain vertical wall under non-impulsive conditions [Eq. (16.3)].
0.10
0.09 RC=0
dimensionless discharge q/(gHm0 )0.5
3
3 0.5
0.08 q/(g Hm0 ) = 0.062 ± 0.006
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0.00
0.00 0.01 0.02 0.03 0.04 0.05 0.06
(fictitious) wave steepness sm-1,0
1.E+03
1.E-01
1.E-02
1.E-03 5%
1.E-04
5%
1.E-05
1.E-06
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
(impulsive) dimensionless freeboard h* Rc/Hm0
Fig. 16.10. Mean overtopping at a plain vertical wall under impulsive conditions [Eq. (16.4)].
the data about the mean prediction: log10 (qmeasured )−log10 (qpredicted ) is taken as a
normally distributed stochastic parameter with a mean of 0 and a standard deviation
σ = 0.37 (i.e., 68% of predictions lie within the range of ×/÷ 2.3). For probabilistic
calculations, Eq. (16.4) should be taken together with these stochastic coefficients.
For predictions of measurements or comparison with measurements also Eq. (16.4)
should be taken with, for instance, 5% upper and lower exceedance curves. For
deterministic design or safety assessment, a coefficient of 2.8 × 10−4 should be
used in Eq. (16.4), instead of 1.5 × 10−4 :
−3.1
q −4 Rc Rc
= 1.5 × 10 h∗ valid over 0.03 < h∗ < 1.0 . (16.4)
h∗ ghs
2 3 H m0 H m0
For h∗ Rc /Hm0 < 0.02 arising from hs reducing to very small depths (as opposed
to from small relative freeboards), there is evidence supporting an adjustment
downward of the predictions of the impulsive formulae due to the observation
that only broken waves arrive at the wall.4 The mean prediction is then described
by Eq. (16.5). The reliability of this equation is described by considering the
scatter in the logarithm of the data about the mean prediction: log10 (qmeasured ) −
log10 (qpredicted ) is taken as a normally distributed stochastic parameter with a mean
of 0 and a standard deviation σ = 0.15 (i.e., 68% of predictions lie within the range
of ×/ ÷ 1.4):
−2.7
q Rc Rc
= 2.7 × 10−4 h∗ valid for h∗ < 0.02, broken waves.
h2∗ gh3s Hm0 Hm0
(16.5)
August 14, 2009 10:4 9.75in x 6.5in b684-ch16 FA
1.E+04
Bruce et al (2003) data
0.5
dimensionless discharge q / h *2(ghs3)
1.E+02
5%
1.E+01
5%
1.E+00
0.000 0.005 0.010 0.015 0.020 0.025 0.030
(impulsive) dimensionless freeboard h* Rc / Hm0
Fig. 16.11. Mean overtopping discharge for lowest h∗ Rc /Hm0 (for broken waves only arriving at
wall) with submerged toe (hs > 0). For 0.02 < h∗ Rc /Hm0 < 0.03, overtopping response is ill-
defined — lines for both impulsive conditions (extrapolated to lower h∗ Rc /Hm0 ) and broken wave
only conditions (extrapolated to higher h∗ Rc /Hm0 ) are shown as dashed lines over this region.
1.E-03
0.5
.(sm-1,0 x m)
1.E-04
0.5
dim'less discharge q/(gHm03)
1.E-05 5%
1.E-07
0.0 1.0 2.0 3.0 4.0 5.0 6.0
0.33
dimensionless freeboard Rc/Hm0 x sm-1,0 xm
Fig. 16.12. Mean overtopping discharge with emergent toe (hs < 0).
valid for
Rc
2.0 < m s0.33
m−1,0 < 5.0; 0.55 ≤ Rc /Hm0,deep ≤ 1.6; sm−1,0 ≥ 0.025 .
Hm0,deep
(16.6)
Fig. 16.13. Battered walls: typical cross-section (left), and Admiralty Breakwater, Alderney
Channel Islands (right, courtesy G. Müller).
1.E+01
dimensionless discharge q / h *2(ghs3)0.5
1.E-01
1.E-02
1.E-03
1.E-04
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
(impulsive) dimensionless freeboard, h* Rc/Hm0
freeboards. Multiplying factors are given in Eqs. (16.17a) and (16.7b) (plotted in
Fig. 16.14).
10:1 battered wall: q10:1batter = qvertical × 1.3 , (16.7a)
5:1 battered wall: q5:1batter = qvertical × 1.9 , (16.7b)
where qvertical is arrived at from Eq. (16.4). The uncertainty in the final estimated
overtopping discharge can be estimated as per the plain vertical cases.
No dataset is available to indicate an appropriate adjustment under non-
impulsive conditions. An alternative method, however, may be to calculate
August 14, 2009 10:4 9.75in x 6.5in b684-ch16 FA
overtopping for a vertical structure and for a steep smooth slope 1:1. By using the
cot α of the battered wall as the parameter (between the limits 0 and 1), a fair guess
will be reached (see also Fig. 14.2 in Chap. 14 of this handbook).
(i) Small toe mounds which have an insignificant effect on the waves approaching
the wall — here, the toe may be ignored and calculations proceed as for simple
vertical (or battered) walls.
(ii) Moderate mounds, which significantly affect wave-breaking conditions, but are
still below water level. Here, a modified approach is required.
(iii) Emergent mounds in which the crest of the armor protrudes above still water
level. Prediction methods for these structures may be adapted from those for
crown walls on a rubble mound (Chap. 15 of this handbook).
1.E+03
VOWS data
dimensionless discharge q / d *2(gd3)0.5
1.E+02
composite vertical - probabilistic (Eq. 8)
1.E+00
1.E-01
1.E-02
5%
1.E-03
5%
1.E-04
0.0 0.2 0.4 0.6 0.8 1.0 1.2
(impulsive, composite) dimensionless freeboard d* Rc/Hm0
1.E+01
dimensionless discharge q / h *2(ghs3)0.5
1.E-01
1.E-02
1.E-03
1.E-04
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
(impulsive) dimensionless freeboard, h* Rc/Hm0
the “worst case” conditions at locations along the wall where the discharge is the
greatest:
Rc
for β = 150 , h∗ ≥ 0.2, as per impulsive β = 0◦ (Eq. 16.4), (16.12a)
Hm0
Rc
for β = 300 , h∗ ≥ 0.07, as per impulsive β = 15◦ (Eq. 16.11b), (16.12b)
Hm0
Rc
for β = 600 , h∗ ≥ 0.07, as per non-impulsive β = 0◦ (Eq. 16.3). (16.12c)
Hm0
Fig. 16.17. An example of a modern, large vertical breakwater with wave return wall (left) and
cross-section of an older seawall with recurve (right).
Fig. 16.18. A sequence showing the function of a parapet/wave return wall in reducing over-
topping by redirecting the uprushing water seaward (back to right).
August 14, 2009 10:4 9.75in x 6.5in b684-ch16 FA
Fig. 16.19. Parameter definitions for assessment of overtopping at structures with parapet/wave
return wall.
• the case where a wall is chamfered backward at the crest [normally admitting
greater overtopping (α > 90◦ )].
For the latter, chamfered wall case, influence factors γ should be applied to Franco
et al.’s equation10 for non-impulsive mean discharge [Eq. (16.13)] with the value of
γ selected as shown8 :
q 4.3 Rc
0.5 = 0.2 exp − , (16.13)
3
(gHm0 ) γ Hm0
γ = 1.01 for α = 120◦ ,
γ = 1.13 for α = 135◦ ,
γ = 1.07 for α = 150◦ .
Fig. 16.20. “Decision chart” summarizing methodology for tentative guidance. Note that symbols
R∗0 , k23 , m, and m∗ used (only) at intermediate stages of the procedure are defined in the lowest
boxes in the figure. Please refer to text for further explanation.
Given the level of scatter in the original data and the observation that the
methodology is not securely founded on the detailed physical mechanisms/processes,
it is suggested that it is impractical to design for k < 0.05, i.e., reductions in mean
discharges by factors greater than 20 cannot be predicted with confidence. If such
large (or larger) reductions are required, a detailed physical model study should be
considered.
(1) changing the shape of the incident wave crest at the structure resulting in a
possible modification of the dominant regime of wave interaction with the wall;
August 14, 2009 10:4 9.75in x 6.5in b684-ch16 FA
(2) blowing up-rushing water over the crest of the structure (for an onshore wind,
with the reverse effect for an offshore wind) resulting in possible modification
of mean overtopping discharge and wave-by-wave overtopping volumes;
(3) modifying the physical form of the overtopping volume or jet, especially in
terms of its aeration and breakup resulting in possible modification to post-
overtopping characteristics such as throw speed, landward distribution of dis-
charge, and any resulting post-overtopping loadings (e.g., downfall pressures).
10,0
De Waal et al. (1996)
Davey (2004)
Pullen and Allsop (2004)
Eq. (15)
f Wind [-]
5,0
0,0
1,0E-07 1,0E-06 1,0E-05 1,0E-04 1,0E-03
qss [m3/s/m]
Fig. 16.21. Wind adjustment factor fwind plotted over mean overtopping rates qss for three
datasets.
August 14, 2009 10:4 9.75in x 6.5in b684-ch16 FA
1.E+01
0.5
dimensionless discharge q / h *2(ghs3)
1.E-02
1.E-03
1.E-04
1.E-05
1.E-06
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
(impulsive) dimensionless freeboard, h* Rc/Hm0
Fig. 16.22. Large-scale laboratory measurements of mean discharge at 10:1 battered wall
under impulsive conditions showing agreement with prediction line based upon small-scale tests
[Eq. (16.7a)].
August 14, 2009 10:4 9.75in x 6.5in b684-ch16 FA
1.E+04
dimensionless discharge q / d *2(gd3)0.5
1.E+01
1.E+00
1.E-01
1.E-02
5%
1.E-03
5%
1.E-04
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
(impulsive, composite) dimensionless freeboard d* Rc/Hm0
Fig. 16.23. Results from field measurements of mean discharge at Samphire Hoe, UK, plotted
together with Eq. (16.8).
16.4.1. Introduction
While the prediction of mean discharge (Sec. 16.3) offers the information required to
assess whether overtopping is slight, moderate, or severe, and make a link to any pos-
sible flooding that might result, the prediction of the volumes associated with indi-
vidual wave events can offer an alternative (and often more appropriate) measure
for the assessment of tolerable overtopping levels and possible direct hazard. First, a
method is given for the prediction of maximum overtopping volumes expected asso-
ciated with individual wave events for plain vertical structures under perpendicular
wave attack (Sec. 16.4.2). This method is then extended to composite (bermed)
structures (Sec. 16.4.3) and to conditions of oblique wave attack (Sec. 16.4.4).
Finally, a short section on scale effects is included (Sec. 16.4.5).
The methods given for perpendicular wave attack are the same as those given
previously in UK guidance,2 but now using the Tm−1,0 period measure (with Weibull
parameters, steepness values, etc., adjusted accordingly). The extension to oblique
wave attack is new.
August 14, 2009 10:4 9.75in x 6.5in b684-ch16 FA
Rc
Now = Nw exp −1.21 (for h∗ > 0.3) . (16.17)
Hm0
b
V
PV = 1 − exp − , (16.19)
a
where PV is the probability that an individual event volume will not exceed V ; a and
b are Weibull “shape” and “scale” parameters, respectively. Thus, to estimate the
largest event in a wave sequence predicted to include (e.g.) Now = 200 overtopping
events, Vmax would be found by taking PV = 1/200 = 0.005. Equation (16.19) can
then be rearranged to give
1/b
Vmax = a (ln Now ) . (16.20)
The Weibull shape parameter a depends upon the average volume per over-
topping wave Vbar , where
qTm−1,0 Nw qTm Nw
Vbar = 0.8 or Vbar = . (16.21)
Now Now
For non-impulsive conditions, there is a weak steepness-dependency for the
scale and shape parameters a and b:
0.74Vbar 0.66 for sm−1,0 = 0.024
a= b= for h∗ > 0.3 . (16.22)
0.90Vbar 0.82 for sm−1,0 = 0.048
(Note that the fictitious steepness values sm−1,0 = 0.024 and 0.048 in Eq. (16.22)
correspond to the values of sop = 0.02 and 0.04 quoted by Besley2 .)
For impulsive conditions2,16 :
1.E+00
measured max. individual overtopping
1.E-02
1.E-03
1.E-04
1.E-04 1.E-03 1.E-02 1.E-01 1.E+00
3
predicted max. individual overtopping volume, Vmax [m /m]
Fig. 16.24. Predicted and measured maximum individual overtopping volume — small- and large-
scale tests.16
The effectiveness of the predictor under impulsive conditions can be gauged from
Fig. 16.24. Note that all the a values are fairly similar, leading to similar steep
distributions (see also Sec. 14.2.2 in Chap. 14 of this handbook). A further step in
development could be to choose a fixed value for a and modify b accordingly.
1 Rc
Now = Nw exp − 2 for h∗ > 0.3 . (16.24)
C Hm0
August 14, 2009 10:4 9.75in x 6.5in b684-ch16 FA
Table 16.1. Summary of prediction formulae for individual overtopping volumes under
oblique wave attack. Oblique cases valid for 0.2 < h∗ Rc /Hm0 < 0.65. For 0.07 <
h∗ Rc /Hm0 < 0.2, the β = 00 formulae should be used for all β.
C is given by
0.91 for β = 0◦
C = 0.91 − 0.00425β for 0◦ < β < 40◦ for h∗ > 0.3 . (16.25)
0.74 for β ≥ 40◦
7.0
max. dim'less vertical throw speed, u z / ci
6.0
5.0
4.0
3.0
2.0
1.0
0.0
0.0 0.1 0.2 0.3 0.4 0.5
h* parameter
Fig. 16.25. Speed of upward projection of overtopping jet past structure crest plotted with h∗
parameter (after Bruce et al.5 ).
• the pressures that may arise due to the downfalling overtopped jet impacting on
the structure’s crown deck.
1
fraction of discharge fallen by distance x/L
0.9
0.8
c. 95% of discharge
0.7 falls within
0.25 × wavelength
0.6
0.5
c. 90% of discharge falls
0.4 within 0.2 × wavelength
0.3
0.2
approx. lower (worst case)
envolope
0.1
0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
dimensionless landward distance x/L0
Fig. 16.26. Landward distribution of overtopping discharge under impulsive conditions. Curves
show proportion of total overtopping discharge which has landed within a particular distance
shoreward of seaward crest.
Under violent (impulsive) overtopping conditions, the idea of spatial extent and
distribution has a greater physical meaning — where does the airborne overtopping
jet come back to the level of the pavement behind the crest? The answer to this
question, however, will (in general) depend strongly upon the local wind conditions.
Despite the difficulty of directly linking a laboratory wind speed to its prototype
equivalent (see Sec. 16.3.6), laboratory tests have been used to place an upper
bound on the possible wind-driven spatial distribution of the “fall back to ground”
footprint of the violently overtopped volumes.3,17 Tests used large fans to blow
air at gale-force speeds (up to 28 m/s) in the laboratory. The lower (conservative)
envelope of the data (Fig. 16.26) gives the approximate guidance that 95% of the
violently-overtopped discharge will land within a distance of 0.25 × Lo , where Lo is
the offshore (deepwater) wavelength. No data is available for the case of a shallow
foreshore with no clear definition of Lo .
this distance was observed to increase to ∼ 2 × Hm0 . Over all small-scale tests,
pressure maxima were measured over the range
p1/250
2< < 17 with a mean value of 8. (16.27)
ρ gHm0
The largest downfall impact pressure measured in large-scale tests was 220 kPa
(with a duration of 0.5 ms). The largest downfall pressures were observed to result
from overtopping jets thrown upward by very-nearly breaking waves (the “flip
through” condition). Although it might be expected that scaling small-scale impact
pressure data would overestimate pressure maxima at large scale, approximate com-
parisons between small- and large-scale test data suggest that the agreement is good.
Acknowledgments
This chapter was based on the EurOtop Overtopping Manual,14 which was funded
in the United Kingdom by the Environmental Agency, in Germany by the German
Coastal Engineering Research Council (KFKI), and in the Netherlands by Rijkswa-
terstaat and Netherlands Expertise Network (ENW) on Flood Protection.
The Project Team for the creation, editing, and support of the manual; the
Project Steering Group for guidance and supervision; and a number of individual
persons, have been listed and acknowledged in Chap. 14.
References
Chapter 17
Iribarren number is widely used by coastal engineers for classifying the type of
breaking waves and for estimating armor weight, run-up height, and even for esti-
mating wave overtopping. Iribarren number is the ratio of breakwater slope or
beach slope to the square root of wave steepness. Another existing surf parameter
is the ratio of breakwater slope to the wave steepness. In the present chapter
new surf parameters, which are called “first order wave action slope” and “second
order wave action slope,” are introduced for representing the local wave condi-
tions in shallow waters by employing local values of wavelength as well as wave
height. The use of linear wave theory on a flat bed of the depth at the front of
breakwater might be considered far better than the simple adoption of the deep-
water wavelength for characterizing surfing waves at a shoaling depth. The wave
action slopes are formed by the product of the breakwater slope and the celerity
ratio to the wave height. The run-up height is related to the first order wave
action slope, and the optimum or minimum weight of armor unit is related to the
second order wave action slope.
17.1. Introduction
When waves approach a coast, wave particles transform from elliptic movement to
horizontal movement and energy transport velocity (group velocity) continues to
reduce to near zero. Then wave heights increase up to a certain limit, that is, until
they lose their stability of wave formation. The type of breaking waves is one of
the major factors for the determination of wave forces affecting beaches and coastal
structures.
Iribarren1 suggested a wave breaking parameter to determine the type of
breaking waves and the Iribarren number is in recent years widely employed for
the estimation of run-up height and unit weight of armor block. Iribarren number
is the ratio of beach slope to the square root of wave steepness with introducing
441
July 30, 2009 11:4 9.75in x 6.5in b684-ch17 FA
442 D. H. Yoo
Using local values of wavelength and wave height, the first order Iribarren number
I is represented by1 :
S
I= √ , (17.1)
M
July 30, 2009 11:4 9.75in x 6.5in b684-ch17 FA
where M = H/L (wave steepness) and S = tan θ (beach slope or breakwater slope).
Offshore Iribarren number Io uses the values of wave height Ho and wavelength Lo
at deepwater for the estimation of wave steepness. Inshore Iribarren number Ii uses
the local value of wave height but still uses the value of wavelength at deepwater
condition. On the other hand, the local Iribarren number I takes the local values
of both wave height and wavelength for the estimation of wave steepness.
The second order Iribarren number J is also, using local values of wavelength
and wave height, represented by:
S
J= . (17.2)
M
The second order Iribarren number is directly related to the wave steepness, while
the first order Iribarren number is related to the square root of the wave steepness.
The former is associated with the square of wave celerity, and the latter with the
wave celerity with no power. Similarly, offshore second order Iribarren number Jo
uses the values of wave height Ho and wavelength Lo at deepwater for the estimation
of wave steepness. Inshore second order Iribarren number Ji uses the local value of
wave height but still uses the value of wavelength at deepwater condition. On the
other hand, the local second order Iribarren number J takes the local values of both
wave height and wavelength for the estimation of wave steepness.
Wave action slope is the product of the beach slope S and the Froude number
associated with wave celerity C and wave height H as suggested by Yoo and Na.2
Two types of wave action slopes are suggested; first order and second order. The
first order wave action slope SX is represented by:
SX = FH S, (17.3)
√
where FH = C/ gH. Similarly, the offshore first order wave action slope SXo uses
the values of wave height Ho and wave celerity Co at deepwater. And the inshore
first order wave action slope SXi uses the local value of wave height but still uses
the value of wave celerity at deepwater condition. On the other hand, the local first
order wave action slope takes the local values of both wave height and wave celerity.
The second order wave action slope SY (Ref. 3) is represented by:
SY = FH
2
S. (17.4)
Offshore second order wave action slope SY o uses the values of wave height Ho
and wave celerity Co at deepwater. And inshore second order wave action slope
SY i uses the local value of wave height but still uses the value of wave celerity at
deepwater condition. On the other hand, the local second order wave action slope
takes the local values of both wave height and wave celerity.
Using dispersion relation as follows:
444 D. H. Yoo
SXo SXi SX SY o SY i SY
2 2
√Co S √Co S √C S
Co Co C2
Definition gHo gH gH gHo
S gH
S gH
S
q
Relationship √1 Io √1 Ii tanh kh
I 1
J 1
J tanh kh
J
2π 2π 2π 2π o 2π i 2π
where σ is the angular frequency (σ = 2π/T ) and k is the wave number (k = 2π/L),
the local first order wave action slope SX is related with the local first order Iribarren
number I as follows:
C tanh kh
SX = √ S=I . (17.6)
gH 2π
√
For√the offshore condition tanh kh ≈ 1, and hence SXo = Io / 2π and SXi =
Ii / 2π. That is, the first order wave action slopes are found simply proportional
to the first order Iribarren numbers. Similarly, the local second order wave action
slope SY is related with the local second order Iribarren number J as follows:
C2 tanh kh
SY = S=J . (17.7)
gH 2π
For the offshore condition tanh kh ≈ 1, and hence SY o = Jo /2π and SY i = Ji /2π.
That is, the second order wave action slopes are found simply proportional to the
second order Iribarren numbers.
The first order wave action slope is closely related to the first order Iribarren
number and the second order wave action slope is closely related to the second
order Iribarren number. It is, however, found that the relations do not have simple
linear proportionality. The wave action slopes are related to the Iribarren numbers
as presented in Table 17.1. As summarized in the table, the offshore numbers and
the inshore numbers are linearly proportional to each other, but the local numbers
are not directly proportional to each other. Further effect of depth dispersion is
reflected in the wave action slopes through dispersion relation.
The run-up height is the vertical length from the mean surface to the top of run-
up reach, and the run-up height ratio is defined by the ratio of run-up height to
wave height. Saville4 conducted laboratory experiments on a concrete slope with
various conditions of beach slope and wave steepness for the monochromatic waves,
and presented two separate graphs of run-up height ratio to the slope with various
inshore wave steepness (Mi = H/L0 ) as shown in Fig. 17.2(a). It is found that the
run-up height ratio is almost proportional to the beach slope up to a certain limit,
from where reflection might be a significant factor on the wave particle movement
on the slope. Depending on the water depth condition, the run-up height ratio is
July 30, 2009 11:4 9.75in x 6.5in b684-ch17 FA
Fig. 17.2. Run-up height ratio ηR versus (a) S or (b) SXi using the laboratory data of Saville.4
related to the Iribarren number or first order wave action slope as follows:
ηR = 2.64SXi , (SXi < 0.427Mi−0.25, 1 < h/H < 3), (17.8a)
ηR = 2.39SXi , (SXi < 0.674Mi−0.16, 3 < h/H), (17.8b)
where ηR = hR /H (run-up height ratio), hR is the run-up height, H is the wave
height. As shown in Fig. 17.2(b), the effect of wave period is found well implied in
the expression of inshore wave action slope SXi . Further simplification can be made
by the use of local wave action slope, and the result is given by:
ηR = 2.80SX (SX < 0.394M −0.25). (17.9)
When SX < 0.394M −0.25, the influence of reflection is considered very small. Oth-
erwise, the effect of reflection becomes strong and then the run-up height becomes
almost irrespective of slope or the wave action slope.
Mase5 conducted laboratory experiments of run-up on smooth concrete slope
(S = 0.033 − 0.25) for irregular waves, and suggested empirical equation of run-up
height related to the offshore Iribarren number as follows:
ηR−n = αIoβ , (17.10)
July 30, 2009 11:4 9.75in x 6.5in b684-ch17 FA
446 D. H. Yoo
hR−n α β α β α β δ
hR−max 2.32 0.77 — — 0.41 2.2 1
hR−2 1.86 0.71 0.37 1.38 0.44 1.64 0.742
hR−1/10 1.70 0.71 0.34 1.27 0.38 1.52 0.688
hR−1/3 1.38 0.70 0.27 1.04 0.30 1.25 0.568
hR−avr 0.88 0.69 0.17 0.66 0.20 0.79 0.359
where ηR = hR /HS , HS is the significant wave height, and the subscript n indicates
the reaches of different probability. The values of constants α and β for different
probability are presented in Table 17.2.
Hedges and Mase6 re-analyzed the laboratory data of run-up height of irregular
waves of Mase.4 They deducted the set-up height from the total run-up height, and
related the pure run-up height to the inshore Iribarren number as follows:
ηR−n = α + βIi , (17.11)
where α and β are constants, but the values vary depending on the probability
of run-up reaches as presented in Table 17.2. For the case of irregular waves, the
run-up height ratio is determined by using significant wave height HS , that is,
ηR = hR /HS . For the case of 2% reach ηR−2 they suggested α = 0.37 and β = 1.38.
New empirical equation of run-up height ratio is developed using the local Iribarren
number as follows:
ηR−n = α + βI. (17.12)
Comparison of the equations (17.10), (17.11), and (17.12) are shown in Fig. 17.3
against Mase’s laboratory data.5 Up to the value of Ii = 2, Eqs. (17.10) and (17.11)
show very good correlation with the laboratory data. But when the inshore first
order Iribarren number exceeds approximately 2, the results of Eqs. (17.10) and
(17.11) deviate considerably from the laboratory data. On the other hand, when
using the local Iribarren number as expressed by Eq. (17.12), good correlations are
found over the whole region of I.
Another empirical equation of run-up height ratio is developed using the local
wave action slope SX instead of the inshore Iribarren number Ii or the local Irib-
arren number I. Hedges and Mase6 used different separate values of constants
of Eq. (17.10) or (17.11) for different reaches of probability in order to keep the
accuracy. For the engineering use the run-up height of different reaches of proba-
bility ηR−n can be simply related to the maximum run-up height as follows:
ηR−n = δηR−max , (17.13)
where δ is the reduction factor of probability. The maximum run-up height ratio of
irregular waves ηR−max is expressed by:
ηR−max = 0.32 + 6.9SX (17.14)
July 30, 2009 11:4 9.75in x 6.5in b684-ch17 FA
Fig. 17.3. Run-up height ratio versus Io , Ii , I, and SX using the laboratory data of Mase.5
or
ηR−max = 8.1SX . (17.15)
The degree of accuracy of Eq. (17.11) or (17.12) is slightly better than that
of Eq. (17.13), but Eq. (17.13) is also acceptable for an engineering use with less
than 5% endurance. The reduction factor of probability δ = 1 for the maximum
reach, δ = 0.742 for the 2% reach, δ = 0.688 for the 10% reach, δ = 0.568 for
the 1/3 reach, and δ = 0.359 for the average reach as summarized in Table 17.2.
The computation results are presented in Fig. 17.4. As shown in the figure, the
correlation of Eq. (17.13) is generally found very good simply by introducing the
July 30, 2009 11:4 9.75in x 6.5in b684-ch17 FA
448 D. H. Yoo
Fig. 17.4. Various run-up height ratios versus SX using the laboratory data of Mase.5
reduction factor. The correlation coefficients are estimated over 0.97 for all cases,
and the determination coefficients are about 0.93 for most cases.
Comparing both constants of Eqs. (17.9) and (17.15) each other, the maximum
run-up reach of irregular waves is found almost tripple times bigger than the run-up
reach of monochromatic waves. The reduction factor of average reach of irregular
waves is about 0.36, and this is almost the same as the ratio of the two constants
of Eqs. (17.9) and (17.15).
The run-up height can be significantly reduced for the rough slope of breakwaters
from the values given for the smooth concrete slope. Ahrens7 suggested to use
another reduction factor for the rough wall. The run-up height is, hence, finally
computed by:
ηR−n = ξδηR−max , (17.16)
where ξ is the reduction of run-up height for rough wall. ηR−max is the maximum
run-up height ratio on smooth concrete slope. The values of ξ are given in Table 17.3.
Table 17.3. Reduction factor of wall roughness for the estimation of run-up height.
where ηW = W/γH 3 , W is the armor weight, γ is the specific weight of armor block,
s is the specific gravity of armor block, and KD is the empirical constant which is
primarily associated with the type of armor block. Hudson equation indicates that
the weight ratio is solely dependent on the breakwater slope, but various workers
found that the optimum weight of armor block is also significantly influenced by
wave period, percolation of slope, and the formation of irregular waves. van der
Meer9 employed the inshore first order Iribarren number, and suggested a set of
two equations depending on the condition determined by the Iribarren number. The
equation of van der Meer9 is represented by:
0.0042NS Ii1.5
ηW = (0.4 < Ii < 2.5), (17.18a)
(s − 1)3 P 0.54
NS P 0.39 S 1.5
ηW = (2.5 < Ii ), (17.18b)
(s − 1)3 Ii3P
√
where P is the percolation, NS = ( N/D)0.6 , N is the number of incident waves,
and D is the degree of damage which is defined by the ratio of damage area to the
projected area of armor unit. The number of incident waves for a field is normally
between 1000 and 3000, and the degree of damage is between 1 and 3.
The armor weights measured in the laboratory by van der Meer9 are related to
the Iribarren number Ii and the second order wave action slope SY . As shown in
Fig. 17.5, the use of the second order wave action slope gives the best correlation
with the laboratory data. As found by van der Meer,9 the variation of ηW becomes
complicated when Ii is bigger than 2.5. New empirical equation has been developed
using the second order wave action slope as follows:
0.178KY SY
ηW = , (17.19)
(s − 1)3
where KY is a constant depending on percolation, degree of damage, the condition
of reflection, but primarily depending on the type of armor block. The constant
0.178 is determined by adopting the values of N = 3000, D = 3, and P = 0.1. The
empirical constant KY = 1 for the armored slope of three-layered round stones.
July 30, 2009 11:4 9.75in x 6.5in b684-ch17 FA
450 D. H. Yoo
Fig. 17.5. Armor block weight ratio versus Ii and SY (laboratory data of van der Meer9 ).
Sealock is one of the most well-devised concrete armor blocks. It is widely used
in Japan and recently introduced to Korea. The stability of the block is well tested
in the laboratory experiments. The experiments were conducted in the channel of
1.5 m deep, 1.0 m wide, and 30 m long against the breakwater of slope S = 3/4.
The number of incident waves was 500. Waves are generated with four different
wave periods 1.3, 1.5, 1.8, 2.1, and 2.4 s, and six or seven different wave heights
for each wave period between 0.15 m and 0.24 m. The degree of damage (DS ) is
defined by the percentage of failure units, and it varies from 0.0 to the maximum
value 4.9. For the zero damage level Hudson parameter of Sealock was estimated
12.0 for T = 2.4 s, 14.7 for T = 2.1 s, 21.3 for T = 1.8 s, 22.0 for T = 1.5 s, and 17.8
and 28.5 for T = 1.3 s as shown in Fig. 17.6(b). The values of Hudson parameter
are found to become small for longer wave period, that is, longer waves are found
more damageable to the stability of armored breakwaters. The parameter KD of
the Hudson equation is found strongly dependent on the wave period.
From the results Hudson parameter KD is almost proportional to the degree of
damage but strongly dependent on the wave period as follows:
KD = 3.874DS + 23.31 (T = 1.3 ∼ 1.8 s), (17.20a)
KD = 3.529DS + 13.51 (T = 2.1 ∼ 2.4 s). (17.20b)
For the safety with a confidence Samsung Hydraulic Engineering suggested that KD
equals to 10.0 for the armor block Sealock.10
The set of the data is re-analyzed by relating them to the local second order
wave action slope SY . The results are demonstrated in Figs. 17.6(c) and 17.6(d).
The figure demonstrates that even the local second order wave action slope does
not seem to include sufficiently well the effect of wave period. There still exists
some deviation between the regression lines of different wave period. It may be
further refined by considering the effect of set-down and set-up for the computation
July 30, 2009 11:4 9.75in x 6.5in b684-ch17 FA
Fig. 17.6. Weight ratio of armor block Sealock versus S with KD and SY with KY .
of the local wavelength or wave celerity. The variation of the parameter KY is not
ignorable as shown in Fig. 17.6(d). Nevertheless, the variation of the parameter KY
becomes almost irrespective of wave period and it is given by:
KY = −0.010DS + 0.140. (17.21)
When the degree of damage defined by Samsung Hydraulic Engineering is zero,
KY for the armor unit Sealock is approximately 0.14. The values of the parameter
KY for various armor blocks are suggested in Table 17.4. The values of three-layered
rounded quarry stones and Sealocks are determined by analyzing the laboratory
data, and other values are deduced by relating it with the Hudson parameter KD .
More confident values should be determined by testing each relevant set of labo-
ratory data.
July 30, 2009 11:4 9.75in x 6.5in b684-ch17 FA
452 D. H. Yoo
KD KY
Nonbreaking Breaking (Nonbreaking)
17.6. Conclusions
In the present chapter, several new surf parameters are presented primarily using
the local values of wavelength as well as wave height. The first one has the similar
form to the Iribarren number, and it is called “second order Iribarren number.” The
second order Iribarren number is related to the wave steepness, while the original
(first order) Iribarren number is related to the square root of wave steepness. The
√ It is the product of the beach slope S and
second group is called “wave action slope.”
the Froude number defined by FH = C/ gH, where C is the wave celerity and H is
the wave height. The first order wave action slope is FH S(= SX ), the second order
wave action slope is FH2
S(= SY ). The first order wave action slope is closely related
to the first order(original) Iribarren number and the second order wave action slope
is closely related to the second order Iribarren number. It is, however, found that
the relations do not have simple linear proportionality.
Run-up height is related to the first order wave action slope, and unit weight
of armor block is related to the second order wave action slope. Both relations
have simple linear forms, and excellent agreements are found in comparison with
laboratory data. The effects of wave period or wavelength are found to be well
included in the new parameters. The use of the new parameters are considered to
be suitable particularly when the wave period is relatively big and the water depth
is relatively deep.
Acknowledgment
References
1. R. C. Iribarren, Generalization of the formula for calculation of rock fill dikes and ver-
ification of its coefficients (Revista de Obras Pulicas, Madrid, 1950 WES Translation
N.51-4).
2. D. H. Yoo and D. Y. Na, Hydraulic condition of breaking waves, J. Korean Soc. Civil
Eng. 25(2B), 159–163 (2005) (in Korean).
July 30, 2009 11:4 9.75in x 6.5in b684-ch17 FA
3. D. H. Yoo, S. K. Koo and I. H. Kim, Minimum weight of breakwater armor unit, Proc.
1st Asian and Pacific Coastal Engineering Conf., Dalian, China (2001), pp. 605–612.
4. T. J. P. Saville, Wave run-up on shore structures, Proc. ASCE, Paper No. 925
82(WW2), 359–373 (1956).
5. H. Mase, Random wave run-up height on gentle slope, J. Waterway, Port, Coastal,
and Ocean Eng. 115(5), 649–661 (1989).
6. T. S. Hedges and H. Mase, Modified Hunt’s equation incorporating wave setup,
J. Waterway, Port, Coastal and Ocean Eng. ASCE 130(3), 109–113 (2004).
7. J. P. Ahrens, Approximate Upper Limit of Irregular Waves Run-Up on Riprap,
Department of the US Army Corps of Engineers (1988).
8. R. V. Hudson, Laboratory investigation of rubble-mound breakwater, ASCE, Trans.,
Paper No. 3213, 126 (1961).
9. J. W. van der Meer, Rock slopes and gravel beaches under wave attack, Delft
Hydraulics Communication No. 336 (1988).
10. Samsung Hydraulic Engineering, Hydraulic experiments of Sealock, Report No. 11
(1978) (in Japanese).
July 30, 2009 11:4 9.75in x 6.5in b684-ch17 FA
Chapter 18
Shigeo Takahashi
Port and Airport Research Institute
3-1-1, Nagase, Yokosuka 239-0826, Japan
takahashi s@pari.go.jp
The experiences of caisson breakwater failures are explained and the current
design of caisson breakwaters is described to prevent such failures. The design
method of conventional caisson against wave forces is especially explained in this
chapter. The design method of caisson breakwaters is still under development
including a new design method of performance design. The future direction will
be discussed in the final section.
18.1. Introduction
455
August 5, 2009 14:18 9.75in x 6.5in b684-ch18 FA
456 S. Takahashi
to dissipate the wave energy and reduce the wave force, and is now frequently being
constructed especially in shallow surf zone in Japan.
Various new caisson breakwaters have been invented and commercialized in order
to mitigate the drawbacks associated with conventional composite breakwaters. One
of typical new caisson breakwater is the perforated wall caisson breakwater invented
by Jarlan2 in 1961. The caisson dissipates wave energy by front perforated wall
and wave chamber. Many breakwaters of this type were subsequently constructed
throughout the world with further improvement.
In this chapter, the experiences of caisson breakwater failures are explained and
the current design of caisson breakwaters is described to prevent such failures. The
design method of conventional caisson breakwaters against wave forces is especially
described in this chapter. The design method of caisson breakwaters is still devel-
oping including an integrated method of performance design. The future direction
will be discussed in the final section.
It is very important to learn from the actual failures. The design methods of the
breakwaters have been developed significantly from such experiences, and the devel-
opment of the design methods has prevented them from suffering failure. It should
be noted that the total length of breakwaters in Japan is more than 800 km and the
failures of the breakwaters are now very limited compared with the total amount
of caissons in the country.
Before explaining each failure, it should be noted that the most important cause
of failures of caisson breakwaters is the caisson sliding and breakage due to impulsive
wave pressures. It should also be noted that the rubble mound/rubble foundation
of composite breakwaters is vital to prevent the failure of the upright section by
scouring, as well as stabilizing the foundation against the wave force and caisson
weight.3
458 S. Takahashi
by another caisson which was designed to resist such severe wave pressures, but
unfortunately not in place before the storm.
Similar failure due to impulsive wave pressure occurs when breaking waves acting
on a caisson installed on a high/wide rubble foundation. Impulsive wave pressures
occur when the vertical wall is attacked by an almost vertical wave front, and
therefore larger vertical wave front due to plunging or surging breakers gives larger
impact pressures. Such caisson failures due to impulsive pressures had occurred
frequently in history but have recently been greatly reduced using accumulated
knowledge about impulsive wave pressures including impulsive pressure coefficient
as discussed in Sec. 18.3.2.
Scouring
Oblique waves caused a strong wave-induced current along the breakwater caissons,
although the wave height is not large.
An estimation method of the wave-induced current in front of the caisson due
to normal as well as oblique waves was already proposed and was included in the
current design.4
460 S. Takahashi
It should be noted that protecting the caisson is essential even though some part
of the rubble foundation is scoured. The rubble foundation is usually designed with
an enough length considering its deformation due to front erosion.
462 S. Takahashi
Use of an asphalt mat is very effective in reducing the necessary width of the
caisson, and has been widely applied in recent designs. Care must be taken to check
foundation stability, however.
Foot protection blocks are placed to prevent through-wash of the rubble foun-
dation and the sand below, while shape-designed concrete blocks are installed to act
as the armor layer of the rubble foundation. The water depth d above the rubble
mound including the armor layer is 10 m at LWL, and the berm width BM of the
rubble mound is 12.8 m. For scour protection, additional gravel is placed, being
called a “gravel mat.” A vinyl sheet is also used to prevent scouring of the sand
under the rubble foundation.
This breakwater is designed to withstand a wave of HD = 11.6 m (significant
wave height H1/3 = 6.66 m), significant wave period T1/3 = 13 s, and wave angle
θ of 22◦ at a water depth h of 21 m. The design significant wave in deepwater H0
is 12.2 m at a return period of 50 years. The design wave at the breakwater site
was evaluated using a wave transformation calculation, with wave pressures on the
caisson being evaluated by the extended Goda pressure formula.
464 S. Takahashi
466 S. Takahashi
mound, i.e.,
αI = αI0 αI1 , (18.10)
H/d when H ≤ 2d,
αI0 = (18.11)
2.0 when H > 2d,
and αI1 represents the effect of the mound shape (shown by the contour lines). This
term can be evaluated using
cos δ2 when δ2 ≤ 0,
αI1 = cosh δ1 (18.12)
0.5 −1
[(cosh δ1 )(cosh δ2 ) ] when δ2 > 0,
20δ11 when δ11 ≤ 0,
δ1 = (18.13)
15δ11 when δ11 > 0,
4.9δ22 when δ22 ≤ 0,
δ2 = (18.14)
3δ22 when δ22 > 0,
BM h−d
δ11 = 0.93 − 0.12 + 0.36 − 0.6 ,
L h
(18.15)
BM h−d
δ22 = −0.36 − 0.12 + 0.93 − 0.6 .
L h
The value of αI reaches a maximum of 2 at BM /L = 0.12, d/h = 0.4, and H/d > 2.
When d/h > 0.7, αI is always close to zero and is less than α2 . It should be noted
that the impulsive pressure significantly decreases when the angle of incidence θ is
oblique.
August 5, 2009 14:18 9.75in x 6.5in b684-ch18 FA
468 S. Takahashi
where the wave height fluctuates along the breakwater alignment and significantly
increases near the breakwater head. If waves exceeding the design wave height attack
the breakwater, and if the breakwater caissons slide, the shape of these caissons will
subsequently form a meandering shape.
stones or concrete blocks. Figure 18.11 illustrates the movement of the armor stones
in observed sections, where in the upper figure, the stones are scattered from the
slope of the mound when incident waves approach perpendicular to the breakwater
alignment. Note the occurrence of heavy scattering around the breakwater head.
In the middle figure where θ = 45◦ , the armor stones not only scatter away from the
slope, but in the flat berm near the caisson as well. The stones also scatter around
the breakwater head on its downstream side and especially at the caisson edge. In
the lower figure with θ = 60◦ , the stones on the slope do not move, though they
significantly scatter in the flat berm, especially near the caisson.
It is postulated that the stability of armor stones or blocks is mainly threatened
by the water particle velocity induced by the waves, i.e., by the drag and uplift forces
produced by the water particle velocity. In Fig. 18.11, the water particle velocity is
high where the stones moved, which supports this hypothesis.
Figure 18.12 shows a hodograph of the water particle velocity at the breakwater
trunk. When θ = 0, the water particle only moves perpendicular to the breakwater
alignment and the velocity is almost zero near the vertical walls, being largest at the
node of the clapotis (standing wave). As θ increases, the water velocity component
parallel to the breakwater alignment increases, with the velocity near the vertical
wall also significantly increasing.
Figure 18.13 shows the distribution of the peak water particle velocity around
the breakwater head, where it is revealed that a high velocity occurs around the
edge of the upright section.
August 5, 2009 14:18 9.75in x 6.5in b684-ch18 FA
470 S. Takahashi
Fig. 18.13. Water particle velocity distribution around the breakwater head.
August 5, 2009 14:18 9.75in x 6.5in b684-ch18 FA
1−κ h (1 − κ)2 h
Ns = max 1.8, 1.3 + 1.8 exp −1.5
κ1/3 H1/3 κ1/3 H1/3
(18.19)
in which
κ = κ1 (κ2 )B , (18.20)
2kh
κ1 = , (18.21)
sinh 2kh
(κ2 )B = max{αs sin2 θ cos2 (kBM cos θ), cos2 θ sin2 (kBM cos θ)}, (18.22)
August 5, 2009 14:18 9.75in x 6.5in b684-ch18 FA
472 S. Takahashi
Fig. 18.14. The calculated necessary weight of armor stones for rubble foundation.18
where max {a, b} denotes the maximum of a and b; h , the water depth above the
rubble mound foundation; L , the wavelength corresponding to the significant wave
period at the depth h ; k, the wave number (= 2πL ); and BM is the berm width as
shown in Fig. 18.14. αs is a correction factor obtained using wave tank experiments
and is 0.45.
Equation (18.19) is also extended to include the stability of the rubble mound
armor layer in the breakwater head. In the breakwater head, the term (κ2 )T is used
instead of (κ2 )B to represent the water particle velocity at the breakwater head,
where
κ = κ1 (κ2 )T , (18.23)
1
(κ2 )T = (αs τ 2 ), (18.24)
4
where τ expresses the ratio of the water particle velocity at the breakwater head
to that of the incident wave and is determined to be 1.4 for the wave angle less
than 45◦ .
Figure 18.14 shows sample calculations to determine the necessary weight of
armor stones. The shape of the breakwater is indicated and the necessary weight
August 5, 2009 14:18 9.75in x 6.5in b684-ch18 FA
for different wave conditions and water depths in front of the vertical wall are
calculated. When θ = 60◦ , the weight is the largest, whereas that for θ = 30◦ is the
smallest. It is obvious to see that the weight increases as the wave period increases
or the water depth in front of the wall decreases. When the weight is large, armor
concrete blocks must be used to cover the rubble mound.
474 S. Takahashi
caisson remains intact and the foot-protection blocks and armor blocks of the rubble
mound still function as designed. In actuality, one of the primary roles of the rubble
mound and the foot-protection blocks is to protect the caisson from such scouring.
The cause of toe area scouring is probably due to the strong wave-induced
velocity. These velocities are especially increased by reflected waves from the ver-
tical wall. However, the scouring mechanism is very complicated and has not yet
been properly explained. Generally, two types of toe scouring exist: local scouring
and large-scale sand movement.
Local scouring was investigated by Xie,21 Irie et al.,5 Oumeraci,22 and others.
Irie found the occurrence of N- and L-type scours, in which the former type scour is
due to suspension of relatively fine sand which causes accretion at the node, while
the later type is due to “bed load” of relatively coarse sand that causes erosion
at the node and accretion at the loop. L-type scour appears to be predominant in
proto type seas where erosion at the node is usually found.
Scour is usually inevitable for vertical breakwaters built on a sandy sea bottom.
However, scour is not a fatal problem due to the protection features provided by the
rubble mound. Nevertheless, scour protection should be included in breakwaters in
which severe scouring is expected. There are several scour protection methods, e.g.,
the use of gravel, geotextile, or asphalt mats. These methods can prevent scouring
to some extent, though no fully sufficient method has yet been realized.
Over the last 15 years improved awareness of wave impact induced failures3,23,24
has focused attention on the need to account for the dynamic response of maritime
structures to wave impact loads. Also, the attention was focused on the probabilistic
consideration in the design of breakwaters.25–28 The most relevant collaborative
research projects carried out in the recent past in the European Union on wave
loads on caisson breakwaters and seawalls include:
(1) PRObabilistic design tools for VERtical BreakwaterS, see Oumeraci et al.29 and
references therein;
August 5, 2009 14:18 9.75in x 6.5in b684-ch18 FA
476 S. Takahashi
(2) Violent Overtopping of Waves at Seawalls, see Cuomo30 and references therein;
(3) Breaking Wave IMpacts on step fronted COastal Structures, see Bullock et al.31
and references therein.
Recently, a new innovative design method is being developed for the design of
breakwater caissons which is called the performance design method. Performance
design can be considered as a design process that systematically and clearly defines
performance requirements and respective performance evaluation methods. In other
words, performance design allows the performance of a structure to be explicitly
and concretely described.32
A new performance design for the stability of breakwater caissons was being
developed which is called deformation-based reliability design.33–37 Sliding distance
is selected as the performance evaluation item and the probabilistic nature is fully
considered. Performance design requires a reliable performance evaluation method.
Thus, in deformation-based reliability design of a breakwater caisson, a calculation
method to determine the sliding distance due to wave actions is used, employing
Monte Carlo simulation to include the probabilistic nature of waves and response
of the breakwater caisson.
Since breakwaters are designed to withstand wave heights having a particular
return period, such as 50 years, a high probability exists that higher waves than the
design wave will attack them and there still remain some failures by heavy storms.
Recently, it becomes very important to evaluate the extent of damage/deformation
as well as the consequence including restoration and the total economic losses.
The performance design is just the design process to consider these items for more
reasonable and economical design of breakwaters. Also, the performance design is
becoming increasingly important for the design of coastal defenses against storm
surges and tsunami.38
References
1. S. Takahashi, Design of vertical breakwaters, Reference Document No. 34, Port and
Harbour Research Institute (1996) 85 pp.
2. G. E. Jarlan, A perforated vertical breakwater, The Dock and Harbour Authority
41(488), 394–398 (1961).
3. S. Takahashi, K. Shimosako, K. Kimura and K. Suzuki, Typical failures of com-
posite breakwaters in Japan, Proc. 27th Int. Conf. Coastal Eng., ASCE (2000),
pp. 1899–1910.
4. K. Kimura, S. Takahashi and K. Tanimoto, Stability of rubble mound foundations
of caisson breakwater under oblique wave attack, Proc. 24th Int. Conf. Coastal Eng.
Kobe, ASCE (1994).
5. I. Irie, Y. Kuriyama and H. Asakawa, Study on scour in front of breakwaters by
standing waves and protection method, Rep. Port. Harbour Res. Inst. 25(1), 3–86
(1986) (in Japanese).
6. K. Suzuki, S. Takahashi and Y. H. Kang, Experimental analysis of wave-induced
liquefaction in a fine sand bed, Proc. 26th Int. Conf. Coastal Eng. (1998),
pp. 3643–3654.
August 5, 2009 14:18 9.75in x 6.5in b684-ch18 FA
478 S. Takahashi
26. M. A. Losada and J. Llorca, Breakwaters under the action of sea oscillations. Spanish
recommendations ROM 1.1, Coastal Structures (2007).
27. J. W. van der Meer, Deterministic and probabilistic design of breakwater armor layers,
Proc. American Society of Civil Engineers, J. Waterways, Coastal Ocean Eng. 114(1),
66–80 (1988).
28. J. K. Vrijling, H. G. Vorrtman, H. F. Burcharth and J. D. Soren-sen, Design phi-
losophy for a vertical breakwa-ter, Proc. Coastal Structures ’99 (1999), pp. 631–635.
29. H. Oumeraci, A. Kortenhaus, N. W. H. Allsop, M. B. De Groot, R. S. Crouch,
J. K. Vrijling and H. G. Voortman, Probabilistic Design Tools for Vertical Breakwaters
(Balkema, Rotterdam, 2001), 392 pp.
30. G. Cuomo, Dynamics of wave induced loads and their effects on coastal structures,
PhD. dissertation, University of Roma TRE, Italy (2005).
31. G. N. Bullock, D. H. Peregrine, H. Bredmose, C. Obhrai, G. Walter and G. Muller,
Characteristics and design implication of breaking wave impacts, Proc. 29th Int. Conf.
Coastal Engineering (ASCE, 2004), pp. 3966–3978.
32. SEACO, Vision 2000-Performance-based seismic engineering of bridges (1995).
33. K. Shimosako and S. Takahashi, Application of deformation-based reliability design
for coastal structures, American Society of Civil Enginners (ASCE), Coastal Struc-
tures ’99 (1999), pp. 363–371.
34. T. Takayama, S. Ikesue and K. Shimosako, Effect of directional occurrence distri-
bution of extreme waves on composite breakwater reliability in sliding failure, Proc.
27th Int. Conf. Coastal Engineering (ASCE, 2000), pp. 1738–1750.
35. Y. Goda and H. Takagi, A reliability design method for caisson breakwaters with
optimal wave heights, Coastal Engineering 42(4), 357–387 (2000).
36. M. Hanzawa, N. Yamagata, T. Nishihara, Y. Umezawa, T. Takayama and
S. Takahashi, Performance of seawall against wave overtopping, American Society of
Civil Engineers, Proc. 29th Int. Conf. Coastal Eng. (2004), pp. 4314–4325.
37. J. A. Melby and N. Kobayasi, Damage progression on breakwaters, Proc. 26th Int.
Conf. Coastal Eng. (1998), pp. 1884–1897.
38. S. Takahashi, H. Kawai, T. Tomita and T. Takayama, Performance design concept
for storm surge defenses, American Society of Civil Engineers, Proc. 29th Int. Conf.
Coastal Eng. (2004), pp. 3074–3086.
August 4, 2009 9:33 9.75in x 6.5in b684-ch19 FA
Chapter 19
Krystian Pilarczyk
(Former) Hydraulic Engineering Institute
Public Works Department, Delft, The Netherlands
HYDROpil Consultancy, 23 Nesciohove, 2726BJ Zoetermeer, The Netherlands
krystian.pilarczyk@gmail.com
19.1. Introduction
There has been an increasing need in recent years for reliable information on design
methodology and stability criteria of revetments exposed to wave and current action.
The use of revetments, such as riprap, blocks and block mats, various mattresses,
and asphalt in civil engineering practice is very common. The granular filters,
and more recently the geotextiles, are more or less standard components of the
revetment structure.6,14,20,22,23,26,28−30 However, the proper design rules are not
always available.
Within the scope of the research on the stability of open slope revetments, much
knowledge has been developed about the stability of rock under wave and current
load and stability of placed (pitched) stone revetments under wave load.5,7,8 Until
recently, no or unsatisfactory design tools were available for a number of other
(open) types of revetment and for other stability aspects. This is why the design
methodology for placed block revetments has recently been extended in applicability
by means of a number of desk-studies for other (open) revetments:
479
August 4, 2009 9:33 9.75in x 6.5in b684-ch19 FA
480 K. Pilarczyk
• The downward flowing water will exert a drag force on the cover layer and the
decreasing phreatic level will coincide with a downward flow gradient in the filter
(or in a gabion). The first mechanism can be schematized by a free flow in the
filter or gabion with a typical gradient equaling the slope angle. It may result in
sliding.
• During maximum wave run-down there will be an incoming wave that a moment
later will cause a wave impact. Just before the impact, there is a “wall” of water
giving a high pressure under the point of maximum run-down. Above the run-
down point the surface of the revetment is almost dry and therefore there is a
low pressure on the structure. The high pressure front will lead to an upward
flow in the filter or a gabion. This flow will meet the downward flow in the run-
down region. The result is an outward flow and uplift pressure near the point of
maximum wave run-down (Fig. 19.1).
The schematized situation can be quantified on the basis of the Laplace equation
for linear flow. After complicated calculations, the uplift pressure in the filter or a
gabion can be derived.9,26 The uplift pressure is dependent on the steepness and
height of the pressure front on the cover layer (which is dependent on the wave
height, period, and slope angle), the thickness of the cover layer, and the level of
the phreatic line in the filter or a gabion. In case of riprap or gabions, it is not
dependent on the permeability of the cover layer, if the permeability is much larger
than the sublayer and/or subsoil. For semi-permeable cover layers the equilibrium
of uplift forces and gravity forces (defined by components of a revetment) leads to
the following (approximate) design formula26 :
0.67
Hscr D bDk
=f with Λ = (19.1a)
∆D Λξop k
August 4, 2009 9:33 9.75in x 6.5in b684-ch19 FA
or
0.33
bk Hscr D k −0.67
Λ/D = and =f ξop (19.1b)
Dk ∆D b k
or
Hscr −0.67 Hs
= F ξop with maximum = 8.0 and ctgα ≥ 2, (19.1c)
∆D ∆D cr
where √ Hscr = significant wave height at which blocks will be lifted out (m); ξop =
tan α/ (Hs /(1.56Tp2)) = breaker parameter; Tp = wave period at the peak of the
spectrum (s); Λ = leakage length (m), ∆ = (ρs − ρw )/ρw = relative volumetric
mass of cover layer, with: ρs = density of the protection material and ρw = density
of water (kg/m3 ); b = thickness of a sublayer/filter (m), D = thickness of a top
(cover) layer (m), k = permeability of a sublayer (m/s), k = permeability of a top
layer (m/s), f = stability coefficient, mainly dependent on structure type, tan α
and friction; F = total (black-box) stability factor. The leakage length (Λ) and
stability coefficient (F ) are explained more in detail for specific applications in the
next sections.
There are two practical design methods available: the analytical model based on
Eq. (19.1a) and the black-box model based on Eq. (19.1c). In both cases, the final
form of the design method can be presented as a critical relation of the load (H)
compared to strength (∆D), depending on the type of wave attack. More general
form (also applicable for riprap and ctgα ≥ 1.5) is provided by Pilarczyk24,26 as:
Hs F cos α tan α
= , where ξop = (19.2)
∆D cr ξop
b
Hs /Lop
482 K. Pilarczyk
(–), Lop = gTp2 /(2π), deep-water wavelength at the peak period (m), and Tp =
wave period at the peak of the spectrum (s) and b = exponent; 0.5 ≤ b ≤ 1.0. For
porous top layers, such as sand mattresses and gabions, the relative density of the
top layer must be determined, including the water-filled pores:
ρs − ρw
∆= and ∆t = (1 − n) · ∆ (19.3)
ρw
in which ∆t = relative density including pores and n = porosity of the top layer
material. D and ∆ are defined for specific systems such as:
• for rock: D = Dn = (M50 /ρs )1/3 (= nominal diameter) and ∆t = ∆ =
(ρs − ρw )/ρw ,
• for blocks: D = thickness of block and ∆t = ∆,
• for mattresses: D = d = average thickness of mattress and ∆t = (1 − n)∆,
where n = bulk porosity of fill material and ∆ = relative density of fill material.
For sand and common quarry stone (1 − n) ∆ ∼ 1.
The approximate values of stability factor F are: F = 2.25 for riprap, F = 2.5
for pitched stone of irregular shape and placed geobags, F = 3.0–3.5 for pitched
basalt, F = 4.0 for cement geomattresses, 3.5 ≤ F ≤ 5.5 for block revetments
(4.5 as an average/usual value), 4.0 ≤ F ≤ 6.0 for block mats (higher value for
cabled systems), 6.0 ≤ F ≤ 8.0 for gabions, and 6.0 ≤ F ≤ 10 for (asphalt or
concrete) slabs. Exponent b refers to the type of wave–slope interaction and its value
is influenced by the roughness and the porosity of a revetment. The following values
of exponent b are recommended: b = 0.5 for permeable cover layers (i.e., riprap,
gabions, pattern grouted riprap, open block mats, and geobags), b = 2/3 for semi-
permeable cover layers (i.e., pitched stone and placed blocks, block mats, concrete-
or sand-filled geomattresses, and 2/3 < b ≤ 1.0 for slabs).
The advantage of this black-box design formula is its simplicity. The disad-
vantage, however, is that the value of F is known only very roughly for many types
of structures.
The analytical model is based on the theory for placed stone revetments on a
granular filter (pitched blocks). In this calculation model, a large number of physical
aspects are taken into account. In short, in the analytical model, nearly all physical
parameters that √ are relevant to the stability have been incorporated in the “leakage
length”: Λ = (bDk/k ). The final result of the analytical model may, for that
matter, again be presented as a relation such as Eq. (19.1) where F = f (Λ).
A system without a filter layer is (directly on sand or clay and geotextile) not
the permeability of the filter layer, but the permeability of the subsoil (eventually
with gullies/surface channels) is filled in. For the thickness of the filter layer, it is
examined to which depth changes at the surface affect the subsoil. One can fill in
0.5 m for sand and 0.05 m for clay. The values for D and ∆ depend on the type of
revetment.
In the case of a geotextile situated directly under the cover layer, the perme-
ability of the cover layer decreases drastically. Since the geotextile is pressed against
the cover layer by the outflowing water, it should be treated as a part of the cover
layer. The water flow trough the cover layer is concentrated at the joints between the
August 4, 2009 9:33 9.75in x 6.5in b684-ch19 FA
blocks, reaching very high flow velocities and resulting in a large pressure head over
the geotextile. The presence of a geotextile may reduce k by a factor 10 or more.
To be able to apply the design method for placed stone revetments under wave
load to other systems, the following items may be adapted:
the revetment parameter F ;
the (representative) strength parameters ∆ and D;
the design wave height Hs ;
the (representative) leakage length Λ;
the increase factor Γ (friction/interlocking between blocks) on the strength.
Only such-like adaptations are presented in this summarizing review. The basic
formulae of the analytical model are not repeated here. For these, the reader is
referred to Refs. 9 and 26.
The wave attack on a slope can be roughly transformed into the maximum
velocity component on a slope during run-up and run-down, Umax , by using the
following formula:
Umax = p gHs ξop (19.4)
484 K. Pilarczyk
Stability parameter Φ
The stability parameter Φ depends on the application. Some guide values are:
Turbulence factor K T
The degree of turbulence can be taken into account with the turbulence factor KT .
Some guide values for KT are:
• Normal turbulence of rivers: KT ≈ 1.0
• Increased turbulence:
river bends: KT ≈ 1.5
downstream of stilling basins: KT ≈ 1.5
• Heavy turbulence
hydraulic jumps: KT ≈ 2.0
strong local disturbances: KT ≈ 2.0
sharp bends: KT ≈ 2.0 to 2.5
• Load due to water (screw) jet: KT ≈ 3.0 to 4.0
Depth parameter Kh
With the depth parameter Kh , the water depth is taken into account, which is nec-
essary to translate the depth-averaged flow velocity into the flow velocity just above
the revetment. The depth parameter also depends on the development of the flow
profile and the roughness of the revetment. The fully developed profile can be found
in natural rivers. In case of local civil engineering works as bottom protection or
slope protection, the nondeveloped profile is usually present. The following formulae
August 4, 2009 9:33 9.75in x 6.5in b684-ch19 FA
486 K. Pilarczyk
Slope parameter Ks
The stability of revetment elements also depends on the slope gradient under which
the revetment is applied, in relation to the angle of internal friction of the revetment.
This effect on the stability is taken into account with the slope parameter Ks , which
is defined as follows:
2 2
sin α tan α
Ks = 1 − = cos α 1 − (19.8a)
sin θ tan θ
or
Ks = cos αb , (19.8b)
where θ = angle of internal friction of the revetment material, α = transversal slope
of the bank (◦ ), and αb = slope angle of river bottom (parallel along flow direction)
(◦ ). The following values of θ can be assumed as a first approximation: 40◦ for
riprap, 30◦ –40◦ for sand-filled systems, and 90◦ for stiff and anchored mortar-filled
mattresses and (cabled) block mats (Ks = cos α). However, for flexible nonanchored
mattresses and block mats (units without contact with the neighboring units) this
value is much lower, usually about 3/4 of the friction angle of the sublayer. In case
of geotextile mattress and block mats connected to geotextile lying on a geotextile
filter, θ is about 15◦ –20◦.
Pilarczyk’s formula was verified using Neill’s data (see Fig. 19.3). The advantage
of this general design formula is that it can be applied in numerous situations.
The disadvantage is that the scatter in results, as a result of the large margin in
parameters, can be rather wide.
With a downward flow along a steep slope and small h/ks values, it is difficult
to determine or predict the flow velocity because the flow is very irregular. In such
case formulae based on the discharge are developed.17,26
Fig. 19.3. Comparison of Pilarczyk’s equation with the experimental data collected by Neill.19
August 4, 2009 9:33 9.75in x 6.5in b684-ch19 FA
It should be noted that Pilarczyk formula can be transformed into similar form
as Maynord formula, which is more or less a standard formula in the USA.
Maynord formula4,18 :
1/2 2.5
2.5
γw U U
D30 = Sf Cs Cv CT h ⇒ CM · h
γs − γw (Ks gh) (Ks g∆h)
(19.9)
where D30 = characteristic riprap size of which 30% is finer by weight, Sf = safety
factor (minimum = 1.1), Cs = stability coefficient for incipient failure: = 0.30 for
angular rock and = 0.375 for rounded rock, Cv = velocity distribution coefficient: =
1.0 for straight channels, 1 < Cv < 1.28 for river bends = 1.25 downstream of
concrete channels, CT = blanket thickness coefficient (CT = 1 for blanket thickness
t = 1.5D50 ), Ks = side slope correction factor, and h = local depth (use depth at
20% upslope from toe for side slopes). D30 can be translated (approximately) into
D50 using gradation ratio of rock (D85 /D15 ) as: D50 = D30 (D85 /D15 )1/3 . At the
first approximation one may assume: D30 = 0.75D50 ≈ 0.90Dn50. The minimum
value of the total stability coefficent CM is about 0.3.
Transformation Pilarczyk formula:
1.25
2.5
2.5
0.035ΦKT U U
Dn50 = h = CP · h .
2Ψcr (Ks g∆h) (Ks g∆h)
(19.10)
The original minimum value of CP is about 0.3; however, for practical applications
CP = 0.4 is recommended. Equation (19.10) has the same structure as Eq. (19.9)
by Maynord. Comparison of the total stability coefficients CP and CM provides
a similar result when using comparable input data and taking into account the
difference between Dn50 and D30 .
488 K. Pilarczyk
Fig. 19.4. Schematized development of S-profile and possible local sliding in sand.
Fig. 19.5. Geotechnical stability; design diagram for block mats and wave steepness Hs /
Lop = 0.03.
19.2.4. Filters
Granular and/or geotextile filters can protect structures subjected to soil erosion
when used in conjunction with revetment armor such as riprap, blocks and block
mats, gabions and mattresses, asphalt or concrete slabs, or any other conventional
armor material used for erosion control.11,14,21,22
However, there is still a misunderstanding about the function of geotextiles in
the total design of these structures, especially in comparison with the granular
filters. In this section, the general principles of designing revetments incorporating
granular or geotextiles are reviewed. Attention is paid to the replacing of a granular
filter by a geotextile, which may often lead to geotechnical instability. Furthermore,
it appears that a thicker granular filter gives larger geotechnical stability, but lower
cover layer stability (uplift of blocks). The conclusion is therefore that the wave
loads must be distributed (balanced) adequately over the sand (shear stress) and
the cover layer (uplift pressure).
Filters have two functions: erosion prevention and drainage. Traditional design
criteria for filters are that they should be “geometrically tight” and that the filter
permeability should be larger than the base (soil) permeability. However, it results
in a large number of layers which are often unnecessary, uneconomical, and difficult
to realize. In several cases, a more economical filter design can be realized using the
concept of “geometrically open filters” (e.g., when the hydraulic loads/gradients are
too small to initiate erosion). Recently, some criteria for “geometrically open” filters
including geotextiles were developed (and are still under further development).
However, the application of these criteria requires the knowledge/prediction of the
hydraulic loads.
In the cases when the erosion exceeds an acceptable level, a filter construction is
a proper measure for solving this problem. In revetment structures, geotextiles are
mostly used to protect the subsoil from washing away by the hydraulic loads, such
as waves and currents. Here, the geotextile replaces a granular filter. Unfortunately,
the mere replacing of a granular filter by a geotextile can endanger the stability of
other components in the bank protection structure. The present section shows that
designing a structure is more than just a proper choice of geotextile.
Filter structures can be realized by using granular materials (i.e., crushed stone),
bonded materials (i.e., sand asphalt, sand cement), and geotextiles, or a combi-
nation of these materials. The choice between the granular filter, a bonded filter,
or geotextile depends on a number of factors. In general, a geotextile is applied
because of easier placement and relatively lower cost. For example, the placement
of granular filter underwater is usually a serious problem; the quality control is very
difficult, especially when placement of thin layers is required.
When designing with geotextiles in filtration applications, the basic concepts
are essentially the same as when designing with granular filters. The geotextile
must allow the free passage of water (permeability function) whilst preventing the
erosion and migration of soil particles into the armor or drainage system (retention
function). In principle, the geotextile must always remain more permeable than
the base soil and must have pore sizes small enough to prevent the migration of
the larger particles of the base soil. Moreover, concerning the permeability, not only
August 4, 2009 9:33 9.75in x 6.5in b684-ch19 FA
490 K. Pilarczyk
the opening size but also the number of openings per unit area (Percent Open Area)
is of importance.27
It has to be stressed that geotextiles cannot always replace the granular filter
completely (see Sec. 19.2.3). A granular layer can often be needed to reduce (damp)
the hydraulic loadings (internal gradients) to an acceptable level at the soil interface.
After that, a geotextile can be applied to fulfill the filtration function.
In respect to the filters for erosion control (granular or geotextile), the distinction
can be made between:
Only geometrically tight filters are discussed. For other type of filters, the reader is
guided to Refs. 6, 26, and 27.
Df 15
≤ 4 to 5, (19.11)
Db85
where Df 15 is the grain size of the filter layer (or cover layer) which is exceeded by
15% of the material by weight in m; Db85 is the grain size of the base material (soil)
which is exceeded by 85% of the material by weight in m.
• Internal stability
In the case of very “wide” gradation, the situation requires an additional check with
respect to the internal migration. In this respect, an important parameter is the
so-called “uniformity coefficient” Cu , defined by Eq. (19.13) and the shape of the
sieve curve:
Db60
Cu = , (19.13)
Db10
where Cu is the coefficient of uniformity. Usually, if this ratio is smaller than 6
(to 10), the soil or filter structure is considered internally stable. However, internal
stability can be judged more carefully by the following rules [Eq. (19.14)]:
• Permeability criterion
Df 15
> 5. (19.15)
Db15
492 K. Pilarczyk
in the sand tightness criterion for a geotextile in strong cyclic loading situations.
O98 is also referred to as Omax .
Of filtration opening size (FOS). Of is comparable with O95 (hydrodynamic
sieve method).
AOS is the apparent opening size (acc. to ASTM method), also called EOS
(effective opening size). The AOS is determined by sieving spherical glass particles
of known size through a geotextile. The AOS, also frequently referred to as O95 (dry
sieve method), is defined as a standard sieve size, x, mm, for which 5% or less of
the glass particles pass through the geotextile after a specified period of sieving.
Dw effective opening size which corresponds to the sand diameter of the fraction
of which 10%, determined by the wet sieve method, passes through the geotextile.
Dw is comparable with O95 .
The transport of soil particles within a grain structure is possible when there is
enough space and a driving force (groundwater pressure, hydraulic gradients within
the soil). In most cases, it is the intention to prevent the transport of small-sized
soil particles in the subsoil and therefore the term soil tightness is used and not the
term space for transport or pore volume (in the case of the transport of water the
terms pore volume and water permeability are used). The relation between pore
magnitude and grain diameter can be characterized by: pore diameter ≈ 20% of
the grain diameter. Just as for the characterization of the performance of a grain
structure with regard to the transport of soil particles, for geosynthetics, too, the
term soil tightness is used.
As was mentioned before (Fig. 19.6), in a theoretical case when the soil is com-
posed of spheres of one-size diameter, all spheres can be retained if all apertures
in the geosynthetic are smaller than the diameter of the spheres. Usually, the soil
consists of particles with different diameters and shapes, which is reflected in the
particle-size distribution curves. Smaller particles can disappear straight across the
geosynthetic by groundwater current. In this case, the retained soil structure can
function as a natural filter; see Fig. 19.7. The better the soil particles are distributed,
the better the soil tightness of the soil structure is effected. Smaller soil particles get
stuck into the spaces between larger ones and the soil structure prevents the flow
of fine particles. When certain particle-size fractions are lacking, the soil structure
is not stacked very well and cavities develop through which erosion can occur. The
displacement of soil particles not only depends on the soil tightness but also on
the hydraulic gradient in the soil structure. Moreover, the dynamic effects due to
heavy wave loading may not allow the forming of a natural filter, and the process
of washing-out may continue.
According to some researchers the forming of a natural filter is only possible
for stationary flow.6 However, for small values of the hydraulic gradients, this is
also possible for nonstationary flow. For heavy wave attack (i.e., exposed break-
waters, coastal revetments), this is usually not the case. In extreme situations, soil
liquefaction is even possible. In such situations the soil particles can still reach the
surface of a geotextile and be washed out.
In order to judge the risk of wash-out of soil particles through the geosynthetics,
some aspects have to be considered. An important factor is the internal stability
of the soil structure. In the case of a loose particle stacking of the soil, many small
soil particles may pass through the geosynthetic before a stable soil structure is
developed near the geosynthetic. Also, a proper compaction of soil is very important
for the internal stability of soil. The internal stability is defined by the uniformity
coefficient Cu [see Eq. (19.13)]. In case of vibration, for instance, caused by waves or
by traffic, the stable soil structures can be disturbed. To avoid such situations, the
subsoil has to be compacted in advance and a good junction between geosynthetic
and subsoil has to be guaranteed and possibly, a smaller opening of geotextile must
be chosen.
The shape of the sieve curve also influences the forming of a natural filter.
Especially, when Cu > 6, the shape of the base gradation curve and its internal
stability must be taken into account.27 For a self-filtering linearly graded soil, the
representative size corresponds to the average grain size, Db50 . For a self-filtering
gap graded soil, this size is equal to the lower size of the gap. For internally unstable
soils, this size would be equivalent to Db30 in order to optimize the functioning of the
filter system. It is assumed that the involved bridging process would not retrogress
beyond some limited distance from the interface.
Soil tightness
With respect to the soil tightness of geotextiles, many criteria for geometric soil
tightness have been developed and published in the past.2,6,10,12,15,20,27 An example
of such design criteria, based on Dutch experience, is presented in Table 19.1. An
additional requirement is that the soil should be internally stable. The internal
stability of a grain structure is expressed in the ratio between Db60 and Db10 . As a
rule, this value has to be smaller than 10 to guarantee sufficient stability.
However, in many situations additional requirements will be necessary,
depending on the local situation. Therefore, for design of geometrically tight
geotextiles the method applied in Germany can be recommended.2,13,20,27 In this
method a distinction is made between the so-called stable and unstable soils.
August 4, 2009 9:33 9.75in x 6.5in b684-ch19 FA
494 K. Pilarczyk
Table 19.1. Design requirements for geosynthetics with a filter and separation function.
Soil type
If Ip is unknown at the preliminary design stage, then the soil may be regarded
as a problem soil if the clay size fraction is less than 50% of the silt size fraction.
The design criteria are presented in Table 19.2.27 In the case of fine sand or
silty subsoils, however, it can be very difficult to meet these requirements. A more
advanced requirement is based on hydrodynamic sand tightness, viz. that the flow
is not capable of washing out the subsoil material, because of the minor hydrody-
namical forces exerted (although the apertures of the geotextile are much larger
than the subsoil grains).
the case of a riverbank protection, it means that the permeability of the geotextile
has to be larger than the permeability of the soil on which the geotextile has to
be applied. In the case of a dike slope or dike foundation, the geosynthetic is often
applied on an impermeable layer of clay. Proper permeability of geotextiles is very
important in respect to the stability of relatively less permeable cover layers as, for
example, block and block mats. When a geotextile lies directly under the cover layer
it considerably reduces the open area of the cover layer, and as a result the uplift
forces increase (see example in Sec. 19.3.2). The water permeability of woven fabrics
and nonwovens may decrease in the course of time owing to the fact that fine soil
particles, which are transported by the groundwater flow from the subsoil, block the
openings in the geotextile, or migrate into the pores of the geosynthetic (clogging).
To prevent mineral clogging, the pore size of the geotextiles has to be chosen
as large as possible; but, of course, this pore size has still to meet the requirements
for soil tightness. The danger of clogging increases when the soil contains more
than 20% of silt or in the case of gap-grading of a soil. On the other hand, there
usually is no danger of clogging when the total hydraulic gradient (over the subsoil
and geotextile together) is less than 3, or when the subsoil is well graded. In all
situations it holds that the soil must be internally stable. For less critical situations
no clogging can be expected if:
496 K. Pilarczyk
a pitched block revetment are placed tightly together in a smooth pattern. This
ensures that external forces such as waves and currents can exert little drag on the
blocks and also that blocks support each other without any loss of flexibility when
there are local subsoil irregularities or settlement.
A (concrete) block mat is a slope revetment made of (concrete) blocks that are
joined together to form a “mat”; see Fig. 19.8. The interconnection may consist of
cables from block to block, of hooks connecting the blocks, or of a geotextile on
which the blocks are attached with pins, glue, or other means. The spaces between
the blocks are usually filled with rubble, gravel, or slag.
The major advantage of block mats is that they can to be laid quickly and
efficiently and partly underwater. Block mats are more stable than a setting of
loose blocks, because a single stone cannot be moved in the direction perpendicular
to the slope without moving other nearby stones. It is essential to demand that
already with a small movement of an individual stone a significant interactive force
with the surrounding stones is mobilized. Large movements of individual blocks are
not acceptable, because transport of filter material may occur. After some time,
this leads to a serious deformation of the surface of the slope. The block mats are
vulnerable at edges and corners. If two adjacent mats are not joined together, then
the stability is hardly larger than that of pitched loose stones.
498 K. Pilarczyk
than with a small leakage length (effect of kf /k in the leakage length formula). The
effect of the leakage length on the dimensions of the critical wave for semi-permeable
revetments is apparent from the following equations:
0.67 0.33
Hscr D Hscr D k −0.67 Hscr −0.67
=f =f ξop = = F ξop , (19.17)
∆D Λξop ∆D b k ∆D
where Hscr = √ significant wave height at which blocks will be lifted out (m);
ξop = tan α/ (Hs /(1.56Tp2)) = breaker parameter; Tp = wave period (s); ∆ =
relative volumetric mass of cover layer = (ρs − ρ)/ρ; f = stability coefficient mainly
dependent on structure type and with minor influence of ∆, tan α, and friction, and
F = total (black-box) stability factor.
These equations indicate the general trends and have been used together with
measured data to set up the general calculation model.9,26 This method works
properly for placed/pitched block revetments and block mats within the following
range: 0.01 < k kf < 1 and 0.1 < D/bf < 10. Moreover, when D/Λ > 1 use
D/Λ = 1, and when D/Λ < 0.01 use D/Λ = 0.01. The range of the stability coeffi-
cient is: 5 < f < 15; the higher values refer to the presence of high friction among
blocks or interlocking systems. The following values are recommended for block
revetments:
f = 5 for static stability of loose blocks (no friction between the blocks),
f = 7.5 for static stability of a system (with friction between the units),
f = 10 for tolerable/acceptable movement of a system at design conditions.
• If the slope angle is reduced from 1:3 to 1:4 (tan α from 0.33 to 0.25), Hscr is
increased by about 20% (due to the breaker parameter, ξop );
• An increase of 20% in the thickness of the cover layer, D, Hscr increases by about
27%;
• A 30% reduction in the leakage length, Λ, Hscr increases by about 20%. This can
generally be achieved by halving the thickness of the filter layer or by doubling
the k /kf value. The latter can be achieved by approximation by:
— reducing the grain size of the filter by about 50%, or
— by doubling the number of holes in (between) the blocks, or
— by making hole sizes 1.5 times larger, or
— by doubling joint width between blocks.
Example: In 1983, the Armorflex mat on a slope 1:3 was tested on prototype
scale at the Oregon State University: closed blocks with thickness D = 0.12 m and
open area 10% on two types of geotextiles and very wide-graded subsoil (d15 =
0.27 mm, d85 = 7 mm). In the case of a sand-tight geotextile, the critical wave height
(instability of mat) was only Hscr = 0.30 m. In the case of an open net geotextile
(opening size about 1 mm), the critical wave height was more than 0.75 m (maximum
capacity of the wave flume). The second geotextile was 20 times more permeable
than the first one. This means that the stability increased by factor 200.33 = 2.7.
In most cases, the permeability of the cover layer and sublayer(s) are not exactly
known. However, based on the physical principles as described above, the practical
“black-box” method has been established where parameter Λ and coefficient “f ”
are combined to one stability factor “F ”. F depends on the type of structure,
characterized by the product of the ratios of k /kf and D/bf . With the permeability
formulae9 it is concluded that the parameter (k /kf ) × (D/bf ) ranges between 0.01
and 10, leading to a subdivision into three ranges of one decade each. Therefore,
the following types are defined:
(a) Low stability: (k /kf )(D/bf ) < 0.05–0.1,
(b) Normal stability: 0.05–0.1 < (k /kf )(D/bf ) < 0.5–1,
(c) High stability: (k /kf )(D/bf ) > 0.5–1.
For a cover layer lying on a geotextile on sand or clay, without a granular filter,
the leakage length cannot be determined because the size of bf and k cannot be
calculated. The physical description of the flow is different for this type of structure.
For these structures there is no such a theory as for the blocks on a granular filter.
However, it has been experimentally proved that Eq. (19.1) or (19.17) are also valid
for these structures.
It can be concluded that the theory has led to a simple stability formula
[Eq. (19.17)] and a subdivision into four types of (block) revetment structures:
• (a1) cover layer on granular filter possibly including geotextile, low stability;
• (a2) cover layer on granular filter possibly including geotextile, normal stability;
• (a3) cover layer on geotextile on sand;
• (a4) cover layer on clay or on geotextile on clay.
August 4, 2009 9:33 9.75in x 6.5in b684-ch19 FA
500 K. Pilarczyk
Fig. 19.11. Example of stability function for type a1.2 (loose closed blocks on granular filter);
low stability.
for loose blocks in order to avoid the mechanical abrasion of geotextiles by moving
blocks. The lower boundary of stability of cabled mats can be increased by a factor
of 1.25 (or 1.5, if additionally washed-in with granular material) in comparison with
loose blocks. Such an increase of stability is only allowable when special measures are
taken with respect to the proper connection between the mats. The upper boundary
of stability (F = 8) remains the same for all systems. Application of this higher
stability requires optimization of design (including application of geometrically open
but stable filters and geotextiles).6,9
To be able to apply the design method for placed stone revetments under wave
load to other semi-permeable systems, the following items may be adapted: the
revetment parameter F , the (representative) strength parameters ∆ and D, the
August 4, 2009 9:33 9.75in x 6.5in b684-ch19 FA
502 K. Pilarczyk
design wave height Hs , and the (representative) leakage length Λ. Table 19.4 gives
an overview of usable values for the revetment stability parameter F in the black-
box model for linked blocks (block mats).
The terms “favorable,” “normal,” and “unfavorable” refer to the composition
of the granular filter and the permeability ratio of the top layer and the filter
layer.9 In the case of fine granular filter and relatively permeable top layer, the total
composition can be defined as “favorable.” In a case of very coarse granular layer and
less permeable top layer, the composition can be defined as “unfavorable.” In a case
of blocks connected to a geotextile and concrete-filled mattresses on a filter layer, the
construction can be usually defined as between “unfavorable” and “normal,” and the
stability factor F = 3.0–3.5 (max. 4.0) can be applied. For blockmats and permeable
mattresses on sand F = 5 (max. 6.0) can be applied. The higher values can also be
used in cases that the extreme design loading is not very frequent or when the system
is (repeatedly) washed in by coarse material providing additional interlocking. This
wide range of recommended values for F only gives a first indication of a suitable
choice. Furthermore, it is essential to check the geotechnical stability with the design
diagrams (see Fig. 19.5 and for a full set of diagrams see Refs. 26 and 27).
The permeability of the mattress is one of the factors that determine the stability.
It is found that the permeability given by the suppliers is often the permeability
of the geotextile, or of the so-called Filter Points (Fig. 19.13). In both cases, the
permeability of the whole mattress is much smaller. A high permeability of the
mattress ensures that any possible pressure buildup under the mattress can flow
away, as a result of which the uplift pressures across the mattress remain smaller.
In general, with a subsoil of clay and silty sand, the permeability of the mattress
will be higher than the permeability of the subsoil. Therefore, the water under
the mattress can usually be discharged without excessive lifting pressures on the
mattress. The permeability of the mattress will be lower than the permeability of
the subsoil or sublayers if a granular filter is applied, or with sand or clay subsoil
having an irregular surface (gullies/cavities between the soil and the mattress). This
will result in excessive lifting pressures on the mattress during wave attack.
• First, cavities will form under the mattress as a result of uneven subsidence of
the subsoil. The mattress is rigid and spans the cavities.
August 4, 2009 9:33 9.75in x 6.5in b684-ch19 FA
504 K. Pilarczyk
• With large spans, wave impacts may cause the concrete to crack and the spans to
collapse. This results in a mattress consisting of concrete slabs which are coupled
by means of the geotextile.
• With sufficiently high waves, an upward pressure difference over the mattress will
occur during wave run-down, which lifts the mattress (Fig. 19.1).
• The pumping action of these movements will cause the subsoil to migrate, as a
result of which an S-profile will form and the revetment will collapse completely.
It is assumed that local settlement of the subsoil will lead to free spans of the
concrete mattress. Then, the wave impact can cause the breaking of these spans, if
the ratio of Hs /D is too large for a certain span length. A calculation method is
derived on the basis of an empirical formula for the maximum wave impact pressure
and the theory of simply supported beams. The collapsing of small spans (less than
1 or 2 m) is not acceptable, since these will lead to too many cracks.
The empirical formula for the wave impact is16 :
Fimpact
= 7.2Hs2 tan α, (19.18)
ρg
Fig. 19.14. Calculation results for concrete mattresses (Hs /∆D < 4 because of acceptable crack
distance due to impacts on spans).
Taking into consideration the above failure mechanisms, the following design
(stability) formula has been derived for the mattresses [similar to Eq. (19.1c)]:
Hs F Hs
= 2/3 with = 4, (19.19)
∆D ξop ∆D max
where
D = mass per m2 /ρs (which can be called Deffective or Daverage ),
∆ = relative volumetric mass of the mattress (−) = (ρs − ρw )/ρw ,
ρs = volumetric mass of concrete (kg/m3 ),
F = stability factor (see below).
For an exact determination of the leakage length, one is referred to the analytical
model.9,16 However, besides the mattresses of a type as, for example, the tube mat
(Crib) with relative large permeable areas, the other types are not very sensitive to
the exact value of the leakage length. It can be recommended to use the following
values of F in design calculations:
F = 2.5 or (≤ 3) — for low-permeable mattresses on (fine) granular filter,
F = 3.5 or (≤ 4) — for low-permeable mattress on compacted sand,
F = 4.0 or (≤ 5) — for permeable mattress on sand or fine filter (Df 15 < 2 mm).
The higher values can be applied for temporary applications or when the soil is
more resistant to erosion (i.e., clay), and the mattresses are properly anchored.
August 4, 2009 9:33 9.75in x 6.5in b684-ch19 FA
506 K. Pilarczyk
19.5.1. Introduction
Gabions are made of rectangular baskets of wire mesh, which are filled with stones.
The idea of the protection system is to hold the rather small stones together with
the wire mesh. Waves and currents would have easily washed away the small stones,
but the wire mesh prevents this. A typical length of gabions is 3–4 m, a width of
1–3 m, and a thickness of 0.3–1 m. The gabions with small thickness (less then 0.5 m)
and large length and width are usually called Reno-mattresses or stone-mattresses.
An important problem of this protection system is the durability. Frequent wave
or current attack can lead to a failure of the wire mesh because of the continuously
moving grains along the wires, finally cutting through. Another problem is the
corrosion of the mesh. Therefore, meshes with plastic coating or corrosion resistant
steel are used. On the other hand, the system is less suitable where waves and
currents frequently lead to grain motion.
Hs −2/3
= F · ξop with 6 < F < 9 for slope of 1:3 (tan α = 0.33) (19.20a)
∆D
where
Hs = significant wave height of incoming waves at the toe of the structure (m),
∆ = relative density of the gabions (usually: ∆ = ∆t ≈ 1),
D = thickness of the gabion (m),
F = stability factor, √
ξop = breaker parameter = tan α/ (Hs /(1.56Tp2)
Tp = wave period at the peak of the spectrum (s).
August 4, 2009 9:33 9.75in x 6.5in b684-ch19 FA
508 K. Pilarczyk
Fig. 19.15. Summary of test results (Ashe1 ) and (Brown3 ) and upper-limit design curve.
Hs F
= , (19.21)
∆f D f ξop
August 4, 2009 9:33 9.75in x 6.5in b684-ch19 FA
Toe protection consists of the armoring of the beach or bottom surface in front of a
structure which prevents it from scouring and undercutting by waves and currents.
Factors that affect the severity of toe scour include wave breaking (when near the
toe), wave run-up and backwash, wave reflection, and grain size distribution of the
beach or bottom materials. Toe stability is essential because failure of the toe will
generally lead to failure throughout the entire structure. Toe scour is a complex
process. Specific (generally valid) guidance for scour prediction and toe design based
on either prototype or model results have not been developed as yet, but some
general (indicative) guidelines for designing toe protection are given in Refs. 4, 5,
22, and 31.
The maximum scour force occurs where wave downrush on the structure face
extends to the toe and/or the wave breaks near the toe (i.e., shallow water
structure). These conditions may take place when the water depth at the toe is less
than twice the height of the maximum expected unbroken wave that can exist at
that water depth. The width of the apron for shallow water structures with a high
reflection coefficient, which is generally true for slopes steeper than about 1 on 3,
can be planned based on the structure slope and the expected scour depth. The
maximum depth of a scour trough due to wave action below the natural bed is
about equal to the maximum expected unbroken wave at the site. To protect the
stability of the face, the toe soil must be kept in place beneath a surface defined by
an extension of the face surface into the bottom to the maximum depth of scour.
This can be accomplished by burying the toe, when construction conditions permit,
thereby extending the face into an excavated trench the depth of the expected scour.
Where an apron must be placed on the existing bottom, or can only be partially
buried, its width should not be less than twice the wave height.
If the reflection coefficient is low (slopes milder than 1 on 3), and/or the water
depth is more than twice the wave height, much of the wave force will be dissipated
on the structure face and a smaller apron width may be adequate, but it must
be at least equal to the wave height (minimum requirement). Since scour aprons
generally are placed on very flat slopes, quarrystone of the size (diameter) equal
to 1/2 of the primary cover layer probably will be sufficient unless the apron is
exposed above the water surface during wave action. Quarrystone of primary cover
layer size may be extended over the toe apron if the stone will be exposed in the
troughs of waves, especially breaking waves. The minimum thickness of cover layer
over the toe apron should be two quarrystones. Quarrystone is the most favorable
material for toe protection because of its flexibility. If a geotextile is used as a
secondary layer it should be folded back at the end, and then buried in cover stone
and sand to form a Dutch toe. It is recommended to provide an additional flexible
August 4, 2009 9:33 9.75in x 6.5in b684-ch19 FA
510 K. Pilarczyk
edge (at least 1 m) consisting of loose material which may easily follow the scour at
the toe. Some alternative designs of toe protection are shown in Fig. 19.16.
The size of toe protection against waves can be roughly estimated by using the
common formulae on slope protection and schematizing the toe by mild slopes (i.e.,
1 on 10) or by using formulae developed for breakwaters.
August 4, 2009 9:33 9.75in x 6.5in b684-ch19 FA
Toe protection against currents may require smaller protective stone, but wider
aprons. The necessary design data can be estimated from site hydrography and/or
model studies. Special attention must be given to sections of the structure where
scour is intensified; i.e., to the head, the areas of a section change in alignment,
the channel sides of jetties, and the downdrift sides of groynes. Where waves and
reasonable currents (>1 m/s) occur together, it is recommended to increase the
cover size at least by a factor of 1.3.
Note that the conservatism of the apron design (width and size of cover units)
depends on the accuracy of the methods used to predict the waves and current
action and to predict the maximum depth of scour. For specific projects, a detailed
study of scour of the natural bottom and at nearby similar existing structures should
be conducted at a planned site, and/or model studies should be considered before
determining a final design. In all cases, experience and sound engineering judgment
play an important role in applying these design rules.
Hs 1.5 cos αi
= , (19.22)
∆Dn Rc
ΦT ξ 2b 1 − R n
where
Hs = significant wave height,
ξ = breaker index; ξ = tan α(Hs /Lo )−0.5 ,
α = slope angle,
αi = angle of crest or inner slope,
L0 = wavelength,
b = coefficient equal to 0.5 for smooth slopes and 0.25 for riprap,
Rc = crest height above still water level,
Rn = wave run-up on virtual slope with the same geometry (see Fig. 19.17),
August 4, 2009 9:33 9.75in x 6.5in b684-ch19 FA
512 K. Pilarczyk
The length of protection in the splash area, which is related to the energy decay,
depends on the permeability of the splash area. However, it can be roughly assumed
as equal to:
ψ
Ls = T g(Rn − Rc ) ≥ Lmin , (19.23)
5
with a practical minimum (Lmin ) equal at least to the total thickness of the
revetment (including sublayers) as used on the slope. ψ is an engineering-judgment
factor related to the local conditions (importance of structure), ψ ≥ 1, usually
1 < ψ < 2.
Stability of rockfill protection of the crest and rear slope of an overtopped or
overflowed dam or dike can also be approached with the Knauss formula.17 The
advantage of this approach is that the overtopping discharge, q, can be used directly
as an input parameter for calculation. Knauss analyzed steep shute flow hydraulics
(highly aerated/turbulent) for the assessment of stone stability in overflow rockfill
dams (impervious barrages with a rockfill spillway arrangement). This kind of flow
seems to be rather similar to that during high overtopping. His (simplified) stability
relationship can be re-written to the following form:
√
q = 0.625 g(∆Dn )1.5 (1.9 + 0.8Φp − 3 sin αi ), (19.24)
August 4, 2009 9:33 9.75in x 6.5in b684-ch19 FA
where
q = maximum admissible discharge (m3 /s/m),
g = gravitational acceleration (9.81 m/s2 ),
Dn = equivalent stone diameter, Dn = (M50 /ρs )1/3 ,
∆ = relative density; ∆ = (ρs − ρw )/ρw ,
αi = inner slope angle, and
Φp = stone arrangement packing factor, ranging from 0.6 for natural dumped
rockfill to 1.1 for optimal manually placed rock; it seems to be reasonable to
assume Φp = 1.25 for placed blocks.
Note: When using the Knauss formula, the calculated critical (admissible) dis-
charge should be identified with a momentary overtopping discharge per overtopping
fraction of a characteristic wave, i.e., volume of water per characteristic wave divided
by overtopping time per wave, roughly (0.3–0.4)T (T = wave period), and not with
the time-averaged discharge (q).
514 K. Pilarczyk
As a general principle, one can state that the transition should be of strength
equal to or greater than the adjoining systems. Very often it needs reinforcement in
one of the following ways:
Top edge and flank protection are needed to limit the vulnerability of the
revetment to erosion continuing around its ends. Extension of the revetment beyond
the point of active erosion should be considered but is often not feasible. Care should
therefore be taken that the discontinuity between the protected and unprotected
areas is as small as possible (use a transition roughness) so as to prevent under-
mining. In some cases, open cell blocks or open block mats (eventually vegetated)
can be used as transition (i.e., from hard protection into grass mat). The flank
protection between the protected and unprotected areas usually needs a thickened
August 4, 2009 9:33 9.75in x 6.5in b684-ch19 FA
Fig. 19.19. Construction aspects of revetments; examples of toe protection and placing of block
mats (mattresses), and some methods of anchoring.
or grouted cover layer, or a concrete edge strip with some flexible transition,
i.e., riprap.
516 K. Pilarczyk
19.10. Conclusions
The newly derived design methods and stability criteria will be of help in preparing
the preliminary alternative designs with various revetment systems. However, there
are still many uncertainties in these design methods. Therefore, experimental veri-
fication and further improvement of design methods are necessary. Also, more prac-
tical experience at various loading conditions is still needed.
References
518 K. Pilarczyk
Limits
φ(rock) = 2.25
Criterion ctgα ≥ 2
Hs cos α cos α
= ψu · φ b = ψu 2.25 b 0.5 ≤ b ≤ 1.0
∆m D ξp ξp
9.75in x 6.5in
Ref. 1.0 Riprap (2 layers) Granular
Rock 1.33 Riprap (tolerable damage) Granular
Pitched 1.00 Poor quality (irregular-)stone Granular
Stone 1.33 Good quality (regular-)stone Granular
1.50 Natural basalt Granular
Blocks/Blockmats 1.50 Loose closed blocks; Hs < 1.5 m Geotextile on sand
1.50 Loose (closed-)blocks Granular
1.50 Blocks connected to geotextile Granular
2.00 Loose closed blocks Geotextile on clay
2.00 Cabled blocks/Open blocks (>10%) Granular
≥2.50 Grouted (cabled-) blocks/Interlocked blocks Granular
adequately designed
b684-ch19
Grout 1.05 Surface grouting (30% of voids) Granular
1.50 Pattern grouting (60% of voids) Granular
(Continued )
519
FA
August 4, 2009 9:33
520
Appendix A. (Continued )
Limits
φ(rock) = 2.25
Criterion ctgα ≥ 2
Hs cos α cos α
= ψu · φ b = ψu 2.25 b 0.5 ≤ b ≤ 1.0
9.75in x 6.5in
∆m D ξp ξp
Open
K. Pilarczyk
Stone 2.00 Open stone asphalt; Up ≤ 6 m/s Geotexile on clay
Asphalt 2.50 Open stone asphalt; Hs < 4 m Sand asphalt
Gabions 2 + 3.0 Gabion/mattress as a unit, Hs < 1.5 Geotexile on sand or Geotextile on clay
2 + 2.5 Stone-fill in a basket, dmin = 1.8 Dn
Fabric 1.00 Pm 1 less permeable mattress Sand or Clay and ev. Geotextile
Containers 1.50 Pm ≈ 1 (Pm = ratio permeab. top/sublayer
2.00 Pm 2 Permeable mattress of special design
Grass Grass-mat on poor clay; Up < 2 m/s Clay (Up = permiss. velocity)
Grass-mat on proper clay; Up < 3 m/s/s
b684-ch19
Note: F = ψu Φ; Φ = stability factor for riprap, and Ψu = upgrading factor in comparison with riprap.
FA
August 1, 2009 12:50 9.75in x 6.5in b684-ch20 FA
Chapter 20
Krystian Pilarczyk
(Former) Hydraulic Engineering Institute
Public Works Department, Delft, The Netherlands
HYDROpil Consultancy, 23 Nesciohove
2726BJ Zoetermeer, The Netherlands
krystian.pilarczyk@gmail.com
Erosion control and coastal stabilization are common problems in coastal engi-
neering. A brief overview of some available alternative systems for shore sta-
bilization and beach erosion control is presented. Special attention is paid to
artificial reefs and geosystems. Geosystems (geotubes, geocontainers, etc.) have
gained popularity in recent years because of their simplicity in placement, cost
effectiveness, and environmental aspects. However, all these systems have some
advantages and disadvantages, which have to be recognized before application.
20.1. Introduction
Some coastal environments may be regarded as rather stable (rock and reef coasts)
while others are more vulnerable (sand and mud coasts, soft cliffs). In this last
case, coastal users and managers all over the world are frequently faced with serious
erosion of their sandy coasts. Possible causes of erosion include natural processes
(i.e., action of waves, tides, currents, sea level rise, etc.) and sediment deficit due to
human impact (i.e., river regulations, sand mining, and coastal engineering works).
Countermeasures for beach erosion control function depend on local conditions of
shore and beach, coastal climate, and sediment transport. Continuous maintenance
and improvement of the coastlines, together with monitoring and studies of coastal
processes have yielded considerable experience on various coastal protection mea-
sures all over the world.
This contribution presents an overview of the various available methods for
shore stabilization and beach erosion control, with special emphasis on the
novel/alternative systems in various design implementations. More detailed infor-
mation on other coastal protection systems and measures applied nowadays
throughout the world can be found in extensive list of references.15,16,51,52,54,80
521
August 1, 2009 12:50 9.75in x 6.5in b684-ch20 FA
522 K. Pilarczyk
the downdrift end where the uncontrolled drift rate is re-established. Without
an adequate supply of beach material, groins are of limited value. In addition
to controlling the rate of drift, groins are also used extensively to control the
distribution of material along a frontage and to limit the temporary effects of
drift reversal. There are unfortunately many examples where either bad design or
failure to provide for the downdrift consequences has resulted in an adverse effect
August 1, 2009 12:50 9.75in x 6.5in b684-ch20 FA
524 K. Pilarczyk
526 K. Pilarczyk
Functional requirements and design outline are depicted in Fig. 20.3. It is seen
that the design in various stages is verified through the use of simulation models
at different levels of complexity. Boundary conditions (bottom) constitute input to
both design considerations and the models employed, while the functional require-
ments (top) ensure evaluation of the suitability of the design and provide design
objectives at the same time.
August 1, 2009 12:50 9.75in x 6.5in b684-ch20 FA
The starting point in the design process consists of the identification of the
beach erosion problem, followed by the selection of the type of protection measure;
the final design can incorporate the risk analysis. Attention is drawn to the proper
choice of the shore protection measure. The selection is usually affected by the cost.
For example, beach nourishment can be cost-effective for low sediment deficit but
might be comparable with offshore breakwaters for higher erosion rates. Aside from
the cost, many other aspects must also be taken into account upon selection of the
shore protection measure. Not shown in the drawing are legal restrictions, regional
constraints and priorities, construction, operation and maintenance aspects, etc.
Coastal structures are constructed to protect life and property against storm
surges, to combat erosion and/or to create (often artificial) beaches for recreational
purposes, and to preserve the natural environment. However, the absolute safety of
an area or structure is nearly impossible to achieve. Therefore, it is much better
to speak about the probability of failure (or safety) of a certain protection system.
To implement this concept, all possible causes and outcomes of failure have to be
analyzed. This concept is actually being developed for breakwaters52 and the dike
and dune design, mostly in the Netherlands (see, www.enwinfo.nl).18,53,55 The “fault
tree” is a handy tool for this aim. In the fault tree all possible modes of failure
of elements, which can eventually lead to the failure of a structure section and to
inundation are included. They can also badly affect the behavior of the structure,
even if the latter is properly designed on the whole. Although all categories of
August 1, 2009 12:50 9.75in x 6.5in b684-ch20 FA
528 K. Pilarczyk
events which may cause the inundation of a land or damage of structure are equally
important for the overall safety, the engineer’s responsibility is mainly limited to
the technical and structural aspects. In the case of coastal structures the following
major events can be distinguished:
530 K. Pilarczyk
However, broad-crested structures are much more expensive and their use should
be supported by a proper cost-benefit studies. On the other hand, the development
in alternative materials and systems, for example, the use of sand-filled geotubes as
a core of such structures, can reduce effectively the cost.24,45,56 The upgrading of
(integrated/multidisciplinary) design criteria for LCS structures recently took place
in the scope of European project DELOS22 ; (see also: www.delos.unibo.it).
The relatively new innovative coastal solution is to use artificial reef struc-
tures called “Reef Balls” as submerged breakwaters, providing both wave atten-
uation for shoreline erosion abatement, and artificial reef structures for habitat
enhancement.7 An example of this technology using patented Reef BallTM is
shown in Fig. 20.7. Reef Balls are mound-shaped concrete artificial reef modules
that mimic natural coral heads. The modules have holes of many different sizes
in them to provide habitat for many types of marine life. They are engineered
to be simple to make and deploy and are unique in that they can be floated
to their drop site behind any boat by utilizing an internal, inflatable bladder.
Worldwide a large number of projects have already been executed by using this
system.
The first applications were based purely on experience from previous smaller
projects. Since recently, more well-documented design criteria are available. Sta-
bility criteria for these units were determined based on analytical and experimental
studies. For high energetic wave sites the units can be hydraulically anchored with
cables to the sea bed. Wave transmission was studied in Canada.4 Technical design
aspects are treated by Harris.31
532 K. Pilarczyk
may be artificially induced by beach watertable drainage. The state of the art of this
technique is presented by Vesterby.77 An extensive evaluation of drainage systems is
provided by Curtis and Davis18 and actual information can be found on the website:
http://shoregro.com/. It is concluded that the drainage system has, in general, a
positive effect on diminishing the beach erosion; however, its effectiveness is still
difficult to control. The idea to achieve lowering of the water table without pumps
by enhancing the beach’s own drainage capacity or hydraulic conductivity through
the use of strip drains has been applied in Australia and in Japan.39
534 K. Pilarczyk
Fig. 20.11. General transmission characteristic for Aquareef33 (Ht is the transmitted wave height
recorded on the landward side, H1/3 and L1/3 are the significant wave height and wavelength at
the toe of the rubble mound, B is the crown width, and R is the submergence of the crown).
of the salient or tombolo varies in accordance with reef dimensions. Of course, one
can expect this kind of morphological change only if the sediment is available (from
natural sources or as sand nourishment).
The examples of simple geometrical empirical criteria for the layout and
shoreline response of the detached, exposed (emerged) breakwaters can be found in
Refs. 20, 32, and 53. To include the effect of submergence (transmission) Pilarczyk57
proposes, at least as a first approximation, adding the factor (1− Kt) to the existing
rules. Then the rules for low-crested breakwaters can be modified to (for example):
• Tombolo:
• Salient:
For salients where there are multiple breakwaters: GX/L2s > 0.5(1 − Kt ), where
Ls is the length of a breakwater and X is the distance to the shore, G is the gap
width, and the transmission coefficient Kt is defined for annual wave conditions.
The gap width is usually L ≤ G ≤ 0.8Ls , where L is the wavelength at the structure
defined as: L = T (gh)0.5 ; T = wave period, h = local depth at the breakwater. As
first approximation, Kt = 0 for emerged breakwaters and Kt = 0.5 for submerged
breakwaters can be assumed for average annual effects.
These criteria can only be used as preliminary design criteria for distinguishing
shoreline response to a single, transmissive detached breakwater. However, the range
of verification data is too small to permit the validity of this approach to be assessed
for submerged breakwaters.
August 1, 2009 12:50 9.75in x 6.5in b684-ch20 FA
536 K. Pilarczyk
In general, one may conclude that these simple geometrical design rules, both for
emerged as well as submerged breakwaters, are of limited value for design calcula-
tions because they do not include the effect of the rate of sediment transport, which
can be very different for a specific coast. It is supported by the studies on effect of
offshore breakwaters along UK coast.6,67 On the other side, it can be stated that
numerical models (i.e., Genesis, Delft 2D-3D, Mike 21, etc.) can already be treated as
useful design tools for the simulation of morphological shore response to the presence
of offshore structures. Examples can be found in Refs. 13, 17, 28–30, 42, and 70.
As mentioned above, while considerable research has been done on shoreline
response to exposed offshore breakwaters, very little qualitative work has been done
on the effect of submerged offshore reefs, particularly outside the laboratory. Thus,
within the Artificial Reefs Program9 (www.asrltd.co.nz), Andrews2 examined aerial
photographs seeking cases of shoreline adjustment to offshore reefs and islands. All
relevant shoreline features in New Zealand and eastern Australia were scanned and
digitized, providing 123 different cases. A range of other statistics, particularly reef
and island geometry, was also obtained. Some of these results are repeated below.
To examine the effects of wave transmission on limiting parameters, data for
reefs and islands were considered separately. The data indicated that tombolo for-
mation behind islands occurs with Ls /X ratio of 0.65, and higher and salients form
when Ls /X is less than 1.0. Therefore, for islands the Ls /X ratios determining the
division between salients and tombolos are similar to those from previously pre-
sented breakwater research. Similarly, data resulting from offshore reefs indicate
that tombolo formation occurs at Ls /X ratios of 0.6 and higher, and salients most
commonly form when Ls /X is less than 2. The data suggests that variation in wave
transmission (from zero for offshore islands through to variable transmission for off-
shore reefs) allows a broader range of tombolo and salient limiting parameters. Thus,
a reef that allows a large proportion of wave energy to pass over the obstacle can be
(or must be) positioned closer to the shoreline than an emergent feature. Thus, from
natural reefs and islands the following general limiting parameters were identified:
• Islands: Tombolos form when Ls /X > 0.65; Salients form when Ls /X < 1.0;
• Reefs: Tombolos form when Ls /X > 0.60; Salients form when Ls /X < 2.0;
• Nondepositional conditions for both shoreline formations occur at Ls /X < ≈0.1.
Fig. 20.12. Examples of application of geosystems: (a) Geotube as a breakwater, (b) BEROSIN
curtain.
can be made between: bags, mattresses, tubes, containers, and inclined curtains.56
All of which can be filled with sand or mortar. Some examples are shown in
Fig. 20.12. Bags are suitable for slope protection and retaining walls or toe pro-
tection but the main application is the construction of groynes, perched beaches,
and offshore breakwaters. The tubes and containers are mainly applicable for con-
struction of groynes, perched beaches, and (offshore) breakwaters, and as bunds for
reclamation works (Fig. 20.13).
Geotubes can form an individual structure in accordance with some functional
requirements for the project but they can also be used complementary to the arti-
ficial beach nourishment to increase its lifetime. Especially for creating the perched
beaches, the sand tubes can be an ideal, low-cost solution for constructing the sub-
merged sill.34,45,56 Geotubes and geocontainers can also be used to store and isolate
August 1, 2009 12:50 9.75in x 6.5in b684-ch20 FA
538 K. Pilarczyk
contaminated materials from harbor dredging, and/or to use these units as bunds
for reclamation works.
An interesting application for shore erosion control is the geocurtain known
under the name BEROSINTM,71 [Fig. 20.12(b)]. The BEROSIN curtain is a flexible
structure made of various woven geotextiles, which after placing by divers near the
shore and anchoring to the bed catches the sand transported by currents and waves
providing accretion on a shore and preventing the erosion. The horizontal curtain
(sheet) can be easily spread by a small workboat and two divers. The upper (shore-
side) edge, equipped with some depth-compensated floaters, should be properly
anchored at the projected line. The sea-side edge is kept in position by the workboat.
By ballasting some of the outside pockets at the lower edge with sand or other mate-
rials and with the help of divers, the lower edge is sinking to the required position.
The proper choice of permeability of geotextile creates the proper conditions for
sedimentation of suspended sediment in front/or under the curtain and at the same
time allowing the water to flow out without creating too high forces on the curtain
and thus, on the anchors. In case of Pilot project at the coast of Vlieland (NL), some
of the horizontal curtains placed in the intertidal zone have provided a growth of a
beach/foreshore of 0.5–1.0 m within a week while others within a few weeks.56 These
geocurtains can also be applied for the construction of submerged sills and reefs.
In the past, the design of geotextile systems for various coastal applications
was based mostly on vague experience rather than on the general valid calcu-
lation methods. However, the increased demand in recent years for reliable design
methods for protective structures have led to the application of new materials and
systems (including geotextile systems) and to research concerning the design of
these systems. Contrary to research on rock and concrete units, there has been
no systematic research on the design and stability of geotextile systems. However,
past and recent research in the Netherlands, USA, Germany, and in some other
countries on a number of selected geotextile products has provided some useful
results which can be of use in preparing a set of preliminary design guidelines
for the geotextile systems under current and wave attack.31,56,58 The recent, large
scale tests, with large geobags, can be found on the website: http://sun1.rrzn.uni-
hannover.de/fzk/e5/projects/dune prot 0.html.
August 1, 2009 12:50 9.75in x 6.5in b684-ch20 FA
When installing geosystems, one should see to it that this does not take place on
a rough foundation. Sharp elements may easily damage the casing of the element.
Geosystems must usually not be filled completely. With a fill ratio of approximately
75% an optimum stability of the elements is reached. A sound soil protection is
necessary if gravel (sand) sausages are used in circumstances where they are under
attack of flow or waves. Stability criteria are discussed by Pilarczyk.56
The main (large) fill-containing geosystems (geobags, geotubes, and geocon-
tainers filled with sand or mortar) and their design aspects are briefly discussed
below. For more detailed information, the reader will be guided to the relevant
manuals and publications (see references and websites).
20.5.2. Geobags
Geobags can be filled with sand or gravel (or cement, perhaps). The bags may have
different shapes and sizes, varying from the well-known sandbags for emergency
dikes to large flat shapes or elongated “sausages” (see Fig. 20.14). The most common
use for sandbags in hydraulic engineering is for temporary structures. Uses for sand-
or cement-filled bags are, among other things:
• repair works;
• revetments of relatively gentle slopes and toe constructions;
• temporary or permanent groynes and offshore breakwaters;
• temporary dikes surrounding dredged material containment areas.
540 K. Pilarczyk
Because this material is easy to use and cheap, it is extremely suitable for temporary
structures. A training groyne is a good example. The working of a groyne is difficult
to predict in advance. That is why it is a good procedure to make such a construction
using a relatively cheap product first, to see how one thing and another works
out, and subsequently either make improvements or, after some time, a permanent
structure. However, it is recommended not to use sand-filled geobags for a wave
height above 1 m and/or a flow velocity above 2 m/s. The geosystems filled with
sand cannot be used at these more extreme conditions because the sand in the
systems is no longer internally stable.
Sandbags can be placed as follows:
(1) As a blanket: One or two layers of bags placed directly on the slope. An “inter-
locking” problem arises if the bags are filled completely. The bags are then too
round. A solution is not to fill the bags completely, so that the sides flatten out
somewhat, as a result of which the contact area becomes larger.
(2) As a stack: Bags stacked up in the shape of a pyramid. The bags lie half-
overlapping with (usually) the long side parallel to the shoreline. The stacking of
geotubes increases their stability due to interlocking with the neighboring bags.
More recently, prototype experience and large scale tests with large geobags, are
described by Restall,58 and more detailed information can be found on the website:
http://sun1.rrzn.uni-hannover.de/fzk/e5/projects/dune prot 0.html.
its desired shape when filled up to about 70–80% (or a height equal to about a half
of the flat width of the tube); a higher filling grade is possible but it diminishes the
friction resistance between the tubes.
Tube can be filled on land (e.g., as dikes for land reclamation, bunds, toe
protection, or groyns) or in water (e.g., offshore breakwaters, sills of perched
beaches, dikes for artificial islands, etc.). The tube is rolled out along the intended
alignment with inlets/outlets centered on top. When a tube is to be placed in
water, the effects of buoyancy on the tube geotextile prior to filling as well as on
the dredged material’s settling characteristics must be considered. In order to max-
imize inlet/outlet spacing, an outlet distant from the inlet may be used to enhance
the discharge of dredged slurry and thereby encourage and regulate the flow of fill
material through the tube so that sufficient fill will flow to distant points. Com-
monly, the filter geotextile on both sides of a tube (against scour) and flat main
tube are fully deployed by floating and holding them in position prior to beginning
the filling operation. The geotextile filter is often furnished with small tubes at
the edges which when filled with sand holds the filter apron in place. The required
length of apron is usually two times the local wave height.
542 K. Pilarczyk
Fig. 20.16. Example of split barge and filling and placing of geocontainer.
August 1, 2009 12:50 9.75in x 6.5in b684-ch20 FA
• Containers can be filled with locally available soil, which may be available from
simultaneous dredging activities.
• Containers can be relatively accurately placed regardless of weather conditions,
current velocities, tides, or water depths.
• Contained material is not subject to erosion after placing.
• Containers can provide a relatively quick system buildup.
• Containers are very cost competitive (for larger works).
The following design aspects are particularly of importance for the design of
containers:
(a) change of shape of units in function of perimeter of unit, fill-grade, and opening
of split barge,
(b) fall-velocity/equilibrium velocity, velocity at bottom impact,
(c) description of dumping process and impact forces,
(d) stresses in geotextile during impact and reshaping,
(e) resulting structural and executional requirements, and
(f) hydraulic stability of structure.
544 K. Pilarczyk
(1) The required perimeter of geotextile sheet must be sufficient enough to release
geocontainer through the given split width b0 for a required cross-sectional area
of material in the bin of barge Af (or filling-ratio of fill-material in respect to
the max. theoretical cross-section). The derivation of the required minimum
length of perimeter of geotextile sheet is given by Pilarczyk.56 After opening of
the split of a barge the geocontainer is pulled out by the weight of soil but at
the same time the friction forces along the bin side are retarding this process.
Due to these forces the tension in geotextile is developing at lower part and
both sides of the geocontainer. The upper part is free of tension till the moment
of complete releasing of geocontainer. The question is how far we are able to
model a friction and the release process of geocontainer.
(2) Geocontainer will always contain a certain amount of air in the pores of soil
and between the soil and the top of (surplus) geotextile providing additional
buoyancy during sinking. The amount and location of air pockets depends on
soil consistency (dry, saturated) and uniformity of dumping. The air pockets
will exert certain forces on geotextile and will influence the way of sinking. The
question is how to model in a proper way the influence of soil consistency and
air content on shape and stresses in geotubes/geocontainers.
(3) The forces due to the impact with the bottom will be influenced by a number
of factors:
• consistency of soil inside the geocontainer (dry, semi-dry, saturated, cohesion,
etc.) and its physical characteristics (i.e., internal friction);
August 1, 2009 12:50 9.75in x 6.5in b684-ch20 FA
• amount of air;
• permeability/air tightness of geotextile;
• strength characteristics of geotextile (elasticity/elongation versus stresses,
etc.);
• fall-velocity (influenced by consistency of soil; saturated soil diminish amount
of air but increases fall speed);
• shape and catching surface of geocontainer at impact including effect of not
horizontal sinking (i.e., catching of bottom with one end);
• type of bottom (sand, clay, soft soil, rock, soil covered with rockfill mattress,
etc.) and/or type of sublayer (i.e., layer of previous placed containers).
During the impact the cross-sectional shape of geocontainer will be undergoing
a continuous reshaping; from cone shape, first probably into a transitional cylin-
drical shape, and through a certain relaxation, into a semi-oval shape or flat
triangular/rectangular shape dictated by soil type, perimeter, and elongation
characteristics of geotextile. The question is how far we are able to model
this impact phenomena and resulting forces/stresses in geotextile. The impact
forces with the bottom are a function of the fall velocity (dump velocity) of a
geocontainer.
(4) In final situation the geocontainers will perform as a core material of various
protective structures or as independent structure exposed to loading by currents
and waves, and other loadings (ice, debris, ship collision, vandalism, etc.). In
most cases the geocontainers will be filled by fine (loosely packed) soils. The
question is how these structures will behave in practice under various types of
external and internal loadings.
Practical note: The prototype experience indicates that geocontainers with volume
up to 200 m3 and dumped in water depth exceeding 10 m have been frequently
damaged (collapse of seams) using geotextile with tensile strength lower than
75 kN/m, while nearly no damage was observed when using the geotextile with
tensile strength equal to or more than 150 kN/m. This information can be of use
for the first selection of geocontainers for a specific project. The placing accuracy
for depths larger than 10 m is still a problem.8
546 K. Pilarczyk
20.7. Conclusions
The author does not intend to provide the new design rules for alternative coastal
structures. However, it is hoped that this information will be of some aid to
designers looking for new sources, which are considering these kinds of structure
and improving their designs.
Offshore breakwaters and reefs can be permanently submerged, permanently
exposed or inter-tidal. In each case, the depth of the structure, its size, and its
position relative to the shoreline determine the coastal protection level provided by
the structure. To reduce the cost some alternative solutions using geosystems can
be considered. The actual understanding of the functional design of these structures
needs further improvement but may be just adequate for these structures to be
considered as serious alternatives for coastal protection.
Continued research, especially on submerged breakwaters, should further
explore improved techniques predict shore response and methods to optimize
breakwater design. A good step (unfortunately, limited) in this direction was
made in a collective research project in the Netherlands.17 Research and prac-
tical design in this field is also the focus of the “Artificial Reefs Program”
in New Zealand (www.asrltd.co.nz), the International Society for Reef Studies
(ISRS) (www.artificialreefs.org), the European Project DELOS (Environmental
Design of Low Crested Coastal Defence Structures; http://www.delos.unibo.it),
and recent US National Shoreline Erosion Control Development and Demon-
stration Program (Section 227). Some useful documents can also be found
on the website: http://www.citg.tudelft.nl/live/pagina.jsp?id=4de0d195-5207-4e67-
84bb-455c5403ae47&lang=en; www.hydraulicengineering.tudelft.nl.
Also, the past and recent research in the Netherlands, USA, Germany, and in
some other countries on a number of selected geotextile products (geosystems) has
provided some valuable results, which can be of use in preparing a set of preliminary
design guidelines for the geotextile systems under current and wave attack.
The following conclusions can be drawn on application of geosystems based on
the actual developments and experience.
548 K. Pilarczyk
These new efforts will bring future designers closer to more efficient application
and design of these promising coastal solutions. The more intensive monitoring of the
existing structures will also help in the verification of new design rules. The intention
of this literature search is to uncover the technical information on these systems and
make them available for the potential users. It will help to make a proper choice
for specific problems/projects and it will stimulate the further developments in this
field. International cooperation in this field should be further stimulated.
References
550 K. Pilarczyk
60. T. Sawaragi, I. Deguchi and S. K. Park, Reduction of wave overtopping rate by the
use of artificial reef, Proc. 21st Int. Conf. Coastal Eng., ASCE (1988).
61. T. Sawaragi, Detached breakwaters; short course on design and reliability of coastal
structures, Venice, 1–3 October 1992.
62. T. Sawaragi, Coastal Engineering-Waves, Beaches, Wave–Structure Interactions
(Elsevier, 1995).
63. S. R. Seabrook and K. R. Hall, Wave transmission at submerged rubble mound break-
waters, 26th Int. Conf. Coastal Eng., Copenhagen (1998).
64. R. Silvester and J. R. Hsu, Coastal Stabilization, (Prentice Hall Inc., Englewood Cliffs,
1993), http://www.worldscibooks.com/engineering/3475.html.
65. D. Smit et al., Submerged-crest breakwater design, Advances in Coastal Structures
and Breakwaters’95, London (1995).
66. D. K. Stauble and J. R. Tabar, The use of submerged narrow-crested breakwaters
for shoreline erosion control, J. Coastal Res. 19(3), 352 (2003), http://chl.erdc.usace.
army.mil/CHL.aspx?p=m&a=MEDIA.
67. F. Thomalla and Ch. Vincent, Designing offshore breakwaters using empirical rela-
tionships: a case study from Norfolk, UK, J. Coastal Res. 20(4) (2004).
68. T. Uda, Function and design methods of artificial reef (in Japanese), Ministry of
Construction, Japan (1998) (see also, Coastal Zone’93).
69. US Army Corps, Engineering design guidance for detached breakwaters as shoreline
stabilization structures, WES, Technical Report CERC–93–19, December 1993.
70. S. C. Van der Biezen, J. A. Roelvink, J. Van de Graaff, J. Schaap and L. Torrini, 2DH
morphological modelling of submerged breakwaters, 26th ICCE, Copenhagen (1998).
71. van der Hidde, BEROSIN, Bureau van der Hidde, Harlingen, P.O.B. 299,
the Netherlands (1995).
72. J. W. van der Meer, Stability of breakwater armour layers, Coastal Eng. 11 (1987).
73. J. W. van der Meer, Rock slopes and gravel beaches under wave attack, Doctoral
thesis, Delft University of Technology (1988) (Also: Delft Hydraulics Communication
No. 396).
74. J. W. van der Meer, Low-crested and reef breakwaters, Delft Hydraulics Rep. H 986
(1990).
75. J. W. van der Meer, Data on wave transmission due to overtopping, Delft Hydraulics
(1990).
76. J. W. van der Meer and K. d’Angremond, Wave transmission at low-crested structures,
Coastal Structures and Breakwaters (Thomas Telford, London, UK, 1991), pp. 25–40.
77. H. Vesterby, Beach drainage — state of the art — Seminar on Shoreline Management
Techniques, Wallingford, 18 April 1996.
78. C. Vidal, M. A. Losada, R. Medina, E. P. D. Mansard and G. Gomez-Pina, A universal
analysis for the stability of both low-crested and submerged breakwaters, 23rd Coastal
Engineering, Venice (1992).
79. C. Vidal, I. J. Losada and F. L. Martin, Stability of near-bed rubble-mound structures,
26th Coastal Engineering, Copenhagen (1998).
80. N. Von Lieberman and S. Mai, Analysis of an optimal foreland design, 27th Coastal
Engineering 2000, Sydney (2000).
81. T. Wamsley and J. Ahrens, Computation of wave transmission coefficients at detached
breakwaters for shoreline response modelling, Coastal Structures’03, Portland, USA
(2003).
82. K. Yoshioka, T. Kawakami, S. Tanaka, M. Koarai and T. Uda, Design manual for
artificial reefs, in Coastlines of Japan II, Coastal Zone’93 (ASCE, 1993).
August 1, 2009 12:50 9.75in x 6.5in b684-ch20 FA
Chapter 21
21.1. Introduction
In view of the increasing storminess associated with climate change and its effect
on coastal flood and erosion, more versatile materials and solutions are required for
the design of new, cost-effective shore protection structures as well as for the rein-
forcement of existing threatened coastal barriers and structures, including dune rein-
forcement and scour protection. In search for low-cost, soft, and reversible solutions,
the concept of geotextile sand containers (GSCs) as “soft rock” for coastal defense
structures was introduced for the first time in the 1950s in the form of “sand bags”
for the construction of a dike to close the inlet “Pluimpot” in the Netherlands.53
553
August 4, 2009 9:34 9.75in x 6.5in b684-ch21 FA
At the same time some test groins were constructed by using “sand bags” at the
German North Sea coast.10,57 Later, GSCs have mainly been used as temporary
shore protection measures due to the difficulties associated with the assessment of
the long-term performance. Meanwhile, significant advances have been made with
respect to the following issues:
(i) they are more versatile and can be adapted to build any type of structures;
(ii) they can better fulfill any requirements with respect to structure slope and
geometry (better tolerance);
(iii) maintenance and remedial work are much easier in case of vandalism or failure
by wave action;
(iv) less tensile strength is required and less change of shape will be experienced,
thus resulting in longer lifetime;
(v) higher density of the sand fill can be achieved;
(vi) less risk of liquefaction of the sand fill is expected and thus less GSC-
deformations;
(vii) simpler mobilization of the required equipment; and
(viii) the smaller the containers, the larger the self-healing capacity of the structure.
However, larger containers may be required, for instance, in the case of higher
waves forces and for temporary structures. A brief illustration of the container sizes
used in coastal engineering is given in Fig. 21.2.
August 4, 2009 9:34 9.75in x 6.5in b684-ch21 FA
Fig. 21.2. Range of nonwoven geotextile sand container sizes applied in coastal engineering.
Table 21.1. Properties of nonwoven and woven geotextiles (After Lawson and Kemplin, 1995)∗ .
Nonwovens
Heat-bonded 3–25 20–60 0.02–0.35 10–200 60–350
Needle-punched 7–90 30–80 0.03–0.20 30–300 100–3000
Resin-bonded 5–30 25–50 0.01–0.25 20–100 130–800
Wovens
Mono-filament 20–80 20–35 0.07–4.0 80–2000 150–300
Multi-filament 40–1200 10–30 0.05–0.90 20–80 250–1500
Flat tape 8–90 15–25 0.10–0.30 5–25 90–250
Knitted
Weft 2–5 300–600 0.20–2.0 60–2000 150–300
Warp 20–800 12–30 0.40–1.5 80–300 250–1000
Stitch-bonded 30–1000 10–30 0.07–0.50 50–100 250–1000
∗ C. R. Lawson and G. T. Kempton, Geosynthetics and Their Use in Reinforced Soils (Terram
As shown in Table 21.1, nonwoven and woven fabrics have significantly different
properties which can be exploited to produce the best solution for each specific
need, including composite fabrics to combine the advantages of both types.
The lower tensile strength of nonwoven geotextiles as compared to woven geo-
textiles might represent a drawback, if the containers have to accommodate very
large stresses during installation without failure. Generally, however, the stresses on
the containers are much lower in service than during installation.
The higher capacity for elongation of nonwoven geotextiles can to some extent
compensate the disadvantage of lower tensile strength in the sense that it allows
to accommodate large strains without failure. This is particularly important when
the container is required to reshape during installation and in service (adaptation
to scour development and settlement, self-healing effect). Project experiences have
shown that the final dimensions of sand-filled containers made of nonwoven geo-
textile are comparable to those made of woven geotextile. Based on tests with both
materials — woven and nonwoven — the final height of hydraulically filled geo-
textile containers (tubes) was approved to be 80% of the theoretical diameter.17
The hydraulic permeability of geotextile used for sand containers is particularly
important when subject to cyclic wetting and drying (e.g., in tidal regime). Water
should be drained from the sand container fast enough to avoid excess pressure
build up and to ensure overall stability. Therefore, it is generally required that
the geotextile should have a much higher permeability than the sand fill without
losing the finer fractions. Alternatively, the geotextile can be selected to fulfill the
commonly used filter criteria. Nonwoven geotextiles have a higher permeability and
a higher fine retention capability than their woven counterparts, but it should be
stressed that the permeability is a function of the fabric thickness, and thus depends
on its compressibility under normal stresses.
The abrasion resistance is particularly important in the surf environment where
coarse angular sand, shell fragments, or coral debris are present. The larger thickness
of nonwoven geotextiles and the ability of their structure to retain sand material
August 4, 2009 9:34 9.75in x 6.5in b684-ch21 FA
make them more resistant. For testing abrasion resistance of different geotextile
products, a special test setup has been introduced by the Federal Waterways Engi-
neering and Research Institute (BAW, Germany) already about 30 years ago.47 For
geotextiles being used in the surf environment abrasion tests, standard test setup (as
described in BAW Guidelines for Testing Geotextiles for Navigable Waterways4,47)
are recommended.
The puncture resistance is important in the case of vandalism, drift ice, drift
wood, or dropped rock material during construction. The catalogue for geo-
textile testing in waterway engineering established by the German Institute for
Waterway Engineering (BAW) also includes a puncture test setup which allows a
1:1 test according to the actual stone drop energy expected for the specific job
conditions.23,47 Due to their high elongation capacity and to the retained sand
within the fabric structure, nonwoven geotextiles are more susceptible to limit
damage from puncture, including vandalism.
A higher friction angle between sand containers is desirable to enhance the
hydraulic stability against wave and current actions. Due to their structure, non-
woven geotextiles provide a higher friction (see also Sec. 21.4.25).
The resistance against UV-radiation and oxidation still represents one of the
most critical issues which can limit the service time of exposed GSCs. The resins,
fibers, and yarns of the used woven and nonwoven fabrics need special stabilization
packages to avoid deterioration already during production and to guarantee long-
term stability for more than 50 years. As they are reasonable cost factors, the special
stabilization packages need to be described in the specifications or material certifi-
cates with, e.g., confidential information to the certification institute. In general,
increasing the thickness of fibers and yarns and increasing weight of the fabric
improve UV resistance and long-term stability.49 Dug-up operations from woven
and nonwoven geotextiles carried out in the seventies have already shown up to
25 years long-term stability with no indication of further concern to the upcoming
years. This is also linked to the fact that the design parameters for surviving han-
dling and installation are giving much reserve for the container simply lying on
the sea bed after successful installation. Well-designed and well-installed woven and
nonwoven geotextiles for coastal engineering applications have indeed shown that
they can be resistant over a long term.9 Despite the significant progress in the use
of UV-stabilizers, coating or/and armoring of the exposed geotextile containers still
remain the sole alternative to achieve a satisfactory lifetime without damage. Simul-
taneously, the coating/armoring will also protect the geotextile against abrasion,
vandalism, drift ice, and drift wood.
The resistance against chemical and microbiological attack of polymers (e.g.,
polypropylene) is very high, so that no significant loss of strength is expected during
the design lifetime.
The ability of geotextiles to enhance marine growth and to attract/support
diverse invertebrate communities also becomes an important issue, if maximizing
the biodiversity of the recruiting communities is relevant for the choice of the
type of geotextile. First, results of investigations comparing woven and nonwoven
geotextiles in Australia have shown that the latter are more favorable in this
respect.5 The marine growth may represent an enhancement of the resistance against
August 4, 2009 9:34 9.75in x 6.5in b684-ch21 FA
UV-radiation and abrasion. However, it is still not clear whether it may cause serious
changes or detrimental effects on the mechanical properties of the geotextile over
lifetime.
Ideally, a degradation curve over the design lifetime for each relevant property of
the geotextiles such as tensile strength, specific mass in g/m2 , elongation capacity,
and hydraulic permeability should be determined, together with the associated
acceptable degradation limits at which the geotextile cannot perform its primary
function (Fig. 21.4). The most authoritative evidence for long-term durability is
generally obtained from monitoring of the degradation under real service conditions.
In fact, the empirical evidence of long-term durability from retrieved (non-
exposed) geotextiles has shown that the reduction of tensile strength and other
important engineering properties strongly depends on the prevalent service con-
ditions and therefore significantly differs from one site to another. Samples of
PP (polypropylene) and PET (polyester) nonwoven geotextiles retrieved from
25 sites in France lost up to 30% of their tensile strength after 10–15 years service
time as filters, separators, and drains, while no chemical/biological effects were
identified.51
A further interesting case was reported by Lefaive,21 showing that the reduction
in tensile form strength after 17 years of the same PET straps embedded in a
concrete facing wall and in the backfill was completely different: 2% in the backfill
with pH = 8.5 and up to 40% at the transition between the wall and the backfill
where pH values of 13–14 and temperature up to 30◦ C prevailed. This case well
illustrates the contribution of alkaline surface attack (25%), internal hydrolysis
(5–10%), and mechanical damage to degradation of PET fibers or yarns.
The effect of UV-radiation is illustrated by a case reported by Troots et al.,52
where woven PET samples were retrieved after 13 years from an earth embankment:
while along the section of the slope covered by vegetation and bitumen to provide
protection against UV-attack no significant changes of the geotextile properties
occurred; a reduction up to 50% of the tensile strength was identified in the non-
protected part.
These and further numerous examples from the literature show that the results/
data from retrieved (essentially nonexposed) fabrics, although very valuable, have
serious limitations when intended to be used for the prediction of long-term dura-
bility and lifetime of permanent structures. Moreover, the obtained data are often
incomplete and relate to conditions that are generally far from those for which the
prediction/assessment is being made (Fig. 21.5).
In order to make the best use of field data some brief recommendation for future
site monitoring are given in Plate 21.1.
These limitations and the urgent necessity for both users and manufacturers to
predict lifetime of geotextiles have led to the development of accelerated tests which
also have serious limitations (Fig. 21.5).
(i) generally only one dominant degradation mechanism can be considered, thus
ignoring the interaction with other mechanisms and
(ii) the approach cannot be applied to all degradation mechanisms.
August 4, 2009 9:34 9.75in x 6.5in b684-ch21 FA
Fig. 21.7. Seawall made of geotextile sand containers (principle sketches): (a) plan view and
(b) cross-section.
Fig. 21.8. Geotextile containment for dune reinforcement, Sylt/Germany (extended and modified
from Ref. 26).
More flexible and much simpler in both engineering design and installation, but
equally efficient are smaller volume containers (Fig. 21.9).
Moreover, smaller containers have many advantages over larger containers and
tubes (see Sec. 21.1 and Ref. 2). A comparative analysis of containers with V =
0.75 m3 and containers with a 30 times larger volume for dune reinforcement which
confronted the pros and cons of both methods from the client and contractor view
point clearly resulted in the selection of the smaller containers.2
August 4, 2009 9:34 9.75in x 6.5in b684-ch21 FA
(a) (b)
(c)
Fig. 21.9. Beach and dune reinforcement with geotextile sand containers. (a) Beach reinforcement
in Australia, (b) Stockton Beach Revetment,48 and (c) Dune Reinforcement in Wangerooge, North
Sea/Germany.54
(a)
(b)
Fig. 21.10. Wave reflection and overtopping performance of GSC-seawalls.28 (a) Overtopping
performance and (b) reflection performance.
August 4, 2009 9:34 9.75in x 6.5in b684-ch21 FA
fabric is made in a color that blends naturally with the beach environment.24
Heavy-duty UV stabilized nonwoven needle-punched geotextiles with high tenacity
polyester thread in all seams (min. 80% of basic fabric strength) have also been suc-
cessfully used.44 Despite large displacement that occurred during storms, the groin
continued to provide protection and even withstood abrasion and UV attack over
10 years.48 To reduce the costs in the case of larger groins and larger projects, the
groin core can be made of smaller container with lower requirements while strength-
ening the outer protective layer.24 Further example applications are provided by
Saathoff et al.48 and Restall et al.44,45
On the other hand, the use of sand as a quasi-impervious core instead of quarry
stone would result in an increase of
Fig. 21.13. Class of geocore structures in comparison to conventional rubble mound structures.
August 4, 2009 9:34 9.75in x 6.5in b684-ch21 FA
and the results are published by Oumeraci et al.30 Prediction formulae for wave
reflection, wave runup and overtopping, wave transmission, and armor stability
have been determined for the nonconventional breakwater in comparison with the
conventional rubble mound breakwater.30
Depending on the prevailing wave loads and degradation mechanisms, GSCs may
experience different types of failure modes: (i) hydraulic failure modes, (ii) geotech-
nical failure modes, and (iii) failure modes related to the geotextile itself (Fig. 21.14).
In the following, only the first type which is related to hydraulic stability under
wave loads will be addressed.
First, simple stability formulae without explicit account of the effect of defor-
mation will be proposed separately for the containers on the slope and the con-
tainers on the crest of the structure (called hereafter “slope containers” and “crest
containers”).
Second, the necessity of a better understanding of the processes responsible
for the deformation of the sand containers under wave loads as well as their
effect on the hydraulic stability is illustrated by some selected results from recent
research.
Finally, more detailed stability formulae are proposed which can also explicitly
account for the deformation effects, including a comparison with the simple stability
formulae.
Hs Cw
Ns = =√ . (21.1)
ρE
−1 ·D ξ0
ρW
With the surf similarity parameter ξ0 = tan α/ Hs /Lop expressed in terms of the
deepwater length Lop = gTp2 /2π (Tp = peak period of wave spectrum) the following
stability formula is obtained in terms of the characteristic size D of the container:
3/4 1/2
Hs · Tp · (tan α)1/2
D= 2π 1/4 ρE . (21.2)
Cw · d ρW − 1
The empirical parameter Cw was determined by stability tests in a large and a small
wave flume to Cw = 2.75. In Fig. 21.16 only the results of the large-scale model
tests are plotted to illustrate that the threshold curve between stable and unstable
containers is obtained for Cw = 2.75.
August 4, 2009 9:34 9.75in x 6.5in b684-ch21 FA
(a)
(b) (c)
Fig. 21.15. Hydraulic failure modes of slope containers. (a) Wave uprush and downrush on slope
containers, (b) pull-out effect in the FZK-large wave flume, and (c) pull-out effect in a dune
reinforcement (Courtesy of Heerten).
3/4
Hs · Tp
lc = (21.4)
ρE
1.74 · ρW
− 1 · sin(2α)
with
Fig. 21.16. Stability of slope containers from large-scale model tests (modified from Ref. 28).
August 4, 2009 9:34 9.75in x 6.5in b684-ch21 FA
The stability formula in Eq. (21.4) expressed in terms of the required container
length lc is plotted in Fig. 21.17 for a structure slope angle α = 45◦ in order to
illustrate the sensitivity to the significant wave height Hs and to the peak period
Tp . To compare the results with those of the Hudson-formula for rock armor, the
volume of the tested slope containers (V = lc ·0.46lc ·0.14lc = 0.065lc3) is considered.
For Hs = 1.5 m and Tp = 4 s, about the same required weight for slope containers
with ρE = 18,000 kg/m3 is obtained as for a rock armor with ρE = 2700 kg/m3
and Kd = 2.0. Using a 10 times higher Kd -value (Kd = 20) and applying Hudson-
formula in this specific case for slope containers would indeed provide the same
result as the proposed formula. However, such an approach is not applicable as
the stability of slope container is very sensitive to the wave period and the Kd -
value is expected to be a function of both wave height and wave period, due
to the deformation effect caused by wave action. In fact, if for the same wave
height Hs = 1.5 m the wave period is increased from Tp = 4 s to Tp = 6 s the
required length lc will increase by more than 20% (Fig. 21.17). Moreover, due
to the effect of deformation of the slope containers on the long-term stability, it
is not advisable to use unprotected slope containers for design wave heights of
Hs > 1.5–2.0 m.
(i) uplifting during the wave uprush process and shoreward displacement by the
wave overtopping flow,
(ii) dislodgement and pull-out effect similar to the mechanism observed for the slope
containers.
Due to the boundary conditions of the crest containers which are more critical
than those of the slope containers (no overburden from upper layers), it is expected
that the stability of the crest containers will be more critical than that of the
slope containers if the crest level of the structure is not high enough. The relative
freeboard Rc /Hs , therefore, represents the most important influencing parameter.
In fact, it was difficult to identify a noticeable effect of the surf similarity parameter
August 4, 2009 9:34 9.75in x 6.5in b684-ch21 FA
ξ0 on the stability number Ns as clearly observed for slope containers (Fig. 21.19).
However, plotting at the top left of Fig. 21.19, the stability number Ns against the
relative freeboard Rc /Hs shows that the stability number of the crest containers
increases with increasing relative freeboard Rc /Hs according to the following linear
relationship (Fig. 21.19):
Hs Rc
Ns = = 0.79 + 0.09 , (21.5)
ρE
−1 ·D Hs
ρW
Hs
lc = . (21.6)
ρE Rc
ρW
−1 0.79 + 0.09 Hs
sin α
In a similar way as in Fig. 21.17 for slope containers, the required container length
lc in Eq. (21.6) is plotted against the design wave height Hs for different relative
freeboard Rc /Hs = 0 − 2.0 (Fig. 21.20). Considering exemplarily a typical value
Rc /Hs = 1.2 and Hs = 1.5 m, a container length lc = 3.15 m and a container weight
W = 36.6 kN are obtained which are much larger than those obtained for the slope
containers in Fig. 21.17.
August 4, 2009 9:34 9.75in x 6.5in b684-ch21 FA
Fig. 21.20. Required length of crest containers for the hydraulic stability.
For this purpose it is necessary to address the following aspects: (i) hydraulic per-
meability of GSC-structures and its effect on the stability, (ii) wave loads and identi-
fication of the most critical loading case and location of the containers, (iii) internal
movement of sand fill and its effect on the stability, (iv) effect of the friction angle
between geotextile containers on the stability, and (v) effect of the container defor-
mations on the stability.
The content of this section and next section represents a brief summary of
selected key results which have essentially been obtained from comprehensive labo-
ratory studies, including several types of experiments in combination with numerical
August 4, 2009 9:34 9.75in x 6.5in b684-ch21 FA
(iii) The mode of placement of GSCs generally affects the permeability of the
GSC-structure (Fig. 21.22). However, randomly placed GSCs and longitu-
dinally placed GSCs have approximately the same permeability coefficient:
k = 2.4 cm/s and k = 2.3 cm/s, respectively. To achieve a lower permeability an
“interlaid” placement (gaps of the longitudinally placed containers are blocked
by transversally placed containers) is required, allowing to reduce the perme-
ability up to k = 1 cm/s and even to k = 0.5 cm/s.
(iv) A conceptual model and a systematic procedure to predict the hydraulic per-
meability is proposed by Recio and Oumeraci.38,41 Validation by experimental
data show that uncertainties less than 30% would be expected when applied
to prototype containers.
to illustrate that the highest horizontal force in seaward direction and the highest
vertical force in upward direction occur on the container located just below still
water level.
Using pressure measurements within the gaps between slope containers for both
breaking wave impacts and nonbreaking waves, it is found that the former are
less critical for the hydraulic stability than the nonbreaking waves rushing up and
down the slope and causing more damage. In fact, the impact pressure induced by
breaking waves is of much shorter duration. Moreover, it is strongly damped when
propagating inside the gap (Fig. 21.24).
(i) Noticeable sand movements occur only for larger waves that are capable to
substantially move the front part of the container up and downward during the
wave runup and rundown process [Fig. 21.26(b)]. After about 30 wave cycles the
internal sand movement decreases significantly due to the accumulated sand at
the seaward front of the container [Fig. 21.26(c)].
(ii) Due to the sand fill redistribution at the seaward front of the container, the
latter deforms, thus offering a larger impact surface area to the mobilizing wave
forces and reducing the contact area with neighboring containers [Fig. 21.26(c)].
August 4, 2009 9:34 9.75in x 6.5in b684-ch21 FA
Fig. 21.24. Impact pressure propagation inside a gap between geotextile sand containers.42
Fig. 21.25. Permeable transparent container for the investigation of internal movement of sand.38
August 4, 2009 9:34 9.75in x 6.5in b684-ch21 FA
Fig. 21.26. Internal movement and redistribution of sand fill in the transparent container.42
With the increased mobilizing forces and the decreased resisting forces, an incre-
mental lateral seaward displacement of the deformed container occurs (pull-out
effect) which causes again the start of the internal sand movement in a similar way
as during the first wave cycles [Fig. 21.26(a)].
These results have considerable practical implications as the internal sand
movement is responsible for the deformations of the container which affect
• the hydraulic stability by reducing the contact area between GSCs and by
increasing the drag and lift forces due to the increased exposed areas,
• the crest level of a GSC-structure. Even in the laboratory it is observed that the
height of sand containers is reduced by 4% after placement under water and by
further 6% due to wave action. As a result, a total reduction of the height of the
GSC-structure of about 10% was observed.38,41
Since these effects are strongly dependent on the adopted sand fill ratio, future
research and design guidance should be directed toward the definition of an optimal
sand fill ratio by accounting for the deformation properties of the geotextile and
August 4, 2009 9:34 9.75in x 6.5in b684-ch21 FA
by balancing the advantages and drawbacks of high sand fill ratio. Moreover,
this issue should also be explicitly addressed in future standards and guidelines
due to the considerable effect of the sand fill ratio on stability and long-term
performance.
(i) Reduction of the stability against sliding caused (a) by the increase of the drag
and uplift forces as a result of the increased exposed areas AS and AT (Fig.
21.28) as well as (b) by the decrease of the resisting force as a result of the
decreased friction area between the containers (Figs. 21.28 and 21.29).
(ii) Reduction of the stability against overturning caused by the increase of the
mobilizing drag and uplift forces as mentioned above, but also by the seaward
August 4, 2009 9:34 9.75in x 6.5in b684-ch21 FA
Fig. 21.28. Increase of exposed areas of drag and uplift forces and decrease of friction areas due
to container deformation.
Fig. 21.31. Reduction of effective contact areas between containers during wave action.37,42
which contribute to the resistance against dislodgment by friction) has shown that
the “effective” contact areas
(i) decrease due to the upward movement of the front part of the containers
(Fig. 21.31).
(ii) increase with increasing slope angle of the GSC-structure.
the deformation effect, and then correction factors are introduced to account for
the deformation effect. Finally, a comparative analysis between the results of the
stability formulae with and without deformation effect is provided for both slope
and crest containers.
It should be underlined that the new formulae proposed below were derived
analytically for the geometry of the containers commonly used in coastal structures,
i.e., with a container length lc which is twice the container width (lc /2) and five
times the container height (lc /5) resulting in the following relationships for the
volume of the container V and the application area of the drag and uplift forces AS
and AT , respectively [Fig. 21.32(a)].
V = 0.1lc3 ,
AS = 0.1lc2 , (21.7)
AT = 0.5lc2 .
These relationships provide the geometrical parameters that govern the resisting
forces (weight) and the mobilizing forces (drag, inertia, and uplift forces), thus
allowing to express the stability formulae in terms of the container length lc (see also
Sec. 21.4.1). If, however, other container geometries, and thus other relationships
which differ from those in Eq. (21.7) are adopted, the stability formulae can be
modified accordingly. Further indications on how to proceed with such modifications
and on the limitations of the proposed stability formulae will also be given below.
August 4, 2009 9:34 9.75in x 6.5in b684-ch21 FA
And with the relative density parameter ∆ = (ρE /ρW ) − 1 (see Plate 21.2):
∂u
µ ∆gV − 0.5CL AT · u 2
≥ 0.5CD AS u + CM V
2
. (21.9)
∂t
Given the considered container geometry, relationships similar to those in Eq. (21.7)
which provide V , AT , and AS as a function of the container length lc can be obtained
and substituted in Eq. (21.9) which is then solved to obtain either the required
length lc or the required mass WGSC of the container. Using, for instance, the
geometry described by Eq. (21.7), the stability formulae are given in Fig. 21.32 in
terms of the required length lc or mass WGSC of the container.
(b) Stability against overturning: A sand container is stable as long as the stabilizing
moment induced by the weight of the container under buoyancy FGSC remains larger
than the mobilizing moment induced by the drag, inertia, and uplift forces FC , FM ,
and FL (Fig. 21.33):
FGSC · rs ≥ FD · rh + FM · rh + Fc · rs (21.10)
(ρE − ρw )gV · rs ≥ 0.5ρw CD u2 As · rh
∂u
+ ρw CM V · rh + 0.5ρW CL u2 AT · rs . (21.11)
∂t
Given the considered container geometry, relationships similar to those in Eq. (21.7)
which provide V , As , and AT , but also the lever arms rs and rh as a function of
the container length lc can be obtained and substituted in Eq. (21.11). The latter is
then solved to obtain either the required length lc or the required mass WGSC of the
container. Using, for instance, the geometry described by Eq. (21.7), the overturning
stability formulae are obtained in terms of the required length lc or mass WGSC of
the container (Fig. 21.33).
More details on the force coefficients CD , CM , and CL as well as on further input
parameters required in the stability formulae summarized in Figs. 21.32 and 21.33
will be given in Sec. 21.4.3.2.
modes37,43 : sliding and overturning. The deformation factors are obtained as cor-
rection factors, describing the changes of (Plate 21.3):
The resisting force FR against sliding. The correction factor KSR is obtained as the
ratio of the effective weight contribution to the frictional force FR with and without
deformation.
The resisting moment against overturning. The correction factor KOR is obtained
as the ratio of lever arm rs of the container weight FGSC under buoyancy with and
without deformation effect.
The mobilizing drag, lift, and inertia forces contributing to sliding. The correction
factors of the drag force FD and lift force FL (KSCD and KSCL ) are obtained as the
ratios of the areas AS and AT with and without deformation effect. The correction
factor KSCM is assumed to be 1.0 since the container volume V remains constant.
The mobilizing moment induced by the drag, lift, and inertia forces. The correction
factor for the moments induced by the drag and lift forces (KOCD and KOCL ) area
is obtained as the ratio of describing the changes of both surface areas (AS and AT )
and lever arms (rm and rs ) of the drag and lift forces FD and FL . The correction
factor KOCM is obtained as the ratio of lever arms rsn of the inertia force with and
without deformation effect.
The values of the correction factors suggested in Table 21.4 are derived on the
basis of a number of simplifying assumptions (see Ref. 37 for more details).
August 4, 2009 9:34 9.75in x 6.5in b684-ch21 FA
Plate 21.4. Defining of parameters and typical values to be used in the stability formulae.
Among the most important assumptions the following are noteworthy: sand fill
ratio of 80% and a slope angle of the GSC-structure of 45◦ . Indications to account
for other slope angles and recommendations on further research to overcome most
of the simplifying assumptions are given by Recio.37
Moreover, the force coefficients CD , CL , and CM are also given as a function of
the Reynolds number for different locations and boundary conditions which may
represent different practical applications (scour protection on the sea bed, artificial
reef, slope containers, and crest container of a surface piercing structure such as
revetments, seawalls, groins, etc.). The proposed values of CD , CM , and CL have
been determined on the basis of systematic laboratory experiments.35
In Plate 21.4 the parameters used in the stability formulae described in
Figs. 21.32–21.35 are defined and typical values are also given, including some
remarks on the limitations of the suggested values.
in Sec. 21.4.1 are also considered to illustrate the difference with the results from
the more process-based formulae.
Depending on the range of the design wave height Hs , the following remarks may
be drawn from the comparison of the formulae for the slope containers (Fig. 21.34):
(i) For smaller design waves (Hs ≤ 1.5 m): the simple stability formulae proposed
in Sec. 21.4.2 are too conservative and the deformation effect on the stability
obtained from the comparison of the new formulae proposed on Sec. 21.4.3 are
relatively small.
(ii) For larger design waves (Hs ≥ 2.5 m): the simple stability formulae become
more unsafe with increasing wave height. The effect of deformation also
increases with the increase of the design wave height.
Depending on the range of the design wave height Hs the following remarks
may be drawn from the comparison of the stability formulae for the crest containers
(Fig. 21.35):
(i) For smaller design waves (Hs ≤ 1.5 m): the simple stability formula proposed in
Sec. 21.4.2 is slightly conservative and becomes unsafe as soon as the significant
wave height exceed 1.5 m. It can therefore be used instead of the more process-
based formula only for Hs ≤ 1.5 m. However, the effect of deformation is higher
than in the case of slope containers.
(ii) For larger design waves (Hs ≥ 2.5 m): the simple stability formula by Oumeraci
et al. (2003) becomes more unsafe with increasing wave height Hs . The effect
of deformation on the stability also increases with increasing wave height.
Comparing Figs. 21.34 and 21.35 also confirms that for commonly used relative
freeboards Rc /Hs in the order of 1.2, much larger containers are required for the
crest than for the slope of the structure. Moreover, it also shows that the effect
of deformation on the stability dramatically increases with increasing design wave
height and is much more pronounced for the crest containers than for the slope
containers.
However, several problems still need to be solved in order to make use of the full
potential of GSCs, particularly including (i) the long-term durability and lifetime
prediction and (ii) the hydraulic stability under severe wave action.
The facts, limitations and research needs related to the durability and lifetime
issue may be summarized as follows:
(i) Although field evidence (mostly nonexposed geotextile) is available over about
50 years, useful information extracted from nonretrieved samples is often very
incomplete or very limited as the results can hardly be transferred to other sites,
to present geotextile products and to time durations and conditions beyond
the experienced service time/conditions.
(ii) Results of accelerated tests — even in combination with those from site mon-
itoring and retrieved samples — are still very limited when trying to predict
lifetime of more than about 25 years. In fact, a systematic methodology to
combine both laboratory and field monitoring approaches for this purpose is
still missing.
(iii) Recommendations for future research in the mid-term and long-term should
focus on two directions: (a) improvement of the understanding of the degra-
dation mechanisms, including physical, biological, and chemical processes, both
isolated and in combination, and (b) development of a consistent framework
for the assessment of long-term durability and lifetime (up to 100 years and
more) based on the results of the above and including site monitoring, labo-
ratory testing, and theoretical/numerical modeling.
(iv) Meanwhile, in order to contribute to solve the present problems in practice,
the following two recommendations might be helpful. (a) Apply engineering
judgment based on the present knowledge of degradation mechanisms rather
than relying on “extrapolation approaches” to predict lifetime and (b) apply
where feasible and necessary well-established measures to enhance long-term
performance, including, for instance, appropriate stabilizers and additives (e.g.,
against UV-radiation), more robust geotextiles (e.g., multi-layer), and geo-
textile coating (e.g., against abrasion and vandalism), sand covering (e.g.,
against UV-radiation and vandalism and to enhance aesthetical aspect), rock
covering (e.g., against ice loads, debris, very high waves, UV, and vandalism)
and setup of a consistent maintenance plan. Moreover, the collaboration with
experienced companies with regard to geotextile production, manufacturing of
sand containers, filling, handling, and placing is also strongly recommended.
Regarding the hydraulic stability under severe wave action, the present state of
knowledge, the limitations, and the needs for future research may be summarized
as follows:
(i) The large experience and stability formulae available for rock and concrete
armor units cannot simply be transferred to GSCs, essentially due to the
deformation of the GSCs under very severe wave attack.
(ii) The deformations of the GSCs are essentially induced by the internal
movement of the sand fill of the containers.
August 4, 2009 9:34 9.75in x 6.5in b684-ch21 FA
(iii) The effect of the deformations of the GSCs on the hydraulic stability rapidly
increases with the severity of wave attack, depending on the size and sand
fill ratio of the containers as well as on the degree of exceedance of the wave
loads required for the inception of the internal movement of the sand fill.
(iv) Besides, the effect on the hydraulic stability, the internal sand movement in
submerged geotextile containers may lead to a substantial reduction of the
height of GSC-structures (up to about 10%) when subject to severe wave
attack.
(v) The deformation of GSCs affects the hydraulic stability, essentially due to
two mechanisms: (a) reduction of the contact areas between GSCs caused by
the uplift of the containers by wave action, thus decreasing the stabilizing
forces/moments and (b) increase of the surface areas exposed to drag and lift
forces which represent the main destabilizing forces/moments. Therefore, the
deformation effect should be accounted for explicitly in the stability formulae.
(vi) Friction between GSCs affects the hydraulic stability much more than com-
monly assumed in past and present design practice. Therefore, and due to
the possible changes in service life, friction should be incorporated explicitly
in stability formulae.
(vii) The hydraulic permeability of structures made of GSCs is not only important
for the prediction of the hydraulic performance (e.g., wave transmission,
runup, overtopping, etc.), but also slightly affects the hydraulic stability.
However, no clear correlation could be found between stability and perme-
ability for the range of practical permeability coefficients of GSC-structures in
the order of k = 1–3 cm/s. The permeability of a GSC-structure is essentially
determined by the gaps between the containers, so that the flow through the
sand fill itself can be neglected. Therefore, the hydraulic permeability essen-
tially depends on the mode of placement of the containers. For randomly
placed containers and longitudinally (in wave direction) placed containers,
the permeability coefficient is in the order of k = 2.5 cm/s.
(viii) The effect of breaking wave impact on sliding and overturning stability
of slope containers has been found much less than expected, due to the
potential of the GSCs to effectively damp impact pressure propagation
inside the gaps. More efficient to destabilize the slope containers are the
uprush and downrush of the longer nonbreaking waves and partially breaking
waves.
(ix) The proposed simple stability formulae derived in Sec. 21.4.1 on the basis
of the Hudson-formula take additionally into account the effect of the wave
period for the slope containers and that of the relative freeboard (Rc /Hs )
for the crest containers. In both cases, the effect of container deformations is
taken into account implicitly through the empirical parameters determined
from laboratory experiments. These formulae are conservative for waves up
to about Hs = 1.5 m and can thus be used for design wave heights not larger
than about 2 m. For higher waves (Hs > 2 m) the effect of deformation on
the stability becomes more important and must therefore be considered more
explicitly in order to ensure long-term stability performance.
August 4, 2009 9:34 9.75in x 6.5in b684-ch21 FA
(x) The more detailed stability formulae proposed in Sec. 21.4.3 do not only
take into account the most relevant processes, but also allow us to better
quantify the effect of deformation on the required size of the container as a
function of the incident wave height for both slope containers (Fig. 21.34) and
crest containers (Fig. 21.35). In fact, these new process-based formulae clearly
highlight the effect of container deformation which dramatically increases with
the design wave height and is much more pronounced for the stability of the
crest containers than for that of the slope containers. Moreover, they also show
that even for the commonly used relative freeboard of about Rc /Hs ≈ 1.2
much larger containers are required for the crest than for the slope of a GSC-
structure to ensure hydraulic stability.
(xi) Although significant advancement has been achieved in the understanding of
the processes and mechanisms responsible for the hydraulic failure of GSCs,
more systematic research is further needed to investigate and better control the
friction between GSCs, the effect of the sand fill ratio, the effect of the slope
steepness of the GSC-structure, the internal movement of the sand fill, the con-
tainer deformations and their more explicit consideration in both stability for-
mulae and numerical simulations. A fully coupled CFD and CSD model system
well-validated by experimental data will be needed as a necessary tool in combi-
nation with new laboratory experiments to achieve these goals.
Acknowledgments
References
1. ACT, Innovative Technology for Coastal Erosion Control — Subsurface Dune Stabi-
lization, Advanced Coastal Technologies, LLC (2006), http://bcs.dep.state.fl.us/ inno-
vative/report/appedic c/06 Subsurface Dune Protect.pdf.
2. J. Buckley and W. Hornsey, Woorin beach protection–chasing the tide sand fill tubes
versus sand filled containers, Proc. Int. Conf. Geosynthetics, Yokohama, Japan (2006),
pp. 761–764.
3. CEN, Guide to Durability of Geotextiles and Geotextile Related Products (European
Normalization Committee (CEN/TC189/WGS/N210/Paris), 1998).
4. EAG-CON, German recommendations on geotextile containers in hydraulic engi-
neering, WG-UG5 “Geotextile Containers of AK 5.1” Geosynthetics, Geotechnics and
Hydraulic Engineering (German Geotechnic. Society (DGGT), 2008).
August 4, 2009 9:34 9.75in x 6.5in b684-ch21 FA
23. MAG, Code of Practice — Use of Geotextile Filters on Waterways (Federal Waterways
Engineering and Research Institute (BAW), Karlsruhe, 1993), www.baw.de.
24. A. McClarty, J. Cross, L. Gilbert and O. M. James, Design and construction of coastal
erosion protection groyne using geocontainers, Langebaan, South Africa. Proc. Int.
Conf. Geosynthetics, Yokohama, Japan (2006).
25. NAUE, Scherversuche im Labor der Firma NAUE für Soil Filters Australia, NAUE
Fasertechnik (2002).
26. H. Nickels and G. Heerten, Objektschutz Haus Kliffende, HANSA, Band 137(3), 72–75
(2000) (in German).
27. H. Oumeraci, M. Bleck, M. Hinz and S. Kübler, Large-scale model test for hydraulic
stability of geotextile sand containers under wave attack, Leichtweiss-Institute for
Hydraulic Engineering and Water Resources, LWI Report No. 878 (2002) 62 and
Annexes.
28. H. Oumeraci, M. Hinz, M. Bleck and A. Kortenhaus, Sand-filled geotextile containers
for shore protection, Proc. COPEDEC, Colombo, Sri Lanka (2003).
29. H. Oumeraci, J. Grüne, H. Sparboom, R. Schmidt-Koppenhagen and Z. Wang,
Investigations on scour and scour protection for monopile foundation of wind off-
shore turbines, Forschungszentrum Küste (FZK), Res. Report (2007) 79 and Annexes
(in German).
30. H. Oumeraci, A. Kortenhaus and K. Werth, Hydraulic performance and armour sta-
bility of rubble mound breakwaters with core made of geotextile sand containers —
Comparison with conventional breakwaters, Proc. Int. Conf. Coastal Structures,
Venice, ASCE (2007).
31. K. W. Pilarczyk, Geosynthetics and Geosystems in Hydraulic and Coastal Engineering
(A.A. Balkema, Rotterdam, The Netherlands, 2000).
32. J. Recio and H. Oumeraci, Effect of the deformation on the hydraulic stability of revet-
ments made of geotextile sand containers, Proc. Int. Symp. “Tsunami Reconstruction
with Geosynthetics”, Bangkok, Thailand (2005), pp. 53–68.
33. J. Recio and H. Oumeraci, A numerical study on the hydraulic processes asso-
ciated with the instability of GSC-structures using a VOF-RANS model, Leichtweiss-
Institute for Hydraulic Engineering and Water Resources, LWI Report No. 941 (2006).
34. J. Recio and H. Oumeraci, Processes affecting the stability of revetments made
with geotextile sand containers, Proc. Int. Conf. Coastal Engineering, ICCE 2006,
San Diego, USA (2006).
35. J. Recio and H. Oumeraci, Geotextile sand containers for coastal structures, hydraulic
stability formulae and tests for drag, inertia and lift coefficients, Leichtweiss-Institute
for Hydraulic Engineering and Water Resources, LWI Report No. 936 (2006).
36. J. Recio and H. Oumeraci, Preliminary experiments and numerical simulations of
solitary wave acting on a submerged-filter-reef, Leichtweiss-Institute for Hydraulic
Engineering and Water Resources, LWI Report No. 945 (2006).
37. J. Recio, Hydraulic stability of geotextile sand containers for coastal structures —
Effect of deformations and stability formulae, PhD-Thesis, Leichtweiss-Institute for
Hydraulic Eng., Technical University Braunschweig (2007).
38. J. Recio and H. Oumeraci, Permeability of GSC-structures, model tests and analyses,
Leichtweiss-Institute for Hydraulic Engineering and Water Resources, LWI Report
No. 943 (2007).
39. J. Recio and H. Oumeraci, “Coupled” numerical simulations on the stability of coastal
structures made of geotextile sand containers, Leichtweiss-Institute for Hydraulic
Engineering and Water Resources, LWI Report No. 942 (2007).
August 4, 2009 9:34 9.75in x 6.5in b684-ch21 FA
40. J. Recio and H. Oumeraci, Effect of the deformations on hydraulic stability of coastal
structures made of geotextile sand containers, Geotex. Geomemb. J. 25, 278–292
(2007).
41. J. Recio and H. Oumeraci, Hydraulic permeability of GSC-structures: Laboratory
results and conceptual model, Geotex. Geomemb. 26, 473–487 (2008).
42. J. Recio and H. Oumeraci, Processes affecting the hydraulic stability of GSC struc-
tures: Experimental and numerical investigations, Coastal Eng. 56, 632–658 (2009).
43. J. Recio and H. Oumeraci, New stability formulae for coastal structures made of
geotextile sand containers, Geotex. Geomemb. 56, 260–284 (2009).
44. S. J. Restall, L. A. Jackson, G. Heerten and W. P. Hornsey, Case studies showing the
growth and development of geotextile sand containers — An Australian perspective,
Geotex. Geomem. 20(5), 321–342 (2002).
45. S. J. Restall, W. P. Hornsey, H. Oumeraci, M. Hinz, F. Saathoff and K. Werth,
Australian and German experiences with geotextile containers for coastal protection,
Proc. Euro-Geo 3, Munich, Germany (2005), pp. 141–146.
46. A. L. Rollin, Long term performance of geotextiles, 57th Canadian Geotechnical Conf.,
Session 4D (2004), pp. 15–20.
47. RPG, Guidelines for Testing Geotextiles for Navigable Waterways (Federal Waterways
Engineering and Research Institute (BAW), Karlsruhe, 1994), www.baw.de.
48. F. Saathoff, H. Oumeraci and S. Restall, Australian and German experiences on the
use of geotextile containers, Geotex. Geomem 25, 251–263 (2007).
49. H. F. Schroeder, H. Bahr, P. Herrmann, G. Kneip, E. Lorenz and I. Schmuecking,
Durability of polyfine geosynthetics under elevated oxygen pressure in aqueous liquids,
Proc. Second European Geosynthetics Conf. EUROGEO, Bologna, Italy (2000).
50. S. K. Shukla and J. H. Yin, Fundamentals of Geosynthetics Engineering (Routledge,
2006).
51. M. Solton, B. Leclerq, J. L. Paute and D. Fayoux, Some answer’s components on
durability problems of geotextiles, Proc. 2nd Conf. Geotechnics 2, 553–558 (1982).
52. G. H. Troost, G. Den Hoedt, P. Risseeuw, W. Voskamp and H. M. Schmidt, Dura-
bility of a 13 year old embankment reinforced with polyester woven fabric, 5th
Int. Conf. Geotextiles, Geomembranes and Related Products, Singapore, IGS (1994)
pp. 1185–1190.
53. G. van Santvoort (ed.), Geotextiles and Geomembranes in Civil Engineering (A.A.
Balkema, Rotterdam, The Netherlands, 1994).
54. J. Vohlken, H. Lind and J. Witte, Dune reinforcement with geotextile sand containers,
HANSA 4, 60–62 (2003) (in German).
55. J. D. M. Wisse, C. J. M. Broos and W. H. Boels, Evaluation of the life expectancy
of polypropylene geotextiles used in bottom protection structures around the Ooster
Schelde storm surge barrier — A case study, Proc. 4th Conf. Geosynthetics (1990),
pp. 697–702.
56. J. Wouters, Slope Revetment — Stability of Geosystems, Delft Hydraulics Report H.
1939 (1998) Annex 7 (in Dutch).
57. F. F. Zitscher, Kunststoffe für den Wasserbau, Bauingenieur-Praxis, Heft 125, Verlag
von Wilhelm Ernst & Sohn, Berlin, München, Düsseldorf (1971).
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
Chapter 22
The chapter describes typical features of low crested breakwaters, their hydraulic
stability, effects on waves, induced circulation, erosion, and problems related to
construction and maintenance, providing basic tools for design.
22.1. Introduction
As low crested breakwaters (LCBs) we mean breakwaters that are frequently over-
topped or weakly submerged. They are in any case breakwaters, i.e., the height over
the bed of these structures is sufficient to cause systematic wave breaking on the
crest (submergence less than one significant wave height), but they are also low,
i.e., not so high to make overtopping rare (freeboard less than one significant wave
height). In this sense they are distinguished from artificial reefs (deployed mainly to
enhance fisheries) or bed protections on one hand, that are deeply submerged and
do not cause breaking, and by usual harbor breakwater on the other hand that are
high enough not to cause significant overtopping.
LCBs mimic in some manner the functioning of coral reefs, that however have
usually a much wider crest and a narrower range of submergence/emergence due
to their generation mechanism. As a consequence they are sometimes named also
“reefs.”
LCBs are usually built with natural rock for the defence of beaches (Fig. 22.1),
even if some examples of LCBs protecting the main breakwater from highest waves
or protecting an external harbor from excessive agitation and currents (Fig. 22.2)
can also be found.
Beach defence breakwaters are usually built in a few meter water depth within
the surf zone at design conditions. The freeboard (positive or negative) is usually of
the same order of magnitude as the tidal range, i.e., in contrast with most harbor
breakwaters, freeboard variability in lifetime is substantial.
601
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
Fig. 22.1. Detached LCBs protecting the beach of Skagen (DK, 57◦ 44 07 N 10◦ 37 50 E; coordi-
nates are given to let the reader obtaining a plan view from Google-Earth).
Fig. 22.2. Jetties protecting Ijmuiden entrance (NL, 52◦ 27 55 N 4◦ 32 18 E) during a storm surge
event. From http://www.vandermeerconsulting.nl/.
As it will be shown later, the height of the structure in this case is not greater
than a few armor stones, posing serious problems in building the traditional filter
structure separating the armor from the sandy bed.
LCBs are usually located in the surf zone where design wave height is depth
limited and, therefore, lasts for long periods. It is therefore unsafe to accept any,
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
even small, permanent damaging rate in design conditions, since the damage will
cumulate in a few years at a significant level.
For all the reasons sketched above, LCBs cannot be treated as a small scale
version of rubble mound harbor breakwaters, even if they share some evident
common features. Due to their wide use in beach defence, however, they deserve
some special attention.
This contribution is dedicated to highlight differences between high and low
breakwaters. We summarize the experience accumulated in the last two decades
on the subject, making use of several results focused within DELOS research
project, that we had the opportunity to coordinate; we acknowledge at the same
time the contribution of project partners, evident to everyone will read one at
least of the two major deliverables of the project: the special issue of Coastal
Engineering8,24,28,29,35,46,57,68 and the Design Guidelines.5
The sections are arranged as follows: Section 22.2 — typical features,
Section 22.3 — hydraulic stability, Section 22.4 — wave reflection and transmission,
Section 22.5 — overtopping, Section 22.6 — induced circulation, Section 22.7 —
scour and erosion, Section 22.8 — construction and maintenance problems.
LCBs were used since long time as detached emergent breakwaters for the defence
of beaches. The low crest was motivated by economy reason in conditions where
some agitation in the protected area could be accepted and sometimes required in
order to avoid silt sedimentation. The problem of visual impact was not posed at
that time.
Later during the last two decades of the 20th century, in Europe particularly
along the Mediterranean coasts subject to intensive tourism exploitation, the aim
of reducing visual impact of defence works promoted the use of submerged break-
waters.
Figure 22.3 shows one of the most deeply submerged breakwater constructed
in Pellestrina (near Venice, IT) and completed in 1997.29 It is a 9 km long defence
work, where a 4.5 × 106 m3 capital nourishment of 0.2 mm sand was protected by
a LCB system. The offshore part of the LCB is a submerged breakwater defending
the beach against offshore sand losses. The groins limit the long-shore displacement
of sand and are characterized by an emerged part in the aerial beach zone and a
submerged continuation to reach the submerged longitudinal breakwater. Average
water depth at structure is 4.5 m. Design submergence is −1.5 m a.m.s.l. Spring
tidal range near Venice is 1.0 m, but high tide combined with storm surge caused
by Scirocco wind may reach 1.5 m a.m.s.l. Incident significant wave height with
return period 10 years is 3.70 m, therefore, extreme waves at the structure are depth
limited.
Since the construction of Venice outlet jetties (end of the 19th century), the
beach, deprived of any natural sand nourishment, suffered a slow but continuous
erosion. The beach, after the construction of the defence system, showed to be rather
stable with low southerly directed littoral drift.
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
Fig. 22.3. Pellestrina (IT, 45◦ 15 57 N 12◦ 18 10 E) offshore submerged breakwater and groins.
Courtesy of Consorzio Venezia Nuova.
Fig. 22.4. Lønstrup defense section (DK, 57◦ 28 37 N 9◦ 47 44 E). From Burcharth et al.6
Figure 22.4 shows the section of a Danish beach protection example at Lønstrup
in northern Jutland.29 Mean tidal range at the site is 0.3 m; exceptional high tide
may reach 1.5 m; wave are usually depth limited. Longshore net transport is great,
approx. 600 × 103 m3 /year. Beach sand size is 0.2 mm. A 1.1 km long segmented
breakwater was constructed to protect the cliff and the village from an ongoing
1.5 m/year erosion. To maintain the shoreline position, a 20–30 × 103 m3 /year nour-
ishment was required even after the breakwater construction.
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
Fig. 22.5. Sea Palling defence section (UK, 52◦ 47 32 N 1◦ 36 11 E). Adapted from Halcrow;21
see also Sea Palling site at http://argus-data.wldelft.nl/index.html, reefs located north of the video
station.
breakwater. In 1972, two years after the construction ended, a replacement of more
than one-third of the placed tetrapods was necessary, and a similar replacement was
required in 1974. In 1979, the last 90 m of the wave wall were destroyed. Breakage of
tetrapods was observed in prototype that, together with the insufficient character-
ization of design waves (regular rather than irregular waves with equal significant
wave height), were recognized as the main causes of the failure. The construction
of some protection works of the roundhead was decided and, after comparison of
several alternative solutions in a hydraulic model, the construction of a submerged
breakwater with section similar to the previous one but shifted 52 m offshore was
selected, aiming to reduce through breaking the wave height incident on the wave
wall. The construction of the selected design (see Fig. 22.6) ended in 1982; the
crest was constructed 1 m higher than shown in the drawing anticipating a 2 m set-
tlement. The behavior of the submerged breakwater was evaluated satisfactory in
1993 either regarding its stability under wave action or regarding the protective
effect on tetrapod stability and wave wall overtopping. The prototype verification
appears significant since during the 11 year monitoring period a 9.0 m peak signif-
icant wave height event occurred (January 1985), reaching almost design intensity.
PERFIL 6 - 6
L = 65,0 m - Solução 1
0 25 50 m + 15,0↓
L = 40,0 m - Solução 2 + 11,5↓
t.
L = 52,0 m - Solução 3 (ADOPTADA) MACIÇO DE PROTECÇÃO OBRA A PROTEGER 0t
4 S4 ↓+7,25
3 DE .
L + 4,0 (PMAV)↓ PO 3t t ↓+3,5 T. O T. ↓+4,0
Á >
12,0 TR .
Z.H.↓ + 1,0↓ TE NR T. O T. EN
E SELECIONADO EN R . Z.H.
1 1
4
R. > 3 3
4
2 > tt
0, 5 . 2
2 PARALELEPÍPEDOS DE BETÃO ~ 90 tt.
2 PARALELEPÍPEDOS DE BETÃO ~ 90 tt. 1 1 T. O T. tt.
7,0
1 5
4
- 12,0 ↓ ENR. 1 - 5 tt. ENR. < 1tt.
6
−1
− 15
4
−1
6
- 105,0
( - 12,0)
− 13
− 12
(0,0)
L
(0,0)
(+ 5,0)
,5)
(+ 7
(+ 15,0)
(+ 7,25)
(+ 0,5)
(+ 6,8)
(- 5,0)
6 PLANTA
0 50 100 m POSTO A
Fig. 22.6. Terminal section and layout of the Leixões harbor breakwater (PT, 41◦ 10 22 N
8◦ 42 30 W). From Vera-Cruz and Reis De Carvalho.58
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
A low breakwater compared to a high one is less exposed to breaker impact and
to wave downrush, since the breaker falls on the crest often covered with water
and the impact finds vent shorewards in the absence of a wave wall. This section
is dedicated to describe this effect and to provide proper tools for armor design.
Systematic laboratory experiments regarding rock armor stability in low crested
breakwaters were carried out by several authors. Givler and Sørensen20 provided
results of 2D experiments with regular waves, van der Meer51 and Burger7 with
irregular waves. Vidal et al.59,60 Kramer and Burcharth,27 and Kramer et al.28
performed experiments with long crested and short crested irregular waves in a
wave basin observing also damages at roundheads. A synthetic presentation of these
results with a critical analysis is presented in Burcharth et al.6 on which this section
is essentially based. Experiments are described in detail in the referred papers. All
experiments are carried out on a fixed bed, where no settlement or erosion can take
place, and therefore they do not provide information on these failure modes.
Results refer to where the damage takes place and how severe the damage is.
Assuming equal armor stone size at trunk and roundhead, all authors agree on
where first and how damage occurs, see Fig. 22.7.
• For submerged or zero freeboard breakwaters (Rc ≤ 0), hydraulic damage occurs
first at trunk crest and results in a shorewards displacement of stones; roundhead
in not more critical than the trunk and damage is distributed all around the head.
• For slightly emerged breakwaters (0 < Rc < Dn50 ), the crest remains the most
critical part of the trunk with some stones thrown shorewards and other dragged
1
Least stable section, Eq. (22.1)
0
-3 -2 -1 0 1 2
Fig. 22.7. Partial and global start of damage curve, from Burcharth et al.5
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
down the offshore slope; leeward roundhead becomes the first damaged part of a
breakwater.
• For emerged LCBs (Dn50 < Rc ), the seaward slope is the most critical part of
the trunk with most stones dragged down the slope and the leeward roundhead
is the first damaged part of a breakwater.
• The crest width, built with stones of the same size as the armor slope, has no
evident effect on hydraulic stability.
• In all the mentioned tests, slope was 2:3 or 1:2 and no significant effect of the
slope was observed.
Burger7 showed that gradation and stone shape within reasonable limits
(D85 /D15 < 2.5, L/D < 3) have not an evident effect on stability.
Based on all the available information Burcharth et al.6 proposed the following
formula representing the initiation of damage in some part of the breakwater
2
Hs Rc Rc
= 0.06 − 0.23 + 1.36 (22.1)
∆Dn50 Dn50 Dn50
for relative crest freeboard less than 1.9. Above this value the stability number
remains constant and equals to 1.14, the minimum of Eq. (22.1).
This value is significantly lower than the one that can be derived from Hudson
formula (∼=1.5). This is due partially to the fact that no damage is assumed in
this formula against a moderate damage (5%) accounted for in Hudson formula
and partially to the deliberate use of Hs as the characteristic wave height (Hudson
formula is used sometimes with H1/10 and some other with Hmax ).
Most LCBs are placed in the littoral zone and in design conditions are subjected
to depth breaking waves. For many of these the mean water level varies in a relatively
wide range, regularly due to tide and occasionally due to storm surge. If waves
are depth limited and the breaking limit is represented as a limit ratio γ between
significant wave height and the local water depth, among the incident wave height
Hi , the water depth at structure toe hs , the structure height hc , and structure
freeboard Rc the relation holds
Hi = γ · hs = γ · (hc − Rc ), (22.2)
where the breaker index γ depends on seabed slope, wave steepness, and the char-
acteristic wave height in use (significant, extreme . . . ). Figure 22.8 shows a typical
situation at failure boundary. The shape of the two domains (failure domain and
possible sea conditions) is such that stability is assured if breaking limit does not
cross the start of damage curve or is tangent to it as a limit.
This leads to the hydraulic stability condition under variable water level and
depth limited waves characterized by the significant wave height:
Dn50 γs /∆
≥ . (22.3)
hc 1.36 − (γs /∆ − 0.23)2 /(4 · 0.06)
If the tangency point does not lie within the possible sea level range or breaking
conditions are unlikely, the condition contains an implicit safety margin.
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
Fig. 22.8. Failure representation under variable water level, from Burcharth et al.6
Results for the minimum stone size according to Eq. (22.3) for relative stone
density ∆ = 1.6 and different values for significant breaker index γs are reported
in Table 22.1. The breaker index values are evaluated according to van der Meer52
with sop = 0.03 and hs /Lop = 0.05.
For a gentle foreshore slope the following simple and cautious rule of thumb is
found
The ratio Dn50 /hc imposes serious restrictions to the structure design. A double
armor layer with Dn50 /hc > 0.4 require that at least part of the filter is placed
below the natural bed level; this could be beneficial in case a bed erosion might
occur. A value Dn50 /hc > 0.3 implies that there is no sufficient space for a core and
the two layer armor stones rest directly on a coarse filter. Only when Dn50 /hc < 0.2
there is space for a conventional core and filter layer.
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
It should be noted that Eq. (22.1) is based on model tests with sea-side slope
1:2 (some 1:1.5). For milder slopes, smaller armor stones can presumably be stable.
According to Hudson formula, for a slope 1:5, an increase of the stability number for
emerged breakwaters by a factor 1.36 is expected; a smaller increase in stability is
argued in the case of submerged crest because the crest (including edges) is the most
damaged part in this case. Also, some laboratory test cases support this conclusion
(Ostia IT, Ferrante et al.18 ; Sea Palling UK, Leasowe Bay UK, Barber and Davies1 ).
Offshore and inshore the armor slopes, a toe berm is normally constructed with
the function of supporting armor and preventing the propagation of damage resulting
from scour or settlement. Regarding berm stability no specific analysis was carried out
on low breakwaters. Design rules suggested by van der Meer et al.56 were evaluated as
reasonably safe for the offshore berm of low breakwaters in Burcharth et al.6 For a suffi-
ciently wide crest, the breaker will fall on the crest and the inshore toe may be designed
with the same material as the offshore one. For narrow and submerged breakwaters,
the breaker caused by the offshore slope can fall beyond the crest causing a severe solic-
itation of the inshore toe berm; care should be paid in this case.
It is worth noting that the above formulae are derived from fixed bed models,
and do not represent any effect of bed mobility and/or erosion, whereas in most
prototype conditions the fine sand seabed is subjected to intense mobilization and
transport. Indeed, the observed damage of prototype rubble mound (as described
by Lamberti et al.29 ) is often the consequence of geotechnical or morphodynamic
instabilities. First of all, when proper bed protection is not used, dumping of stones
onto bed cause their rapid sinking already during construction followed by a slower
settlement. Moreover, breaking waves and related strong currents produce local
scour (Sumer et al.46 ) which affects toe stability either directly (stone sliding in
scour hole) or indirectly (increase of water depth and breaker height). To avoid
sinking of the rubble stone material into a sandy bed it is necessary to separate
the two materials with an appropriate filter: widely dispersed small stone layers,
geotextiles, or mattresses.
If sinking of stones stops within construction period, the rubble mound volume
used for construction must be oversized and that is all; if sinking last longer, the
structure must be designed and initially built higher (see, e.g., the Leixoes example).
The overheight will correspond to the anticipated settlement, and depends on seabed
characteristics, construction method, and structure height. For instance, a large
rubble mound built on muddy seabed by means of floating equipment demands a
large overheight, whereas a low mound structure placed on a coarse sandy seabed
by landbased equipment running on already placed material demands much less
overheight (but great overvolume) as settlement will be almost completed during
construction.
Burcharth et al.6 describes in detail some settlement cases. Statistics from a
much wider set of cases from Japan is presented in Uda.49 Depending on the nature
of the bed material the statistics in Table 22.2 was observed.
The fraction of settling breakwater increases from 10% on rock (probably degra-
dation of armor units), up to 43% on gravel, 63% on sand and 100% on sylt. The
settlement height is highly variable in the range 0–2.80 m, with average and modal
value around 1.00–1.20 m.
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
Settlement
Bed material Exist Nil Unknown Total
Results are also given for scattering (displacement, hydraulic damage) of armor
units. Globally scattering was observed in 14.2% of the 1,552 cases, but the fre-
quency varies significantly between breakwaters showing evident settlement, for
which the scattering frequency was 17.9%, and those for which no settlement was
observed, for which scattering frequency was 9.4%. The frequency difference is highly
significant, proving that scattering and settlement are related. In our experience
hydraulic damage, i.e., the removal of stones from the breakwater crest, is able to
justify only a minor part of crest settlement. On the other side, bed erosion around
the structure and stone sinking preferably along the structure perimeter are the
cause of block dilation on the slopes and therefore of reduced interlocking and insta-
bility. The mutual influence of settlement and block scattering is therefore bilateral
and can explain the evident correlation observed.
The main effect of a LCB compared to a high one is that a significant fraction of
incident energy can pass over its crest and generate waves behind the structure.
The increase of wave transmission leads, on one hand, to lower beach protection
from erosion or to less calm areas for swimming purposes inshore the LCBs. On
the other hand, it guarantees a high degree of water recirculation and thus a water
quality improvement within the protected cell.
Transmission, as well as reflection, can be characterized by a global intensity
factor, normally the transmission and reflection coefficients, and by a description of
energy distribution within the energy containing frequency range.
The EU-projects DELOS (www.delos.unibo.it) and CLASH (www.clash-eu.org)
collected and generated wide datasets from tests on all kind of structures, where
the transmission and reflection coefficients were available in almost all cases. An
extensive and homogeneous database on wave transmission (van der Meer et al.57 )
and on wave reflection (Zanuttigh and van der Meer64 ) were prepared, whose
analyses produced for LCBs the major outcomes synthesized below.
For rubble mound LCBs, to be able to estimate the transmission coefficient Kt
also in presence of submerged structures with very wide crests, the following two
formulae are given for design conditions (Briganti et al.3 ; van der Meer et al.57 ).
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
with imposed lower and upper limit values Ktmin = 0.075, Ktmax = 0.8 and
−0.65
Rc Bc
Kt = −0.35 + 0.51 (1 − e−0.41ξop ) (22.5b)
Hsi Hsi
with imposed lower and upper limit values Ktmin = 0.075, Ktmax = 0.8, and limi-
tations: 1 < ξop < 3, 1 < Bc /Hsi < 4.
Transmitted spectra are often different from incident spectra. Waves breaking
over a LCB may generate two or more transmitted crests on the lee side per incident
crest. The effect is that more energy is present at higher frequencies than in the
incident spectrum. Usually the peak period is quite close to the incident one, but the
mean period may decrease considerably. If the reduction of wave energy is mainly
led by the dissipations due to the flow through the armor layer, however, higher
frequencies may be cut. A first analysis on this topic can be found in van der Meer
et al.,53 who developed a simple model where in average the 60% of the transmitted
energy is present in the range f < 1.5fp and the other 40% of the energy is evenly
distributed between 1.5fp and 3.5fp. This scheme was analyzed little further by
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
Briganti et al.3 and by van der Meer et al.,57 concluding that the redistribution is
applicable only to submerged rubble mound structures.
Particular care to the representation of the shape of the transmitted wave
spectra was paid by Lamberti et al.30 The transmitted wave is reconstructed from
the incident wave as the sum of the overtopped and filtered components. Incident
waves transform seawards the structure into overtopping events and then regen-
erate, leeward the structure, due to perturbations caused by impulsive overtopping
volumes. The overtopping and filtration discharges are added and converted into
displacements of an ideal wave maker placed leeward the structure. A sample of
the reconstructed transmitted wave spectrum with comparison of an experimental
one is given in Fig. 22.9. The approach describes fairly the spectrum modification.
Related to spectral change is the effect on obliquity described below.
Similarly, for traditional structures several formulae exist that predict the
reflection coefficient Kr as function of the Iribarren–Battjes surf similarity
parameter ξ and are empirically determined for different types of armor units,
of tested conditions and layouts. The only formula that can be applied, with
varying coefficients (a, b) for different kinds of structures is the one by Seelig and
Ahrens42
a1 · ξ 2
Kr = . (22.7)
ξ 2 + b1
The coefficients (a1 , b1 ) are, respectively, equal to (0.6, 6.6) for rubble mound struc-
tures, and to (1.0, 6.2) for smooth slopes.
Fig. 22.9. Measured transmitted spectra (solid line with +), transmitted spectra components
due to filtration (dot line) and to overtopping (dash line), in case of abundant overtopping. From
Lamberti et al.30
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
A more recent formula by Zanuttigh and van der Meer64 exists that
• can be applied to all kinds of structures or revetments in design conditions, pro-
vided only that the roughness factor γf is known,
• represents physical bounds,
• was validated against a wide database.
This formula reads:
1.20
1.00
0.80
K rm/K rc
0.60
UPC
0.40 UCA
UPD
UFI
0.20
Seabrook
0.00
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5
Rc/Hsi
Fig. 22.10. Measured to computed values based on Eq. (22.10) of the reflection coefficient with
varying degree of overtopping.
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
obtained from Eq. (22.6). Similarly to van der Meer et al.,57 Zanuttigh and van der
Meer64 proposed an extension of their formula to rubble mound LCBs by including
a simple linear regression on Rc /Hsi . Reflection is not influenced if the relative crest
height is > 0.5 and decreases to 0 when submergence is twice the wave height.
The formulae (22.5), (22.6), and (22.7) are given for perpendicular waves only.
Based on a limited set of data, van der Meer et al.57 and Wang et al.61 analyzed the
effects of oblique wave attack on wave transmission. With increasing incident wave
angle βi , Kt tends to decrease and the obliquity of the transmitted wave βt is a
little smaller than the incident βi . The influence of wave obliquity on transmission
is more evident for smooth structures, where it results proportional to cos2/3 β, and
hardly observable for rubble LCBs. The deviation of transmitted waves (obliquity
reduced to about 80%) can be interpreted as refraction of the average wave due
to spectral change and reduced nonlinearity of transmitted waves. Explanation of
effects is provided in Wang et al.63
Analysis of the effect due to wave obliquity on Kr can be found in Wang et al.,62
who derived for smooth LCBs a Kr reduction proportional to cos2/3 β.
Overtopping refers to the process of water passing over the structure crest. If the
structure is not continuously submerged the process can be naturally interpreted
and made up of several events strictly related to the passage of wave crests and
characterized by a certain mass and momentum (or volume and velocity). The con-
sequences of overtopping are dependent on specific statistics of the process related
to the considered effect: for instance, the extreme momentum of one overtopping
crest may be responsible for damaging exposed elements, whereas the mean volume
per unit time (or mean overtopping discharge) is responsible for flooding. Never-
theless, the last is the standard parameter for characterizing overtopping intensity,
probably because it is the easiest to measure.
The total elapsed time and the total volume of water passing over the structure
are simply the sum of elapsed time and passed volume within each wave. This
trivial observation is the basis of the relation among mean overtopping discharge
qot , overtopping probability Pot , mean overtopping volume ·Vot , and mean period
of well formed waves Tm :
Volumes Not · Vot Vot
qot = waves = = Pot , (22.9)
waves Periods Nw · T m Tm
where the overtopping probability Pot is the ratio between the number of over-
topping waves Not and the number of waves Nw .
Since every wave whose runup exceeds the structure crest (and only these waves)
cause overtopping, the overtopping probability at the offshore edge of the crest
is equal to the probability that a single wave runup is higher than the crest. For
irregular waves, van der Meer and Stam54 suggested for runup at rubble mound
structures a Weibull probability distribution, with parameters k1 and k2 related to
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
incident waves and slope characteristics, from which the overtopping probability
can be derived:
Prob(Ru ≤ z) = 1 − exp(−(z/k1 )k2 ) ⇔ Pot = exp(−(Rc /k1 )k2 ), (22.10)
where k1 = 0.4Hsi s−0.25 om cot gα−0.2 , som is the mean wave steepness and α the
−0.75
√ slope; k2 = 3.0ξm
structure offshore
−0.3 P
for plunging waves (ξm < 2.5) and k2 =
0.52·P ξm cot α for surging waves (ξm > 2.5), ξm is the Iribarren number based
on mean wave period and P is the notional structure permeability. Later, van der
Meer and Janssen55 provided a simpler expression for the runup distribution at
dikes: the shape parameter is constant k2 = 2, i.e., the distribution is Rayleighian,
and the location parameter is proportional to incident wave height k1 = 0.81 · γh ·
γf · γβ · min(γb · ξop , 2) · Hsi . The Rayleigh distribution fits in any case rather well
the distribution of runup.
In Eq. (22.9) Pot is the factor controlling the order of magnitude of mean over-
topping discharge and is the expression of the main effect of relative crest elevation
Rc /Hsi , as it can easily vary by orders of magnitude as a consequence of modest
variations of structure crest elevation, see the CLASH database (Steendam et al.44 ;
wwwclash-eu.org).
The data obtained in wave basin by Zanuttigh and Lamberti67 for perpendicular
and oblique layouts of LCBs show that the pattern and values of Pot are in good
agreement with predictions from Eq. (22.10) and that the overtopping volumes can
be well approximated by a Weibull distribution, as in van der Meer and Janssen.55
The mean value of such distribution Vot shows a more or less parabolic relation
with the runup Ruot , that is the median potential runup of waves causing over-
topping (Fig. 22.11). Ruot is derived from Eq. (22.10) as the value giving exceeding
probability just half the overtopping probability. As suggested by Pilarczyk,38 when
overtopping is rare, Vot is almost proportional to (Ruot − Rc )2 ; this implies a fixed
shape of the overtopping crests. The scaled volumes, however, increase significantly
as soon as Pot exceeds 0.4–0.5.
The location parameter k1 of the volume distribution (typical overtopping
volume) can be obtained from the mean value as k1 = Vot /Γ(1 + k2−1 ).
The shape parameter k2 of the distribution increases with increasing Pot
(Fig. 22.12). Its values range from something below 1 up to 3; the lowest values
refer to the case of “rare” overtopping and the highest values to very frequent
overtopping. If due attention is paid to difference in overtopping frequency and
parameter uncertainties, the lowest figures are not substantially different from the
value 3/4 suggested by van der Meer and Janssen.55
In conclusion, for LCBs that are overtopped by most waves, not just by the
highest (Pot > 0.4 or Rc < k1 ), the crest level is so low in the approximate Rayleigh
distribution of runup that the shape of all overtopping characteristics distributions
become more symmetric and less variable (the shape parameter in a Weibull distri-
bution becomes greater) and the overtopping events become longer. The combined
model of a Weibull (or Rayleigh) distribution of runup, providing the overtopping
probability, and a shape model for overtopping crests, as the one suggested by
Pilarczyk,38 represents properly the process for high and low freeboards if the shape
parameters are assumed variable with relative freeboard or overtopping probability.
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
8
F=0, Layout 1, Narrow berm
F=0, Layout 1, Wide berm Ru
7
F=3, Layout 1, Narrow berm F
F=3, Layout 1, Wide berm
6
F=0, Layout 2, Narrow berm
<Vot>/(Ruot-Rc)2
0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Pot
Fig. 22.11. Mean overtopping volumes Vot versus overtopping probability Pot , emergent and
zero-freeboard structures. In the plot, F = Rc . From Cacéres et al.8
3.0
F=0, Layout 1, Narrow berm
F=0, Layout 1, Wide berm
2.5 F=3, Layout 1, Narrow berm
F=3, Layout 1, Wide berm
Rayleigh
F=0, Layout 2, Narrow berm
2.0 F=3, Layout 2, Narrow berm
F=0, Layout 2, Wide berm
k2
Exponential
1.0
0.5
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Pot
Fig. 22.12. Shape parameter k2 of the Weibull volume distribution versus overtopping probability
Pot , emergent and zero-freeboard structures. From Cacéres et al.8
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
where the symbol γ denotes the runup reduction factors accounting for slope
roughness (γf ) and wave obliquity (γβ ), whose values and expressions can be found
in TAW,47 tends to overestimate the overtopping discharge for LCBs, even with the
indeed small correction by Kofoed and Burcharth.26 Equation (22.11) can be thus
used only when overtopping is rare, i.e., well-emerged LCBs or weak wave attack
with respect to the crest freeboard.
Unlike high barriers, the flow rate over LCBs is abundant and related to the rear
piling-up. Overtopped water accumulates behind the structure, establishing a higher
mean water level, or piling-up, which forces return flows along different paths: water
may return offshore through gaps, originating the so-called rip currents, and, if the
crest is submerged, also over the structure itself. In this latter case, the flux over
the crest during the wave cycle is alternately directed inshore and offshore, driven
by waves and piling-up. Since LCBs are typically made of permeable rubble mound,
return flow occurs also through the structure, due to the unbalance between the
steady hydraulic gradient induced by piling-up and the breaking wave thrust.
Gaps are normally shorter but proportionate to breakwater length (∼ = 1/3),
but breakwaters, even when permeable, are far less permeable than the open gap,
therefore, when the breakwater is emerged the overtopping discharge returns to the
sea almost only through the gaps.
The flux through gaps qG globally equals the net flux across the structure qIS ,
which is the algebraic sum of the inshore directed flux over the structure crest qD
and the return flow across the structure qU , see Fig. 22.13. The return flow through
gaps qG is frequently named also recirculation flow, since it is usually simulated in a
wave flume by a recirculating system. The process can be schematized in analogy to
a pump system (Lamberti et al.31 ; Zanuttigh et al.66 ). The head losses associated to
return flow through gaps are represented by a characteristic curve and the relation
between piling-up and net mass flux across the structure is similarly described
by a barrier pumping curve. The system operational point at equilibrium is the
intersection between the two curves and is the solution of the equation.
The pumping curve was experimentally investigated in wave flumes equipped
with a recirculation system and it was found to be approximately linear for emerged
conditions (Ruol et al.40,41 ; Cappietti et al.10 ). Piling-up reaches its maximum p0 in
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
zO
qD zI
qIS
qU
p ≡ zI − zo
qG
qG ⋅ LG LB
Fig. 22.13. At the left-hand side scheme of planimetric circulation; at the right-hand side, scheme
of the model approach. From Lamberti et al.31
absence of recirculation and decreases to zero when qIS is recirculated without any
flow resistance. For given wave conditions, p in confined conditions is quite greater
than for a structure with gaps: indeed, the overall return flow resistance decreases
with increasing the gap to structure length ratio LG /LB , and consequently the
piling-up required to drive all the return flows is smaller (Martinelli et al.36 ).
Waves and currents around LCBs can be estimated by means of physical or
numerical modeling.
Laboratory experiments are usually dedicated to the analysis of a specific layout,
are time consuming and expensive, and may be affected by scale effects. It is thus
common practice for design purposes to adopt numerical models for simulating
waves and current fields.
When the structure crest is fully submerged, standard wave averaged models
(as) can be used with the assumption of negligible filtration through the structure.
If the breakwater is deeply submerged, its permeability has secondary effects when
compared to the return flow over breakwater crest. For this case (reef barrier)
induced circulation and setup can be evaluated with a 2DH wave-current model.
Depending on the relative depth (depth to wavelength ratio) the user is addressed
to the following models:
recommended for greater relative water depth and/or wider space. The Boussinesq
model is only a little more time consuming than the previous one and it is poten-
tially less robust, but it is specifically designed to represent wave dispersion and it
can be used with confidence almost up to the deepwater limit if it uses higher order
approximations.
• Phase averaged models (as MIKE 21 or LIMCIR, and many others, see Johnson
et al.24 ). These models represent wave statistics (spectra) and current separately,
based essentially on mild slope approximation. They are to be used when the
relative water depth variation in a wavelength is small and the computational
domain is much wider than the wavelength. When they are applied on a break-
water, which is invariably rather steep, they tend to overestimate wave height
decay rate due to breaking, wave setup, and induced currents.
For both submerged and emerged LCBs, accurate predictions with high com-
putational efforts can be obtained from the 2DV RANS-VOF code COBRAS by
Losada et al.,35 which allows to represent the structure porosity.
Some analytical models are also available under simplifying assumptions: both
Calabrese et al.9 and Bellotti2 provided the fluxes over an impermeable structure,
the first including friction effects in a 2D context and the second one excluding
friction effects in 3D conditions. A rapid and reasonable assessment of the 3D fluxes
around and within LCBs can be derived from Zanuttigh et al.66 for all crest free-
boards and wave attacks.
Wave breaking and currents generated by the presence of the LCBs induce in turn
morphological changes in their vicinity within a timescale of the order of a storm,
i.e., within a time frame of hours to days. A detailed description of seabed dynamics
and timescales, together with a classification of morphological changes, can be found
in De Vriend and Ribberink.13
In describing morphological effects of waves and currents, it is usual to distin-
guish between near-field scour and far-field erosion. First of all, scour occurs in
the immediate vicinity of the structure (less than 1/4 the wavelength), and it is
caused by quite different mechanisms from those responsible for far-field erosion.
These mechanisms include wave streaming, horse-shoe vortices, local turbulence,
and possible liquefaction due to vertical pressure gradients. Furthermore, different
timescales are associated with the development of scour and far-field erosion. Similar
considerations apply to the shoreline response induced by LCBs, i.e., the devel-
opment of salient or tombolos, especially with regards to the timescale involved,
which is larger than that associated with far-field erosion and typically ranges from
months to years.
Far-field erosion in the vicinity of LCBs is intimately linked to the onshore flow
over the crest of the structures, which returns offshore mainly through gaps, and
to the deflection and acceleration of longshore and return currents at the round-
heads. Currents flowing offshore through gaps produce irregularly shaped erosion
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
Fig. 22.14. Field surveys performed in 2004 in Lido di Dante. From Zyserman et al.68
conditions at the site and information about bed sediment characteristics. Besides
economical and technical difficulties, prototype observations are difficult to gener-
alize due to the particular conditions of each site (Lamberti et al.29 ).
Physical modeling is another valid alternative approach, and can also provide
data for calibration of numerical models. Few physical model tests of LCS have
been carried out in wave basins (Kramer et al.28 ) and most of them adopted a fixed
bed and dealt only with hydrodynamics around impermeable submerged structures
(Haller et al.22 ; Drønen et al.17 ). So far only two sets of movable bed experiments
were performed in a wave basin with the aim of analyzing morphodynamic changes
induced by LCBs. Both cases included detached rubble-mound breakwaters sep-
arated by gaps of different length, under regular (Van der Biezen et al.50 ) and
irregular (Martinelli et al.36 ) waves. However, restrictions imposed by the limited
extension and boundary conditions of laboratory facilities, and scale effects asso-
ciated with movable bed models must be kept in mind when selecting this method
to investigate far-field erosion.
Wave and flow fields in the vicinity of coastal structures and the associated mor-
phological response are thus usually assessed on the basis of coastal area numerical
models, see, e.g., De Vriend et al.15 and De Vriend14 for a description. Several ref-
erences can be found in the literature regarding applications of two-dimensional,
depth-integrated (2DH), or quasi three-dimensional (Q3D) models to study sites
or simplified cases. A nonexhaustive list of cases includes: large-scale and local
effects of groynes (De Vriend and Ribberink13 ), rip channels and bars evolution
(Damgaard et al.11 ; Ranasinghe et al.39 , where numerical results are compared
with video imaging analysis), bed changes in presence of breakwaters (Leont’yev32 ;
Nicholson et al.37 , which also includes a comparison of different model perfor-
mances), intermediate beach sedimentation inside and close to harbors (Lesser
et al.33 ), morphological response induced by LCBs (Zyserman et al.68 ; Martinelli
et al.36 ; Zanuttigh65 ). Main advantages of numerical models are the lack of scale
effects and the limited amount of work involved in modifying the model setup, which
in turn allows testing a wide range of layouts at limited cost.
Near-field erosion close to LCBs is related to two processes: steady streaming, due
to the superposition of incident and reflected waves at the offshore structure slope,
and plunging breakers inshore the structure. Steady streaming occurs both at the
structure trunk and roundheads. The plunging breaker is basically the key element
of the scour process inshore the roundhead. After its formation, the plunger travels
some distance along the crest width of the breakwater, eventually descends toward the
bottom, impinges on the bed, and mobilizes the sand leading to a scour hole.
Whilst far-field erosion can be generally predicted based on numerical or physical
modeling results, near-field erosion has been extensively analyzed by means of lab-
oratory tests deriving some analytical expressions for the estimate of its depth and
extension.
Figures 22.15 and 22.16 show the scour depth and extension, respectively, due
to steady streaming and plunging breakers. The plotted data are obtained under
regular and irregular waves by Sumer et al.,46 Fredsøe and Sumer,19 Lillycrop and
Hughes34 and Dixen.16 In all these works α = 1:1.5, with the exception of Lillycrop
and Hughes34 where α = 1:2.
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
Fig. 22.15. Depth S (to the left) and width Ly (to the right) of the scour hole induced by steady
streaming. In the plots, F = Rc and H = Hsi . From Sumer et al.46
Figures 22.15 and 22.16 are based on a dimensional analysis, which allows to rep-
resent the characteristics of the steady streaming [Eq. (22.12)] and of the plunging-
breaker [Eq. (22.13)] induced scour by the following nondimensional equations
S h L Rc
=f , , Re, ϑ, α, , (22.12)
Hsi L dn50 h
√
S Tw gH Rc
= fp , ϑ, α, , (22.13)
Hsi h h
where S is the scour depth, H is the significant incident wave height, h is the
bottom depth in front of the structure, L is the incident wavelength, ϑ is the Shields’
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
Fig. 22.16. Depth S (to the left) and width Ly (to the right) of the scour hole induced by plunging
breakers. In the plots, F = Rc and H = Hsi . From Sumer et al.46
parameter,
√ α is the offshore structure slope, dn50 is the nominal sediment diameter,
Tw gH/h is the plunger parameter (Fredsøe and Sumer19 ) and represents the ratio
between the amount of water in the plunging breaker entering in the main body
of water and the distance to be penetrated, Rc is the crest height with respect to
mean sea level (positive if the structure is emerged), Re = aUm /ν (being a the
amplitude of the orbital motion of water particles, Um the maximum value of the
orbital velocity at the bed, Tw is the wave period, ν the kinematic viscosity).
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
Field surveys highlight appreciable erosions along the trunk of LCBs, offshore
and inshore, and severe erosions at LCB roundheads and gaps. The erosion induced
at the offshore slope appears less pronounced and dangerous than in the case of
traditional emerged structures. This can be explained because of the lower reflection
and of sediment transport toward the structure induced by wave overtopping, with
consequent deposition offshore the LCB and flattening or canceling erosion holes.
Erosion may induce displacements of the structure toe and sliding of the armor
layer into the scour holes. The prolongation of such conditions in time can bring to
the structure destabilization. It is therefore necessary to protect the LCBs with a
sufficiently wide toe berm.
Toe protection layer may be constructed in the form of a protection apron. The
apron must be designed so that it will remain intact under wave and current forces,
and it should be “flexible” enough to conform to an initially uneven seabed. With
this countermeasure, scour can be minimized, but not entirely avoided. Some scour
will occur at the edge of the protection layer, and consequently, armor stones will
slump down into the scour hole. This latter process, however, will lead to the for-
mation of a protective slope, a desirable effect for “fixing” the scour. The deter-
mination of the width of the protection layer is an important design concern. The
width should be sufficiently large to ensure that some portion of the protection apron
remain intact, providing adequate protection for the stability of the breakwater.
On the basis of the works by Sumer and Fredsøe45 and Sumer et al.,46 it is
recommended to estimate from the following empirical equation the width of the
protection apron at the trunk section W (from the offshore structure edge)
L
W = − αhs , (22.14)
4
where hs is the structure slope. The validity range of Eq. (22.14) is h/L = 0.1 ÷ 0.2.
On the inshore side, the same W can be cautiously assumed, in order to prevent
also damages induced by wave overtopping.
At the roundhead, the toe protection width can be calculated as
Rc Rc Rc
W = We , if > +0.9; W = 0.29 + 0.74 We , if < +0.9,
Hsi Hsi Hsi
(22.15)
where We is the value of W for emerged breakwaters, which is evaluated following
Fredsøe and Sumer19 as
We = A · Um · Tw , (22.16)
where A is 1.5 for complete scour protection and 1.1 for a scour protection which
allows a scour depth of 1% of the breakwater width at the toe.
The above equation (22.15) is based on experiments where the breakwater slope
was 1:1.5. Therefore, for slopes milder than 1:1.5, the width necessary for protection
might be reduced (and for steeper slopes increased). Furthermore, Eq. (22.15) is
for scour protection against the local scour caused by the combined effect of steady
streaming and stirring up of sediment by waves. Due considerations must be given to
global scour caused by the far-field flow circulations around the breakwater. Finally,
the recommended width is for protection at the offshore side of the head.
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
Experiments show that berm width designed according to Eq. (22.16) is also
able to protect the sand bed against the breaker-induced scour at the inshore side
of the head. However, scour might occur in the protection layer itself due to wave
plunging. Therefore, over-sizing of the protection material is recommended at the
inshore side compared to the usual rule provided, for instance, by formula (22.17),
van der Meer et al.,56
where hb is the minimum water depth over the berm and Nod is the number of
stones displaced from a one stone wide strip of the berm and for standard toe size
(about five stones wide and two to three stones high at trunk) Nod = 0.5, 2, 4,
respectively, for no damage, acceptable, and severe damage.
The volume of the toe berm shall be such that its material is sufficient to protect
the scour/erosion hole from further erosion without destabilizing the armor layer
slope, i.e., its width should be around three times the erosion depth and its thickness
at least four times its maximum stone size (SPM43 ; Burcharth et al.6 ). In this
way, slide berm stones can form although dispersed a stable and continuous slope
covering the sand bed.
The erosion induced at gaps can both cause serious problem of structure stability
and act as sink for sediments inside the protected area, making them first slide
into the hole and then favoring their exit from the gap transported by rip currents.
It is therefore necessary to adequately protect the gaps with a stable and flexible
plateau that may follow bottom movements, usually consisting of the same material
at the barrier toe. The objective must be to shift erosion from the structure at such
a distance not to compromise structure stability. Gap protection shall be extended
more in offshore than in inshore direction, although it is not realistic an offshore
protection to the limit of the eroded area. The amount of material must exceed
the strictly necessary quantity in order to fill the holes that invariably form at
the protection boundaries. Maintenance works for restoring toe protection before
structure damage occur should be planned.
Land-based construction is more accurate and can be carried out usually under
sea conditions that do not allow construction with the aid of barges. Moreover, the
risk of damaging concrete units is lower. When both type are possible land-based
is normally preferred.
In many sites, however, depending on local conditions, different combinations of
land and floating equipment are actually used. Where the tidal range is large, one
can rather safely assume that for half of the daylight period water level is lower
than the structure basement level (higher than m.s.l.); during the summer period
this allows 8 hrs working period per day, that is normally considered quite sufficient
to carry out safely a work on the upper beach, where land-based equipment are
normally used.
In beaches exposed to frequent wave activity, so that a barge cannot operate safely
for long periods, land equipment are used also in the lower part of the intertidal beach
and even below. LCBs are constructed by dumping rocks from lorries moving on the
mound crest and placing armor with land-based cranes. If the structure is to be sub-
merged, the mound is constructed emergent and advancing from land, with equipment
moving on it; the emergent crest is lowered in a second phase, when the crane is
retreating and crest material is dumped at the sides of the mound. In some other case
the access causeway is temporary, and is removed at the end of works.
Barge-mounted equipment are normally used in calm waters. Barges are rather
insensitive to the short period waves (2–4 s) that are generated by local winds
blowing in bays or among islands. The direct dumping from barges with the assis-
tance of a floating crane for armor placement is the most common construction
method in this environment. Material shuttle barges and crane barge for placement
are usually separate, allowing crane barge to remain on station. Self unloading
split, bottom-door, tilting, or side unloading barges are usually used to dump core
material. For bedding layers, scour protection and berms flat deck barges will a
bulldozer for discharge can also be used. For the placement of filter layers only
side unloading or flat deck barges can offer a good placement precision. Thin layers
(0.50 m) can be placed only with barges operating with a high precision positioning
system in one or multiple passage.
LCBs pose a challenging separation problem between bed, foundation mound,
and armor, requiring a good placement precision.
For operations the following site conditions shall be considered: current, wind,
waves, tidal level variation, maneuvring space, water depth, and visibility, including
seasonal variation.
Positioning of barges is obtained by a roundabout anchoring system (usually
six anchors) or by a dynamic and computerized positioning system only for large
structures.
Generally, in sheltered water (no severe currents and waves), a horizontal
accuracy of 1 m can be achieved. In exposed conditions the accuracy will be less
and will decrease with increasing water depth.
Down-time caused by waves and wind is often determined by the influence on
the positioning accuracy and on crane loading, rather than by operational limitation
of the barge. Critical are the barge movements; short wind waves have less impact
than long swell. Generally, wind waves should not exceed 1.0–1.5 m height, whereas
a 0.5 m high swell might be critical for positioning and accurate dumping. More
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
severe are the limits for placing heavy armor units with cranes: cranes are normally
not designed to sustain lateral forces caused by swinging loads due to barge motions,
and heavy concrete units must be laid on bed with a minimal velocity. For these
reasons maximum allowable tilts during placement should not exceed a few degrees.
Often it is not possible to build a perfect filter separating the mound breakwater
from the bed, particularly under water. Some movement of sand in the mound pores
under heavy waves can therefore be anticipated in these cases. Moreover, if the bed
is composed of fine sand some liquefaction or partial liquefaction of the upper layers
are possible in storm conditions.
The sand body sustaining the breakwater weight should not be affected by, not
even partial, liquefaction. For all the mentioned reasons wide toe berms are useful:
they reduce filtration velocity below the breakwater body, they rest on and let sand
at rest; if bed erosion occurs, material sliding from the berm protects the bed from
further erosion. The state of berms is therefore an indicator of structure safety
that should be monitored. If scour holes of the order of twice the stone diameter
are shown by bathymetric surveys, or the berm width is reduced below 2–4 stones
diameter, toe berm stability may be compromised and toe protection should be
reinforced and widened.
Since the functionality of a LCB is very sensitive to crest elevation, the crest
settlement should be also monitored. As soon as the crest is lowered on the average
more than half the armor stone size or anywhere more than one stone size, the
armor layer should be recharged.
The design should not foresee a too frequent maintenance (below 10 years) since
this will probably induce risk for structure stability and cause disturbance to the
surrounding environment due to repeated working activity.
Nevertheless, a limited settlement of 3–5 cm/year is present also in properly
designed and constructed breakwaters, as those described in Section 22.2. A
settlement rate greater than 10–15 cm/year should be considered instead worth
of reassessing the stability of the breakwater and of analyzing a reinforcement
feasibility.
Acknowledgments
Most of the research on LCBs synthesized and discussed in this study has been
carried out within the framework of the DELOS project (EVK3-CT-2000-00041),
through which the support of EC is gratefully acknowledged.
References
41. P. Ruol, A. Faedo and A. Paris, Physical model study of water piling-up behind low
crested structures, Proc. 29th Int. Conf. Coastal Eng. (2004), pp. 4165–4177.
42. W. N. Seelig and J. P. Ahrens, Estimation of wave reflection and energy dissipation
coefficients for beaches, revetments and breakwaters, CERC technical paper 81-1, Fort
Belvoir, U.S.A.C.E., Vicksburg, MS (1981).
43. SPM, Shore protection manual, 4th edn., US Army Corps of Engrs (Coastal Engng.
Res. Center, US Govt Printing Office, Washington DC, 1984).
44. G. J. Steendam, J. W. van der Meer, H. Verhaeghe, P. Besley, L. Franco and
M. Van Gent, The international database on wave overtopping, Proc. 29th Int. Conf.
Coastal Eng., Lisbon (2004), pp. 4301–4313.
45. B. M. Sumer and J. Fredsøe, Experimental study of 2D scour and its protection at a
rubble-mound breakwater, Coastal Eng. 40, 59–87 (2000).
46. M. Sumer, J. Fredsøe, A. Lamberti, B. Zanuttigh, M. Dixen, K. Gislason and
A. Di Penta, Local scour and erosion around low-crested structures, Coastal Eng.
52(10–11), 995–1025 (2005).
47. TAW, Wave run-up and wave overtopping at dikes, J. W. van der Meer (author),
Tech. Report of the Tech. Advisory Committee on Water Defences in the Netherlands
(2002).
48. F. Thomalla and C. E. Vincent, Designing offshore breakwaters using empirical
relationships: A case study from Norfolk, UK, J. Coastal Res. 20(4), 1224–1230
(2004).
49. T. Uda, Statistical analysis of detached breakwaters in Japan, Proc. 21st Int. Conf.
Coastal Eng. (1988), pp. 2028–2042.
50. S. C. Van der Biezen, J. A. Roelvink, J. Van de Graaff, J. Schaap and L. Torrini,
2DH morphological modelling of submerged breakwaters, Proc. 26th Int. Conf. Coast.
Eng., Copenhagen, DK (1998), pp. 2028–2041.
51. J. W. van der Meer, Rock slopes and gravel beaches under wave attack, Delft
Hydraulics Commun. 396 (1988), 214 pp.
52. J. W. van der Meer, Extreme shallow water wave conditions, Report H198. Delft
Hydraulics Laboratory, The Netherlands (1990).
53. J. W. van der Meer, H. J. Regeling and J. P. de Waal, Wave transmission: Spectral
changes and its effects on run up and overtopping, Proc 27th Int. Conf. Coastal Eng.
(2000), pp. 2156–2168.
54. J. W. van der Meer and C. J. M. Stam, Wave runup on smooth and rock slopes
of coastal structures, J. Waterway, Port, Coastal and Ocean Eng. 118(5), 534–550
(1992).
55. J. W. van der Meer and J. P. F. M. Janssen, Wave run-up and wave overtopping
at dikes, Wave Forces on Inclined and Vertical Structures, eds. N. Kobayashi and
Z. Demirbilek (1995), Ch. 1, pp. 1–27.
56. J. W. van der Meer, K. D. Angremond and E. Gerding, Toe structure stability of
rubble mound breakwaters, Proc. Advances in Coastal Structures and Breakwaters
Conf., Institution of Civil Engineers (Thomas Telford Publishing, London, UK, 1995),
pp. 308–321.
57. J. W. van der Meer, R. Briganti, B. Zanuttigh and B. Wang, Wave transmission and
reflection at low crested structures: Design formulae, oblique wave attack and spectral
change, Coastal Eng. 52(10–11), 915–929 (2005).
58. D. Vera-Cruz and J. Reis de Carvalho, Macico submerso de pre-rebentacao das ondas
como meio de proteccao de obras maritimas: O caso do quebra-mar de Leixoes, LNEC
Memoria no 796, Lisboa (1993), p. 36.
August 6, 2009 13:20 9.75in x 6.5in b684-ch22 FA
Chapter 23
Yu-Cheng Li
School of Civil Engineering
Dalian University of Technology, Dalian, China
liyuch@dlut.edu.cn
Knowledge of the hydrodynamic behavior of net cage under the action of waves
and currents is the basis of the design and management of net cages in the open
sea. Techniques used to investigate the net cage have typically included the use of
scaled physical and numerical models, and, where possible, field measurements. In
this chapter, information on the hydrodynamic behavior of net cages in the open
sea is focused on gravity cages. The main methods used for research into hydro-
dynamic behavior are introduced: physical tests and numerical computation.
23.1. Introduction
Exposed net cages in the open sea are subject to wave and current action. Thus,
knowledge of their hydrodynamic behavior under the action of waves and currents
is important for the design and management of net cages in the open sea. There are
many types of net cages, among which gravity cages are the most popular. Thus, in
this chapter, information on the hydrodynamic behavior of net cages in the open sea
is focused on gravity cages. The main methods used for research into hydrodynamic
behavior are introduced: physical tests and numerical computation.
633
August 4, 2009 9:39 9.75in x 6.5in b684-ch23 FA
634 Y.-C. Li
greater than 20:1. If the model net is designed strictly according to geometric simi-
larity principles, two difficult problems arise. First, the yarn used for the model net
will be very thin, which is difficult to manufacture. Second, because the model yarn
is so thin, the Reynolds number (Re) for the model net will change significantly rel-
ative to the prototype net, leading to a large difference in hydrodynamic behavior
between the model and the prototype nets. Thus, special simulation methods are
needed. Tauti’s criteria1,2 were developed in the 1930s during research into fishing
net problems and are used only for conditions where a current is present.
LT
LT
S
rS
WS
rS
WS
is applied in the presence of waves. Thus, for model tests of net cages in wave and
current conditions, the following gravity simulation criteria are recommended.
L
b
636 Y.-C. Li
transducer
0.4m
lower bar
sinker
1m
16
Gravity
Tauti
12
Force (KN)
0
0 20 40 60 80 100 120 140
Velocity (cm/s)
Fig. 23.4. Comparison of force values between the gravity similarity and Tauti similarity criteria.
yarn diameter is 1.18 mm and the half-mesh opening is 10 mm. The sinker system
consists of a lower bar and sinkers. The prototype mass of the weight system is
240 kg. To record the net configuration, two bulbs were fixed at the middle and
bottom of the net.
As shown in Figs. 23.4 and 23.5, both the hydrodynamic force and deflection of
the model net are in agreement under the two similarity criteria, but the weakness
in Tauti’s simulation criteria does not exist in the extended gravity simulation
criteria. The
√ latter can be applied in both current and wave conditions, and since
Vm = Vp / λ, the requirement to create a current in the test facility is lower than
for Tauti’s simulation criteria.
800
400
200
Gravity
Tauti
0
0 20 40 60 80 100 120 140
Velocity (cm/s)
Fig. 23.5. Comparison of projected area between the gravity similarity and Tauti similarity
criteria.
This section introduces simulation methods for the main parts of a gravity cage,
including the float collar and fishing net, then the whole gravity cage can be simu-
lated by connecting the main parts of cage.7–15
Fig. 23.6. Sketch of the float system for a gravity cage. (a) Simplified float system and (b) section
of double pipes.
August 4, 2009 9:39 9.75in x 6.5in b684-ch23 FA
638 Y.-C. Li
u
b
v
n
c
a
d u
obtained by summing the forces acting on each mini-segment. Figure 23.7 is a sketch
of a mini-segment of float collar with a local coordinate system n, τ , v defined
for each mini-segment. As to coordinate system, n and τ are in the normal and
tangential direction of the mini-segment, respectively. Thus, v is normal to the
mini-plane (Fig. 23.7).
According to Brebbia and Walker,16 the n component of wave-induced forces on
a mini-segment in local coordinates can be obtained as follows:
1
Fn = CDn ρAn · |(un − Un )| · (un − Un ) + ρV0 an + Cmτ ρV0 (an − U̇n ), (23.8)
2
where CDn and Cmn represent the drag and inertial coefficients of the n component,
respectively; An is the effective projected area in the direction of the n component;
and an and U̇n represent the acceleration vectors for water particles and mini-
segments of the n components, respectively. Other parameters are as described
above. The same expression can be applied to other wave-induced forces (Fτ , Fν )
of v components. Figure 23.8 is a sketch of a simplified pipe model for calculation.
e η n
es urfac
Wav
γ v
where li is the length of the mini-segment. Aτ is related to the arc area of the mini-
segment (r · φi · li ) in the water. φi is the corresponding central angle of the pro-
jected chord-length, which is calculated from cos(φi /2) = 1 − dn /r. For the normal
projected area Av , when dn ≥ r, let dv = 2r.
In addition to wave-induced forces, the float collar is also subjected to gravity,
buoyancy, and mooring line forces. The gravity acting on the mini-segment can be
written as:
Gi = G/N, (23.10)
where G is the total gravity acting on the floating system and N is the number of
mini-segments. The buoyancy acting on a mini-segment can be calculated as:
Ffi = ρg · Vi . (23.11)
The relationship between mooring line forces and elongation can be obtained
directly by experimental measurement.
3
p
r
G 2
z rp
rG 1
o y
x
Fig. 23.9. Schematic diagram of the moving coordinate system for a rigid body.
August 4, 2009 9:39 9.75in x 6.5in b684-ch23 FA
640 Y.-C. Li
Six degrees of freedom are required to describe the motion of a float collar in
general spatial motion, resulting in six equations of motion. According to Newton’s
second law, under fixed coordinates, the three translational equations of motion are
given by:
n n n
1 1 1
ẍG = Fxi , ÿG = Fyi , z̈G = Fzi , (23.12)
mG mG mG
i=1 i=1 i=1
where Fxi , Fyi , Fzi are the components of the external force vector Fi (i = 1, n)
along the fixed coordinate axes xyz, n is the number of external forces, and mG is
the mass of the rigid body.
Axes 1, 2, 3 are principal axes with origin at the center of mass G, and thus
Euler’s equations of motion for a rigid body17 can be applied. Under the body
coordinate system, the three rotational equations of motion are given by:
n
n
I1 ω̇1 + (I3 − I2 )ω2 ω3 = M1i , I2 ω̇2 + (I1 − I3 )ω3 ω1 = M2i ,
i=1 i=1
n (23.13)
I3 ω̇3 + (I2 − I1 )ω1 ω2 = M3i ,
i=1
The kinematic differential equations for the Bryant angles are given by:
ω1 = φ̇1 cos φ3 cos φ2 + φ̇2 sin φ3 , ω2 = −φ̇1 sin φ3 cos φ2 + φ̇2 cos φ3 ,
The motion of the float collar can be described by three translational displace-
ments of the centroid and three Bryant angles, which can be obtained by solving the
simultaneous ordinary differential Eqs. (23.12), (23.13), and (23.16). The equations
can be solved using the Runge–Kutta–Verner fifth-order and sixth-order method.
During calculation, transformation between the fixed coordinates and body coordi-
nates is carried out by matrix [R]. It should be noted that the forces and motions
of the collar are dependent on the net through the mutual mass points attached to
both the net and the collar.
642 Y.-C. Li
mesh knot
d
Fig. 23.10. Schematic diagram of the mass-spring model.
ξ
η
Fig. 23.11. Schematic diagram of the local coordinates for mesh yarns.
where a is the acceleration of mass point i, Tij is the vector of the tension in bar ij
(j is the code for knots at the other end of bar ij), n is the number of knots adjacent
to point i, FD is the drag force, FI is the inertial force, W
is the gravity force, and
B is the buoyancy force. According to our research,9,10 the inertial force FI on a
fishing net is rather small under wave conditions compared to other external forces,
so it can be omitted here.
The mass point at a mesh knot is assumed to be a spherical point at which the
fluid force coefficient is constant in the direction of motion, so the motion equation
is easily set up according to Eq. (23.17). For points at mesh bars, the forces and
motion equation are described in the following.
Since the points at mesh bars are assumed to be cylindrical elements, the
direction of the fluid forces acting on the point masses at each mesh bar should be
considered. Therefore, the motion of point mass i is set at the center of a mesh bar,
and the local coordinates (τ, η, ξ) passing point i are defined to simplify the pro-
cedure (Fig. 23.11). The η axis lies on the plane including the τ axis and V , and the
velocity vector V of the water particles at point i can be divided into τ (tangential)
and η (normal) components, so the local fluid force on the plane can be estimated
by defining the τ and η components of the fluid force.
Under global coordinates, the vectors drawn from i to 1 and 2 are ei1 and
ei2 , respectively. The unit vectors of the τ , η, ξ axes are eτ = (xτ , yτ , zτ ), eη =
(xη , yη , zη ), and eξ = (xξ , yξ , zξ ).
August 4, 2009 9:39 9.75in x 6.5in b684-ch23 FA
The relationship between tension and elongation based on Wilson22 is given by:
T l − l0
= C1 εC2 , ε= , (23.18)
d2 l0
where l0 is the undeformed length, l is the deformed length, d is the diameter of the
mesh bar, and C1 and C2 are constants that define the elasticity of the element.
Therefore, the τ component of tension at point i can be expressed as:
where the mid-dot [·] represents the inner product. The same expression can be
applied to the other tension forces (Tη , Tξ ) of η, ξ components.
The τ, η, and ξ components of the relative velocity of point i are τ̇ − eτ · V ,
η̇ − eη · V , and ξ̇, respectively. Thus, the τ component of the drag force of point i
can be represented as:
1
Fdτ = − ρCdτ Dl|τ̇ − eτ · V |(τ̇ − eτ · V ), (23.20)
2
where Cdτ represents the drag coefficient of the τ component, D is the diameter of
the cylindrical element, and l is its length. The same expression can be applied to
the other drag forces (Fdη , Fdξ ) of η, ξ components.
Dividing the other external forces B and W along the local coordinate axes, the
τ component of the motion equation for point i can be expressed as:
Thus, the motion equations for point i in the η, ξ directions can be expressed in the
same manner.
As the displacements are expressed using local coordinate variables in
Eq. (23.21), we can transform these into the global coordinate system by:
Thus, the motion can be represented by the following system of ordinary differ-
ential equations:
dẋi
= f (xi , yi , zi , ẋi , ẏi , żi , x1 , y1 , z1 , x2 , y2 , z2 ; t)
dt
dẏi
= g(xi , yi , zi , ẋi , ẏi , żi , x1 , y1 , z1 , x2 , y2 , z2 ; t) (23.24)
dt
dżi
= h(xi , yi , zi , ẋi , ẏi , żi , x1 , y1 , z1 , x2 , y2 , z2 ; t).
dt
August 4, 2009 9:39 9.75in x 6.5in b684-ch23 FA
644 Y.-C. Li
The net shape at each time step can be calculated numerically by solving these
ordinary differential equations for a given initial condition.
25 30
24
25
20
20
Time(s)
Force(N)
15
15
10 G0
G4 10
5 4
G16 5
1
0 0
0 15 30 45 60 G0 G4 G16
V(cm/s)
(a) (b)
Fig. 23.13. Comparison of the (a) total drag force and (b) calculation time for different mesh
grouping methods.
August 4, 2009 9:39 9.75in x 6.5in b684-ch23 FA
In this section, we compare the force and movement characteristics of different cage
models. Three models were selected: a gravity cage, a sea station cage, and a quasi-
sea station cage. The sea station cage model has that same enclosed volume as
the gravity cage model, but has a greater cage diameter and height. The quasi-sea
station cage model has the same diameter as the sea station cage model, but is 1.4-
fold greater in cage height. This increases its effective aquaculture volume to 2.2-
fold that of the latter. Model tests to determine the hydrodynamic characteristics of
sea cages were carried out under floating and submerged conditions. Current-only,
wave-only, and wave plus current conditions were set during the tests. Table 23.1
shows the model settings in detail.
646 Y.-C. Li
0.15m
0.637m
0.4m
1.0m
z
Gravity Cage sinker
α
x 1.71m 1.09m
0.1m
1.0m
1.0m
Sea Station
0.25m
z mooring line
α
x 1.71m 1.04m
0.15m
1.0m
0.25m
0.7m
1.0m
z Quasi-Sea Station
0.75m
0.25m
α
x 1.71m 1.02m
Fig. 23.14. Sketch of cage models under floating conditions (black dots on cages are the tracing
points).
Figure 23.14 shows the model settings for different cages under floating condi-
tions. All cage models were moved to 2/5 of the water depth when submerged. The
mooring line forces were acquired by transducers attached to the bottom of each
mooring line. Cage movement data were gathered by tracing diodes fixed on the
rigid ring or float system of each cage model using a CCD camera.
Cages of different styles will exhibit different force and motion characteristics
under the same current and wave conditions. For offshore aquaculture facilities,
security, management, and investment are the main factors to be considered.
Table 23.2 provides a general comparison of the cage models under floating and sub-
merged conditions, for which the gravity cage was selected as the reference. Values
for the mooring line force, aquaculture volume and cost were all set to 1.0, and the
data are reported as ratios compared to values for the gravity cage.
From the viewpoint of forces, the gravity cage has advantages when it is fixed
with heavier sinkers. Management of gravity cages is more convenient and the
investment costs are much lower compared to the other two cages. The sea station
and quasi-sea station systems are semi-rigid cages, which are beneficial in controlling
volume reduction and preventing predator attack. However, the maximum volume
of the gravity cage is nearly three-fold greater than that of the sea station cage
August 4, 2009 9:39
Hydrodynamic Behavior of Net Cages in the Open Sea
Table 23.2. General comparison of the cage models under floating and submerged conditions (ratios).
Wave + currenta
9.75in x 6.5in
Current only Wave only (H = 35 cm)
Cage
Status model 11.5 cm/s 15.5 cm/s 17.2 cm/s T = 1.6 s T = 1.8 s T = 2.0 s T = 1.6 s T = 1.8 s T = 2.0 s Diameter Volume Cost
Floating I 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0
II 0.97 1.02 1.04 1.16 1.0 1.04 1.24 1.04 1.18 1.57 1.03 3∼4
III 2.11 2.08 2.06 1.27 1.06 1.14 1.84 1.58 1.71 1.57 2.27 >4
Submerged I 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0
II 0.84 1.12 1.11 1.43 1.63 1.16 0.64 0.63 0.81 1.57 1.03 3∼4
III 1.96 1.92 1.9 1.4 1.97 1.5 1.41 1.4 1.51 1.57 2.27 >4
a For wave + current, the wave height was H = 35 cm and the current velocity was u = 17.29 cm/s.
b684-ch23
647
FA
August 4, 2009 9:39 9.75in x 6.5in b684-ch23 FA
648 Y.-C. Li
for the same cage diameter and height. In other words, the advantages of the sea
station cage can be attributed to largely losing the original aquaculture volume.
Even for the same original volume as the gravity cage, management of sea station
and quasi-sea station cages is inconvenient. From the viewpoint of cage motion, the
sea station and quasi-sea station cages exhibit better stability, which is considered
beneficial for fish in the cages. However, it should be noted that even if cages can
endure the action of tidal currents and waves, there is no guarantee that fish in such
cages will survive.
Part of the fluid will flow through adjacent areas of the fishing net owing to
its blockage effect. However, it is difficult to determine the flux passing through
adjacent areas because of the complexity of the interaction between the fluid and
the fishing net. In general, the fluid velocity in adjacent areas decreases as the dis-
tance between the fluid and the fishing net increases. The effective adjacent area is
related to the fabric property of the fishing net. Here, the area of the fishing net
is discretized into two parts: the solid projected area of the twines and knots, and
the area of the holes. The latter is treated the same as the effective adjacent area,
which means that the average velocity in holes is assumed to be the same as that
in the effective adjacent area. Therefore, the effective adjacent area is a key factor
for proper evaluation of the velocity reduction behind a fishing net. Figure 23.15
shows the definition of the flowing areas.
It is assumed that the cross-sectional area of the fishing net is A1 with solidity
Sn and the effective adjacent area is A2 . Here, the variation in area introduced by
changes in water level in different sections is neglected. The solid projected area of
the fishing net is Sn A1 and the area of holes is (1−Sn )A1 . Therefore, considering the
unit: cm
200
50 100 50
39.6
Area of net
(A1, u 1)
70
effective adjacent area, the total area affected by fluid passing through the fishing
net is
A = (1 − Sn )A1 + A2 . (23.27)
According to mass conservation, we have:
(A1 + A2 ) · u0 = [(1 − Sn )A1 + A2 ] · u2 , (23.28)
where u0 is the velocity of the far-field approach fluid, which is assumed to be
undisturbed by the fishing net; u2 is the velocity in the adjacent area, which is the
same as that in the holes of the fishing net. A coefficient k is introduced here to
denote the ratio between the adjacent area to the solid projected area. It can be
written as:
k = A2 /(Sn · A1 ). (23.29)
Substituting Eq. (23.29) into Eq. (23.28), we have:
1 + k · Sn
u2 = · u0 , (23.30)
1 − (1 − k) · Sn
As described above, the velocity of the fluid passing through the holes will
decrease due to amplification of the sectional area behind the fishing net. According
to mass conservation within the section of fishing net (excluding the adjacent area),
we have:
(1 + k · Sn ) · (1 − Sn )
u1 = (1 − Sn ) · u2 = · u0 , (23.31)
1 − (1 − k)Sn
i.e.,
u1 (1 + k · Sn ) · (1 − Sn )
= . (23.32)
u0 1 − (1 − k)Sn
Studies on the hydrodynamics of pile groups or double-circle cylinders reveal
that the influence of adjacent piles can be neglected when the gap between them
is four-fold greater than the pile diameter, i.e., the hydrodynamics of each pile can
be treated separately. Therefore, the range influenced by one pile is three-fold the
pile diameter. For the fishing net, we assume that the effective area influenced is in
proportion to the solid projected area of the net. Here, the ratio k is supposed to
be the value of Eq. (23.32), which can be written as:
u1 (1 + 3Sn ) · (1 − Sn )
= . (23.33)
u0 1 + 2Sn
According to Eq. (23.32), if Sn = 0, i.e., there is no fishing net in the flow
field, we have u1 = u0 . If Sn = 1, the fishing net is equivalent to a solid plate and
the velocity behind the fishing net is equal to zero. Results under these two cases
are reasonable. However, when Sn = 0 and k = 0, the fishing net exists in the
whole section, which is similar to a trash rack. It is known that under this case, the
velocity behind it will increase to some extent. However, the result calculated with
August 4, 2009 9:39 9.75in x 6.5in b684-ch23 FA
650 Y.-C. Li
H0 · u0 = H · u , (23.35)
where H0 is the water level in the far-field section in front of the fishing net; H is
the water level behind the fishing net; u is the flow velocity behind the fishing net;
and hj is the local energy loss in this section of fishing net, which can be calculated
according to the trash rack analogy. The general form of the local energy loss is as
follows:
4/3
u2 b u2
hj = ξ · 0 = β · · sin α · 0 , (23.36)
2g t 2g
where ξ denotes the local head loss coefficient and β is the shape factor of the
trash rack. For the fishing net, β is equal to 1.79. b and t refer to the grid width
and the gap between grids, respectively. Thus, b/t is the ratio between the solid
projected area and the gap area. For the fishing net, we have b/t = Sn /(1 − Sn ). α
in Eq. (23.36) refers to the angle between the cross plane of the trash rack and the
horizontal plane, which is 90◦ here. Substituting the value of each parameter into
Eq. (23.36) yields:
4/3 2
Sn u
hj = 1.79 × · 0. (23.37)
1 − Sn 2g
By combining Eqs. (23.34), (23.35), and (23.37), the flow velocity (u ) behind the
fishing net can be determined. Then the value of ∆u is obtained according to:
∆u = u − u0 . (23.38)
50 20
A1 C1
39.6
A2 C2
Flow direction A3 C3
70
A4 C4
Plane net
Measuring Point Measuring point
Table 23.3. Comparison of the velocity reduction between calculated results and
experimental data.
Model tests were carried out to investigate the hydrodynamic behavior of a gravity
cage of a given structure size. The geometric scale λ and model mesh scale λ were
20:1 and 2:1, respectively. The gravity simulation criteria described in Sec. 23.2.1.2
were applied in designing the model net cage. Under current-only, wave-only, and
wave + current conditions, the mooring line force and motion of the net cage were
measured when the net cage was floating or submerged under water. In model tests,
August 4, 2009 9:39 9.75in x 6.5in b684-ch23 FA
652 Y.-C. Li
Table 23.4. Design of experimental conditions (H and T denote the wave height and
period, respectively).
No. 1 2 3 4 5 6 7 8 9 I J K
Wave
H (cm) 20 20 20 25 25 25 30 30 30 35 35 35
T (s) 1.2 1.4 1.6 1.4 1.6 1.8 1.4 1.6 1.8 1.6 1.8 2
Velocity (cm/s) 8 12 16 20
h=1.0m
Transducer Bottom collar
z
Sinker
Sinking ring
x 2.92m
(a)
0.796m
H/3
z Transducer h=1.0m
Bottom collar
Sinking ring Sinker
x 2.92m
(b)
Fig. 23.17. Sketch of the setup for the gravity cage model. (a) Floating condition and (b) sub-
mergence condition.
cage, and the sinking ring was used to control the submergence depth of the cage.
The water mass that was perfused into the pipe was 334.1 g.
Using the data measured, empirical formulae for the mooring line force and
motion were obtained by the least square method.
654 Y.-C. Li
20
Floating
15
Submergence
Inclination/Deg
10
0
0 0.2 0.4 0.6 0.8 1
Velocity/ms-1
Fig. 23.18. Inclination angle for the cage under current conditions.
where ξF denotes the maximum amplitude of the horizontal motion of tracing points
on the lee side of the floating collar and Uh is the submergence depth.
In contrast, the amplitude of the vertical motion of tracing points is greater on
the wave side than on the lee side of the floating collar under wave action. The
empirical formula for the maximum amplitude of the vertical motion of tracing
points on the wave side of the floating collar is (correlation coefficient R = 0.991):
ηF H gT 2 Uh
= 0.321 × + 3.05 × 10−3 × − 0.217 × , (23.46)
D D h h
where ηF denotes the maximum amplitude of the vertical motion of tracing points
on the wave side of the floating collar.
test and numerical simulation are introduced, which will give references of the design
of net cage.
Weight mode A1 A2 A3 B2 B3 B4
H=1.0m
x 2.92m
(a)
H=1.0m
Bottom of net
z Transducer
Sinker
x 2.92m
(b)
Fig. 23.19. Sketch of the model setup. (a) Weight system comprising a bottom-collar sinker and
(b) weight system comprising sinkers.
August 4, 2009 9:39 9.75in x 6.5in b684-ch23 FA
656 Y.-C. Li
60 60
A1 B2
A2 B3
Force /(kN)
40 40
Force/(kN)
A3 B4
20 20
0 0
0.36 0.54 0.72 0.89 1.07 0.36 0.54 0.72 0.89 1.07
-1 -1
Velocity /m s Velocity /ms
(a) (b)
Fig. 23.20. Resultant forces in the two upstream mooring lines. (a) Weight mode A and (b) weight
mode B.
1.0
0.8
Ratio
0.6
0.4 B2/A2
B3/A3
0.2
0.72 0.89 1.07
-1
Velocity /cms
Fig. 23.21. Comparison of the resultant mooring-line forces for two different weight systems.
23.7.1.2. Deformation
Net deformation is an important consideration. A sea cage is a 3D structure and
thus deformation of a fishing net is also a three-dimensional problem under current
August 4, 2009 9:39 9.75in x 6.5in b684-ch23 FA
60 60
Force /(kN)
Force /(kN)
40 40
20 A1 20 B2
A2 B3
A3 B4
0 0
7.2 8.0 8.0 8.9 T/s 7.2 8.0 8.0 8.9 T/s
4m 5m 6m 7m H/m 4m 5m 6m 7m H/m
(a) (b)
Fig. 23.22. Resultant forces in the two wave-side mooring lines. (a) Weight mode A and (b) weight
mode B.
1.5
1.2
Ratio
0.9
B2/A2
B3/A3
0.6
7.2 8.0 8.0 8.9 T/s
4m 5m 6m 7m H/m
Fig. 23.23. Comparison of the resultant mooring-line forces for two different weight systems.
658 Y.-C. Li
90% 90%
A1 B2
Rate of area loss
A2
30% 30%
0% 0%
0.36 0.54 0.72 0.89 1.07 0.36 0.54 0.72 0.89 1.07
-1
Velocity/ms -1
Velocity/ms
(a) (b)
Fig. 23.24. Influence of weight size on net deformation. (a) Weight mode A and (b) weight mode B.
90%
B2
A2
Rate of area loss
B3
60%
A3
30%
0%
0.36 0.54 0.72 0.89 1.07
-1
Velocity/m·s
40% 40%
Maximum Rate of area loss
Maximum Rate of area loss
A1
A2 30%
30%
A3
20% 20%
B2
10% 10% B3
B4
0% 0%
7.2 8.0 8.0 8.9 T/s 7.2 8.0 8.0 8.9 T/s
H/m 4m 5m 6m 7m H/m
4m 5m 6m 7m
(a) (b)
Fig. 23.26. Influence of weight system on net deformation. (a) Weight mode A and (b) weight
mode B.
August 4, 2009 9:39 9.75in x 6.5in b684-ch23 FA
2
Ratio
1
B2/A2
B3/A3
0
7.2 8.0 8.0 8.9 T/s
4m 5m 6m 7m H/m
Net
Current
Sinker
Fig. 23.28. Sketch of the gravity cage model with a bottom-collar sinker-weight system.
August 4, 2009 9:39 9.75in x 6.5in b684-ch23 FA
660 Y.-C. Li
Table 23.6. Properties of the net cages used for numerical simulation.
No. of meshes
Cage no. RDH Mesh style Diameter (m) Height (m) Circumference Height
Mass (g) Bottom collar One sinker (g) Scaled weight system (g)
WA 0 87.3 1396.8
WB 931.2 29.1 1396.8
mesh size of 32 mm and a yarn thickness of 1.8 mm. The netting itself is not scaled,
and ordinary full-scale netting is used. Mounted as square meshes, the solidity ratio
(S) of the netting is 0.225. The net then forms an open vertical cylinder with a
diameter of 1.435 m.
Keeping the cage diameter constant, three RDH values are applied: 2.0, 1.43,
and 1.0. In each RDH mode, two types of knotless net are used and simu-
lated, a square mesh and a diamond mesh. The yarn diameter and mesh size
are set to 1.8 and 32 mm, respectively, for both net types. The properties of
the net cages are presented in Table 23.6. The bottom of the net cage is not
modeled.
In this section, two types of weight system are applied: (1) a sinker (WA), and
(2) a bottom-collar sinker (WB), as shown in Table 23.7.
The gravity cage system is subjected to five different current velocity cases
(Usteady = 0.11, 0.19, 0.26, 0.34, and 0.37 m/s) corresponding to the full-scale cases
(Usteady = 0.3, 0.5, 0.7, 0.9, and 1.0 m/s) according to the gravity simulation cri-
teria. After entering the net cage, the velocity of a fluid particle will decrease
slightly.
Vc
Cvh = , (23.47)
Vc0
where Vc is the volume of the net cylinder exposed to the current and Vc0 is the
initial volume of the net cylinder not exposed to the current.
August 4, 2009 9:39 9.75in x 6.5in b684-ch23 FA
1.0 1.0
Volume holding coefficient [-]
Volume holding coefficient [-]
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
RDH=2.0 RDH=2.0
0.5
RDH=1.43 0.5 RDH=1.43
RDH=1.0 RDH=1.0
0.4
0.4
0.3
0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.10 0.15 0.20 0.25 0.30 0.35 0.40
Velocity [m/s] Velocity [m/s]
(a) (b)
Fig. 23.29. Volume-holding coefficient under (a) the sinker system and (b) the bottom-collar
sinker system at different RDH as a function of current velocity.
0.09
volume holding coefficient[-]
0.08 RDH=2.0
RDH=1.43
Increase in the
0.07 RDH=1.0
0.06
0.05
0.04
0.03
0.02
Fig. 23.30. Increase in the volume-holding coefficient with the bottom-collar sinker system for
different RDH relative to the sinker system.
August 4, 2009 9:39 9.75in x 6.5in b684-ch23 FA
662 Y.-C. Li
Fig. 23.31. Shapes of net cylinders with diamond and square meshes.
(a) (b)
Fig. 23.32. Initial net images (a) with a sinker system and (b) a bottom-collar sinker system.
1.00
0.90
0.75
0.85
0.70
0.80
0.65 0.75
0.60 0.70
0.65
0.55 RDH=2.0
0.60 RDH=1.43
0.50 RDH=2.0
RDH=1.43 0.55
0.45 0.50
0.40 0.45
0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.10 0.15 0.20 0.25 0.30 0.35 0.40
Velocity [m/s] Velocity [m/s]
(a) (b)
Fig. 23.33. Volume-holding coefficient under (a) the sinker system and (b) the bottom-collar
sinker system at different RDH as a function of current velocity.
0.26
volume holding coefficient [-]
0.24
RDH=2.0
Increase in the
0.22 RDH=1.43
0.20
0.18
0.16
0.14
Fig. 23.34. Increase in the volume-holding coefficient with the bottom-collar system at different
RDH relative to the sinker system.
As shown in Fig. 23.34, the bottom-collar sinker system exhibits lower defor-
mation than the sinker system with a lower RDH. For a gravity cage with a diamond
mesh, the structure of the bottom collar is critical when the RDH is low.
664 Y.-C. Li
Fig. 23.35. Comparison image of net deformation with the sinker system (V = 0.37 m/s).
Fig. 23.36. Comparison image of net deformation with the bottom-collar sinker system
(V = 0.37 m/s).
1.00
Volume holding coefficient [-]
0.95 1.00
Volume holding coefficient [-]
0.95
0.90 0.90
0.85
0.85 0.80
0.80 0.75
0.70
0.75 0.65
0.60
0.70
0.55
0.65 Square mesh 0.50
0.45 Square mesh
Diamond mesh
0.60 0.40 Diamond mesh
0.35
0.55
0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.30
0.10 0.15 0.20 0.25 0.30 0.35 0.4
Velocity [m/s] Velocity [m/s]
(a) (b)
Fig. 23.37. Comparison of the volume-holding coefficient with the sinker-weight system.
(a) RDH = 2.0 and (b) RDH = 1.43.
images of net deformation for the sinker and bottom-collar sinker systems with
RDH = 2.0, in which the dashed lines denote net deformation for the square mesh.
More quantitative comparisons are shown in Figs. 23.37 and 23.38.
According to Fig. 23.37, when the sinker system (WA) is applied, net defor-
mation is greater for a diamond mesh than for a square mesh, but the discrepancy
decreases with increasing current velocity. This discrepancy may be induced by the
initial net deformation. Even if the net is not exposed to the current, net defor-
mation for the diamond mesh is significant, whereas deformation for the square
mesh is small.
According to Fig. 23.38, when the bottom-collar sinker system (WB) is applied,
net deformation for the diamond mesh is less than that for the square mesh with
increasing current velocity. This is the reason why rotation of the bottom of the net
is reversed for the two different mesh types, and with increasing current velocity
August 4, 2009 9:39 9.75in x 6.5in b684-ch23 FA
1.00
1.00 0.95
0.95
0.85
0.90 0.80
0.85 0.75
0.70
0.80
0.65
0.75 0.60
0.70
0.55
Square mesh
0.50
0.65 Diamond mesh Square mesh
0.45
0.40 Diamond mesh
0.60
0.35
0.55 0.30
0.10 0.15 0.20 0.25 0.30 0.35 0.40
0.10 0.15 0.20 0.25 0.30 0.35 0.4
Velocity [m/s] Velocity [m/s]
(a) (b)
Fig. 23.38. Comparison of the volume-holding coefficient with the bottom-collar sinker system.
(a) RDH = 2.0 and (b) RDH = 1.43.
the orientation of the bottom of the diamond-mesh net is more helpful in reducing
net deformation.
References
1. M. Tauti, A relation between experiments on model and on full scale of fishing net,
Bull. Jpn. Soc. Scientific Fisheries 3(4), 171–177 (1934).
2. M. Tauti, The force acting on the plane net in motion through the water, Bull. Jpn.
Soc. Scientific Fisheries 3(1), 1–4 (1934).
3. Y. C. Li, F. K. Gui, H. H. Zhang and C. T. Guan, Simulation criterica of fishing
net in aquaculture sea cage experiments, J. Fishery Sciences of China 12(2), 179–187
(2005) (in Chinese).
4. F. K. Gui, Y. C. Li and H. H. Zhang, The proportional criteria for model testing
of force acting on fishing cage net, China Offshore Platform 17(5), 22–25 (2002)
(in Chinese).
5. F. K. Gui, Hydrodynamic behaviors of deep-water gravity cage, Doctoral dissertation,
Dalian University of Technology (2006) (in Chinese).
6. F. K. Gui, Y. C. Li, G. H. Dong and C. T. Guan, Application of CCD image scanning
to sea-cage motion response analysis, Aquacultural Eng. 35, 179–190 (2006).
7. Y. C. Li, Y. P. Zhao, F. K. Gui and B. Teng, Numerical simulation of the hydrody-
namic behavior of submerged plane nets in current, Ocean Eng. 33(17–18), 2352–2368
(2006).
8. Y. P. Zhao, Y. C. Li, G. H. Dong, F. K. Gui and B. Teng, Numerical simulation of the
effects of structures ratio and mesh style on the 3D net deformation of gravity cage
in current, Aquacultural Eng. 36(3), 285–301 (2007).
9. Y. P. Zhao, Y. C. Li, G. H. Dong and F. K. Gui, Numerical and experimental study
of submerged flexible plane nets in waves, Aquacultural Eng. 38, 16–25 (2008).
10. Y. P. Zhao, Y. C. Li, G. H. Dong and F. K. Gui, Numerical simulation of the hydrody-
namic behaviour of gravity cage in waves, China Ocean Eng. 21(2), 225–238 (2007).
August 4, 2009 9:39 9.75in x 6.5in b684-ch23 FA
666 Y.-C. Li
Chapter 24
Subrata Chakrabarti
Civil and Mechanical Engineering Department
University of Illinois at Chicago, 842 West Taylor
Chicago, IL 60607, USA
chakrab@uic.edu
This chapter will review and highlight the research being carried out today to meet
the challenges in the design, and operation of offshore structures. The subject
matter, while general in nature, will focus on one of the most unique areas in the
offshore structure design, namely, the fluid-induced responses of offshore struc-
tures and the associated structural design consequences. Due to the rapid growth
in the offshore field, particularly in deepwaters, this area is seeing a phenomeno-
logical advancement. The chapter will begin with an overview of the historical
development of fixed and floating structures. It will state the design status for
these systems. The fixed structure design is more mature today, even though
many aspects of it still remain empirical. For floating structures the design pro-
cedure is still advancing and more research is ongoing in this area. These will be
highlighted. The treatment of the individual components of the floating structure,
namely, the floater, the mooring system, and the riser system including their
interactive coupling effects with fluid will be discussed. The state-of-the-art in
the treatment of the individual components of the floating structure, namely, the
floater, the mooring system, and the riser system will be briefly described. The
design methods for these offshore components will be included. The basic dif-
ferences between the coupled and uncoupled systems and the complexity of the
later method will be discussed. The chapter will conclude with a discussion of
the present-day deepwater design challenges that remain and the research that is
needed to meet these challenges.
Offshore structures are located isolated in waters of the ocean with no continuous
access to dry land. Their design life-span ranges from a few years to as many as
25 years. In most cases, they may be required to stay in position in all weather
667
August 17, 2009 17:18 9.75in x 6.5in b684-ch24 FA
668 S. Chakrabarti
conditions. Offshore structures are defined by either their function or their config-
uration. The functions of an offshore structure may be one of the following:
• Exploratory Drilling Structures: A Mobile Offshore Drilling Unit [MODU] con-
figuration is largely determined by the variable deck payload and transit speed
requirements.
• Production Structures: A production unit can have several functions, e.g.,
processing, drilling, workover, accommodation, oil storage, and riser support.
• Storage Structures: These structures are used in storing the produced crude
oil temporarily at the offshore site before its transportation to the shore for
processing.
Sometimes a structure may be used for multiple functions.
The configuration of offshore structures may be classified by whether the
structure is a fixed structure, either piled or gravity, a compliant or articulated
structure, or a floating structure. The requirements of a floating structure are that
it be moored in place and that the facility remains under the environment within
a specified distance from a desired location achieved either by mooring lines, or by
dynamic positioning system using thrusters, or a combination of the two.
First, a short description of these structures and their applications will be dis-
cussed. Then the current state-of-the-art in the general hydrodynamic analysis asso-
ciated with the design of these structures, and the future problems that need to be
considered in these areas will be addressed.
Fixed offshore structures composed of small tubular members are mainly used
for the production of oil and gas. These structures may be composed entirely of
Fig. 24.1. Historical development of fixed jacket structures (Courtesy Shell Oil Co.).
August 17, 2009 17:18 9.75in x 6.5in b684-ch24 FA
small battered members which are piled in place. These are called jacket structures.
The evolution of fixed jacket structures with water depth is shown in Fig. 24.1.
Note that, within a short span of 40 years, the fixed structure extended from a
water depth of 6 m to over 400 m.
Exploratory drilling operations may be performed with fixed structures, called
jack-ups, which are mobile. The jack-ups are usually buoyant during transit and
are towed from station to station. Once they reach their destination at the drilling
site, the small-membered legs of the jack-up are set on the ocean bottom and the
deck is jacked up above the water level on these legs for the drilling operation. The
gravity-type fixed structures, on the other hand, are mostly composed of large steel
shell-type members, which may also be used for storage in addition to production
of oil and gas. There have also been many concrete gravity structures in existence,
most of which are located in the North Sea.
(a) (b)
Fig. 24.2. Flow effects on body shapes. (a) Steady flow past airfoil and (b) steady flow past
bluff body.
August 17, 2009 17:18 9.75in x 6.5in b684-ch24 FA
670 S. Chakrabarti
and the presence of vortices, the pressure behind the body in steady flow is low
compared to the forward of the body fluctuating asymmetrically. These forces are
difficult to compute numerically specially at high Re and research in this area is
continuing.
The flow past the body, as shown in Fig. 24.2(b) for a cylinder in steady flow,
produces an asymmetric (fore-aft) pressure distribution around the cylinder, which
traverses the flow axis at a given (Strouhal) frequency at a low or intermediate Re.
This pressure distribution, in turn, generates an oscillating lift (or transverse) force
with near-zero mean, in addition to an inline mean drag force, and an oscillating
inline force of smaller magnitude over the steady force.
Since the vortex shedding behind the cylinder is alternating in nature, the trans-
verse force will be periodic and its frequency (at low Re) will depend on the fre-
quency of vortex shedding. This is demonstrated in Fig. 24.3 computed in a 2D
numerical simulation. The result presented applies to a Reynolds number of 200. At
higher Reynolds numbers, this shedding of vortices is more random and complex
causing a more irregular large transverse force on the body, which, at the present
time, causes significant computational difficulty.
Fig. 24.4. Drag coefficient for a smooth circular cylinder in steady flow.
Fig. 24.5. Lift coefficient for a smooth circular cylinder in steady flow.
August 17, 2009 17:18 9.75in x 6.5in b684-ch24 FA
672 S. Chakrabarti
acceleration. No analytical or numerical solution exists for the forces which (per unit
length) are computed from the well-accepted empirical formula, commonly known
as the Morison equation,29 given as
f = CM AI u̇ + CD AD |u|u, (24.3)
When current is present in the direction of wave (negative sign indicates opposing
current), the Morison equation is modified in the presence of current in terms of a
relative velocity model:
f = CM AI u̇ + CD AD |u ± U |(u ± U ). (24.6)
In this case the values of the hydrodynamic coefficients will be different than those
for the wave alone.
Similarly for inclined cylinders, e.g., those found in the cylindrical bracings of
a jacket structure, the modified Morison equation considers force normal to the
cylindrical member based on the normal component of velocity and acceleration of
the water particle at the point.
S
i
blocked by the forward cylinders. API Guidelines2 give a composite blockage factor
for a dense cylinder group in steady current as follows:
N −1
i=1 (CD A)i
CBF = 1 + (24.7)
4A
6
φ = φo + φs + φRn (24.8)
n=1
674 S. Chakrabarti
Fig. 24.7. Illustrative sketch of the linear force calculation on a vertical cylinder.
in which σ represents the source strength, (a, b, c) represents the source point on
the surface of the structure and (x, y, z) the field point in the fluid, ds is the flat
area of the panel on the submerged surface. The function G represents the near-field
Green’s function given in a series or an integral form. The source strength function
σ, is computed from
∂G
2πσs (x, y, z) − σs (a, b, c) (x, y, z; a, b, c)ds = −4πun (x, y, z), (24.11)
S ∂n
where un = known normal fluid velocity at (x, y, z) due to the incident wave.
Equation (24.11) is solved numerically by assuming the field point to be on the
structure surface and setting it up in a matrix form in terms of the centers of the
panels. An N × N complex matrix is formed to describe Eq. (24.11), where N is
the total number of panels. The solution for the source strengths σ at the center
of each panel is obtained by the inversion of the complex matrix. For a large value
of N , this computation is time consuming.
For the radiated potential the right-hand side of Eq. (24.11) is replaced by the
normal displacement of the body at one of six degrees of freedom. Since the only
difference is the right-hand side, this equation may be solved at the same time once
the inverted matrix is obtained. The radiated potential provides the added mass
and damping coefficients of the structure in six degrees of freedom.
Once the diffraction/radiation potentials are known at the center of each panel,
the external forces on the submerged body due to the total diffraction and radiation
potential, are obtained respectively from the integrals
FkD = −iρω (φo + φs )nk ds (24.12)
S
August 17, 2009 17:18 9.75in x 6.5in b684-ch24 FA
and
FkR = −iρω φRk nk ds, (24.13)
S
Fig. 24.8. Illustrative sketch of the fully nonlinear force calculation on a vertical cylinder.
August 17, 2009 17:18 9.75in x 6.5in b684-ch24 FA
676 S. Chakrabarti
by Boo et al.5 for the simulation of nonlinear regular and irregular waves and by
Isaacson and Cheung21 for wave–current–body interaction problems. There exist
other open-boundary conditions, such as absorbing beaches by artificial damping on
the free surface6,19,27 or matching with linear time-domain solutions at far field.17
It is well known that a properly designed artificial damping on the free surface does
not have to be far from the body and can damp out most of the wave energy, if
its length is greater than two wave lengths. Therefore, it is more effective than the
matching technique and ideal to damp out relatively short waves.24
When the simulation starts, a ramp-function7 at the input boundary may be
applied. The ramp function prevents the impulse-like behavior at the start and
consequently reduces the corresponding unnecessary transient waves, which usually
wastes computational time to die out. As a result, the simulation is more stable and
soon reaches the steady state.
Accurate calculation of the time derivative of velocity potential is very crucial
in obtaining correct pressure and force on the body surface at each time step.
There are several ways to obtain this velocity potential. Backward difference is the
simplest way using the potential values of previous time steps. In case of a stationary
structure, more accurate finite-difference formulae7 can also be used. The wave
force on the body surface is calculated by integrating Bernoulli’s pressure over the
instantaneous wetted surface from the nonlinear wave. While the above development
is shown for a wave force computation, the method can be easily extended to include
moving structures as well.24
• Floating hull: facilitating the space for the operation of the production work, and
storage for supplies,
678 S. Chakrabarti
• Mooring system: providing a connection between the structure and the seafloor for
the purposes of securing the structure (generally called station-keeping) against
environmental loads, and
• Riser system: achieving drilling operation or product transport.
8 admass+1% damping
admass + 15%damp
6 no plate+1%damp
Heave, ft/ft
4
0
10 11 12 13 14 15 16 17 18 19 20
Period, sec
Fig. 24.10. Example heave response of a Truss Pontoon Semisubmersible including additional
heave plate added mass and damping.
Note that Eq. (24.18) is still nonlinear in the motion amplitude due to the third
term. A simple iterative technique is chosen to solve the above linearized equation.
In the first step, a linear harmonic solution for xl0 = 0 (l = 1, . . . , 6) on the left-
hand side is obtained. This first iteration value is introduced on the left-hand side
of solution and the process repeated. Generally, between two and three iterations
produce convergence in the results.
An example of the computed heave motion from Eq. (24.18) of a floater (pictured
on the left of Fig. 24.10) is given in Fig. 24.10. Different responses refer to with and
without additional added mass and nonlinear damping generated from the heave
plates at the bottom of the floater.
680 S. Chakrabarti
body, and the solution is generated by a forward integration scheme. The analysis
can easily accommodate the nonlinearities due to the excitation forces (e.g., wind,
current, and waves), as well as damping and restoring forces typically encountered
in a moored floating system. The waves can be single frequency or composed of
multiple frequency components following a given wave energy spectral model (e.g.,
PM or JONSWAP).
For the series representation of long crested waves for a given wave energy
spectrum, the wave profile η may be decomposed into N components as:
N
η(x, t) = an cos[kn x − ωn t + εn ], (24.19)
n=1
where an represents the wave amplitude having frequency ωn and a randomly chosen
phase angle εn . The corresponding linear and nonlinear wave excitation force, and
structure velocity dependent radiation force for each of these wave components
are computed from the linear diffraction/radiation theory at a given time step. At
each time step, a set of second-order differential equations (similar to Eq. (24.18),
but retaining all the nonlinearities present in the system) is solved to obtain the
accelerations of the system. For a single degree of freedom system, the current value
of acceleration is computed from the equation of motion for the total random force
at time t, F (t):
where the subscripts c and p stand for the current and previous values, respectively.
The force time history F (t) may be composed of nonlinear wind and current forces
including wind spectrum, the linear diffraction forces, second-order steady and oscil-
lating forces, and Morison and lift forces. The solution is initiated with prescribed
values for the displacements and velocities, and these values are calculated for the
next time step from the derived accelerations by the forward integration scheme.
e.g., finite-difference:
20
10
Wave
0
profile
400 500 600 700
-10
WAVE
-20
1 Tower
Motion
0
400 500 600 700
OMEGA
-1
20
15 Tanker
Surge
10
SURGE
5
400 500 600 700
4.E+04
LINE
Line
2.E+04
Tension
0.E+00
400 500 600 700
Fig. 24.11. Time history of motions and load due to a random wave on an SPM.
During the earlier floating structure response analysis, the mooring lines are suc-
cinctly assumed to be merely a nonlinear spring and the effects of the environment
on the lines themselves are ignored having small overall effect on the floater response.
In the design of these lines themselves for a floating system, however, the environ-
mental forces may be an important criterion. In the process of design of an offshore
system, one first selects their layout, geometry, and the initial mechanical and struc-
tural properties.
Different types of mooring and anchoring system for floating offshore structures
of today are shown in Fig. 24.12. The lines (typically 8 to 12) are arranged in a
symmetric arrangement except to make room for risers if needed. There are prin-
cipally two types of mooring system in use today — catenary mooring and taut
mooring system. The catenary lines often consists of chain-wire-chain combination.
The taut lines could be steel wires or synthetic polyester. It is customary to use
drag anchors for the catenary lines. For the taut polyester lines, a vertically loaded
plate or suction anchor may be suitable.
August 17, 2009 17:18 9.75in x 6.5in b684-ch24 FA
682 S. Chakrabarti
For the taut mooring system the restoring force comes primarily from the stretch
in the line. The restoring force for the conventional catenary system results princi-
pally from the submerged weight and changes in the catenary shape.
from the wave-induced vessel motion at wave and slow-drift frequency are com-
puted. Then the maximum line tensions resulting from the total vessel offset for each
possible environmental direction are determined. The line tensions are compared
with the minimum breaking load given the safety factor. DNV16 currently suggests
separate safety factors for the mean and dynamic loads to avoid excessive conser-
vatism, since mean loads are much higher and more reliably predictable. As before,
the maximum peak line loads with 1 line broken are recalculated. If the proposed
mooring specification fails the safety factor test, then a new specification is tried.
The dynamic design of the mooring lines includes additional loads on the
mooring lines themselves. In this case the analysis is much more involved and a
numerical method is needed.
The dynamic mooring system design is a two-step procedure as follows. First,
the motions of the floater independent of the line dynamics are calculated to
determine the top-end oscillation of lines. Loads on the floater are the steady
current, and steady and fluctuating wind, wave and wave drift. For the floating
structure, mooring system is an external nonlinear stiffness term (step 1, Fig. 24.13)
as shown earlier. Once the floater dynamics are known, the motions of the mooring
line attachment (fairlead) points are determined. For the mooring lines, motions of
the floating structure are included at the attachment point as an externally defined
oscillation (step 2). To determine the mooring line dynamics, a time domain analysis
is required here. External effects on the mooring lines include the following: top-
end motion, current load, wave load, seabed lateral friction, and soil spring and
damping effect on the portion of the line on bottom.
A lumped mass, finite element or finite difference scheme may be used to
model the lines. The line is decomposed into a number of straight elements (bars)
with linear shape function. The distributed mass plus hydrodynamic added mass
is lumped at the element nodes. The hydrodynamic damping is included for the
relative motion between the line and the fluid. Damping levels vary significantly
depending on water depth, line makeup, offsets, and top-end excitation. Quite often
a modified Morison equation is used to represent the environmental effect. At each
time step, a standard set of matrix equation is developed composed of the inertia,
damping, and stiffness matrix.
684 S. Chakrabarti
It is often important to check the fatigue life of the mooring line. A mooring line
fatigue analysis may be based on long-term cycles of the dynamic tension on the line
due to time varying loads. In fatigue, the factor of safety (FOS) is recommended1
to be between 3 and 10 depending on the lines being inspectable or not. Because
of the low level of experience of lighter polyester lines in deepwater applicable, the
FOS is taken to be as high as 60.
testing is generally done at the University laboratory water channels, and wave
tanks. They are limited by small scales, modeling problems, and measurements.
Their main purpose is to increase the understanding of the flow problems past
risers and generating empirical coefficients. Field measurements often include in-
situ testing in full scale and face the constraints of cost, practicality, environmental
limitation, and client confidentiality.
In spite of these ongoing developments, today almost all riser design analyses
are performed by empirical methods. Static analysis is generally based on steady
current loads (uniform or shear) and provides the riser deflected shape, stresses at
various points on the riser, top and bottom angles, and its structural mode shape
under load. Dynamic analysis considers both the inline and transverse unsteady
loads. In addition to the riser shapes and stresses, it provides the vortex induced
vibration, and fatigue life of the riser.
686 S. Chakrabarti
a CFD analysis incorporating the fluid and the structure is desirable. Due to high
Reynolds number the problem is quite involved and time-consuming. However, the
recent results look promising.12
The numerical analysis applies the turbulent incompressible Navier–Stokes equa-
tions in describing the conservation of mass and momentum in three-dimensions.
Among the numerical methods, most of the CFD codes are 2D on parallel planes.
They include finite element, finite difference, discrete vortex, vortex-in-cell, and
vortex tracking method. A few examples are FSI-Navsim (Norsk Hydro), USP,
DeepFlow (IFP), VIVIC (Imperial College), TACC, Orcina VT, and Orcina WO
of OrcaFlex. There are also a few large 3D Commercial Codes being adopted for
offshore applications, e.g., Fluent and Acusolve.
A fundamental problem in the simulation of long marine risers is the large
scale of the computational fluid dynamics problem itself.12 The risers may have
lengths of thousands of meters so that using three-dimensional CFD solutions
seems impossible. A practical approach to riser VIV predictions that has been pro-
posed is to combine a series of two-dimensional fluid flow solutions along the riser
axis. These “strips” are pieced together with a structural model of the riser to
obtain a prediction of the fluid–structure response. This method reduces the large
3D CFD problem to a large number of small 2D problems. However, this “strip”
method has some serious shortcomings. In particular, the flow around bluff bodies
is inherently three-dimensional so that the 2D strip solutions can only be expected
to give approximate answers. In addition, the riser may be at a steep angle of
attack to the flow as is often the case with steel catenary risers (SCRs) creating
strong axial flow components. Also, VIV suppression devices like helical strakes
that have a very strong three-dimensional wake cannot be modeled. Finally, the
strip method requires that some kind of interpolation method be used to estimate
the forces between the strips. There is no general rule available to make such
interpolations.
In addition to the sheer scale of the CFD problem, several other difficult
problems remain. Although the structural response of the riser usually does not
involve nonlinear material response, it usually includes nonlinear geometric effects,
such as large displacements and rotations. The presence of these effects means that
a nonlinear structural model must eventually be incorporated in the solution of
riser problems. In addition, large motions of the generated meshes must be accom-
modated in the fluid flow solution. Finally, it cannot be expected that the flow
environment remains constant over the entire length of the riser. Both flow speed
and direction may change. Any general solution must be able to treat these effects.
A simple linear structural model for the solutions is described here.12 The riser
axis is oriented in the z-direction and the eigenmodes are assumed sinusoidal. Then
the eigenvectors have the form:
nπy
ζin = sin , (24.25)
L
where the ζin is the eigenvector associated with the nth mode and the ith index
indicates the x or z directions, L is the riser length, y is the distance along the riser
axis, and n is the mode number. It should be noted that the risers modeled here are
August 17, 2009 17:18 9.75in x 6.5in b684-ch24 FA
688 S. Chakrabarti
tension dominated in that the bending stiffness of the riser is negligible compared
to the stiffness due to tension. Also, the use of a sine shape in Eq. (24.25) implies
that the riser tension is constant along the riser length.
With this approach, the motion of the riser is assumed to be a linear summation
of the vibration modes. The response is found by solving the equation:
where yin are the modal amplitudes, and mni mni kin are the associated mass, linear
damping and stiffness for each mode. The forces fin are computed at each time step
using the corresponding eigenvector. The displacement of the riser for the particular
mode is then computed before moving to the next time step.
Modeling a full scale deepwater riser was possible with this analysis [Eq. (24.26)]
using a medium size computational cluster. Satisfactory experimental validation was
obtained from the solution within the limitations, uncertainties, and assumptions
of the available data and CFD model.12
However, some refining of data is still needed for a successful correlation. This
is illustrated by the following case. A test was performed (Fig. 24.16) by towing a
long riser model in water and measuring the detailed inline and cross-flow response
of the riser model.10 The test setup and input data were supplied to the available
software, both research type and commercial codes, for a blind correlation check.
The results from this analysis are given in Fig. 24.16. It shows the measured inline
and cross-flow response compared with results from these codes. The names of the
codes used in this blind test are shown on the left (see the original paper for their
identification). It is clear that the correlation is quite varied with different degrees
of success without a consistent trends in them. After this test, the results were made
available to the users of the codes for further verification. The second correlation
turned out much better than the first.
Thus this type of analysis tool appears to have the capability to predict the
results reasonably well, but needs further experience and validation. In order that
such validation can be achieved more precisely, more accurate benchmark results
should be available. Only then it is possible to make use of such tools in the design
of deepwater risers with confidence.
Because of the numerous uncertainties in the riser analysis, the factor of safety
in the fatigue life in a riser design is considered large. The guidelines, e.g., API,
suggest values of FS = 3, for inspectable riser applications, and FS = 10–20, for
un-inspectable applications with increased uncertainty (e.g., VIV).
The uncoupled analysis is still the traditional design practice for floating production
systems. As already explained, in the uncoupled formulation the numerical analysis
tool is based on the hydrodynamic behavior of the floater, not influenced by the
nonlinear dynamic behavior of the lines. Two distinct stages of design may be
identified in an uncoupled methodology:
August 17, 2009 17:18 9.75in x 6.5in b684-ch24 FA
Cabin
Norsk Hydro
USP
Water surface in the flume
Riser:28mm diameter,
13.12m long Deep Flow
The bottom of the riser is
connected to a stiff frame
(not shown) mounted on Incident velocity
the carriage. profile at the riser;
maximum velocity 1m/s
VIVIC
Orcina VT
Orcina WO
VIVA
VIVANA
VICoMo
SHEAR7
ABAVIV
(1) The first stage is the design of the floating moored system, employing pro-
grams based on an uncoupled model, to analyze the hydrodynamic model of
the floating hull and determine its motions. The contribution of the risers is
ignored in this analysis, and the behavior of the mooring lines is approximated
by scalar coefficients inserted in the 6-DOF equations of motion of the hull, to
represent the contribution of the lines (in terms of mass, damping, stiffness, and
current loading). In this methodology, the stiffness of the risers is generally not
taken into account.
August 17, 2009 17:18 9.75in x 6.5in b684-ch24 FA
690 S. Chakrabarti
(2) The second stage consists of the design of the risers. The floater motions that
result from the first stage (expressed either as RAOs or time series) are applied
to the top of each individual riser, and the structural response of each riser
(subjected also to the wave and current loads) is assessed employing a finite-
element based program. In the analyses of the first stage, the values for the
coefficients that approximate the behavior of the lines should be estimated by
the engineer, determined via simplified analytical models, or calibrated using
results from experimental tests.
The effects of the appendages, namely mooring lines and risers, become increasingly
important as the structure moves into deeper waters. Thus their effects in the total
system may not be ignored in these cases. In a coupled system analysis, the mooring
lines and risers are included as an integral part in the numerical model simultane-
ously with the floating structure. It has recently been recognized13,18,22,23,26,33,34
that the most accurate design methodology for floating offshore systems should
employ analysis programs based on coupled formulations. Ideally such programs
should incorporate, in a single code and data structure, a hydrodynamic model for
the representation of the vessel, coupled to a 3D finite-element model for the rep-
resentation of the hydrodynamic and nonlinear structural dynamic behavior of the
mooring lines and risers. The characteristics and advantages of this so-called “fully-
coupled” methodology are described by Correa et al.13 One characteristic of this
methodology is its excessive computer costs; therefore, “hybrid” methodologies and
analysis procedures are presented in order to circumvent this problem and to grad-
ually advance toward a fully coupled and integrated design methodology.
In a complete coupled analysis (Fig. 24.17) the 6 DOF motions of the floating
vessel is solved at a given time step. The loads and motions of the top of each of the
riser/mooring line are determined. A finite element (or lumped mass) method for
August 17, 2009 17:18 9.75in x 6.5in b684-ch24 FA
riser and mooring line is applied. For example, to compute the dynamic position of
the entire mooring line at each time step in the simulation, a lumped mass method
is used; see, for example, Derksen and Wichers.15 In this lumped mass method, the
mass of the mooring line is lumped to a finite number of nodes that are connected
by linear springs, corresponding to the axial stiffness of the line. Bending stiffness is
not taken into account. The current (and wave) loads on the mooring line elements
are computed. In the simulation, the mooring forces in the equation of motion are
kept constant over a time step (typically taken as 0.1 seconds). During this interval,
the equation of motion is integrated with a Runga–Kutta method and the position
and velocity of the buoy at the new time step is obtained. The new position of
the buoy is then used in the lumped mass method to compute the position of and
the tension in the mooring lines at this new time step, after which the procedure
is repeated until the simulation reaches its desired convergence. The new values of
mooring line top loads and motions are imposed to the vessel and the 6 DOF of
the vessel is solved again. Iterations continue until a convergence is reached and
proceeded to the next time step.
692 S. Chakrabarti
for which more complete reliable and practical CFD or NWT type solutions are
desirable.
The design of large floating structures has traditionally been carried out using
the uncoupled analysis in which the dynamics of the floater are considered along
with the mooring lines and risers treated as external stiffness terms. Once the floater
dynamics are known, they are applied to the individual mooring line or riser in the
design of these components. Today’s design of risers and mooring line is generally
empirical in nature. Progress is being made in the development of more complete
and sophisticated numerical analysis for their design. This area needs further devel-
opment for improved efficiency and accuracy with less dependence on empiricism.
They need careful validation before their practical application.
Moreover, ultradeep water and relatively smaller size of floaters necessitated the
interaction of the appendages and the environmental effects on themselves to be
incorporated in a coupled analysis with the floater. This is mostly in the devel-
opment stage, even though some limited design verification is currently being carried
out by the offshore industry with this type of tools.
References
694 S. Chakrabarti
Chapter 25
Jiin-Jen Lee
Sonny Astani Department of Civil and
Environmental Engineering
University of Southern California
3620 S. Vermont Avenue, Kaprielian Hall 200
Los Angeles, CA 90089-2531, USA
jjlee@usc.edu
Xiuying Xing
Sonny Astani Department of Civil and
Environmental Engineering
University of Southern California
3620 S. Vermont Avenue, Kaprielian Hall 229
Los Angeles, CA 90089-2531, USA
xiuyingx@usc.edu
Harbors are built to provide a sheltered environment for the mooring of ships and
vessels. For some wave periods the semi-enclosed harbor basin acts as a resonator
to amplify the wave motions in the harbor due to the combined effects of wave
diffractions, refractions, and multiple reflections from the boundaries. This unde-
sirable wave motion could induce significant ship motions, damage ships and dock
facilities, and delay loading and unloading activities if the resonant wave periods
are close to that of the ship mooring system. Harbor planners and engineers need
to model the wave induced oscillations as new harbor layouts are contemplated.
This chapter presents a computer model to be used for predicting the response
characteristics of arbitrary shape harbors with variable depth. The model incor-
porates the effects of wave reflection, refraction, diffraction, and dissipation losses
due to boundary absorption, bottom friction, and energy losses due to the flow
separation at the entrances. The model is applied to four real harbors and the
model results have been shown to agree surprisingly well with the field data
obtained from tsunami-genic events as well as hurricane induced wave motions.
The computer model is shown to be an effective engineering tool for harbor
695
August 14, 2009 9:52 9.75in x 6.5in b684-ch25 FA
planning and design to derive ways of eliminating or altering the harbor response
so that the harbor may indeed provide a sheltered environment for moored ships
and vessels.
25.1. Introduction
Harbors are built to provide a sheltered environment for the mooring of ships and
vessels. In order to accomplish this goal, marine structures such as breakwaters and
jetties are constructed (either detached or undetached to the shore) to reduce the
wave energy incident from the open sea.
The effectiveness of the breakwaters and jetties in reducing the incident wave
energy must be ascertained in the planning and design of harbors. Normally, the
breakwaters and jetties are effective in reducing the incident wave amplitude for
waves with shorter wave periods (in the order of 16 s or less). As the wave period
increases, the effectiveness of the breakwaters or jetties in reducing the incident wave
amplitude progressively decreases. For waves of shorter wave period the effect of
wave reflection from the harbor boundaries is quite small, thus only wave refraction
due to changing water depth and wave diffraction around the breakwaters and jetties
need to be considered.
When the wave period increases (thus the wavelength increases), the combined
effect of wave diffraction, wave refraction, and wave reflection from the harbor
boundaries is very significant. In fact, it is possible that for certain semi-enclosed
harbor the combined effect of wave diffraction, refraction, and multiple reflections
from the boundaries can cause significant increase in the wave amplitude compared
with the incident wave amplitude. This is commonly referred to as “harbor reso-
nance” due to long waves.36,37
Resonances or oscillations due to long period waves in bays and harbors have
often been observed. For example, in response to the daily flooding and ebbing
tide, the Bay of Fundy (at the border of eastern Canada and northeast of the
United States) has produced the largest tidal range in the world (approximately
50 ft). At the Crescent City harbor region in northern California, larger than
usual water surface elevations have been observed in response to a tsunami-genic
event whether it is distant tsunami or local tsunami. The records at the tide
gauge station located in Crescent City harbor will be discussed later in this
chapter.
The resonance in a harbor could result in large fluctuation of water level in
certain areas and produce strong currents throughout the harbor. In addition, it
may also induce large ship motions for moored ships, especially if the fundamental
resonant period of the harbor is close to that of the ship mooring system. Such oscil-
lations could last as long as several days, delaying the cargo loading and unloading
activities, breaking the mooring lines, and even damaging the moored boats and
dock facilities. As a result, some ships or boats may have to move out to the
open sea to avoid those large oscillations within the harbor, resulting in significant
economic loss.
August 14, 2009 9:52 9.75in x 6.5in b684-ch25 FA
(a) What are the important wave periods that the harbor would respond to?
(b) Are these wave periods close to the resonant periods of the ship mooring system?
(c) What amplification of wave and current can be expected?
(d) How will the harbor respond to different forcing functions such as tsunamis or
hurricanes (called typhoons in the Pacific region) as opposed to long period
swells?
(e) Will the response characteristics change when the harbor layout is changed or
modified?
In this chapter, we have attempted to show how one can use the computer model
to arrive at the answers to those questions listed above. Even though no one can
completely eliminate the need for physical models, the use of computer models can
be shown to provide fast and reliable answers to the questions in such a way that
the physical models, if absolutely needed, can be performed more efficiently and
effectively to save time and money.
functions on each node of the whole system are calculated by solving a series of dif-
ference equations with certain boundary conditions. The finite difference method is
not restricted to any governing equations and the approximation is straightforward.
But the calculation might not always be stable for time dependent problems and the
curved boundary can only be approximated by rectangular grids. The earliest use of
the method was conducted by Leendertse.25 The two-dimensional depth-averaged
governing equations containing nonlinear terms were solved to study the long wave
propagation mechanism. Other examples can be found in Refs. 8 and 40.
in which ω and k are wave frequency and wave number. c = ω/k is the wave celerity,
and cg = ∂ω/∂k is group velocity.
The theory is restricted to irrotational and inviscid linear harmonic waves, and a
slowly varying bathymetry condition. The loss of energy due to friction or breaking
is not taken into account. It was found that the results of regular mild slope equation
tend to overestimate the amplification factor of the harbor resonance because of
ignoring the separation and dissipation losses.34 The effects of bottom friction and
boundary absorption on wave scattering were examined by Chen.6 The regular mild
slope equation (25.1) was modified by including a parameter λ:
∂ ∂φ
λccg + k2 ccg φ = 0. (25.3)
∂xj ∂xj
Similar approaches were developed by Lejeune et al.24 and Yu.43 The regular
mild slope equation was modified through a parameter to account for the bottom
friction. Shorling and breaking effects were included in many works such as by Balas
and Inan1 and Massel.28 The nonlinear effects of higher order were investigated
by Mei.29
Various extended approaches (so-called extended mild slope equation) were gen-
erated to improve the performance of the regular mild slope equation for abrupt
and undulating topography. Terms of bottom curvature and slope were included
into the regular mild slope equation by groups of researchers.4,5,31 Although dif-
ferent approaches were used to derive the equations, they all obtained equivalent
formulae as
in which f1 and f2 are functions of kh, and h is the water depth. The effects of
curvature and slope terms were examined by Chandrasekera and Cheung5 and Lee
and Yoon.17
Although the mild slope equation was obtained with the assumption of “mild
slope,” the work by Booji3 showed that the regular mild slope equation is applicable
for bottom slope as large as 1/3. Since the mild slope equation can be conveniently
implemented in a finite element model, we will apply it to real harbors in this
presentation. The basics of the model will be shown in the following section.
A hybrid finite element model using the mild slope equation will be briefly described
herein. The model solves the mild slope equation over arbitrary shape harbor
of variable water depth. It incorporated the effects of wave reflection, refraction,
diffraction, and dissipation losses due to boundary absorption, bottom friction, and
energy losses due to the flow separation at the entrances. An earlier version of
the model has been applied successfully for the modeling of harbor responses of
Los Angeles and Long Beach harbors by Lee et al.22 It has also been applied to
model a large coastal region for assessing the potential coastal impacts of tsunamis
surrounding the island of Taiwan.21
Cg ω 2
∇ · (CCg ∇φ) + φ = 0. (25.5)
C
It can also be written in the form of (25.1). In which φ = φ(x, y) is the horizontal
variation of the velocity potential Φ as shown in Eq. (25.2), C = ω/k is the wave
celerity, and the group velocity is,
C C 2kh
Cg = (1 + G) = 1+ (25.6)
2 2 sinh 2kh
φR = φI + φS , (25.7)
where φI , φR , and φS are incident wave potential (including incident wave and
reflected wave), outer region wave potential, and scattered wave potential, respec-
tively. For the far-field boundary, the Sommerfeld radiation condition is applied.
√ ∂
lim r − ik φS = 0. (25.8)
r→∞ ∂r
The scattered wave potential function satisfying the Helmholtz equation and the
radiation condition can be expressed as
∞
φS = Hn (kr)(αn cos nθ + βn sin nθ), (25.9)
n=0
absorbed and partially reflected. The energy flux out through the boundary is
formulated using a second-order scheme as
∂φ iα ∂ 2 φ
= −iαkφ − , (25.10)
∂n 2k ∂s2
where α is the absorption coefficient with a range of 0 ≤ α ≤ 1.
It should be mentioned that for partially absorbing boundaries, the changing of
absorption coefficient α may represent different boundary conditions. When α = 0,
it represents fully reflecting condition, i.e., ∂φ/∂n = 0. When α = 1, it represents
fully absorbing condition. The relation between absorption coefficient and reflection
coefficient R for an incident wave angel θi is
(α cos θ i )2 − 2α cos θi + α2
R=− . (25.11)
(α cos θ i )2 + 2α cos θi + α2
∂φT ∂φi
= KT = ikKT φi , (25.12)
∂n ∂n
where KT is the transmission coefficient through the breakwater.
U2 U
∆H = fe = fe |U0 | = Ke U (25.13)
2g 2g
in which,
fe
Ke = |U0 |, (25.14)
2g
where fe represents the dimensionless entrance loss coefficient. |U0 | is the averaged
velocity at the harbor entrance computed considering no entrance loss and U is the
new entrance velocity to be computed considering the entrance loss in the model.
Thus the relationship of the complex velocity potentials at the entrance can be
written as
g U
φ1 = φ2 + ∆φ = φ2 + fe |U | . (25.15)
iω 2g
August 14, 2009 9:52 9.75in x 6.5in b684-ch25 FA
Ef = τb Ub , (25.16)
where τb is the instantaneous complex shear stress at the bottom, which can be
formulated by the water particle velocity near the bed,
1
τb = ρKb |Ub |Ub , (25.17)
2
where Kb is a dimensionless friction coefficient. The particle velocity near the bed is
∂φ 1
Ub = = ∇φ exp(−2iωt). (25.18)
∂s b cosh kh
where αb = a/sinh kh, Cbf is the Nikuradse roughness height, a is wave amplitude.
φ = N i φi (25.21)
∇φ = ∇N i φi = B i φi (25.22)
∂φS ∂NSi i
= φ = PSi φiS . (25.24)
∂r ∂r S
August 14, 2009 9:52 9.75in x 6.5in b684-ch25 FA
in which,
T Cg ω 2 T
M = CCg B B − N N dxdy − iωCg αN T N ds
A C ∂B
2
2 T i 1
+ iωCg Kt N N ds − fω B T Bdxdy
∂T A ω cosh kh
g
− iωCCg (1 − Kj )N T B − Ke B T B ds (25.27)
∂E iω
M1 = CCg PST NS ds (25.28)
∂A
M2 = − CCg N T PS ds (25.29)
∂A
∂φI
Q1 = − CCg N T ds (25.30)
∂A ∂nA
Q2 = CCg PST φI ds. (25.31)
∂A
The quadratic Lagrange elements are used for the region shoreward of the imaginary
common boundary ∂A. For the open sea region an eigenfunction expansion is used
to represent its solution. Solutions for the two regions are matched at the common
boundary ∂A.
A substructure technique is used in the model for which the whole calculation
domain is divided in several small domains. The matrix equations are generated
in each subdomain. The solutions in each subdomain are solved separately based
on their own boundary values. The solution for the entire domain is obtained
by matching the results at the imaginary boundaries between those subdomains.
This technique largely decreases the matrix size, thus effectively accelerating the
calculation.
earthquake generated tsunamis, such as the Crescent City harbor. Some of them
were caused by the typhoons and winter storms, such as the Hualien harbor in
Taiwan and the Pohang harbor in Korea. The oscillation modeling in the four basins
were examined by the numerical model discussed in the previous section. The results
will be presented one by one, and then conclusions will be drawn at the end of this
section.
Fig. 25.3. Simulation domain for Crescent City harbor (left) and locations of special interest as
A, B, C, D, E, and the tide gauge (right).
in the harbor was caused by the tsunami produced by the Kuril Islands earthquake
on November 15, 2006.
The model region for the computer model is shown in Fig. 25.3 (left). Only the
major grid blocks are shown, the model contains 9,709 finite elements and 39,688
nodes with eight incident wave directions (indicated as 1 to 8). The Crescent City
harbor is shown in Fig. 25.3 (right) with five locations of special interest as A, B,
C, D, E, and the tide gage noted.
Figure 25.4 presents the computer simulation response curves with the two dis-
tinct resonant periods at the tide gauge station of Crescent City harbor under dif-
ferent scenarios of wave directions. The ordinate is the amplification factor defined
as the wave height at the tide gauge station divided by the incident wave height.
The abscissa is the dimensionless wave number kl (where k is the wave number, 2π
divided by wavelength, and l is the characteristic length of the harbor which is the
length from the outer harbor entrance to the facing coastal line about 4,363 feet
in the present model). It clearly shows that 22.0 min and 10.3 min resonant periods
existed at the tide gauge station.
Figure 25.5 shows the response curves at locations A–E and tide gauge station
with incident wave coming from direction 2 (oriented toward north). It provides a
clear indication that the waves are amplified as the inner harbor region is approached
for both the 22.0 min and 10.3 min resonant wave periods. Thus, it appears that the
more inner the location is, the more vulnerable for tsunami hazards. It explained
the damage occurred in the small inner harbor on November 15, 2006 during
which strong currents were invoked by the large oscillations. This large oscillation
pushed the boats against the berth facilities and colliding with other neighboring
boats.
An examination of the tide record associated with several different tsunami
events indicates that water levels at the recording station at Crescent City harbor
have been amplified from waves originated from near-field as well as that from
August 14, 2009 9:52 9.75in x 6.5in b684-ch25 FA
Fig. 25.4. Response curves at tide gauge location for different incoming wave directions.
Fig. 25.5. Response curves at locations noted by A–E and tide gauge station with incident wave
coming from direction 2.
August 14, 2009 9:52 9.75in x 6.5in b684-ch25 FA
Fig. 25.6. Historical records at the Crescent City harbor for events on (a) June 15, 2005,
(b) November 15, 2006, (c) January 13, 2007, and (d) August 15, 2007.
far-field. Figures 25.6(a)–25.6(d) show tide gauge records of water surface eleva-
tions during the near-field earthquake generated in offshore of northern California
(June 15, 2005, Magnitude 7.2), and the three far-field earthquakes. Two generated
in Kuril Islands, Japan (November 15, 2006 and January 13, 2007, with Magnitude
8.3 and 8.1), and one generated in offshore of Peru (August 15, 2007, Magnitude 7.9).
The spectral analysis of the records shown in Fig. 25.6 has been performed, and
the spectral density distributions are correspondingly shown in Fig. 25.7. It can
be seen that the dominant waves in those events, which have the highest energy
density, are all around 21 to 22 min.
The incoming wave directions in the four events are roughly from southwest
for the northern California earthquake, from west for the two Kuril Islands earth-
quakes, and from south for the Peru earthquake. The simulated response curves
for these incoming wave directions superimposed with the observed resonant waves
corresponding to those events are plotted in Fig. 25.8. The results indicate that
the actual resonant periods are very close to those computed, especially the funda-
mental mode at 22 min.
Water depth affects the value of wavelength for a certain wave period. Since
the actual water depth may be varied due to the tide condition, the wave period
may change for a certain wavelength in different events. The slight variation of
August 14, 2009 9:52 9.75in x 6.5in b684-ch25 FA
Fig. 25.7. Corresponding spectral density of the historical records at the Crescent City harbor
on (a) June 15, 2005, (b) November 15, 2006, (c) January 13, 2007, and (d) August 15, 2007.
the resonant wave period is reasonable. The results presented clearly indicate the
effect of local topography in amplifying the incident tsunami waves whether they
are near-field or far-field tsunamis.
Fig. 25.8. Simulated response curves with field observed dominant waves superimposed at
Crescent City harbor for events on (a) June 15, 2005, (b) November 15, 2006, (c) January 13,
2007, and (d) August 15, 2007.
Fig. 25.9. Air photos of San Pedro bay with simulation domain imposed (a) and LA/LB port (b).
directly otherwise the reflection and diffraction due to the coast and breakwaters
will weaken the wave field.
Similar to the modeling for Crescent City harbor, the simulation results were
compared with the tide gauge (LA Berth60) records. The ports of Los Angeles
and Long Beach have experienced modifications and expansion over the years. The
harbor layout is now different from that which existed several decades ago. Thus
August 14, 2009 9:52 9.75in x 6.5in b684-ch25 FA
Fig. 25.10. Response curve at LA Berth60 for different incoming wave directions.
we focused on the recorded data for recent tsunamis generated by: the earthquake
in offshore of northern California on June 15, 2005; two major earthquakes near
the Kuril Islands, Japan on November 15, 2006 and January 13, 2007; the Solomon
Islands earthquake on April 1, 2007, with a magnitude of 8.1. Both local and distant
tsunamis generated by those earthquakes reached San Pedro bay, causing water level
oscillations. The tide records at station LA Berth60 were analyzed to obtain the
periods of the dominant modes of oscillation at that location during those events.
The tide gauge records and the corresponding spectral density distributions at LA
Berth60 are shown in Figs. 25.11 and 25.12. One or two dominant waves occurred
during those events, all of them are around 60 min.
The recorded data on April 1, 2007 was not used in the modeling examination for
Crescent City harbor because the tide record was not available during that period
for Crescent City harbor.
For further comparison, the observed dominant waves are superimposed on the
response curves simulated with the corresponding incoming wave direction for each
event. This is shown in Fig. 25.13. Roughly, the incident wave came from the
northwest direction for the northern California offshore earthquake, from west for
the Kuril Islands earthquakes, and from southwest for the Solomon Islands earth-
quake. Apparently, all the observed dominant waves have all been found to corre-
spond to the first mode of the response curves, indicating that the computer model
catches the resonance characteristics of the basin.
It is interesting to note that the four tide records for Crescent City harbor shown
in Figs. 25.6 and 25.7 had the dominant wave period at 22 min which was predicted
by the computed response curves. However, the same tide records at San Pedro
bay LA/LB harbor presented in Figs. 25.11 and 25.12 showed that the dominate
August 14, 2009 9:52 9.75in x 6.5in b684-ch25 FA
Fig. 25.11. Historical records at the LA Berth60 for events on (a) June 15, 2005, (b) November 15,
2006, (c) January 13, 2007, and (d) April 1, 2007.
Fig. 25.12. Corresponding spectral density of the historical records at LA Berth60 on (a) June 15,
2005, (b) November 15, 2006, (c) January 13, 2007, and (d) April 1, 2007.
August 14, 2009 9:52 9.75in x 6.5in b684-ch25 FA
Fig. 25.13. Simulated response curves with field observed dominant waves superimposed at LA
Berth60 for events on (a) June 15, 2005, (b) November 16, 2006, (c) January 13–14, 2007, and
(d) April 1–2, 2007.
resonant modes are at 60 min. This again implies that the local topography and the
plan form strongly affect the resonant response. Thus there is a need for performing
the computer modeling whenever modifications of harbor layouts are planned.
The model for the present condition is simulated and compared with the field
data measured by HMTC during typhoon Tim in July 1994.38 The model results
and field data at station #22 and #8 are presented in Fig. 25.16. It shows the
amplification factor as a function of the incident wave period. It is seen that the
comparison is quite good with respect to resonant periods, resonant bandwidth, and
peak amplification factors. The results clearly indicate that there exists a broadband
of resonant response for wave periods between 100 and 160 s and the computer
model results appear to have captured the resonant modes correctly.
Fig. 25.15. Air photos of Hualien harbor, Taiwan with simulation domain imposed (left) and
stations #8 and #22 indicated (right).
Fig. 25.16. Simulated response curves and the observed field data at stations #22 (left) and #8
(right).
Fig. 25.17. Layout of simulation domain with mesh and incoming wave directions superimposed
(left) and the close up photo of the Pohang new harbor (right).
August 14, 2009 9:52 9.75in x 6.5in b684-ch25 FA
Due to its location and the coastline orientation, Pohang new harbor has been
found to provide fairly good protection against typhoons coming from the south.
However, seiche motion due to long period waves have occurred frequently which
produced undesirable wave and ship oscillations in the harbor, especially during the
season with waves coming from the northeast direction. Field measurement program
has shown the existence of the long period wave oscillation within the harbor.15
Figure 25.18(a) presents the simulated response curve at Station T1 in Pohang
new harbor at various wave frequencies for the existing harbor configuration. It
covers the wave periods from 100 min to 60 s. The ordinate is the amplification factor
defined as the wave height at Station T1 divided by the incident wave height. The
abscissa is the wave frequency with unit at 10 cycles per minute (the right-hand limit
corresponds to 60 s wave). Several resonant modes are clearly seen from Fig. 25.18.
Figure 25.18(b) presents the spectral density curve for Station T1 covering the same
frequency range based on the data obtained in the field measurement program.33
The first four resonant periods indicated by the four vertical lines in Fig. 25.18(a) are
the resonant periods obtained from the field data shown in Fig. 25.18(b), i.e., 4,800 s,
1,650 s, 490 s, and 260 s. It is seen that the present computer results compare very
well with the field data, especially for the first few resonant modes. It is also noted
Fig. 25.18. Results of (a) numerical simulation and (b) field data at Station T1.
August 14, 2009 9:52 9.75in x 6.5in b684-ch25 FA
Fig. 25.19. Results of (a) numerical simulation and (b) field data at Station T2.
that several resonant modes also exist for wave coming from the north direction for
wave period between 60 and 120 s.
Similar results for Station T2 are shown in Figs. 25.19(a) and 25.19(b) for com-
puter results and field measurements. Again, the computer model appears to capture
the major resonant response of the harbor region very well.
25.4.5. Summary
The good agreements between the simulated results and the field observations in
these harbor resonance studies prove that the numerical model is a very useful
tool to investigate the resonance mechanism for long waves. The differences of the
amplification factor caused by various incident wave directions are the results of
wave reflection from the coast or outside jetties, as well as wave diffraction due to
the natural obstacles or manmade breakwaters, both of which weaken the wave field
inside the harbors or bays.
Tide gauge data in three earthquake events (northern California earthquake on
June 15, 2005, Kuril Islands earthquakes on November 15, 2006 and January 13,
August 14, 2009 9:52 9.75in x 6.5in b684-ch25 FA
2007) was used in the result examinations for both San Pedro bay and Crescent City
harbor. Although the incoming waves in the same events may be similar before they
reached the west coast of the USA, the dominant waves in LA/LB harbor (about
1 h) are totally unrelated with those in Crescent City harbor (about 22 min). This
provides proofs that harbor resonance is locally induced due to the harbor layout
and the local geographical configuration.
Fig. 25.20. Air photos of Pier J with the surrounding area included.
August 14, 2009 9:52 9.75in x 6.5in b684-ch25 FA
Fig. 25.21. Response curves before and after the construction of the breakwater at location #81
(left) and #82 (right), by Lee et al.22
Fig. 25.22. Layouts of Pohang harbor with the old one on the left and new one on the right.
Fig. 25.23. Response curves before and after the modification at Station T1 (left) and T9 (right)
in Pohang harbor.
August 14, 2009 9:52 9.75in x 6.5in b684-ch25 FA
that the amplification factor is largely diminished for the wave period ranging from
100 to 200 s, a region in which severe ship motions were found to occur. Thus,
the modification of harbor layout will be helpful in reducing the seiche problem in
Pohang harbor.
Different types of ships may be oscillated by different wave periods. Generally,
the “troublesome” waves for ships have periods ranging from 100 to 300 s. The ship
motion study, such as the modeling done by Tzong et al.41 will help determine the
resonant waves for ships inside a particular harbor.
References
9. M. Ganaba, L. C. Wellford and J. J. Lee, Finite element methods for boundary layer
modeling with application to dissipative harbor resonance problem, Chapter 15 of
Finite Elements in Fluid, Vol. 5 (John Wiley & Sons, 1984), pp. 325–346.
10. S. T. Grilli, J. Skourup and I. A. Svendsen, An efficient boundary element method for
nonlinear water waves, Eng. Analysis with Boundary Elements 6(2), 97–107 (1989).
11. K. Horikawa and H. Nishimura, On the function of tsunami breakwaters, Coastal
Eng. J. 13, 103–112 (1970).
12. J. R. Houston, Interaction of tsunamis with the Hawaiian Islands calculated by a finite
element numerical model, J. Physical Oceanography 8(4), 93–102 (1978).
13. L. S. Hwang and E. O. Tuck, On the oscillations of harbors of arbitrary shape, J. Fluid
Mech. 42, 447–464 (1970).
14. Y. Ito, Head loss at tsunami-breakwater opening, Proc. 12th ASCE Conf. Coastal
Eng. (1970), pp. 2123–2131.
15. W. M. Jeong, S. B. Oh, J. W. Chae and S. I. Kim, Analysis of the wave induced
downtime at Pohang new harbor, J. Korean Soc. Coastal and Ocean Engineers 9(1),
24–34 (1997).
16. I. G. Jonsson and N. A. Carlsen, Experimental and theoretical investigations in an
oscillatory turbulent boundary layer, J. Hydraul. Res. 14, 45–60 (1976).
17. C. Lee and S. B. Yoon, Effect of higher-order bottom variation terms on the
refraction of water waves in the extended mild slope equation, Ocean Eng. 31, 865–882
(2004).
18. J. J. Lee, Wave induced oscillations in harbors of arbitrary shape, Report No. KH-R-
20, W. M. Keck Laboratory, Caltech (1969), p. 266.
19. J. J. Lee, Wave-induced oscillations in harbors of arbitrary geometry, J. Fluid Mech.
45(2), 375–394 (1971).
20. J. J. Lee, C. Chang and F. Zhuang, Interactions of transient nonlinear waves with
coastal structures, 23rd Conf. Coastal Eng. (1992).
21. J. J. Lee and C. P. Lai, Assessing impacts of tsunamis on Taiwan’s and China’s
southeast coastlines, Proc. ICCE 2006, San Diego (2006).
22. J. J. Lee, C. P. Lai and Y. Li, Application of computer modeling for harbor resonance
studies of Long Beach and Los Angeles harbor basins, Proc. 26th Int. Conf. Coastal
Eng. (1998), pp. 1196–1209.
23. J. J. Lee and F. Raichlen, Oscillations in harbors with connected basins, J. Waterways,
Ports, Coastal and Ocean Eng. Div. 98(WW3), 311–332 (1972).
24. A. Lejeune, M. Lejeune and M. Sahloul, Wave plan computation method in study
of the Calvi Bay erosion in Corsica, France, Int. J. Numer. Meth. Eng. 27, 71–85
(1989).
25. J. J. Leendertse, Aspects of a computational model for long-period wave propagation,
Memo KM-5294-PR RAND Corp., Santa Monica, CA (1967).
26. T. G. Lepelletier and F. Raichlen, Harbor oscillations induced by non-linear transient
long waves, J. Waterway, Port, Coastal and Ocean Eng. 113(4), 381–400 (1987).
27. O. T. Magoon, Structural damage by tsunamis, Coastal Engineering, Santa Barbara
Specialty Conf., ASCE (1965), pp. 35–68.
28. S. R. Massel, Ocean Surface Waves: Their Physics and Prediction (World Scientific
Publication, Singapore, 1995).
29. C. C. Mei, Mild-slope approximation for long waves generated by short waves, J. Eng.
Math. 35, 43–57 (1999).
30. C. C. Mei, M. Stiassnie and D. K. Yue, Theory and Applications of Ocean Surface
Waves (World Scientific Publishing Company, 2005).
August 14, 2009 9:52 9.75in x 6.5in b684-ch25 FA
31. S. R. Massel, Expanded refraction-diffraction equation for surface waves, Coastal Eng.
19, 97–126 (1993).
32. J. W. Miles and Y. K. Lee, Helmholtz resonance of harbors, J. Fluid Mech. 67(3),
445–464 (1975).
33. Ministry of Construction, The Report of Wave Measurement and Numerical Exper-
iment in Pohang New Harbor (1987), pp. 108–116.
34. M. Okihiro, R. T. Guza and R. J. Seymour, Excitation of seiche observed in a small
harbor, J. Geophys. Res. 98(C10), 18201–18211 (1993).
35. D. M. Powers, The Raging Sea: The Powerful Account of the Worst Tsunami in U.S.
History (Kensington Publishing Corporation, 2005).
36. F. Raichlen, Long period oscillations in basins of arbitrary shapes, Chapter 7, Coastal
Engineering, Santa Barbara Speciality Conf. (1965), pp. 115–145.
37. F. Raichlen and J. J. Lee, Oscillation of bays, harbors, and lakes, Chapter 13, Handbook
of Coastal and Ocean Engineering, ed. J. B. Herbich (Gulf Publishing Company, 1992).
38. C. H. Su and T. K. Tsay, Numerical simulation on harbor oscillations in Haw-Lien
harbor, Report NO. MOTC-IOT-IHMT-NB9001-1, Institute of Harbor and Marine
Technology Institute of Transportation, Tai-Chung, Taiwan (2002).
39. G. M. Terra, W. J. Berg and L. R. M. Maas, Experimental verification of Lorentz’ lin-
earization procedure for quadratic friction, Fluid Dynamics Res. 36, 175–188 (2005).
40. V. V. Titov and C. E. Synolakis, Numerical modeling of tidal wave runup,
J. Waterways, Port, Coastal and Ocean Eng. 124(4), 157–171 (1998).
41. T. J. Tzong, C. P. Lai, J. J. Lee and F. Zhuang, Ship motion modeling in Los Angeles
and Long Beach harbors, Proc. ICCE 2006, San Diego (2006).
42. Ü. Ünlüata and C. C. Mei, Effects of entrance loss on harbor oscillations, J. Waterways
Harbors and Coastal Eng. Div. 101(WW2), 161–179 (1975).
43. X. Yu, Finite analytic method for mild-slope wave equation, J. Eng. Mech. 122(2),
109–115 (1996).
44. J. A. Zelt and F. Raichlen, A Lagrangian model for wave-induced harbour oscillations,
J. Fluid Mech. 213, 203–225 (1990).
August 18, 2009 17:48 9.75in x 6.5in b684-ch26 FA
Chapter 26
Michael J. Briggs
Coastal and Hydraulics Laboratory
US Army Engineer Research and Development Center
3909 Halls Ferry Road, Vicksburg, MS 39180-6199, USA
michael.j.briggs@usace.army.mil
Marc Vantorre
Ghent University, IR04, Division of Maritime Technology
Technologiepark Zwijnaarde 904, B 9052 Gent, Belgium
marc.vantorre@ugent.be
Klemens Uliczka
Federal Waterways Engineering and Research Institute
Hamburg Office, Wedeler Landstrasse 157
D-22559 Hamburg, Germany
klemens.uliczka@baw.de
Pierre Debaillon
Centre d’Etudes Techniques Maritimes Et Fluviales
2 bd Gambetta, BP60039, 60321 Compiegne, France
pierre.debaillon@equipement.gouv.fr
This chapter presents a summary of ship squat and its effect on vessel underkeel
clearance. An overview of squat research and its importance in safe and efficient
design of entrance channels is presented. Representative PIANC empirical for-
mulas for predicting squat in canals and in restricted and open channels are dis-
cussed and illustrated with examples. Most of these formulas are based on hard
bottoms and single ships. Ongoing research on passing and overtaking ships in
confined channels, and offset distances and drift angles is presented. The effect
of fluid bottoms or mud is described. Numerical modeling of squat is an area of
future research and some comparisons are presented and discussed.
723
August 18, 2009 17:48 9.75in x 6.5in b684-ch26 FA
26.1. Introduction
When a ship travels through shallow water it undergoes changes in its vertical
position due to hydrodynamic forces from the flow of water and wave-induced
motions of heave, pitch, and roll. The focus of this chapter is on the former mech-
anism of ship squat. Squat is the reduction in underkeel clearance (UKC) between
a vessel at-rest and underway due to the increased flow of water past the moving
body. The forward motion of the ship pushes water ahead of it that must return
around the sides and under the keel. This water motion induces a relative velocity
between the ship and the surrounding water that causes a water-level depression in
which the ship sinks. The effect of shallow water and channel banks only exacerbates
these conditions. The velocity field produces a hydrodynamic pressure change along
the ship similar to the Bernoulli effect in that kinetic and potential energy must be
in balance.1 This phenomenon produces a downward vertical force (sinkage, pos-
itive downward) and a moment about the transverse axis (trim, positive bow up)
that can result in different values of squat at the bow and stern (Fig. 26.1). This
combination of sinkage and change in trim is called ship squat.
Most of the time squat at the bow, Sb , represents the maximum value, especially
for full-form ships, such as supertankers. In very narrow channels or canals and
for high-speed (fine-form) ships, such as passenger liners and containerships, the
maximum squat can occur at the stern Ss . The initial trim of the ship also influences
the location of the maximum squat. The ship will always experience maximum squat
in the same direction as the static trim.2 If trimmed by the bow (stern), maximum
squat will occur at the bow (stern). A ship trimmed by the bow or stern when static
will remain that way and will not level out when underway to offset the sinkage at
the bow or stern due to squat.
So why do we care about ship squat? For one thing, ship squat has always
existed, but was less of a concern with smaller vessels and with relatively deeper
channels. The new supertankers and supercontainerships have smaller static UKC
and higher service speeds. Secondly, the goal of all ports is to provide safe and effi-
cient navigation for waterborne commerce. Since operation and maintenance costs
continue to escalate and can easily exceed $3M per vertical meter, it is imperative
to minimize required channel depths and associated dredging costs. Finally, even
though we have a pretty good handle on squat predictions, accidents continue to
occur. Barrass3 noted that there have been 12 major incidents between 1987 and
2004. In 2007, this number of ship incidents had increased to as many as 82 that
are partially attributable to ship squat.4 The luxury passenger liner QEII grounded
off Massachusetts in 1992 with a repair cost of $13M and another $50M for lost
passenger bookings.
In the early 1990s, the Maritime Commission (MarCom) of the Permanent Inter-
national Association of Navigation Congresses (PIANC) formed a working group
(WG30) to provide information and recommendations on the design of approach
channels.5 In the past 10 years since the WG30 report, research in squat predic-
tions was a dynamic area in naval architecture with new experiments to study
the effects of fluid bottoms and passing and overtaking vessels, especially with the
increasing size of the shipping fleet. Time domain Reynolds Average Navier–Stokes
Equation (RANSE) numerical models are being developed to predict squat, but
these models are still being validated. In 2005, the PIANC MarCom formed a new
working group Horizontal and Vertical Dimensions of Fairways (WG49) to update
the WG30 report on design of deep draft navigation channels.6
A summary of ship squat is presented in this chapter. In the second section,
factors governing squat including ship characteristics, channel configurations, and
combined factors are discussed. Some empirical formulas from the PIANC WG30
report are presented and compared in the third section. The fourth section presents
some recent research on the effect of squat on passing and overtaking ships in con-
fined channels by the Federal Waterways Engineering and Research Institute (BAW)
in Hamburg, Germany, and the Flanders Hydraulic Research (FHR) Laboratory in
Antwerp, Belgium. It also includes numerical modeling by Delft University of Tech-
nology and laboratory modeling by FHR on the effect of ship offset and drift on
squat. The fifth section summarizes the recent studies at FHR on the effect of fluid
bottoms (i.e., mud) on squat. The development of numerical models to predict ship
squat is an ongoing research area. The current status of this development at Centre
d’Etudes Techniques Maritimes Et Fluviales (CETMEF), France, is discussed in the
sixth section. Finally, a summary and conclusions of ship squat issues is presented
in the last section.
squat varies as the square of the speed. Therefore, doubling the speed quadruples
the squat and vice versa.
There are two calculated ship parameters that are based on the basic ship dimen-
sions. The ship’s displacement volume ∇ (m3 ) is defined as
The CB can be determined from the ∇ if the other ship dimensions are known. The
underwater midship cross-sectional area AS is generally defined as
AS = 0.98BT. (26.2)
The “0.98” constant accounts for reduction in area due to the keel radius.7 Some
researchers ignore this and use a constant of “1.00” since the error is small relative
to other uncertainties in the squat calculations.
Finally, the bulbous bow and stern-transom are two other characteristics of a
ship that affect squat. Many of the early squat measurements were made before
bulbous bows were in use. Newer designs of bulbous bows, although mainly to reduce
drag and increase fuel efficiency, also have an effect on squat. The newer “stern-
transoms” on some ships are “blockier” (i.e., wider and less streamlined) than earlier
ship designs and affect squat as they become more fully submerged with increases
in draft.8
Fig. 26.2. Schematic of three channel types: unrestricted or open, restricted or confined, and
canal.
August 18, 2009 17:48 9.75in x 6.5in b684-ch26 FA
be classified as unrestricted channels if they are wide enough. The second type
of channel is the restricted channel with an underwater trench that is typical of
dredged channels. The restricted channel is a cross between the canal and unre-
stricted channel type. The trench acts as a canal by containing and influencing the
flow around the ship, and the water column above the hT allows the flow to act
as if the ship is in an unrestricted channel. The last type of channel is the canal.
These channels are representative of channels in rivers with emergent banks. The
sides are idealized as one slope when in reality they may have compound slopes with
revetment to protect against ship waves and erosion. The canal may or may not
be exposed to tidal fluctuations. For instance, the Panama and Suez Canals have a
constant water depth.
Many channels can be characterized by two or three of these channel types as the
different segments or reaches of the channel have different cross-sections. Finally,
many real-world channels look like combinations of these three types as one side may
look like an open unrestricted channel and the other side like a canal or restricted
channel with side walls. Most of the PIANC empirical formulas are based on ships
in the center of symmetrical channels, so the user has to use “engineering judgment”
when selecting the most appropriate formulas. New data are being collected for
some of these more realistic channel shapes, so future formulas may account for
these differences in channel shapes.
Other important parameters necessary to describe restricted channels and canals
are the channel width at the bottom of the channel W , trench height hT from the
bottom of the channel to the top of the trench, and inverse bank slope n (i.e.,
run/rise = 1/ tan θ). The value of n, although not necessarily an integer, typically
has a value such as 1, 2, or 3 representing side slopes of 1:1, 1:2, and 1:3, respectively.
How does one define the width of an unrestricted or an open channel since there
are no banks or sides? In 2004, Barrass had defined an effective width Weff for the
unrestricted channel as the artificial side boundary on both sides of a moving ship
where the ship will experience changes in performance and resistance that affect
squat, propeller RPMs, and speed.3 His width of influence FB is defined for h/T
values from 1.10 to 1.40 as
7.04
FB = Weff = B. (26.3)
CB0.85
Mean values of FB are of the order of 8B to 8.3B for supertankers (CB range from
0.81 to 0.87), 9B to 9.5B for general cargo ships (CB range from 0.68 to 0.80), and
10B to 11.5B for containerships (CB range from 0.57 to 0.71).
The calculated cross-sectional area AC is the wetted cross-section of the canal
or the equivalent wetted area of the restricted channel by projecting the slope to
the water surface. It is given by
AC = W h + nh2 . (26.4)
For an unrestricted channel, use Barrass’s effective width Weff for channel width W
and set n = 0 in the equation for AC .
August 18, 2009 17:48 9.75in x 6.5in b684-ch26 FA
26.3.1. Background
In 1997 the PIANC WG30 report included 11 empirical formulas and one graphical
method from nine different authors for the prediction of ship squat.5 They were
based on physical model experiments and field measurements for different ships,
channels, and loading characteristics. The formulas included the pioneering work of
Tuck11 , Tuck and Taylor,12 and Beck et al.,13 and the early research by Hooft,14
Dand,15 Eryuzlu and Hausser,16 Römisch,17 and Millward.18,19 The PIANC recom-
mends that channels be designed in two stages. The first is the “Concept” Design
where a “quick” or “ballpark” answer is desired. The WG30 report recommended the
International Commission for the Reception of Large Ships (ICORELS) formula20
in this phase. The second stage is the “Detailed” Design phase where more accurate
and thorough predictions and comparisons are required. The WG30 recommended
the formulas by ICORELS, Huuska,7 Barrass,21,22 and Eryuzlu et al.23 in this
second stage.
All of these formulas give predictions of bow squat Sb , but only the Römisch
formula gives predictions for stern squat Ss for all channel types. The Barrass
August 18, 2009 17:48 9.75in x 6.5in b684-ch26 FA
formula gives Ss for unrestricted channels, and for canals and restricted channels
depending on the value of CB . Each formula has certain constraints that it should
satisfy before being applied, usually based on the ship and channel conditions under
which it was developed. Caution should be exercised if these empirical formulas are
used for conditions outside those for which they were developed.
In 2005 the PIANC MarCom formed WG49,6 which is in the process of reviewing
and revising these formulas for an updated report on channel design (expected to be
completed in 2010). There have been some new formulations since the WG30 report
that are being evaluated. Barrass has continued to develop and refine his formulas
and now has predictions for both Sb and Ss . Ankudinov et al.24 proposed the Mar-
itime Simulation and Ship Maneuverability (MARSIM) 2000 formula for maximum
squat based on a midpoint sinkage and vessel trim in shallow water. It is one of the
most thorough and the most complicated formulas for predicting ship squat. The
St. Lawrence Seaway (SLS) Trial and Very Large Crude Carriers (VLCC) formulas
are based on the prototype measurements in the SLS by Stocks et al.25 Briggs26
developed a FORTRAN program to calculate squat using most of these formulas.
It is not possible to include all the formulas in this chapter. We have selected
a representative sample of formulas that can be used for both phases of design.
Some are the “old tried and true” formulas and some are based on new research.
The Concept Design phase is by definition the simplest, of course this does not
necessarily mean that these formulas are any less accurate than some of the more
complicated formulas. In the Detailed Design phase, it is usually a good practice to
evaluate the squat with several of the formulas and calculate some statistics such
as average and range of values. In some cases, the maximum squat values might be
used in design for the case of dangerous cargo and/or hard channel bottoms.
The user should always be mindful for the original constraints. Some of these
constraints are very restrictive (especially for the newer vessels coming on line) as
they are based on the limited set of conditions tested in physical models by the
individual researchers. This does not mean that the particular formula would not
be applicable if the constraints are exceeded by a reasonable amount. Therefore,
the user should exercise Engineering Judgment when deciding the applicability of
those predictions. Table 26.1 summarizes the applicable channel configurations and
Table 26.1. Channel configurations and parameter constraints for PIANC squat formulas.
Configuration Constraint
Code ID Code ID
Formulas U R C CB B/T h/T hT /h L/B L/h L/T
parameter constraints according to the individual testing conditions for the formulas
in this chapter.
where CS = 2.4 and the other factors have been previously defined.
The Finnish Maritime Administration (FMA) uses this formula with different
values of CS depending on the ship’s CB .29,30
1.7 CB < 0.70
CS = 2.0 0.70 ≤ CB < 0.80 . (26.8)
2.4 C ≥ 0.80
B
The BAW, however, recommends a value of CS = 2.0 for the larger containerships
of today which may have a CB < 0.70. Their research is based on many measure-
ments along the restricted channel (side slope n varies from 15 to 40), 100-km long,
River Elbe.31 The wider stern-transom ships (see Sec. 4.3) require CS = 3 because
of the increased bow squat. The FHR has found CS ≥ 2.0 for modern container-
ships. They typically travel at much higher speeds than the ICORELS formula was
originally developed, even in shallow and restricted waters. The Fnh are higher and
in this speed range the effect of blockage S on the critical ship speed is considerable.
For example, a very small S = 0.01 results in an important decrease in critical
speed.10
26.3.2.2. Barrass
The Barrass4,27 formula is one of the simplest and “user friendly” and can be
applied for all channel configurations. Based on his earlier work in 1979,21 1981,22
and 2004,3 the maximum squat SMax at the bow or stern is determined by the value
of ship’s CB and Vk as
KCB Vk2
SMax = . (26.9)
100
According to Barrass,2 the value of CB determines whether SMax is at the bow
Sb or stern SS (requires even keel when static). He notes that full-form ships with
CB > 0.7 tend to squat by the bow and fine-form ships with CB < 0.7 tend to
squat by the stern. The CB = 0.7 is an “even keel” situation with squat the same
August 18, 2009 17:48 9.75in x 6.5in b684-ch26 FA
at both bow and stern. Of course, for channel design, one is mainly interested in
the maximum squat and not necessarily whether it is at the bow or stern.
This formula is based on a regression analysis of more than 600 laboratory and
prototype measurements. Stocks et al.25 found that the Barrass formulas gave the
best results for New and Traditional Lakers in the Lake St. Francis area (unre-
stricted channel) of the SLS. The BAW feels that the Barrass restricted formula is
conservative for their restricted channel applications in the Elbe River.
The coefficient K 4 is defined in terms of blockage factor S as
26.3.2.3. Yoshimura
The Overseas Coastal Area Development Institute of Japan32 and Ohtsu et al.33
proposed the following formula for Sb as part of their new Design Standard for
Fairways in Japan. This formula was originally developed by Yoshimura28 for open
or unrestricted channels typical of Japan. The range of parameters for which this
formula is applicable is shown in Table 26.1. In 2007, Ohtsu34 proposed a small
change to the ship velocity term Vs (last factor in the equation is now Ve ) to include
S to improve its predictions in restricted channels and canals:
Vs Unrestricted
Ve = Vs . (26.12)
Restricted, canal
(1 − S)
Their Sb predictions generally fall near the average for most of the other PIANC
bow squat predictions, regardless of ship type:
3 2
1 CB 1 CB Ve
Sb = 0.7 + 1.5 + 15 . (26.13)
h/T Lpp /B h/T Lpp /B g
August 18, 2009 17:48 9.75in x 6.5in b684-ch26 FA
26.3.3.2. Huuska/Guliev
The next empirical formula in the Detailed Design phase is by Huuska.7 This Finnish
professor extended Hooft’s work for unrestricted channels to include restricted
channels and canals by adding a correction factor for channel width Ks that Guliev36
had developed. The Spanish ROM 3.1-99 (Recommendations for Designing Mar-
itime Configuration of Ports, Approach Channels, and Floatation Areas37 ) and the
FMA recommend the Huuska/Guliev formula for all three channel configurations.
In general, this formula should not be used for Fnh > 0.7. The FMA29 also includes
some additional constraints for lower and upper limits as follows (Table 26.1):
• CB 0.60 to 0.80
• B/T 2.19 to 3.50
• Lpp /B 5.50 to 8.50
• hT /h 0.22 to 0.81
August 18, 2009 17:48 9.75in x 6.5in b684-ch26 FA
2
∇ Fnh
Sb = CS Ks . (26.16)
L2pp 1 − Fnh
2
The squat constant CS = 2.40 is typically used as an average value in this formula.
The value for Ks for restricted channels and canals is determined from
7.45s1 + 0.76 s1 > 0.03
Ks = (26.17)
1.0 s1 ≤ 0.03
S
s1 = . (26.18)
K1
26.3.3.3. Römisch
Römisch17 developed formulas for both bow and stern squat from physical model
experiments for all three channel configurations. His empirical formulas are some of
August 18, 2009 17:48 9.75in x 6.5in b684-ch26 FA
the most difficult to use, but seem to give good predictions for bow Sb and stern
squat Ss given by
Sb = CV CF K∆T T
Ss = CV K∆T T (26.19)
where CV is a correction factor for ship speed, CF is a correction factor for ship
shape, and K∆T is a correction factor for squat at ship critical speed. The value for
CF is equal to 1.0 for the stern squat. The values for these coefficients are defined as
2
4
V V
CV = 8 − 0.5 + 0.0625 (26.20)
Vcr Vcr
2
10CB
CF = (26.21)
Lpp /B
K∆T = 0.155 h/T . (26.22)
The ship critical or Schijf-limiting speed Vcr is the speed that ships cannot exceed
due to the balance between the continuity equation and Bernoulli’s law.9,38,39 For
economic reasons, maximum ship speeds are typically only 80% of Vcr . The Vcr
(m/s) varies as a function of the channel configuration given by
CKU
Unrestricted
Vcr = Cm KC Canal . (26.23)
CmT KR Restricted
The mean water depth hm (m) is a standard hydraulic parameter that is used for
canals and restricted channels. It is defined as
AC
hm = (26.25)
WTop
where WTop (m) is the projected channel width at the top of the channel equal to
The relevant water depth hmT (m) is for restricted channels and is defined as
hT
hmT = h − (h − hm ).
h
August 18, 2009 17:48 9.75in x 6.5in b684-ch26 FA
1/S 1 6 10 20 30 ∞
KC 0.0 0.52 0.62 0.73 0.78 1.0
Note that the KR for the restricted channel is a function of both KU and KC .
Table 26.2 lists Römisch’s limited dataset for KC as a function of 1/S (i.e., AC /AS ).
Appendix 26.A contains more detailed descriptions of KC and some additional
equations for defining it relative to Schijf’s limiting speed and his limiting Froude
number FHL .
0.0
Bow Squat for BAW Hansa Container Ship - Unrestricted
R
Barrass
R Eryuzlu
N
R Huuska
-0.5 N R ICORELS
N R R Romisch
R
Yoshimura
N N Numerical
R BAW
Sb , m
N
-1.0 R
N
-1.5
Example R
N
Fig. 26.4. Comparison of BAW’s experimental measurements, empirical formulas, and numerical
model of bow squat for a Post-Panamax containership in an unrestricted channel (open water).
0.0
Bow Squat for FHR Tanker G, Condition C - Canal
Barrass
Huuska
R Romisch
Yoshimura
N N Numerical
-0.5 FHR
R
R N
Sb, m
R
Example
R N
R
-1.0
R
N R
Canal Bottom
-1.5 R
7 8 9 10 11 12
Vk, knots
Fig. 26.5. Comparison of FHR’s experimental measurements, empirical formulas, and numerical
model of bow squat for a Tanker “G”, in Condition C in a canal with vertical sides.
August 18, 2009 17:48 9.75in x 6.5in b684-ch26 FA
0.0
Bow Squat for Tothil Canadian Laker - Canal
R
R
N R
R
N R
R
N
-0.5
R
Sb , m
R
Barrass
Huuska
R Romisch
-1.0
Yoshimura
N Numerical N
Tothil Example R
4 5 6 7 8
Vk, knots
Fig. 26.6. Comparison of Tothil’s experimental measurements, empirical formulas, and numerical
model of bow squat for a Canadian Laker in a canal.
400 50 17 0.84 19
conservative (i.e., larger than measured). The Römisch is slightly smaller than the
measured values, but follows the trend very well. Appendix 26.B contains worked
examples for the Concept Design formulas of Yoshimura and ICORELS and the
Detail Design formulas of Eryuzlu and Römisch.
Lpp (m) B (m) T (m) CB h (m) hT (m) W (m) WTop (m) n (deg)
(i.e., larger than measured). Appendix 26.B contains worked examples for the
Yoshimura, Barrass (Concept), and Huuska (Detail). The Römisch is not included
in the worked examples for this case as it has already been demonstrated. The
Barrass is a little small, especially for higher ship speeds. The Huuska formula is
conservative for all ship speeds.
So far we have discussed the PIANC empirical formulas for predicting ship squat.
These are based on “idealized” conditions with single vessels that are sailing along
the centerline of symmetrical channels. Unfortunately, real-world channels and ship
transits are seldom this simple. This section discusses some recent research in lab-
oratory and field measurements of ship head-on passing encounters and overtaking
maneuvers in two-way traffic, stern-transom effects, abrupt sills, and offset and drift
angle effects for ships sailing off the centerline with drift angles.
When two ships pass or overtake each other, the water flow and corresponding
squat is affected as a function of the other ship’s size, speed, and direction of
travel, and the channels configuration. Dand44 was one of the first to study this
215.6 22.9 7.77 0.86 9.33 72.3 105.9 1.8 (29 deg)
August 18, 2009 17:48 9.75in x 6.5in b684-ch26 FA
phenomenon. He found increases in bow squat of 50–100% during passing and over-
taking encounters.
During the past 10 years, the BAW has conducted many field and labo-
ratory studies to investigate ship–waterway interactions, especially head-on passing
encounters and overtaking maneuvers of ships in restricted channels within German
federal waterways. Preliminary studies of the dynamic response of large container-
ships in laboratory models have shown tendencies of reduced squat.40,41,45 These
results were confirmed by additional model tests in restricted and unrestricted
channels and field measurements along the Elbe River.31 The FHR (in cooperation
with the Ghent University) has conducted laboratory experiments to study passing
and overtaking in their automated towing tank as part of a larger study to improve
their ship simulator for traffic in Flemish waterways.46 Finally, the Delft University
of Technology47 had conducted some numerical modeling of the effects of ship offset
and drift angles on ship squat. Thus, this section presents a summary of recent
laboratory, field, and numerical investigations of ship squat in real-world situations
including head-on passing encounters, overtaking maneuvers, wider stern-transoms,
and ships with offset and drift angles.
Fig. 26.7. Laboratory measurement of the effect of head-on passing on bow and stern squat for a
PM containership passing a large bulk carrier in the River Elbe. The dark blue curves represent the
single runs of the containership; the light blue curves the encounters with the large bulk carrier.
Fig. 26.8. Laboratory measurement of the effect of head-on passing on bow and stern squat for
a large bulk carrier passing a PM containership in the River Elbe. The dark red curves represent
the single runs of the large bulk carrier; the light red curves the encounters with the containership.
Table 26.7. Characteristics of the PPM containerships in the BAW field measurements.
(YM) supported four journeys on ships of the 5500 TEU Class with Tollerort Con-
tainer Terminals (TCT) acting as the intermediary.8 Table 26.7 presents selected
characteristics of the vessel types as well as the range of mean draft T and draft-
dependent CB during these journeys.
In Hamburg Harbor, the containerships were equipped with four autonomous
digital global positioning systems (DGPS) on the bow and the bridge and one data
collection system on the bridge. Vessel dynamics data were collected from Container
Terminal Altenwerder (CTA) or from TCT until just north of Scharhörn (about
120 km from Hamburg Harbor). The width of the channel in this section ranged
from 250 to 400 m. Current, temperature, and conductivity were measured by a very
fast, small ship at six cross-sections of the lower River Elbe just before the passing
encounters. Head-on and passing situations were recorded and documented.8 Vessel
movement, nautical maneuvers, local squat, trim, heel, and net maneuvering lane
were calculated using special water gage evaluations, precise DGPS measurements,
and calculations of virtual reference positions. Vessel data included propeller speed,
rudder position, etc. Maximum differences between water level interpolation and
DGPS zero measurements of <1 cm were obtained.48
August 18, 2009 17:48 9.75in x 6.5in b684-ch26 FA
Fig. 26.9. Cumulative distribution of the increase in squat for 125 head-on passing encounters of
large PPM containerships (HLCL and YM) at the channel of the lower and outer River Elbe.
0.2
0.4
SINKAGE (m)
0.6
0.8
1
sinkage aft - VT = 12 knots
sinkage fore - VT = 12 knots
1.2 sinkage aft - VT = 8 knots
sinkage fore - VT = 8 knots
1.4
Fig. 26.10. Effect of passing encounter on ship bow and stern squat as a function of ship speed
in FHR tow tank for containership and bulk carrier.49
sign when the midship sections of both ships are at the same position. During the
second part of the meeting, the sinkage aft is increased while the bow is lifted. In
the given examples, the sinkage aft of the containership increases from an initial
value of 0.6 to about 0.9 m, if the bulk carrier has a speed of 8 kt and to about 1.2 m
when both ships have a speed of 12 kt. This corresponds to an increase in squat of
50% for the 8-kt case and 100% for the 12-kt case.
Fig. 26.11. Laboratory measurements of the effect of overtaking on bow and stern squat for
General Cargo (VG3) and Feeder containership (VG4) at the western Kiel Canal. The bow and
stern squat values for the VG3 are shown in red, and the VG4 are shown in blue.
Figure 26.11 is similar to Figs. 26.7 and 26.8 for head-on passing ships. It
shows bow and stern squat for both ships as single ships and during the over-
taking maneuver. Squat during overtaking is larger than for single ships. The
increase in stern squat for the VG3 was ∆S ≈ 0.6 m and ∆S ≈ 0.8 m for the
VG4 at a speed of Vk = 8.1 kt (15 km/h). This increase in squat is caused by
the effect of the additional hydrodynamic mass and channel blockage of each ship.
Since both ships experienced a common speed-dependent long wave, they had
the same order of magnitude of total stern squat Ss = 1.0 m at Vk = 8.1 kt
(15 km/h) (light blue and light red curves at left side of Fig. 26.11). The shorter
VG3 squatted with even keel in the long wave of the larger VG4 (light red
curves).
Fig. 26.12. Effect of overtaking maneuver on bow and stern squat as a function of lateral distance
between ship centerlines for a containership and bulk carrier in the FHR tow tank.
v
vSS c.W.
d.W. [kn]
[Kn]
0 2 4 6 8 10 12 14 16 18 20
0,00
0,50
1,00
Fig. 26.13. Field measurements of bow squat for nine container vessels of HLCL (blue/light blue
squares) and YM (red/yellow triangles) in the channel of the lower and outer River Elbe.8
between the ship and the channel. The ship acts as a lifting surface as it moves
asymmetrically through the water. Drift angles are usually the result of trying to
compensate for large wind forces, especially on containerships.
Fig. 26.14. Influence of offset on squat of a containership (Lpp = 331.3 m, B = 42.8 m, and
T = 14.5 m) sailing at constant speed in a channel with h = 19.6 m. Scale 1:80 towing tank tests,
no propeller action. Open symbols: stern; closed symbols: bow.49
August 18, 2009 17:48 9.75in x 6.5in b684-ch26 FA
Fig. 26.15. Influence of drift angle on squat of a containership at constant speed in a rectangular
channel of 565-m width with h = 16 m. Ship test specifications same as Fig. 26.14.49
Figure 26.15 shows the bow and stern sinkage (squat) of a containership as a
function of speed for several drift angles (0, 5, and 10 deg). The bow sinkage increases
significantly with the drift angle, whereas the stern sinkage decreases slightly.
(1) The pressure field around the moving hull causes undulations of the water–
mud interface that themselves modify the distribution of vertical forces over the
length of the ship and, therefore, sinkage and trim.
(2) If ship’s keel penetrates into the mud layer, the hydrostatic (buoyancy) force
acting on the submerged hull increases due to the higher density of the mud.
• At very low speed (first speed range), the interface remains practically
undisturbed.
• At intermediate speed (second speed range), an interface sinkage is observed under
the ship’s bow if the fluid mud layer is relatively thick. At a certain time, an
internal hydraulic jump, perpendicular to the ship’s longitudinal axis, is observed.
The front of this internal jump moves aft with increasing speed.
• At higher speeds, the internal or interface jump occurs behind the stern (third
speed range).
It can be shown by means of a simplified theory that the critical speed separating
the second and third speed ranges is a function of the mud ρ2 to water density ρ1
ratio and the water depth h1 (Fig. 26.17).
8 ρ1
Ucrit = gh1 1 − (1 − S1 )3 (26.30)
27 ρ2
where S1 = AS /Ac1 is similar to blockage factor S except that the Ac1 is the cross-
sectional area of the channel to the top of the mud layer. This equation is based on
ideal fluid assumptions, and appears to underestimate the critical speed for mud
layers of higher viscosity.
Fig. 26.16. Mud–water interface undulations for second speed range (top) and third speed range
(bottom).51
August 18, 2009 17:48 9.75in x 6.5in b684-ch26 FA
10
8 h1 = 25 m
h1 = 20 m
Ucrit (knots)
6 h1 = 15 m
4 h1 = 10 m
0
1 1.05 1.1 1.15 1.2 1.25 1.3
ρ2/ρ1
Fig. 26.17. Critical speed separating second and third speed ranges as a function of mud–water
density ρ2 /ρ1 ratio for different water depths h1 .51
The description above is typical for motions of the mud–water interface occurring
when a ship moves with a positive UKC above a fluid mud layer of low viscosity
(black water). In case of a negative UKC (i.e., when the keel penetrates the mud
layer), a second internal wave system, comparable to the Kelvin wave system in the
water–air interface, interferes with the hydraulic jump. This may result in either
an interface rising amidships or a double-peaked rising along the hull. Figure 26.18
illustrates the effect of speed (5 and 10 kt), UKC (−12% to +10%), and mud density
(1100–1250 kg/m3 ) on the interface undulation pattern.
Due to the vertical motion of the interface and the ship, contact between the
ship’s keel and the mud layer can occur even if, initially at rest, the UKC of the
ship is positive relative to the mud–water interface. Figure 26.19 shows the initial
UKC required to avoid contact between mud and keel as a function of Depth Froude
number Fn (speed) for different mud characteristics.
bottom (m)
-5
-4.5
solid bottom
-4
above solid
-3.5
-3
above
-2.5
position
-2
position
-1.5
Interface motions
bottom (m)
-5
-4.5
solid bottom
-4
abovesolid
-3.5
-3
above
-2.5
position
-2
position
-1.5
-5
(m) (m)
-4.5
bottom
-4
solid bottom
-3.5
abovesolid
-3
above
-2.5
position
-2
position
1150
-1 UKC to interface: 10 % Density (kg/m³): 1180
-0.5
1210
Ship speed: 5 5 knots 1250
0
-1 -0.5 0 0.5 1
(m)
-5
bottom
-4.5
(m)
-4
solidbottom
-3.5
above solid
-3
above
-2.5
position
-2
position
1150
-1 UKC to interface: 10 % Density (kg/m³): 1180
-0.5
1210
Ship's speed: 10 10 knots 1250
0
-1 -0.5 0 0.5 1
Fig. 26.18. Influence of speed, UKC, and mud density on undulations of the interface.52
August 18, 2009 17:48 9.75in x 6.5in b684-ch26 FA
(a)
(b)
Fig. 26.19. Critical UKC h1,crit /T for different mud layers with stopped propeller for model
ship D, where mud layer thickness to draft ratio h2 /T (a) 0.11 and (b) 0.22 as a function of Depth
Froude Number Fn .52
relative to the condition in which the mud layer is replaced by a solid bottom. This
is because the ship can “feel” the hard bottom more than the softer, less dense,
mud layer. If the mud layer is replaced by water (normal conditions without a mud
layer), however, the sinkage would decrease relative to the condition with the mud
layer. However, this does not take into account the effect of extra buoyancy (i.e.,
mud is denser than water), but this is only important in very dense mud layers
and/or important penetration. In general, the influence on trim is more important
than sinkage since the mud layer causes the ship to be dynamically trimmed by
the stern over its complete speed range. Thus, the effect of mud layers on average
sinkage is only marginal as trim is much more important.
August 18, 2009 17:48 9.75in x 6.5in b684-ch26 FA
Fig. 26.20. Effect of mud layers on sinkage and trim (a) no interface contact, (b) contact with
mud interface, (c) no contact with interface, and (d) negative UKC. The blue line represents the
water surface, the brown line the mud layer interface, and black, the solid bottom.53
The effect of the decrease of UKC is shown in Fig. 26.22. In a range of small
positive to negative UKC, the trim is mostly affected in a moderate speed range
(second speed range, as defined above). A large negative UKC (keel into the bottom
mud–water interface) causes trim by the stern in the complete speed range. The
effect of mud on the average sinkage is less important, but the combination of trim
and sinkage results in an increase of the sinkage aft in some conditions.
Figures 26.21 and 26.22 are valid for slender ships (CB < 0.7) that tend to trim
by the stern above a solid bottom. Full-formed ships, on the other hand, usually
trim by the bow. In muddy navigation areas, such vessels will experience a reduced
trim by the bow — or even trim by the stern — when they have sufficient UKC in
the second speed range. In the third speed range, this effect will be reduced again.
Figure 26.23 shows this effect of midships sinkage and trim as a function of UKC
for a full-form trailing suction hopper dredge.
Many different numerical methods can be used to calculate the ship squat. Their
only common point is that they calculate the velocity components and the pressure
of the flow surrounding the ship. Depending on whether the fluid is modeled as
viscous, a potential velocity function can be used or a more sophisticated flow model
has to be applied. Some models are based on slender body theory, whereas others
use the boundary elements method (BEM) or the finite element method (FEM).
August 18, 2009 17:48 9.75in x 6.5in b684-ch26 FA
(a) 0.6
0.5
sinkage FP (m)
0.4
0.3
0.2
0.1
0
0 2 4 6 8 10 12
ship speed (knots)
(b) 0.6
0.5
sinkage AP (m)
0.4
0.3
0.2
0.1
0
0 2 4 6 8 10 12
ship speed (knots)
(c) 0.6
sinkage midships (m)
0.5
0.4
0.3
0.2
0.1
0
0 2 4 6 8 10 12
ship speed (knots)
(d) 0.2
0
trim (mm/m)
-0.2
-0.4
-0.6
-0.8
-1
0 2 4 6 8 10 12
V (knots)
solid 15% mud F mud G mud H
mud E mud C mud D solid 26%
Fig. 26.21. Sinkage (a) fore, (b) aft, (c) and midships, and (d) trim as a function of ship speed
for Containership D (LOA = 300 m, B = 40.3 m, h = 13.5 m) sailing above a mud layer of 1.5 m
thickness with 15% clearance referenced to mud-water interface (26% to solid bottom). Note the
legends are the same for all plots.52
August 18, 2009 17:48 9.75in x 6.5in b684-ch26 FA
Fig. 26.22. Sinkage (a) fore, (b) aft, (c) and midships, and (d) trim as a function of ship speed
for Containership D (LOA = 300 m, B = 40.3 m, h = 13.5 m) sailing above a mud layer of 3.0 m
thickness, ρB = 1,180 kg/m3 , ρD = 1,100 kg/m3 .52
4
SHIP SPEED (knots)
0 1 2 3 4 5 6
1
SOLID BOTTOM - UKC 10%
0.8 MUDDY BOTTOM - UKC 10%
MUDDY BOTTOM - UKC -4%
0.6
0.4
TRIM (mm/m)
0.2
-0.2
-0.4
Fig. 26.23. (a) Mean midships sinkage and (b) trim as a function of UKC for a full-form trailing
suction hopper dredger (115.6 × 23.0 × 8.0 m3 , scale 1/40) above a simulated mud layer (ρ2 /ρ1 =
1.22, h2 /T = 0.175). Positive trim is equivalent to increased stern squat.54
body method works by vertical cross-sections of the flow, so it is also called the
one-dimensional (1D) theory of squat.
Gourlay58 extended the slender body theory of Tuck with the unsteady slender
body theory. This improvement allows one to consider a ship moving in a non-
uniform depth since the coordinate system is now earth-fixed, whereas it is ship-
fixed for classic numerical methods. The 1D system still uses vertical cross-sections
and decomposition into an inner and outer expansion. The pressure integration is
only made on the ship length based on the ship section B(x) at each x along the
hull. Resolution of the 1D equation is made with the finite difference method. Com-
parison with experimental results for soft squat situations (h/T > 4) showed good
agreement with numerical results. No tests were made for hard squat conditions
(i.e., shallow depths) where flow around the ship is affected.
According to Bernoulli’s principle, the pressure will decrease which will make the
ship sink more. Numerical models have to take into account this “over squat” to
precisely calculate ship squat in all channel configurations. So when a first squat
result has been found, the model has to check that this squat is not disturbing
the hydrodynamics in such a way that squat could increase more. This checking is
important to ensure a reliable result from the numerical model. As these commercial
numerical models do not perform squat checking, they may not be very efficient in
restricted water. The user has to be very careful and take the result with reservations
since the numerical model could in these conditions underestimate ship squat.
no
Displacements < ε
yes
V=V+ ∆ V
Fig. 26.24. Coupling for the ship squat numerical modeling system.
Wb − FPZ
∆Sb = α (26.32)
ρgSf
where Sf is the floating surface of the ship, and α (fixed at 0.9) is a relaxation
coefficient to limit the variation of squat in a calculation cycle for the mesh updating
model.
Pitching θ1 and rolling θ2 angles are estimated in the second model from the x
and y components of the moment equations:
Wb (xG − xP )
tan θ1 = α (26.33)
ρg x∈Sf y∈Sf x(x − x0 )dx dy
Wb (yG − yP )
tan θ2 = α . (26.34)
ρg x∈Sf y∈Sf y(y − y0 )dx dy
The equilibrium (second) model has to calculate the position of the center of vertical
thrust P (xP , yP ) by integration of the pressure over the hull, and O(x0 , y0 ) the
center of the floating surface Sf . All nodes of the ship are then vertically translated
by ∆z and rotated by θ1 around the y-axis and by θ2 around the x-axis.
The third model moves all nodes of the free surface per the results of Bernoulli’s
relation, and the hull node displacements in x and y directions. Then the inner
nodes of the mesh are moved proportionally according to boundary modifications
and the distance from those boundaries.
Since 2006 the University of Compiègne has been working to improve Debaillon’s
squat system. As in Gourlay’s model, the coordinate system will be earth-fixed to
allow ship passing or crossing and bridge pile crossings. The system will also be able
to take into account a nonuniform water depth along the channel.
This chapter has focused on some of the latest advances in predicting ship squat
and its effect on underkeel clearance for channel design. Several of the more popular
PIANC empirical formulas were presented for Concept and Detailed Design phases.
In general, the simpler and more “user friendly” formulas were recommended for
the Concept Design phase, but this does not preclude them being used in the Detail
Design phase and vice versa. Ultimately, the designer wants the maximum squat
value possible (bow or stern) in the Concept Design phase and a more realistic value
in the Detail Design phase. All empirical formulas have certain constraints based
on the field and laboratory data used in their development. It is up to the user to
exercise Engineering Judgment when applying these formulas as they give a range
of squat values.
The PIANC formulas were developed for “idealized” channel and sailing con-
ditions for single ships. Recent research has been conducted to investigate more
“real-world” conditions for the latest generation of larger ships. The BAW and FHR
have conducted extensive laboratory and field investigations of head-on passing
and overtaking maneuvers, where squat is a function of ship speed and lateral sep-
aration distance. Their results indicate that squat can increase 50–100% relative
to a single ship. The BAW measured maximum additional squat during passing
encounters of 0.6 m in the laboratory and 0.4 m in the field. Similarly, the FHR
measured maximum additional squat of 0.3–0.6 m in the laboratory. For overtaking
maneuvers in the laboratory, the BAW recorded additional squat of 0.6–0.8 m and
August 18, 2009 17:48 9.75in x 6.5in b684-ch26 FA
the FHR measured 0.2–0.3 m. These additional squat values represent the effect of
these more realistic ship and channel interactions for a range of conditions.
The BAW noted that ships with wider “transom” sterns experience more bow
squat than most ships with more streamlined transoms due to the additional
buoyancy of the stern. They measured additional bow squat of 0.2–0.5 m as a
function of ship speed.
The Delft and FHR have conducted research on ships sailing with offsets and
drift angle from the channel centerline. Delft found that both offsets and drift angles
increase squat, in a quadratic manner. The FHR found that offsets can increase
squat by 20% and drift angles produce significant increase in bow squat and slight
decrease in stern squat.
The average sinkage of a vessel navigating in muddy channels is generally
reduced by the presence of mud layers. The dynamic trim is affected significantly
by the generation of interface undulations. For ships navigating above mud layers,
the maximum sinkage is comparable to or slightly less than the values occurring if
the mud layer were replaced by a solid bottom. Compared to the situation in which
the mud layer is not present (i.e., replaced by water), the muddy bottom interface
always increases the maximum sinkage, even in case of contact with the mud layer.
This means that the mud layer will increase sinkage even if the ship “plows” through
it. The maximum sinkage (bow or stern) always increases when the lower part of
the water column is replaced by fluid mud.
The designer and harbor pilots should be aware that all of these special influences
can increase squat. Sometimes, it is not economically feasible to design a channel
for all of these eventualities, but it is always possible to slow down as the conditions
warrant.
Numerical modeling of ship squat is just beginning to be developed. Histori-
cally, ship modeling has been concentrated in the areas of wave resistance models.
Squat modeling requires time domain models that are very computer intensive. The
increasing cost of dredging and the larger ships coming on line have motivated many
institutions around the world (such as USACE, BAW, FHR, CETMEF, and FMA)
to begin a more active development of ship squat models. The improvements in
computer speed and storage have made these types of models much more promising.
The CETMEF numerical model matched measured laboratory values very well in
the three examples presented in this chapter.
Numerical models will continue to be improved, but field measurements and
laboratory models will still be necessary to investigate the highly nonlinear dynamic
behavior of the newer and larger ships. They will respond differently than existing
ships and continued study of passing and overtaking, bank and bottom effects, and
sailing alignment will insure optimum and safe navigation design of channels and
waterways.
Acknowledgments
PIANC WG49 for authorizing publication of this chapter. Particular thanks goes to
Bryan Barrass, the WG49 Chair Mark McBride, original WG30 and WG49 member
Werner Dietze, and WG49 vertical subcommittee members Martin Boll, Hans Moes,
Terry O’Brien, and Kohei Ohtsu who assisted with the ship squat and UKC research.
Other members of the WG49 making contributions to this effort included Larry Cao,
Don Cockrill, Rink Groenveld, Jarmo Hartikainen, Jose Iribarren, Susumu Naruse,
Sahil Patel, Carlos Sanchidrian, Esa Sirkiä, and Jos Van Doorn.
References
39. V. Balanin et al., Peculiarities of navigation on canals and restricted channels, orig-
inating hydraulic phenomena associated with them and their effect on the canal
bed; measurements preventing slope deterioration, Paper S.I-3, 24th Int. Nav. Cong.,
Leningrad, Russia (1977).
40. G. Flügge and K. Uliczka, Dynamisches Fahrverhalten und Wechselwirkungen mit der
Fahrrinnensohle von sehr großen Containerschiffen unter extremen Flachwasserbedin-
gungen, Proceedings HTG-Kongress 2001, Hamburg (2001) (in German).
41. G. Flügge and K. Uliczka, Dynamisches Fahrverhalten mit der Fahrinnensohle
von sehr grossen Containershiffen unter extremen Flachwasserbedingungen. Hansa
(2001).
42. M. Vantorre, E. Verzhbitskaya and E. Laforce, Model Test Based Formulations of
Ship-Ship Interaction Forces, Ship Technology Research/Schiffstechnik, Band 49, Heft,
3 August 2002.
43. J. T. Tothil, Ships in Restricted Channels, A Correlation of Model Tests, Field Mea-
surements and Theory, National Research Council of Canada Mechanical Engineering
Report MB264, January 1966.
44. I. W. Dand, Some Measurements in Interaction between Ship Models Passing on
Parallel Courses, NMI R108, August 1981.
45. K. Uliczka and G. Flügge, Squat-Untersuchungen für sehr große Post-Panamax-
Containerschiffe, HTG/STG-Sprechtag FA Seeschifffahrststraßen, Hafen und Schiff,
Hamburg (2001) (in German).
46. M. Vantorre, E. Laforce, G. Dumon and W. Wackenier, Development of a probabilistic
admittance policy for the flemish harbours, 30th PIANC Cong., Sydney, September
2002.
47. H. J. de Koning Gans and H. Boonstra, Squat effects of very large container ships with
drift in a harbor environment, MTEC2007 Conference, Singapore, 26–28 September
2007.
48. Ch. Maushake and S. Joswig, Messung von Squat, Trimm und Krängung sehr
großer Containerschiffe im Rahmen von Grundsatzuntersuchungen auf der Elbe,
Hydrographische Nachrichten Nr. 072, Deutsche Hydrographische Gesellschaft (2004)
(in German).
49. M. Vantorre and G. Dumon, Model test based requirements for the under keel
clearance in the access channels to the flemish harbours, 2nd Squat Workshop Aspects
of Under Keel Clearance in Analysis and Application, Elsfleth, March 2004.
50. National Ports Council, Ship Behavior in Ports and their Approaches — Part 2:
Additional Sinkage Caused by Sailing in the Proximity of Channel Bank, Research
Transport Headquarters, London, U.K. (1980).
51. M. Vantorre, Nautical Bottom Approach – Application to the Access to the Harbour of
Zeebrugge, HANSA — Schiffahrt — Schiffbau — Hafen, 138. Jahrgang, Nr. 6 (2001),
pp. 93–97.
52. G. Delefortrie, Maneuvering behavior of container vessels in muddy navigation areas,
Ph.D. thesis, Ghent University (2007).
53. K. Van Craenenbroeck, M. Vantorre and P. De Wolf, Navigation in Muddy
Areas: Establishing the Navigable Depth in the Port of Zeebrugge, Proceedings
CEDA/PIANC Conference 1991 (incorporating CEDA Dredging Days): Accessible
Harbours, Paper No. E4, Amsterdam (1991).
54. M. Vantorre, Systematische proevenreeksen met het zelfaangedreven schaalmodel van
een sleephopperzuiger boven een mengsel petroleum¬trichloorethaan als slibsimulatie-
materiaal experimentele waarnemingen en theoretische interpretaties. Rijksuniver-
siteit Gent/Waterbouwkundig Laboratorium Borgerhout. Gent/Antwerpen (1990).
August 18, 2009 17:48 9.75in x 6.5in b684-ch26 FA
55. I. W. Dand and A. M. Ferguson, The squat of hull ships in shallow water, Roy. Inst.
Nav. Arch. 115, 237–247 (1973).
56. R. F. Beck, Hydrodynamic forces caused by ship in confined waters, Proc. Am. Soc.
Civ. Eng. 107, ASCE, EM3, June 1981, pp. 523–546.
57. B. V. Korvin-Kroukovsky, Investigations of ship motions in regular waves, Trans.
SNAME 63, 386–435 (1955).
58. T. Gourlay, Mathematical and computational techniques for predicting the squat of
ships. Thesis of the University of Adelaide, Australia (2000).
59. H. Bühring, Prediction of Squat by Fast Boundary Elements, 2nd Squat Workshop,
Oldenburg, Germany (2004).
60. P. Debaillon, Système de modélisation de l’enfoncement dynamique des bateaux,
Thesis of University of Compiègne, France (2005).
26.A. Appendix
K1 = a0 + a1 S + a2 S 2 + a3 S 3 . (26.A1)
Table 26.A1 lists the correlation coefficient R2 and the polynomial coefficients for
each of the hT /h curves. This allows one to program the value for K1 without having
to manually read a plot.
Polynomial coefficients
hT /h R2 a0 a1 a2 a3
Note that the right-hand side is equivalent to KC for canals and restricted channels,
the same as Eqs. (26.28) and (26.A2).
Finally, Briggs26 determined the formula for KC from a least square fit of
Römisch’s limited set of discrete data points in Table 26.2 with an R2 = 0.97. It
gives the same results as the other formulas for KC :
Several worked examples are contained in this appendix. They are the same
examples described in Sec. 26.3. The input ship and channel characteristics were
described in the text and listed in Tables 26.3–26.5.
Since Fnh ≤ 0.70, it is acceptable for all methods. The full form CB = 0.84 is
acceptable for all methods, but slightly exceeds Huuska’s upper limit of CB < 0.8.
The ratio of B/T = 50/17 = 2.94, slightly exceeds Eryuzlu and Römisch’s criterion.
Next, h/T = 19/17 = 1.12 is acceptable, but minimal UKC. The ratio Lpp /B =
400/50 = 8.0 is larger than Eryuzlu’s upper limit of 6.8. Finally, Lpp /T = 400/17 =
23.53 is slightly larger than the upper limits of Huuska and Römisch. Bottom line:
probably acceptable to use the different formulas for this case, but remember that
this is a much larger vessel than the criteria for which most of these formulas were
developed.
August 18, 2009 17:48 9.75in x 6.5in b684-ch26 FA
The second step is to substitute this value into the equation for bow squat Sb
from Table 26.3 and Sec. 26.B.1.1.
This value is too large compared to the measured value of Sb = 0.70 m. Of course,
some institutions use a smaller value of the constant CS . Thus, this value could be
reduced by using a smaller CS , but it would need to be of the order of 1.4 which is
much smaller than commonly recommended.
Thus, this Concept Design application is on the high side, but at least it is a
conservative estimate relative to the measured Sb = 0.70 m.
This value is also larger than the measured value of Sb = 0.70 m, but it is conser-
vative and similar to Yoshimura’s value.
The third step is the calculation of the correction factor for ship shape CF
2
2
10CB 10(0.84)
CF = = = 1.10. (26.B10)
Lpp /B 8.0
The fourth step is the calculation of the correction factor for squat at the critical
speed K∆T
√
K∆T = 0.155 h/T = 0.155 1.12 = 0.16. (26.B11)
The last step is to substitute these values into the equation for bow squat Sb
Sb = CV CF K∆T T = 0.22(1.10)(0.16)(17) = 0.67 m. (26.B12)
This value, although slightly small, is in excellent agreement with the BAW mea-
sured value of Sb = 0.70 m.
Since Fnh ≤ 0.70, it is acceptable for all methods. The full form CB = 0.85 is
acceptable for all the methods, but is greater than Huuska’s upper limit of CB < 0.8.
The ratio of B/T = 33/13 = 2.54 is good for all methods. Next, h/T = 14.5/13 =
1.12 is acceptable. The ratio Lpp /B = 180/33 = 5.45 is acceptable, although a little
low for most methods. Finally, Lpp /T = 180/13 = 13.85 is on the low side, but
acceptable. Bottom line: probably acceptable to use the different formulas for this
case, but remember that some of the original constraints are exceeded.
The second step is to calculate the channel cross-sectional area AC . The zero
value of slope n is to account for the vertical sides of the flume.
AS 420.42
S= = = 0.083. (26.B16)
AC 5075
This relatively small value of S indicates that the channel should be considered as
an unrestricted channel for Barrass application.
The fourth step is to calculate the correction coefficient K given by
The value of K = 0.87 is replaced by K = 1.00 since this is the minimum value
that Barrass intended for relatively wide channels.
The last step is to substitute the values above into the equation for Sb
already been calculated for the Barrass application above. Therefore, Ve is given by
Vs 5.14
Ve = = = 5.61 m/s. (26.B19)
(1 − S) (1 − 0.083)
The second step is to substitute values from above, Table 26.4, and Sec. 26.B.2.1
into the Yoshimura equation for Sb
3
1 0.85 1 0.85 (5.61) 2
Sb = 0.7 + 1.5 + 15 = 1.19 m. (26.B20)
1.12 5.45 1.12 5.45 9.81
S 0.083
s1 = = = 0.083. (26.B21)
K1 1.0
The third step is to calculate the correction factor for channel width Ks , which
depends on the value of s1 . The first equation for Ks is used since s1 > 0.03.
The last step is to substitute these values above into Huuska’s equation for Sb
This value is a little large compared to the measured value. The values of both
Yoshimura and Römisch are Sb = 1.19 m, which are excellent matches to the mea-
sured value of Sb = 1.18 m.
Vs 3.59
Fnh = √ = = 0.38. (26.B25)
gh 9.81(9.33)
Since Fnh ≤ 0.70, it is acceptable for all methods. The full form CB = 0.86 is
acceptable for all the methods, although it is slightly larger than the upper CB limit
of both Barrass and Huuska. The ratio of B/T = 22.9/7.77 = 2.95, slightly exceeds
Eryuzlu’s and Römisch’s upper limits. Next, h/T = 9.33/7.77 = 1.2 is acceptable.
The ratio Lpp /B = 215.6/22.9 = 9.41 is larger than most upper limits. Finally,
Lpp /T = 215.6/7.77 = 27.75 is slightly larger than all upper limits. Bottom line:
probably acceptable to use the different formulas for this case, but remember that
some of the original constraints are exceeded.
The second step is to calculate the channel cross-sectional area AC . The slope
n (i.e., run/rise = 1.8 = (105.9 − 72.3)/(2 ∗ 9.33)) is equivalent to an angle of
θ = 29 deg (i.e., θ = arctan(1/n)).
AS 174.37
S= = = 0.21. (26.B28)
AC 831.25
The last step is to substitute the values above into the equation for Sb
This value is a little low compared to the measured value of Sb = 0.93 m, but is a
good first estimate.
AC 831.25
hm = = = 7.85 m (26.B31)
WTop 105.9
Cm = ghm = 9.81(7.85) = 8.78 m/s (26.B32)
1.5
Arc sin(1 − S)
Kc = 2 sin
3
1.5
Arc sin(1 − 0.21)
= 2 sin = 0.46 (26.B33)
3
Vcr = Cm Kc = 8.78(0.46) = 4.06 m/s. (26.B34)
The second step is to calculate the correction factor for ship speed CV
2
4
3.59 3.59
CV = 8 − 0.5 + 0.0625 = 0.53. (26.B35)
4.06 4.06
The third step is the calculation of the correction factor for ship shape CF
2
2
10CB 10(0.86)
CF = = = 0.83. (26.B36)
Lpp /B 9.41
August 18, 2009 17:48 9.75in x 6.5in b684-ch26 FA
The fourth step is the calculation of the correction factor for squat at the critical
speed K∆T
√
K∆T = 0.155 h/T = 0.155 1.2 = 0.17. (26.B37)
The last step is to substitute these values into the equation for bow squat Sb
Sb = CV CF K∆T T = 0.53(0.83)(0.17)(7.77) = 0.58 m. (26.B38)
This value is 0.35 m too small compared to the measured value of Sb = 0.93 m.
The Römisch trend is pretty good, however.
August 17, 2009 12:3 9.75in x 6.5in b684-ch27 FA
Chapter 27
Mamta Jain
Halcrow Inc., 4010 Boy Scout Blvd.
Tampa, FL 33607, USA
Ashish J. Mehta
Department of Civil and Coastal Engineering
University of Florida, Gainesville, FL 32611, USA
27.1. Introduction
775
August 17, 2009 12:3 9.75in x 6.5in b684-ch27 FA
Fig. 27.1. (a) Newnans Lake in north-central Florida with location of shallow and deep sites.
(b) Typical surface current pattern in the lake (due to two tributary inflows and one outflow)
simulated by the three-dimensional numerical model EFDC (Environmental Fluid Dynamics Code)
of Ref. 11. Current magnitudes are generally a few millimeters per second (smaller than the
arrowheads) (adapted from Ref. 12).
interest are those that are shallow enough to allow waves to “touch” the bottom
and thereby affect its physical state.
Sedimentary matter in many lakes tends to be rich in organic matter. A sig-
nificant example of accumulation of such “black mud” is the large but shallow
Lake Okeechobee in south-central Florida. This lake has an area of 1,750 km2 and
a maximum depth of about 4.5 m below mean sea level. Figure 27.2 shows the
thickness contours of the mud deposit. The volume of mud is estimated to be about
193 × 106 m3 .5,6 Such muds can be the primary carriers of sorbed chemicals that
play a critical role in governing water quality. In these lakes, suspended fine sed-
iment dynamics is influenced by pycnoclines arising from temperature and asso-
ciated hypolimnatic water layer, dissolved substances, or sediment itself. We will
be concerned with sediment-induced pycnoclines, since they are usually the most
important density gradients that influence turbidity in such environments.
When surface waves propagate over a muddy bottom (Fig. 27.3), their height
tends to decrease with distance quite noticeably if the bottom is compliant. Thus,
at a given site two issues must be dealt with, namely prediction of the local wave
amplitude and period based on the wind field, and simulation of sediment resus-
pension and its significance to the turbidity level. We will consider wave damping
first, followed by resuspension.
August 17, 2009 12:3 9.75in x 6.5in b684-ch27 FA
Fig. 27.2. Lake Okeechobee in south-central Florida and contours of black mud thickness in cm
(adapted from Ref. 5).
To simulate the (design) maximum effect of waves due to wind from a given
direction, duration is assumed to be long enough to allow the generated wave train to
reach the downwind shoreline. Also, given the spatial variability in wave amplitude
and frequency, it is convenient to examine the effect of waves at a downwind site
where the waves are high. Prediction of (nonbreaking) waves at a given wind speed,
water depth, and linear distance along the fetch is done in its simplest form via
semi-analytic treatments that rely mainly on natural observations. For a spatial rep-
resentation of wave field, use is presently made of “third-generation” models such
as SWAN (Simulating WAves Nearshore). This model is based on the wave energy
balance equation in the absence of current.7 SWAN has the advantage of enabling
the inclusion of the actual bathymetry of the lake. In Fig. 27.4, an application of
August 17, 2009 12:3 9.75in x 6.5in b684-ch27 FA
Fig. 27.4. Predicted wave heights in Lake Monroe in central Florida using SWAN (adapted
from Ref. 8).
SWAN is illustrated for Lake Monroe in central Florida subject to an easterly wind
of 10 m s−1 . The lake area is 32 km2 and the maximum depth is 3.1 m.8
The semi-analytic approach of Young and Verhagen9 is based on extensive obser-
vations on waves in a lake. However, this and similar treatments for wave prediction
are applicable to noncomplaint, or static, beds only, i.e., there is no allowance for
energy dissipation due to mud. On the other hand, in SWAN the inclusion of an
appropriate dissipation function for mud is relatively straightforward (e.g., Ref. 10).
Given either approach, it is a matter of interest to assess the damping effect of mud
on wave amplitude relative to dissipation over a static bottom. For that purpose,
we will review simple physical descriptions of bed dynamics for the estimation of
wave damping.
The wave energy balance equation (as used in SWAN, for example) is based
on the consideration that the wave timescale, on the order of seconds, is much
shorter than the wind duration over the desired fetch, which is much longer than
the wavelength. That duration can be typically on the order of minutes to tens of
minutes in the fetch of initial wave growth. The steady-state energy conservation
equation is
∂(ECg cos θ) ∂(ECg sin θ)
+ = −εD (27.1)
∂y ∂x
in which x and y are coordinates normal and parallel to the upwind shoreline,
respectively, θ is the angle between the direction of wave approach relative to the
x-coordinate, E is the wave energy per unit water surface area, Cg is the wave
group velocity, and εD is the wave-mean energy dissipation per unit bottom area
per unit time. Wave propagation in the two directions enables the simulation of
wave refraction and damping of the wave crest (e.g., Ref. 7).
Inasmuch as the angle between the wave crest and the shoreline is reduced by
refraction in shallow water, the effect of the y-gradient in Eq. (27.1) tends to be
small compared to the x-gradient. Accordingly, for present purposes this equation
may be simplified by considering waves normally incident to a straight shoreline
August 17, 2009 12:3 9.75in x 6.5in b684-ch27 FA
ax = a0 e−ki x (27.4)
where ki is the wave damping coefficient. By combining Eqs. (27.3) and (27.4),
ki is obtained from
εD
ki = . (27.5)
2Cg E
The magnitude of ki , which in turn determines εD , depends on the rheology
(stress–strain behavior) of the substrate.
Loss of energy in water due to frictional resistance at the static-bed surface arises
within the (laminar or the turbulent) boundary layer. For nonbreaking waves, this is
essentially the only mechanism that operates at the bed surface, and its magnitude
depends on how rough the surface is. All other mechanisms are effective within
the bed, and therefore require descriptions of dissipation that depend on the state
(continuum or two-phased particle–water mixture, solid, or fluid) of the bottom.
Several basic rheological models are available to predict ki (Table 27.1).
August 17, 2009 12:3 9.75in x 6.5in b684-ch27 FA
Property or Constitutive
Bed type medium behavior Constitutive models
Fig. 27.5. Rheological domains based on combined effects of solids volume fraction and inter-
particle interaction (adapted from Ref. 17).
In Fig. 27.5 the domains of different (nonturbulent flow) behaviors are defined
entirely in the qualitative sense in terms of the combined effect of the solids volume
fraction ϕ (or porosity np = 1 − ϕ) and interparticle interaction I. Newtonian
behavior is confined to the lower left-hand corner characterized by low values of ϕ
and very low (almost negligible) values of I. The zone of particular interest (dashed
rectangle) is in the range of moderately low to medium ϕ and low to medium I.
This is part of a larger zone within which the stress-versus-strain rate response of
mud shows thixotropy, i.e., viscosity depends on the strain rate and on ϕ. In other
words, the stress-history of mud plays a role in governing its response to wave action.
As a result, for instance, pre- and post-storm response of marine mud of the same
density and composition to the same wave action may be quite different.
August 17, 2009 12:3 9.75in x 6.5in b684-ch27 FA
τ = τy + K γ̇ n (27.6)
where τy is the yield stress, γ̇ is the flow shear (strain) rate, K and n are the
constants for a given sediment, and n = 1. For fluid muds n typically ranges from 0.3
to 0.9. This variation incorporates effects of mud density, composition, and stress-
history.14 The Casson15 equation is more nonlinear than Eq. (27.6)
√ √
τ = τy + η∞ γ̇ (27.7)
where η∞ is the viscosity at (theoretically) infinite shear rate. Dade and Nowell16
found that their data on dense clay suspensions seemed to better agree with
Eq. (27.7) than Eq. (27.6). This equation can be written as
where
ηm = K γ̇ n−1 + η∞ (27.9)
τ = τB + ηm γ̇ (27.10)
in which the (Bingham) yield stress τy = τB is equal to the threshold stress at and
below which the material is a solid, and above which behaves as a Newtonian fluid.
Thus the Bingham model is a special case of the Herschel–Bulkley equation.
Since the rheological behavior of mud is critical to prediction of wave damping,
the relevant definition of mud must be based on a rheological classification. Accord-
ingly, mud may be defined as, “A sediment–water mixture which consists of particles
that are predominantly less than 63 µm in size, exhibits poroelastic or viscoelastic
rheological behavior when the mixture is particle-supported, and is highly viscous
and non-Newtonian when it is in a fluid-like state”.19
August 17, 2009 12:3 9.75in x 6.5in b684-ch27 FA
This definition allows for the inclusion of some fine sand (>63 µm) as part of
mud, which is often the case. On the other hand, there is no reference to material
composition, i.e., the definition relies solely on the rheological behavior. Thus any
sediment that meets the particle size criterion, including organic matter, is defined
as mud provided it conforms to the stated rheological criteria.
Percolation occurs below the bed surface, as pore water flows (percolates)
through the interstitial space in response to the wave-induced dynamic pressure
gradient. Energy loss is expressed in terms of the well-known Darcy’s law of viscous
flow of pore water (e.g., Ref. 20). This law in the vertical (z) direction is
Ksp dpt
ww = − (27.11)
ηw dz
where ww is the pore water velocity, pt is the total pressure, i.e., the sum of
dynamic and hydrostatic pressures, and Ksp is the specific permeability. The bed
matrix is static and, also, the bed is two-phased. The poroelastic response of
such a bed involves energy loss due to interparticle friction, which is additional
to Darcy viscous (percolation) loss. When the bed heaves, the resulting Coulomb
friction due to interparticle abrasion becomes the dominant cause of energy loss
(compared to percolation). The poroelastic stress (τ )–strain (γ) relationship is
given by
in which G is the shear modulus of elasticity of the bed material and δc is a charac-
teristic energy loss parameter. For compliant continua, a simple (linear) viscoelastic
medium subject to oscillatory forcing is represented by two equivalent represen-
tations, one between shear stress τ and shear strain γ (for a solid), and another
between τ and time-rate of shear strain γ̇ (for a fluid)
in which
√ the equivalent shear modulus G∗ (a complex number characterized by
i = −1) comprises of the storage modulus G and the loss modulus G , and the
equivalent dynamic viscosity η ∗ consists of viscous coefficient η , and a component
η incorporating the material’s elastic response.
The two basic viscoelastic models include the Kelvin–Voigt (K–V) and
the Maxwell elements. The K–V element behaves as a solid when sheared, since the
deformed material regains its initial state after the applied stress is relaxed. The
components of equivalent shear modulus and equivalent viscosity, respectively, are
G = G; G = ησ (27.15)
G
η = ηm ; η = . (27.16)
σ
August 17, 2009 12:3 9.75in x 6.5in b684-ch27 FA
τ τ
Kelvin–Voigt Maxwell
ηm
G ηm
G
τ τ
τ
Jeffreys ηa
G ηb
Fig. 27.6. Mechanical analogs of simple viscoelastic constitutive elements for mud.
The Maxwell element, which represents a fluid because it does not resist defor-
mation under an applied stress, is defined as
ηm σ G
2 2
ηm σG2
G = ; G = (27.17)
(ηm σ + G2 )
2 2 2 σ 2 + G2 )
(ηm
ηm G2 η 2 σG
η = 2 2 ; η = 2 m2 . (27.18)
(ηm σ + G2 ) (ηm σ + G2 )
Mechanical analogs of K–V and Maxwell elements are shown in Fig. 27.6.
Higher-order analogs are combinations of these two types. For example, the Jeffreys
element combines, in series, a Maxwell element (with viscous coefficient ηa ) and a
K–V element (viscous coefficient ηb ).
During wave action, the rheological state of soft mud almost always changes. This
change depends on the relative magnitudes of the applied shear stress τ , the shear
strength with respect to erosion (or critical shear stress) τs , and the plastic yield
stress τy (or the Bingham yield stress τB ). As schematized in Fig. 27.7, fluid mud
is formed in different ways depending on the stresses and accompanying processes.
Case 1: τ ≤ τs < τy . Wave action is too weak to erode the bed surface. However,
wave orbits can penetrate the bed, and mainly by pressure work buildup excess
pore pressure. Interparticle contacts become impermanent and the effective normal
stress, representing the solid-supported part of total normal stress in bed, eventually
vanishes. Mud rigidity decreases even as the particle packing density (i.e., wet bulk
density) may remain largely unaffected. However, the bottom is now in a fluid-
like transitory state, which reverts to bed almost as soon as wave action ceases
and interparticle bonds are reestablished. For transitory fluid mud to change to a
August 17, 2009 12:3 9.75in x 6.5in b684-ch27 FA
Fig. 27.7. Flow chart illustrating the modes of formation of fluid mud, turbidity generation, and
reversion of fluid mud to bed.
200
Shear stress (Pa) 150
100
50
0
-50
-100
-150
-200
-1.2 -0.9 -0.6 -0.3 0 0.3 0.6 0.9 1.2
Shear rate (Hz)
Fig. 27.8. Applied stress versus flow shear rate relationship showing hysteresis.
on the geometric idealization shown in Fig. 27.9, the relevant model equations for
stress τ versus strain rate γ̇ are as given in the following paragraph.
Backbone curve (representing the basic flow curve in the absence of hysteresis):
n
|γ̇|
τ = τp · sign(γ̇); |γ̇| ≤ γ̇c
γ̇c (27.19)
τ = τy · sign(γ̇) + η1 γ̇; |γ̇| > γ̇c .
When γ̇ ≤ γ̇c :
τ = τm0 · sign(γ̇) + η1 γ̇; γ̇γ̈ ≤ 0
(27.21)
τ = −τy · sign(γ̇) + ηm2 γ̇; γ̇γ̈ > 0
n
γ̇mc τm0 + τm1
τm1 = τp ; τm0 = τm1 − η1 γ̇mc ; ηm2 = (27.22)
γ̇c γmc
where γ̇c is a threshold value of γ̇, and η1 and η2 are viscosities when the magnitude
of γ̇ is increasing and decreasing, respectively. Calibration of these sets of equations
requires five mud-specific parameters: i.e., τp , τy , τm0 , γ̇c , and γ̇mc .
Due to their simplicity, basic rheological models merely approximate the actual
stress–strain (or strain rate) behavior of muds. Nevertheless, with proper calibration
they have been found useful in engineering applications requiring knowledge of
wave attenuation. A few illustrative equations for calculating the wave-damping
coefficient are given in Table 27.2.
Consider a 6-s wave of 1-m amplitude propagating in seawater (density
1,030 kg m−3 ) of 4.5 m depth, giving a wave group velocity of 5.14 m s−1 . The bed
consists of mud having a density of 1,200 kg m−3 . Bed surface resistance is charac-
terized by the wave friction factor fw = 0.035, mud dynamic viscosity is 2.5 Pa.s,
and the rigidity modulus is 500 Pa. Assume specific permeability of 1.27 × 10−13 m2 ,
specific energy loss parameter equal to 0.5 and bed porosity 0.5. Calculated wave
damping coefficients ki (given below) using equations in Table 27.3 indicate that
for the selected parameters the viscoelastic K–V model is the only important con-
tributor to wave damping, while bed friction and poroelastic contributions are the
least important. Percolation is limited by low permeability. Poroelastic contribution
is low under given wave conditions, but can be high when the bed is silty.
For soft muds in general, wave damping due to viscoelastic and poroelastic (silty)
bed responses deserves greater consideration than damping due to other mecha-
nisms. From dimensional analysis it can be shown that for the viscoelastic fluid
2
σ d δm ∆ρ
k̃i = ki h = f n , , , (27.23)
gkr h d ρw
where d is mud layer thickness, h is water depth, δm = 2νm /σ is the Stokes
boundary layer thickness in mud, vm is the kinematic viscosity of mud and ∆ρ =
ρm − ρw is the density difference between mud and water.26 The dimensionless wave
damping coefficient k̃i = ki h depends on σ 2 /gkr (i.e., the surface wave dispersion
relationship), the relative mud thickness d/h, the relative thickness of the Stokes
boundary layer δm /d in mud, and the buoyancy of the lower layer represented by
∆ρ/ρw . For a poroelastic bed
2
n σ d Vm
k̃i = f , , ,Pe (27.24)
gkr h Cs
August 17, 2009 12:3 9.75in x 6.5in b684-ch27 FA
Table 27.2. Wave damping coefficients over static and compliant beds.
ρw gηm
Viscoelastic ki = “ ” MacPherson25
2 + G2
4σ ηm
Kelvin– σ2
Voigt bed
kr = wave number (= 2π/L, where L is the wavelength), σ = wave angular frequency, h = water
depth, vw = fluid (water) viscosity, ρw = upper fluid (water) density, ρm = mud density, ηm =
mud dynamic viscosity, Cg = wave group velocity (simply calculated from the Airy linear wave
theory, e.g., Ref. 7), Ksp = specific permeability, a = wave amplitude, g = acceleration due to
gravity, G = shear modulus of elasticity (of mud),pδc = Coulomb specific energy loss, k0 = wave
number for the static (rigid) bed case, σ0 (= k0 2G/ρm ) = poroelastic resonance frequency,
ρs = particle density, and np = bed porosity.
Table 27.3. Wave damping coefficient for various physical system or process.
where Vm /Cs (= M aG ) is the shear Mach number in mud, in which Vm = √G/ρm
is the shear wave velocity, G is the shear modulus of elasticity and Cs = gh is
the shallow water wave celerity. The quantity Pe is a characteristic Péclet number
representing the ratio of rate of advection of flow to its rate of diffusion within the
August 17, 2009 12:3 9.75in x 6.5in b684-ch27 FA
bed. For a given wave condition, a high value of Pe implies low rate of pore water
diffusion, which is conducive to undrained bed condition.27,28 These bed states
are relevant because under drained condition, excess pore pressure generated by
wave action is dissipated via pore water flow, while under undrained condition pore
pressure does not dissipate. This distinction in bed behavior is related to particle
size. Large size correlates with high permeability and easy drainage, while small
size is associated with low flow permeability, i.e., poor drainage. Thus the modality
of wave energy dissipation is dependent on Pe. This number, Pe = σdp /Kp,
is proportional to a characteristic velocity scale (which is defined in terms of
angular wave frequency σ and particle diameter dp ), and inversely proportional to
the isotropic coefficient of consolidation, which varies with flow permeability Kp .
The relation between the flow permeability and specific permeability is given by
Kp = Ksp gρw /ηw . This definition of Péclet number will be apparent in subsequent
reference to mechanisms for wave energy dissipation.
The quantity M aG varies with the solids volume fraction ϕ = 1 − np , where the
porosity np = (ρs − ρm )/(ρs − ρw ) and ρs is the particle density, nominally equal
to 2,650 kg m−3 for typical inorganic sediments. When ϕ ≤ ϕsp , where ϕsp is the
space-filling value of ϕ below which the particles do not form a continuous matrix
(i.e., mud is a fluid, and therefore cannot transmit a shear wave) Vm and therefore
M aG are nil. In the range of ϕ > ϕsp , Vm rapidly increases with ϕ. When waves
begin to liquefy mud, Vm decreases even when ϕ (> ϕsp ) remains constant. In a
flume experiment on a clayey mud bed at ϕ = 0.10, the onset of steady wave action
caused Vm to decrease from 2 to 1.4 m s−1 in one hour.29
The mean range of ρm in which fluid mud occurs is 1,050 to 1,200 kg m−3 , i.e.,
ϕ in the range of 0.03 to 0.12 (taking ρw = 1,000 kg m−3 ). The upper limit rep-
resents the mean value of ϕsp , which has been found to vary between 0.07 and
0.18, depending on mud composition.27 The presence of organic matter can dras-
tically reduce ϕsp . For mud from Lake Okeechobee containing 40% organic matter
by weight, ϕsp was reported to be 0.04.30 Such differences in ϕsp indicate that mud
can be a fluid or a solid depending on its composition, and also its stress history.
with the frequency of wave forcing. These effects on Pe are exemplified by the
order-of-magnitude representations in Table 27.4. In the laboratory or in small lakes
(σ = 6 rad s−1 ) Pe values are generally high and increase ten-fold (from 0.6 to 6) as
the substrate changes from coarse sand to clay. At the other extreme case of open
sea or the ocean (σ = 0.5 rad s−1 ) the Pe values are comparatively low, even as
they increase ten-fold when the substrate changes from coarse to fine. An important
inference to be derived from these effects is that the magnitude of Pe serves to
define the bed as a drained, two-phased medium at one end to an undrained con-
tinuum at the other. An ensuing outcome would be that the viscoelastic model is
appropriate for high values of Pe and poroelastic model for low values.
In Fig. 27.10, the compactness of the bed of given sediment diameter has been
defined by ϕ (or np ). The abscissa is specified by the Péclet number, with values cor-
responding to Table 27.4. Since density ρ has been used in the analysis, its scale is also
included based on particle density of 2,650 kg m−3 and water density of 1,000 kg m−3 .
The boundaries are approximate not only due to the paucity of available information,
Table 27.4. Péclet number for different beds and wave frequencies.
Péclet number, Pe
but also because porosity and permeability depend on particle composition, shape
and texture, in addition to size. This dependence leads to overlaps in the zones of
applicability of the various models. Another factor contributing to uncertainty in
the domain boundaries is due to the characteristically thixotropic response of mud
to stress. As mud microstructure formed during rising stress differs from that under
falling stress, energy dissipation varies with stress history.
Viscoelastic solid and viscoplastic models occupy the lower left-hand region,
because they are important when the bed is dense and cohesive. If modeling is done
within the fluid-flow framework, the viscoelastic model can yield acceptable results
only for short-term response of mud to wave forcing. Similarly, viscoplastics are
modeled correctly only within the framework of the lower layer as a solid at low
stresses.31,32 When the bed is static (noncompliant) and the porosity is low, energy
loss occurs due to friction at the surface. The domain boundary shown for this
mechanism can vary substantially depending on surface roughness. Porous static
beds experience loss due to percolation, whose contribution to overall dissipation
is increasingly subsumed by Coulomb friction as the material becomes finer and
causes the bottom to heave.
Although the choice of the rheological model varies quite significantly with ϕ
and Pe, some broad albeit approximate trends are also evident in Fig. 27.10. Most
importantly, it appears that the poroelastic response is bracketed between Pe =
∼0.4 and ∼2. At higher values of Pe one of the viscoplastic models (including purely
viscous) applies depending on the value of ϕ. At lower values of Pe bed friction and
percolation are the two important dissipation mechanisms.
the turbulent movement of water parcels. In other words, particle motion is not
“iso-kinetic”. This discrepancy, which increases with increasing frequency of turbu-
lence, can be usually ignored for particles less than about 100 µm. In Eq. (27.25) it
is assumed that σT , which depends on particle size and whose value is determined
from empirical evidence, is the same for molecular and turbulent diffusion processes
in the z -direction. Molecular diffusivity is obtained from
κB Tw
εm = (27.26)
6πηw dp
in which κB is the Boltzmann constant (1.38026 × 10−16 erg deg−1 molecule−1 ),
and Tw is the (absolute) temperature of the fluid. Since εm εs , we will ignore
molecular diffusion in the present treatment. In Eq. (27.25) the product ws C is the
settling flux and Ds ∂C/∂z is the diffusive flux, where Ds = εs /σT denotes turbulent
sediment mass diffusivity. Thus the time-rate of change of concentration ∂C(z , t)/∂t
is governed by the net effect of the settling and diffusion. The behavior of C(z , t)
is entirely dependent on the surface and bottom boundary conditions required to
solve Eq. (27.25). At the free surface the usual boundary condition is that there is
no loss or gain of sediment, so the settling and upward diffusive fluxes are equal, i.e.,
∂C
ws C + Ds = 0 @z = h. (27.27)
∂z
At the bottom the condition to be satisfied is
∂C
ws C + Ds = S @z = 0 (27.28)
∂z
where S = Fe − Fs characterizes resuspension (when S > 0), Fe is the erosion flux,
and Fs is the deposition flux. All three quantities have units of mass (of sediment)
per unit volume of suspension and unit time. Dividing each by the bed area gives
the respective fluxes Sf , Fef , and Fsf . Three situations can arise with respect to Sf
(or S): (1) Sf = 0, i.e., there is no change in the concentration profile C(z ), which
represents an equilibrium condition, (2) Sf > 0, i.e., erosion dominates, and C(z , t)
changes in such a way that the depth-mean concentration (or turbidity) increases,
and (3) Sf < 0, i.e., deposition dominates and turbidity decreases. The solution of
Eq. (27.25) (together with Eqs. (27.27) and (27.28)) depends on the choice of the
functions for Fef , Fsf (= ws C at z = 0) and Ds .
where τb is the bed shear stress, τs is the bed shear strength with respect to
erosion and Feo is the erosion flux constant, equal to value of Fef when τb = 2τs .
Typical range of τs values is 0.1 to 0.3 Pa. The constant Feo can be as high as
1 × 10−1 kg m−2 s−1 in the shallow littoral zone. In general, τs and Feo are highly
sensitive to sediment composition and bed density (e.g., Ref. 36). For example, if
mud were significantly palletized and highly cohesive, τs could be higher than the
quoted values. For fluid mud (ρm ≤ ρmf kg m−3 ):
2
Rc
αe ρm uwb − Rg − ws Ca ; Rg < Rc
Fef = Rg (27.30)
0; Rg ≥ Rc
where uwb is the near-bottom velocity amplitude, αe is a sediment-dependent coef-
ficient, Ca is a characteristic near-bottom suspended sediment concentration, and
Rg is a global Richardson number. The quantity Rc is the critical value of Rg below
which no entrainment fluid mud occurs. From experimental data on a natural and
a clayey sediment, Li37 estimated αe = 2 × 10−6 and Rc = 0.043. The Richardson
number is defined as
2
∆ρgδw
Rg = (27.31)
ρw u2wb
where δw = 2νw /σ is the thickness of the Stokes boundary layer in water. We will
return to the method for calculation of uwb and the significance of Ca .
Wolanski et al.40 Cleveland Bay (Australia); field test 0.200 1.40 2.25 2.45
Cleveland Bay; laboratory tests 0.070 1.30 2.50 2.80
Jiang41 Jiaojiang (China); neap tide 0.045 6.00 1.51 1.50
Jiaojiang; spring tide 0.230 10.00 1.80 1.50
Marván42 Ortega River, Florida 0.160 4.50 1.95 1.70
Ganju43 Loxahatchee River, Florida 0.190 5.80 1.80 1.80
symbol wo and termed free-settling velocity) can be taken as constant. For higher
concentrations, Wolanski et al.39 proposed the empiric equation
C nf
ws|γ̇=0 = af (27.34)
(C 2 + b2f )mf
where af is a settling velocity scaling coefficient, nf is a floc-settling exponent, bf is
a hindered settling coefficient, and mf is a hindered settling exponent. Selective
values of these coefficients are given in Table 27.5. When the concentration exceeds
∼ 75 kg m−3 settling is practically arrested and the deposit begins to gel and
consolidate. A representative flow shear rate is
uwb uwb
γ̇ = = . (27.35)
δw 2νw /σ
By assuming that the wave boundary layer δw h, the velocity amplitude uw (z )
can be estimated from the Airy theory as
cosh kr z
uw = aCw kr (27.36)
sinh kr h
in which the wave celerity Cw is given by
g
Cw = tanh kr h. (27.37)
kr
At the bottom, the velocity amplitude uwb is obtained from Eq. (27.36) by setting
z = 0 to yield
aCw kr aσ
uwb = = . (27.38)
sinh kr h sinh kr h
In general,
z
Φs = f n 1 + (27.40)
Lso
in which Lso is known as the Monin–Obukov length scale specified from the steady-
state form of the turbulent kinetic energy balance for the flow system. This length
scale is accordingly defined as
ρu3∗
Lso = (27.41)
κgρ w
where ρ is the fluid density, u∗ is the friction velocity, κ is the Karman constant
(= 0.4 in clear water turbulent flow), ρ w is the component of sediment mass flux
due to turbulent diffusion, ρ is the turbulent fluctuation in density ρ, and w is
the turbulent fluctuation in the fluid vertical velocity. Based on Eq. (27.41), Lso is
dependent on the ratio of kinetic energy associated with flow shear to the potential
energy of the buoyant suspension. When inertia dominates, Lso is large and z/Lso
in Eq. (27.40) approaches zero with the result that Φs = 1 and Ds = Dso . The
higher is the stratification the smaller is the value of Lso , and the greater the effect
of buoyancy on diffusion. Thus when the flow consists of water stratified by fluid
mud, Ds can be significantly different from Dso .
Expressions relating Φs to parameters that can be easily obtained from measure-
ments of velocity and density profiles are commonly used. Munk and Anderson44
proposed a well-known equation of the type
1
Φs = (27.42)
1 + αs Rig
where αs is a sediment-dependent coefficient, whose value typically ranges between
0.5 and 2.30,45 The gradient Richardson number Rig is obtained from
g(∂ρ/∂z)
Rig = − . (27.43)
ρ(∂uw /∂z)2
The related flux Richardson number Rif is defined as
gρ w
Rif = − (27.44)
ρu w duw
dz
which indicates a close relationship between Rif and Lso . In the strict sense, in
Eq. (27.42) Rig must be replaced by Rif . However, Rig is often used because it is easier
to deduce from measurements than Rif . The relationship between Rif and Rig is
Rig = σT Rif (27.45)
where σT is the Prandtl–Schmidt number. Thus in Eq. (27.42) the coefficient αs
includes the effect of σT , which is therefore empirically accounted for in that rela-
tionship. The neutral mass diffusivity Dso is related to the neutral momentum
diffusivity εmo by
εmo
Dso = (27.46)
σT
which makes it convenient to use expressions for εmo derived for the wave boundary
layer, provided σT is known or is estimated. Costa46 reported σT in the range of
August 17, 2009 12:3 9.75in x 6.5in b684-ch27 FA
We will further consider Dso to be independent of height in the water column. Thus,
the solution of Eq. (27.53) is
C wso
= exp − (z − za ) . (27.54)
Ca Dso
August 17, 2009 12:3 9.75in x 6.5in b684-ch27 FA
absence of BNL and BSL, DSL is sustained by erosion of particles from CB. Settling
particles deposit onto the bed, unless the near-bed fluid stresses are high enough
to hinder or prevent deposition. These transport processes are sometimes called
classical erosion and deposition associated with CB.48 When BNL occurs but BSL is
(practically) absent, sediment entrainment occurs due to mixing between fluid mud
and the dilute suspension. The settling parcels, if and when they reach BNL, will
coalesce into BNL. Similarly, coalescence causes downward exchange of sediment
between DSL and BSL. Exchange processes can also occur among BSL, BNL, and
CB that change their thickness and concentration without participation by DSL.
Possible interactions between waves and sediment are indefinitely large, and have
not been explored sufficiently either in the open sea or interior environments.
However, useful pointers can be gleaned from limited studies in some of Florida’s
shallow lakes, which are typically laden with black mud or muck. This material has
a high percentage of amorphous plant debris. Its origins are terrigenous sources in
the watershed, the lacustrine littoral zone, and the lake itself when it is large. A
characteristic feature of the material is that it tends to be refractory, but at the same
time highly reactive with respect to nutrients including nitrogen and phosphorus.5
Organic matter reduces the density of inorganic clayey material in such a way that
the solids volume fraction ϕ of soft beds varies nonlinearly with the fraction of
organic matter. From mass balance,
ρm − ρ D
ϕ=1− (27.55)
ρw
where ρD is the dry density of the bed material (and equivalent to concentration C
when sediment is suspension). Soft bed sediments were sampled at sites shown in
Fig. 27.12. Sampling was carried out using grab-samplers and push-corers, with the
cores varying in thickness between ∼ 0.1 and ∼ 1.5 m. The following approximate
relationships for densities representing mean trends were obtained49
Fig. 27.12. Some lakes and other sites in Florida laden with black mud or muck (adapted from
Ref. 49).
Using characteristic parameters for Newnans Lake (Table 27.6), calculated wave
damping coefficients are given in Table 27.7. Assumed causes of dissipation include
laminar boundary layer, turbulent boundary layer, viscoelastic bed, and poroelastic
bed (Table 27.2). For comparison purposes, percent reductions in wave energy ∆E =
100(a20 − a2 )/a20 are also given. The wave amplitude (a) is conveniently taken as the
value at the (2.1-m deep) site of the measurement platform (Fig. 27.1(a)), and a0 is a
reference wave amplitude at an upwind distance of 5.5 km north of the platform. The
wave-generating wind speed was taken to be 8 m s−1 . Coulomb friction, viscous mud,
and viscoelastic (Jeffreys) models provide competing mechanisms for significant
energy loss.
It should be noted that the poroelastic bed formulation and the Kelvin–Voigt
viscoelastic formulation in Table 27.2 are for infinite mud depth so they are believed
to over estimate energy loss. Wave damping coefficients for viscous model and vis-
coelastic model are obtained from the solution of a semi-analytical model which is
applicable from very thin to thick muds and is presented in Ref. 33.
Low specific permeability (1010 m2 ) suggests that for application in Newnans,
Lake, the poroelastic model may be no less qualitative than the viscoelastic repre-
sentation. At the same time, given low bed elasticity (G = 300 Pa) we may infer that
the viscoelastic bed offers no noteworthy advantage over a viscous bed. In fact the
top ∼ 20 cm active layer of muck was practically a fluid (mean ρm = 1,040 kg m−3 ).
Accordingly, the viscous mud model appears to be is reasonable, which also renders
August 17, 2009 12:3 9.75in x 6.5in b684-ch27 FA
T a h d G ηw ηm ρw ρm Ksp
(s) (m) (m) (m) (Pa) (Pa.s) (Pa.s) (kg m−3 ) (kg m−3 ) (m2 ) δc
1.2 0.05 1.3 0.3 300 1 × 10−3 2.4 1,000 1,200 1 × 10−10 0.5
δw /h δm /d M aG η /ηm
analysis of wave damping easier because viscosity based models are simpler to use
than poroelastic or viscoelastic ones.
In the qualitative sense one can arrive at a similar conclusion as above regarding
the bed by examining the magnitudes of characteristic dimensionless parameters
(Table 27.8). Observe that δw /h = 0.0005 δm /d = 0.09 implies that dissipation
is mainly in the mud, with practically little contribution from water. The Mach
number M aG is about 0.14, which means that wave speed in water would be about
seven times that in mud, a manifestation of the significant role of mud as an energy
dissipater. The very low value (0.002) of η /ηm , the ratio of loss due to mud elasticity
to mud viscosity, corroborates the inference that the muck was more like a fluid
than a solid.
Wave-induced acceleration of bottom mud, total pressure, and pore pressure were
measured at a “deep” site in the lake (Fig. 27.1(a)) 0.2 m below the mud–water
interface.12 The vertical amplitude of acceleration is plotted against the wind speed
in Fig. 27.13. There is an evident trend of increasing amplitude with wind speed.
Acceleration is a manifestation of heave, and as seen, 3.2 m s−1 may be taken as
the threshold speed for the onset of heave. The effective normal stress σv in mud is
August 17, 2009 12:3 9.75in x 6.5in b684-ch27 FA
0.010
Vertical acceleration (m s-2) Data
0.008
0.006
0.004
Critical wind speed for
muck heave = 3.2 m s-1
0.002
0
1 2 3 4 5 6
Wind speed (m s-1)
Fig. 27.13. Vertical acceleration in mud versus wind speed in Newnans Lake.
2.0
Data
Pore water pressure (kPa)
1.5
1.0
0.5
0
Liquefied muck
-0.5 Critical wind speed for
liquefaction 7.1 m s-1
-1.0
-1.5
4 5 6 7 8 9 10
Wind speed (m s-1)
Fig. 27.14. Effective stress in mud versus wind speed in Newnans Lake.
given by
σv = σv − ∆u (27.58)
where σv is the total normal stress (pressure) and ∆u is the pore pressure. Repre-
sentative values of σv are plotted as a function of wind speed in Fig. 27.14. With
increasing wind the resulting bottom shear stress can destabilize the bed particle
matrix and result in a reduction in σv . When the matrix is completely broken σv
vanishes and the bed is liquefied. Liquefaction is seen to occur at wind speeds higher
than about 7 m s−1 . Since the actual wind speed in the lake is often close to (or in
excess of) this value, the top layer of mud is typically found in the liquefied state.
As expected, the mud begins to heave at a lower wind speed (3.2 m s−1 ). As long as
heaving is gentle, the structural integrity of the bed is retained. At the same time,
however, energy dissipation can cause significant wave damping.
The difference between the wave amplitude required for soft mud to heave, and
higher amplitude necessary for mud entrainment can be used as the basis to design
submerged, shore-parallel mud berms for protection of the shoreline against storm
August 17, 2009 12:3 9.75in x 6.5in b684-ch27 FA
waves. Offshore of Mobile Bay in Alabama, the crest elevation of such a berm above
the seabed was so chosen as to enable incoming waves passing over the berm to
heave the mud without entraining it. In this way a stable, wave-energy absorbing
berm was constructed. It was 2.75 km long, 300 m wide, and 6 m high. The ambient
mean water depth was 6.5 m. Wave measurements indicated that during a storm,
incoming wave of 0.75 m amplitude experienced a 46% reduction in amplitude across
the berm, which remained practically uneroded.50,51
27.14. Resuspension
In Newnans Lake the wave period is only on the order of 1 s. Referring to Fig. 27.11,
the consolidating bed participates in the resuspension process only along the shallow
(<∼ 1 m depth) littoral periphery of the lake. In the central portion (>∼ 1 m) the
sediment remains in suspension without participation by the bed.
Parameter Value
0.25
Data
Initial
0.20 90 min
0 min
30 min
60 min
Elevation (m)
0.15 90 min
0.10
0.05
0
0 0.25 0.50 0.75 1.00 1.25 1.50
Concentration (kg m-3)
Fig. 27.15. Concentration profile evolution at the shallow site in Newnans Lake.
they were not made sufficiently close to the bed, is the development of a high-
concentration layer within the first 1–2 cm above bottom. Concentration in this
layer exceeded 1 kg m−3 . Persistent of such a layer can have significant effect on
benthic biota.
Fig. 27.16. Estimation of effective BSL height at the deep site in Newnans Lake.
where ks is the bed roughness and awb is the amplitude of wave-induced water
parcel excursion close to the bed and related to the velocity amplitude uwb by
uwb
awb = . (27.60)
σ
For estimating He using Eq. (27.59) we will choose the following representative
values: a = 0.035 m, T = 1.4 s, and h = 1.6 m (height of water above the assumed
BSL–DSL interface), ks = 0.1 m, ρs = 1,700 kg m−3 , ρw = 1,000 kg m−3 , ws =
1.15 × 10−4 m s−1 , and ϕv = 70/(1,700 × 1,000) = 0.041 × 10−3 . This yields He =
0.68 m, which is within the measured range (0.2 to 1 m).
analyses are fruitful subjects for further research. Efforts are presently being made
to simulate bed erosion as “continuous-phase” modeling, in which the modeling
domain extends to depths at which the mud is entirely stationary. In this physics-
based approach, technical demands remain significant because modeling changes in
state from a bed to a suspension (and vice versa) requires significant treatment
of inter-particle forces, and damping of turbulence by suspended matter. Never-
theless, owing to its simplicity, the treatment presented here is likely to continue
for acquiring estimates of wave damping and resuspension in typical engineering
investigations.
References
40. E. Wolanski, R. Gibbs, P. Ridd, B. King, K. Y. Hwang and A. Mehta, Fate of dredge
spoil, Cleveland Bay, Townsville, Proc. 10th Austalasian Conf. Coast. Ocean Eng.,
Auckland, New Zealand (1991), pp. 63–66.
41. J. Jiang, An examination of estuarine lutocline dynamics, PhD thesis, University of
Florida, Gainesville, FL (1999).
42. F. G. Marván, A two-dimensional numerical transport model for organic-rich cohesive
sediments in estuarine waters, PhD thesis, Heriot-Watt University, Edinburgh, UK
(2001).
43. N. K. Ganju, Trapping organic-rich fine sediment in an estuary, MS thesis, University
of Florida, Gainesville (2001).
44. W. H. Munk and E. A. Anderson, Notes on a theory of the thermocline, J. Marine
Res. 7, 276–295 (1948).
45. M. A. Ross, Vertical structure of estuarine fine sediment suspensions, PhD thesis,
University of Florida, Gainesville (1988).
46. R. C. F. G. Costa, Flow-fine sediment hysteresis in sediment-stratified coastal waters,
MS thesis, University of Florida, Gainesville (1989).
47. Y. Li and T. M. Parchure, Mudbanks of the southwest coast of India VI: Suspended
sediment profiles, J. Coast. Res. 14(4), 1363–1372 (1998).
48. J. C. Winterwerp, A. W. Bruens, N. Gratiot, C. Kranenburg, M. Mory and
E. A. Toorman, Dynamics of concentrated benthic suspension layers, Fine Sediment
Dynamics in the Marine Environment, eds. J. C. Winterwerp and C. Kranenburg
(Elsevier, Amsterdam, 2002), pp. 41–53.
49. J. E. Gowland, A. J. Mehta, J. D. Stuck, C. V. John and T. M. Parchure, Organic-
rich fine sediments in Florida part II: Resuspension in a lake, Estuarine and Coastal
Fine Sediment Dynamics, eds. J. P.-Y. Maa, L. P. Sanford and D. H. Schoellhamer
(Elsevier, Amsterdam, 2007), pp. 167–188.
50. A. J. Mehta and F. Jiang, Some observations on water wave attenuation over nearshore
underwater mudbanks and mud berms. Report UFL/COEL/MP-93/01, Coastal and
Oceanographic Engineering Program, University of Florida, Gainesville (1993).
51. T. N. McLellan, M. K. Pope and C. E. Burke, Benefits of nearshore placement, Proc.
Third Ann. Natl. Beach Preser. Tech. Conf., Florida Shore and Beach Preservation
Association, Tallahassee (1990).
52. S. B. Vinzon and A. J. Mehta, Mechanism for formation of lutoclines by waves,
J. Waterway Port Coast. Ocean Eng. 124(3), 147–149 (1998).
July 27, 2009 9:32 9.75in x 6.5in b684-ch28 FA
Chapter 28
Nobuhisa Kobayashi
Center for Applied Coastal Research
University of Delaware, Newark, DE 19716, USA
nk@coastal.udel.edu
Andres Payo
Graduate School of Science and Technology
University of Kumamoto, 2-39-1, Kurokami
Kumamoto, 860-8555, Japan
Bradley D. Johnson
US Army Engineering Research and Development Center
3909 Halls Ferry Road, Vicksburg, MS 39180-6199, USA
Simple formulas for the cross-shore and longshore transport rates of suspended
sand and bedload on beaches are proposed by synthesizing available data and
formulas. A combined wave and current model based on the time-averaged con-
tinuity, momentum, and energy equations for water is improved and used to
provide hydrodynamic input to the proposed sand transport model. The model is
compared with spilling and plunging wave tests conducted in a large wave basin
using fine sand. The numeric model predicts the measured longshore suspended
sand and total transport rates within a factor of about 2. The longshore bedload
transport rate is predicted to be small. The predicted cross-shore sand transport
rates are relatively small on the quasi-equilibrium beaches in these tests. The
computed beach profile change under 10-h wave action is less than about 5 cm.
The proposed sediment transport model will need to be verified using additional
data but no bedload data is available in surf zones and reliable suspended load
data is scarce.
807
July 27, 2009 9:32 9.75in x 6.5in b684-ch28 FA
28.1. Introduction
The combined wave and current model in the following is an extension of the
time-averaged model developed by Kobayashi et al.1 on the basis of the Dutch
July 27, 2009 9:32 9.75in x 6.5in b684-ch28 FA
Fig. 28.1. Incident irregular waves and wind on beach of alongshore uniformity.
models by Reniers and Battjes2 and Ruessink et al.3 Figure 28.1 shows obliquely
incident irregular waves on a straight shoreline where the cross-shore coordinate
x is positive onshore, and the longshore coordinate y is positive in the downwave
direction. The depth-averaged cross-shore and longshore velocities are denoted
by U and V , respectively. Incident waves are assumed to be unidirectional with
θ is the incident angle relative to the shore normal and uniform in the long-
shore direction. The wave angle θ is assumed to be in the range of |θ| < 90◦ to
ensure that the incident waves propagate landward. The wind speed and direction
at the elevation of 10 m above the sea surface are denoted by W10 and θw ,
respectively.
The governing equations for water under the assumption of alongshore uni-
formity are averaged over a number of wind waves to obtain the time-averaged equa-
tions. The beach is assumed to be impermeable where Kobayashi et al.4 included
a permeable layer for the case of normally incident waves. The depth-integrated
continuity equation of water requires that the cross-shore volume flux is zero in
the absence of wave overtopping where Kobayashi et al.5 analyzed irregular wave
overtopping of permeable slopes. The requirement of zero cross-shore volume flux
is expressed as
gση2
h̄ Ū + cos θ + qr cos θ = 0 (28.1)
Cp
where h̄ is the mean water depth given by h̄ = (η̄ − zb ), with η̄ the mean free surface
elevation and zb the bottom elevation; Ū is the mean cross-shore velocity, which
is negative and offshore because cos θ > 0; g is the gravitational acceleration; ση ,
the standard deviation of the free surface elevation η; Cp , the linear wave phase
velocity in the mean water depth h̄ corresponding to the spectral peak period Tp ;
and qr , the volume flux of a roller on the front of a breaking wave. The second term
on the left-hand side of Eq. (28.1) is the onshore volume flux due to linear waves
July 27, 2009 9:32 9.75in x 6.5in b684-ch28 FA
where the constant value is obtained from the values of θ, h̄, and Tp specified at
x = 0. Reflected waves are neglected in this model.
The cross-shore and longshore momentum equations are expressed as
dSxx dη̄
= −ρg h̄ − τbx + τwx (28.3)
dx dx
dSxy
= −τby + τwy (28.4)
dx
where Sxx is the cross-shore radiation stress; ρ, water density; τbx , cross-shore
bottom stress; τwx , cross-shore wind stress on the sea surface; Sxy , shear component
of the radiation stress; τby , longshore bottom stress; and τwy , longshore wind stress
on the sea surface. The wind shear stresses may not be negligible especially outside
surf zones on natural beaches.7 Linear wave theory for progressive waves are used
to estimate Sxx and Sxy
1
Sxx = (nE + Mr ) cos θ + E n −
2
; Sxy = (nE + Mr ) cos θ sin θ (28.5)
2
with
where Cg is the linear wave group velocity based on h̄ and Tp; E, √ specific wave
energy with the root-mean-square wave height defined as Hrms = 8ση ; and Mr ,
momentum flux of a roller propagating with the phase velocity Cp .
The time-averaged bottom shear stresses are written as
1 1
τbx = ρfb U Ua ; τby = ρfb V Ua ; Ua = (U 2 + V 2 )0.5 (28.7)
2 2
where fb is the bottom friction factor, and the overbar indicates time averaging.
The bottom friction factor fb is of the order of 0.015 and should be calibrated using
longshore current data because of the sensitivity of longshore currents to fb . The
equivalency of the time and probabilistic averaging is assumed to express τbx and
τby in terms of the mean and standard deviation of the depth-averaged velocities U
and V expressed as
with
Ū V̄
FU = U∗ + r cos θ; FV = V∗ + r sin θ; U∗ = ; V∗ = (28.9)
σT σT
July 27, 2009 9:32 9.75in x 6.5in b684-ch28 FA
where Ū and V̄ are the depth-averaged cross-shore and longshore currents; σT , the
standard deviation of the oscillatory (assumed Gaussian) depth-averaged velocity
UT with zero mean; and r, the Gaussian variable defined as r = UT /σT whose
probability density function is given by
2
1 r
f (r) = √ exp − . (28.10)
2π 2
Linear progressive wave theory is used locally to express UT in terms of the oscil-
latory free surface elevation (η − η̄)
Cp
UT = (η − η̄) (28.11)
h̄
which yields the standard deviation σT of the oscillatory velocity UT
σT = Cp σ∗ ; σ∗ = ση /h̄. (28.12)
It is noted that U∗ = Ū /σT and V∗ = V̄ /σT are of the order of unity or less. The
standard deviations of U and V are given by
where cos θ > 0 but sin θ can be negative. Substitution of Eq. (28.8) into Eq. (28.7)
yields
1 1
τbx = ρfb σT2 Gbx ; τby = ρfb σT2 Gby (28.14)
2 2
with
∞ ∞
Gbx = FU Fa f (r)dr; Gby = FV Fa f (r)dr (28.15)
−∞ −∞
τsx = ρa CD W10
2
cos θw ; τsy = ρa CD W10
2
sin θw (28.16)
where ρa is the air density (ρa 1.225 kg/m3 ); CD , the drag coefficient, W10 , the
10-m wind speed; and θw , the wind direction defined in Fig. 28.1. The formula by
Large and Pond8 is used to estimate CD where CD = 0.0012 for W10 < 11 m/s and
CD = (0.00049 + 0.000065 W10) for W10 ≥ 11 m/s. It is noted that the measured
values of CD during tropical cyclones by Powell et al.9 indicated no increase of CD
with the increase of W10 much above 25 m/s. In short, available data is insufficient
to estimate CD for extreme wind conditions.
The wave energy equation is written as
dFx
= −DB − Df ; Fx = ECg cos θ (28.17)
dx
July 27, 2009 9:32 9.75in x 6.5in b684-ch28 FA
where Fx is the cross-shore energy flux based on linear progressive wave theory,
and DB and Df are the energy dissipation rates due to wave breaking and bottom
friction, respectively. The energy dissipation rate DB due to wave breaking in
Eq. (28.17) is estimated using the formula by Battjes and Stive,10 which was mod-
ified by Kobayashi et al.11 to account for the local bottom slope and to extend the
computation to the lower swash zone. The breaker ratio parameter γ in this formula
is typically in the range of γ = 0.5–1.0,1 but should be calibrated to obtain a good
agreement with the measured cross-shore variation of ση if such data is available.
The energy dissipation rate Df due to bottom friction is expressed as
1
Df = ρfb Ua3 . (28.18)
2
Substitution of Ua given in Eq. (28.8) into Eq. (28.18) yields
∞
1
Df = ρfb σT3 Gf ; Gf = Fa3 f (r)dr (28.19)
2 −∞
1
10 30 10
20
10 20 10
Fm
10
0 20 10
20
10 20 10
20 30 10
10
−1
1 20 30 30 20
10 10
20
10
Fm
0 10
10
10
10 20 10
20
20 30
−1 30
1 30 30
20 20
10 20 10
10
Fm
0 10
10 20 10
20
20 30
−1 30
−1 −0.5 0 0.5 1
rm
Fig. 28.2. Contours of maximum percentage error with respect to sin θ as a function of rm and
Fm for Gbx (top), Gby (middle), and Gf (bottom).
analytically to obtain the longshore current V̄ for the computed Gby . This implicit
numeric procedure improves the numeric stability of the computation marching in
x (similar to time).
The combined wave and current model predicts the cross-shore variations of the
hydrodynamic variables used in the following sediment transport model for given
beach profile, water level, and seaward wave conditions at x = 0. The bottom
sediment is assumed to be uniform and characterized by d50 , the median diameter;
wf , sediment fall velocity; and s, sediment specific gravity.
First, the cross-shore variation of the degree of sediment movement is estimated
using the critical Shields parameter ψc 13 which is taken as ψc = 0.05. The instan-
taneous bottom shear stress τb is assumed to be given by τb = 0.5 ρfb Ua2 with Ua
given in Eq. (28.8). The sediment movement is assumed to occur when |τb | exceeds
the critical shear stress, ρg(s − 1)d50 ψc . The probability Pb of sediment movement
can be shown to be the same as the probability of (r − rm )2 > Fb2 = (Rb2 − Fm 2
)
−1 0.5
where Rb = [2g(s − 1)d50 ψc fb ] /σT and rm and Fm are defined in Eq. (28.23).
For the Gaussian variable r given by Eq. (28.10), Pb is given by
1 Fb − rm 1 Fb + rm
Pb = erf c √ + erf c √ for Fb2 > 0 (28.28)
2 2 2 2
July 27, 2009 9:32 9.75in x 6.5in b684-ch28 FA
and Pb = 1 for Fb2 ≤ 0 where erf c is the complementary error function. The value
of Pb computed from x = 0 located outside the surf zone increases landward and
fluctuates in the surf and swash zones, depending on the presence of a bar or a
terrace that increases the local fluid velocity.
Second, the cross-shore variation of the degree of sediment suspension is esti-
mated using the experimental finding of Kobayashi et al.11 who showed that the tur-
bulent velocities measured in the vicinity of the bottom were related to the
energy dissipation rate due to bottom friction. Representing the magnitude of the
instantaneous turbulent velocity by (Df /ρ)1/3 with Df = 0.5ρfb Ua3 in light of
Eq. (28.18), the probability Ps of sediment suspension is assumed to be the same
as the probability of (Df /ρ)1/3 exceeding the sediment fall velocity wf . The prob-
ability Ps is then equal to the probability of (r − rm )2 > Fs2 = (Rs2 − Fm 2
) with
Rs = [(2/fb ) wf /σT ] and is given by
1/3
1 Fs − rm 1 Fs + rm
Ps = erf c √ + erf c √ for Fs2 > 0 (28.29)
2 2 2 2
and Ps = 1 for Fs2 ≤ 0. If Ps > Pb , use is made of Ps = Pb assuming that sediment
suspension occurs only when sediment movement occurs. Fine sands on beaches
tend to be suspended once their movement is initiated.
Third, the suspended sediment volume Vs per unit horizontal bottom area is
estimated by modifying the sediment suspension model by Kobayashi and Johnson14
eB Dr + ef Df dzb
Vs = Ps (1 + Sb2 )0.5 ; Sb = (28.30)
ρg(s − 1)wf dx
where Sb is the cross-shore bottom slope; and eB and ef are the suspension effi-
ciencies for the energy dissipation rates Dr and Df due to wave-breaking and bottom
friction in Eqs. (28.17) and (28.20), respectively. Use is made of eB = 0.002 and
ef = 0.01 adopted by Kobayashi et al.1 but the value of eB is uncertain and should
be calibrated if Vs is measured.15 The sediment suspension probability Ps is added
in Eq. (28.30) to ensure Vs = 0 if Ps = 0. The term (1 + Sb2 )0.5 is the actual bottom
area per unit horizontal bottom area and essentially unity except for very steep
slopes. The cross-shore and longshore suspended sediment transport rates qsx and
qsy are expressed as
qsx = aŪ Vs ; qsy = V̄ Vs (28.31)
where a is the empiric suspended load parameter. The parameter a accounts for
the onshore suspended sediment transport due to the positive correlation between
the time-varying cross-shore velocity and suspended sediment concentration. The
value of a increases to unity as the positive correlation decreases to zero. For the
three small-scale equilibrium profile tests conducted by Kobayashi et al.,11 a was
of the order of 0.2. Use is made of a = 0.2 in the subsequent computation. The
cross-shore suspended sediment transport rate qsx is negative (offshore) because the
return (undertow) current Ū is negative (offshore). On the other hand, the long-
shore suspended sediment transport rate qsy in Eq. (28.31) neglects the correlation
between the time-varying longshore velocity and suspended sediment concentration,
which appears to be very small if the longshore current V̄ is sufficiently large.
July 27, 2009 9:32 9.75in x 6.5in b684-ch28 FA
Fourth, the formulas for the cross-shore and longshore bedload transport rates
qbx and qby are devised somewhat intuitively because bedload in the surf zone
has never been measured. The time-averaged rates qbx and qby are tentatively
expressed as
bPb
qbx = σ 3 (1 + U∗ V∗2 + 2Fm sin θ)Gs (28.35)
g(s − 1) T
bPb
qby = σ3 [V∗ (1 + U∗2 + V∗2 ) − 2rm sin θ] (28.36)
g(s − 1) T
where b is the empiric bedload parameter, and Gs , the bottom slope function. The
sediment movement probability Pb given in Eq. (28.28) accounts for the initiation
of sediment movement. It is noted that b∗ = 1 in Eq. (28.35) to compensate for the
limitations of Eq. (28.32) and the Gaussian distribution of the horizontal velocity
used in Eqs. (28.9) and (28.10) as discussed by Kobayashi et al.17 They calibrated
b = 0.002 using the 20 water tunnel tests of Ribberink and Al-Salem18 and the
four large-scale wave flume tests of Dohmen-Janssen and Hanes.19 However, these
tests were conducted for nonbreaking waves and the assumed value of b = 0.002 is
uncertain in surf and swash zones.
The bottom slope function Gs was introduced by Kobayashi et al.17 to account
for the effect of the steep cross-shore slope Sb on the bedload transport rate and is
expressed as
where φ is the angle of internal friction of the sediment and tan φ 0.63 for sand.20
For Sb = 0, Gs = 1. Equation (28.37) corresponds to the functional form of Gs
July 27, 2009 9:32 9.75in x 6.5in b684-ch28 FA
used by Bagnold21 for steady stream flow on a downward slope with Sb < 0 where
the downward slope increases qbx . Equation (28.38) ensures that Gs approaches
negative infinity as the upward slope Sb approaches tan φ and that Eqs. (28.37)
and (28.38) reduce to Gs (1 − Sb / tan φ) for |Sb | tan φ. Equation (28.35) with
Gs given by Eqs. (28.37) and (28.38) implies that the bedload transport rate qbx
is positive (onshore) for Sb < (tan φ)/2 and negative (offshore) for Sb > (tan φ)/2.
Use is made of |Gs | < Gm = 10 to avoid an infinite value in the computation. The
computed profile change is not very sensitive to the assumed value of Gm because
the beach profile changes in such a way to reduce a very steep slope except in the
region of scarping.
The landward marching computation of the present time-averaged model ends
at the cross-shore location x = xm where the mean water depth h̄ is less than 1 cm.
The probabilistic model by Kobayashi et al.22 could be used to predict the irregular
wave runup distribution but no reliable data exists for suspended sand and bedload
transport rates in the zone which is wet and dry intermittently. Consequently, the
following simple procedure is adopted to deal with the zone with the bottom slope
Sb > tan φ. The cross-shore total sediment transport rate qx = (qsx + qbx ) at
x = xm is denoted by qxm . If qxm is negative (offshore), qx is extrapolated linearly
to estimate qx on the scarped face with Sb > tan φ
where xe is the landward limit of the scarping zone with Sb > tan φ. The extrapo-
lated qx is in the range of qxm ≤ qx ≤ 0 and the scarping zone is eroded due to the
offshore sediment transport. This simple procedure does not allow onshore sediment
transport due to overwash.
Finally, the cross-shore beach profile change is computed using the continuity
equation of bottom sediment for the case of alongshore uniformity
∂zb ∂qx
(1 − np ) + =0 (28.40)
∂t ∂x
where np is the porosity of the bottom sediment which is assumed to be
np = 0.4, and t, slow morphologic time for the change of the bottom elevation
zb . Equation (28.40) is solved using an explicit Lax-Wendroff numeric scheme
(e.g., Ref. 23) to obtain the bottom elevation at the next time level. This compu-
tation procedure is repeated starting from the initial bottom profile until the end
of a profile evolution test. The computation time is of the order of 10−3 of the test
duration.
The present numeric model is compared with the spilling and plunging wave
tests conducted in the large-scale sediment transport facility at the US Army
Engineer Research and Development Center.24 Pumps were used to circulate the
wave-induced longshore current and establish the alongshore uniformity of hydro-
dynamics and morphology on the middle section of the quasi-equilibrium beach
July 27, 2009 9:32 9.75in x 6.5in b684-ch28 FA
Initial t=10 h
40
Spilling
0
zb (cm)
−40
−80
40
Plunging
0
zb (cm)
−40
−80
0 4 8 12 16
x (m)
Fig. 28.3. Quasi-equilibrium beach profile (solid line) and computed profile after 10-h wave action
(dash line) for spilling and plunging wave tests.
comprised of uniform fine sand with d50 = 0.15 mm, s = 2.65, and wf = 1.65 cm/s.
Table 28.1 lists the wave conditions at the seaward boundary x = 0 and the values
of the bottom friction factor fb in Eq. (28.7) and the breaker ratio parameter γ
for DB in Eq. (28.17). These values of fb and γ are the same as those used in
the previous comparison by Kobayashi
√ et al.1 The still water depth d, the root-
mean-square wave height Hrms = 8ση and the incident unidirectional wave angle
θ at x = 0 were practically the same for both tests. The major difference was the
spectral peak period Tp of the incident waves and the resulting breaker patterns.
The quasi-equilibrium beach profiles of the two tests are shown in Fig. 28.3.
Comparisons are made of the measured and computed cross-shore variations of
the mean and standard deviation of the free surface elevation η, cross-shore velocity
U , and longshore velocity V as well as the suspended sediment volume Vs per
unit horizontal bottom area. Vs is predicted using Eq. (28.30) with the sediment
suspension probability Ps and the bottom slope effect, which were not included
previously. These comparisons are assessed in light of those presented by Agarwal
et al.25 and Kobayashi et al.1 The degree of agreement is found to be practically the
same. The modifications presented in this chapter change the computed results very
little. However, the present computation is more efficient and stable because of the
July 27, 2009 9:32 9.75in x 6.5in b684-ch28 FA
0.02
transport rate
(cm2/s)
−0.02 Spilling
0.05
Cross−shore sand
transport rate
(cm2/s)
−0.05
Plunging
−0.1
0 4 8 12 16
x (m)
Fig. 28.4. Computed cross-shore sand transport rates qbx , qsx , and qx = (qbx + qsx ) for spilling
and plunging wave tests.
use of the analytic expressions of Gbx , Gby , and Gf given by Eqs. (28.25)–(28.27).
The roller effect represented by the roller volume flux qr improves the agreement
for the longshore current V̄ as shown in the previous comparisons. The computed
sediment transport rates with the roller effect are presented in the following.
Figure 28.4 shows the computed cross-shore variations of qbx , qsx , and qx for
the spilling and plunging wave tests. The cross-shore bedload transport rate qbx
is positive (onshore), whereas the cross-shore suspended sand transport rate qsx is
negative (offshore). The absolute values of qbx and qsx are larger in the breaker zone
near x = 4 m and near the still water shoreline, especially for the plunging waves.
The computed total sand transport rate qx = (qbx +qsx ) is positive (onshore) except
in the zone near the shoreline where qx < 0. The absolute value of qx is less than
about 0.05 cm2 /s but qx = 0 is required on the equilibrium beach. Consequently,
the profile evolution is computed using the measured quasi-equilibrium profile as
the initial profile. The initial profile is exposed to the wave conditions listed in
Table 28.1 for 10 hours. The computed profile is shown in Fig. 28.3. The computed
change of the bottom elevation zb is less than about 5 cm. The subtle profile change
is difficult to predict and measure accurately. It is noted that the fluid velocity
and suspended sand concentration were not measured synchronously in these tests,
resulting in no measurement of qsx .
Figure 28.5 shows the computed cross-shore variations of qby , qsy , and qy =
(qby + qsy ). The longshore bedload transport rate qby is positive (downwave) and
small in comparison with the longshore suspended sand transport rate qsy . The
longshore total sand transport rate qy is dominated by qsy in these tests using the
fine sand with d50 = 0.15 mm and wf = 1.65 cm/s. The longshore sand transport
rate qy is larger under the plunging waves by a factor of more than 2. The comparison
of Figs. 28.4 and 28.5 indicates that the longshore sand transport qy exceeds the
cross-shore transport rate qx by a factor of about 5.
July 27, 2009 9:32 9.75in x 6.5in b684-ch28 FA
qby qsy qy
Spilling
Longshore sand
0.1
transport rate
(cm2/s)
0.05
0
0.3
Plunging
Longshore sand
transport rate
(cm /s)
0.2
2
0.1
0
0 4 8 12 16
x (m)
Fig. 28.5. Computed longshore sand transport rates qby , qsy , and qy = (qby + qsy ) for spilling
and plunging wave tests.
Measured Computed
0.1
Spilling
qsy(cm2/s)
0.05
0
0.3 Plunging
qsy (cm2/s)
0.2
0.1
0
0 4 8 12 16
x (m)
Fig. 28.6. Measured and computed longshore suspended sand transport rates for spilling and
plunging wave tests.
Figure 28.6 compares the measured and predicted longshore suspended sand
transport rates. The measured qsy was obtained by integrating the product of
the measured mean longshore velocity and mean sand concentration vertically as
explained by Kobayashi et al.1 The numeric model overpredicts qy in the middle of
the surf zone by a factor of about 2 perhaps because Eq. (28.30) does not account
for the downward decay of turbulence generated by broken waves landward of the
breaker zone.
July 27, 2009 9:32 9.75in x 6.5in b684-ch28 FA
Measured Computed
Spilling
0.1
qy (cm2/s)
0.05
0.3 Plunging
qy (cm2/s)
0.2
0.1
0
0 4 8 12 16
x (m)
Fig. 28.7. Measured and computed longshore total sand transport rates for spilling and plunging
wave tests.
Figure 28.7 compares the measured and predicted longshore total sand transport
rates. The measured qy was obtained from the volume of sand collected in each
of 20 downdrift bottom traps. This time-averaged model predicts the large sand
transport rate near the shoreline fairly well. The overall agreement is within a factor
of about 2. The difference in the cross-shore distributions of qy under the spilling
and plunging waves is predicted by the present time-averaged model.
28.5. Conclusions
The combined wave and current model for obliquely incident waves by Kobayashi
et al.1 is improved by including the finite-depth effect on the relationship between
the oscillatory horizontal velocity and free surface elevation as well as the wind
stresses for future field applications. The numeric integrations involved in the
bottom shear stresses and energy dissipation rate are replaced by sufficiently
accurate analytic expressions. These modifications improve the computational effi-
ciency and numeric stability of the wave and current model. The sediment transport
model developed by Kobayashi et al.11,17 for normally incident waves is extended
to include obliquely incident waves and longshore currents. The probabilities of
sediment movement and suspension are proposed to account for the initiation of sed-
iment movement and suspension under irregular waves. The cross-shore and long-
shore suspended sediment transport rates are expressed in terms of the suspended
sediment volume per unit horizontal area and the cross-shore and longshore cur-
rents. The onshore transport due to the correlation between the cross-shore velocity
and suspended sand concentration is taken into account. The cross-shore and long-
shore bedload transport rates are estimated using a formula similar to that by
July 27, 2009 9:32 9.75in x 6.5in b684-ch28 FA
Meyer-Peter and Mueller that reduces to the bedload formula by Kobayashi et al.17
for normally incident waves.
The new sediment transport model coupled with the improved wave and current
model is compared with spilling and plunging wave tests conducted in a large wave
basin using fine sand. The coupled model predicts the measured longshore suspended
sand and total transport rates within a factor of about 2. The longshore bedload
transport rate is predicted to be small in comparison with the corresponding sus-
pended sand transport rate. The cross-shore sand transport rates were not mea-
sured in these quasi-equilibrium beach tests. The coupled model predicts only slight
beach profile changes after 10-h wave action in these tests. However, subtle profile
changes are difficult to predict and measure accurately.
The coupled numeric model will need to be compared with field data to demon-
strate its utility for practical applications. The assumptions of longshore uni-
formity and uniform sediment are too restrictive on natural beaches. The cross-shore
one-dimensional model may be extended to a horizontally two-dimensional model
without much difficulty but available sediment transport data is scarce in surf and
swash zones. For the prediction of long-term morphologic changes, a method for
long-term and spatial averaging will need to be devised to compute morphologic
changes for years.
Acknowledgments
This study was supported by the MORPHOS Project of the US Army Corps of
Engineers, Coastal and Hydraulics Laboratory. The second author was supported
by the Spanish Postdoctoral Scholarship MEC/FULBRIGHT.
References
Chapter 29
John R.-C. Hsu∗,† , Melissa M.-J. Yu, F.-C. Lee and Richard Silvester∗
Department of Marine Environment and Engineering
National Sun Yat-sen University, Kaohsiung 80424, Taiwan, ROC
∗
School of Civil and Resource Engineering
University of Western Australia
Nedlands, WA 6009, Australia
†
jrchsu@faculty.nsysu.edu.tw
29.1. Introduction
825
August 6, 2009 14:1 9.75in x 6.5in b684-ch29 FA
form of salient or tombolo which can be observed in the lee of harbor breakwaters,
detached breakwaters and inclined or composite groins. Despite these beaches have
occupied about 50% of the world coastline1 and are ubiquitous in many coun-
tries in the world, most of them are in dynamic or even unstable condition which
require continuous sediment supply to maintain their stability in a persistent swell
environment.
Although headland-bay beaches have been called in a variety of names since
the 1900s (zeta-curved bays,2 half-heart-shaped bays,3 logarithmic spiral beaches,4
crenulated-shaped bays,5,6 curved or hooked beaches,7 pocket beaches,8 and
headland-bay beaches4,9–11 ; see Refs. 12 and 13), a common feature among them
is the asymmetric shape, which is characterized by a curved shadow zone, a gently
curved transition and or a relatively straight tangential portion at its downdrift
end. In terms of stability, a bay beach may be either in static equilibrium, dynamic
equilibrium, or unstable.12–14 Static equilibrium is reached when the predominant
waves are breaking simultaneously around the whole bay periphery, hence littoral
drift is almost nonexistent and additional external sediment is not required to
maintain its long-term stability. On the other hand, for bay beaches in dynamic equi-
librium, balance in sediment budget is required to maintain their stability; otherwise,
shoreline could retreat as supply reduces. Should supply diminish, then a bay beach
in dynamic equilibrium may recede toward the limit defined by static equilibrium.
Finally for bays in unstable condition, which is often produced by wave sheltering
due to structural addition on a beach, a curved shoreline will result with accretion in
the lee accompanied by erosion downdrift, hence the so-called natural reshaping.13
With a plethora of names and indentations associated with headland-bay
beaches, coastal geologists, geographers, and coastal engineers have attempted to
derive an empiric expression to fit part or whole of the bay periphery, since the
1940s. Most notably these include logarithmic spiral model,4 hyperbolic-tangent
shape model,11 and parabolic bay shape model.15 These models differ in mathe-
matic expressions, coordinate system and origin, controlling parameters, and wave
direction related to wave diffraction point. The former two models consider only the
fitting of geometry to the shoreline rather than stability, without looking at wave
direction and the positioning of a headland in relation to the beach. On the other
hand, the latter (i.e., the parabolic bay shape) relates the shoreline change to the
tip of a headland or wave diffraction point upcoast, which is also the origin of a
quadratic expression in this model, hence relocation of this control point by man-
made extension can be assessed. This parabolic bay shape equation15 has received
recognition in the new Coastal Engineering Manual16 as well as the Spanish Coastal
Modeling System (SMC; Refs. 6 and 17) for project evaluation and design of recre-
ational beaches.
With the immense pressures nowadays from population growth and eco-
nomic activities, coastal strips have become a prime target for industrial and
residential development, harbors, navigation jetties, and recreation. Not only a
natural embayment requiring verification of stability prior to any proposal for
man-made utilization or improvement, but also a plane straight coast may even
be converted into an embayment with an addition of artificial structures. In
fact, artificial bayed beaches have been created for recreation and tourism in the
Mediterranean countries, especially in Spain, France, Italy, and Israel since the 1960s
August 6, 2009 14:1 9.75in x 6.5in b684-ch29 FA
(e.g., Refs. 18–20). By creating a bay beach in static equilibrium with adequate
storm beach buffer, its stability would not be affected by any reduction in sediment
supply, except the onslaught of a fierce storm wave which could erode the berm
with material being transported offshore to form a bar. However, beach recovery is
often ensured if adequate storm buffer width is provided. This is nature’s way of
self-defense for her beach.
The main purpose of this chapter is to demonstrate several geomorphic engi-
neering aspects of the headland-bay beaches based on the concept of the parabolic
bay shape equation,15 using software MEPBAY21 and SMC6,17 that have recently
developed for applying this empiric equation. Examples to be presented include
verification of beach stability on natural and artificial embayment, especially those
at downdrift of navigation jetties, mitigation of beach erosion by headland control,
and the design of a bay beach for recreation.
Hsu and Evans15 have developed an empiric model for bay beach called “parabolic
bay shape equation” from fitting the planform of 27 mixed cases of 14 prototype
and 13 model bays believed to be in static equilibrium. This empiric equation is in
the nondimensional form of a second-order polynomial,
2
Rn β β
= C0 + C1 + C2 . (29.1)
Rβ θn θn
Equation (29.1) has two primary physical parameters, the reference wave
obliquity angle β (measured in radians) and the control line length Rβ (see defi-
nition sketch in Fig. 29.1). The former is the wave obliquity measured at the updrift
Fig. 29.1. Definition sketch of parabolic bay shape equation, showing the primary and dependent
parameters (modified from Refs. 12, 13, and 15).
August 6, 2009 14:1 9.75in x 6.5in b684-ch29 FA
control (diffraction) point from the local incident wave crest to the control line,
which links the diffraction point and a downdrift control point on the more straight
part of the curved beach; or that between the downdrift tangent and the control
line. It is worth noting that any point on or near the straight downdrift part of the
curved beach could be conveniently chosen as a downdrift control point, without
causing sensitive variation in the resulting value of Rn /Rβ . This can be determined
visually from a map, vertical aerial photograph, or on a planning sketch, once an
observer has grasped the physical insight of wave propagation toward a beach with
a headland. In addition, Eq. (29.1) also contains two dependent parameters, which
are radius Rn to any point on the curved shoreline angled θn from the same incident
wave crest line radiating out from the point of wave diffraction. The three C coef-
ficients in this equation were generated by regression analysis to fit the periph-
eries of the 27 prototype and model bays and vary with the reference wave angle
β.12,13 In recent years, the values C coefficients have been modified to include the
straight downdrift part of the embayment,22–24 or can be further simplified to ease
its manual calculation for Rn /Rβ .
Manual application of Eq. (29.1) is straightforward. For example, (1) first
choosing a downdrift control point on a more straight part of the embayment or
an appropriate location on a curved shoreline and linking it to the updrift control
point (i.e., point of wave diffraction) to form the control line; (2) determining the
reference wave angle β and the control line length Rβ ; (3) calculating a set of rays
and polar-angles (Rn , θn ) on the embayed shoreline, starting from θn = β at the
downdrift control point; and (4) finally sketching the resulting shoreline in static
equilibrium.
The user may then visually examine the stability of the bay beach by the
nearness of the existing shoreline to that predicted in static equilibrium. Should they
coincide with each other, or very close, then the existing beach may be classified as
in static equilibrium or in quasi-static equilibrium. In this condition, the stability of
this embayment would not be affected by any reduction in sediment supply exter-
nally or internally, except the onslaught of a fierce storm. If the existing shoreline
is seaward of that predicted, then the beach is said to be in dynamic equilibrium,
in which shoreline could remain unchanged from the balance in sediment budgets.
However, shoreline in dynamic equilibrium may retreat as net supply decreases.
Moreover, an unstable situation which implies the existing shoreline is landward of
the predicted and natural beach reshaping would occur within the embayment due
to wave sheltering by structures, in which accretion in the lee accompanying by
erosion downdrift.
However, manual applications of Eq. (29.1) are rather repetitive and laborious.
To improve the efficiency of application, this empiric equation has been integrated
into software called MEPBAY21 and SMC.6,17 First, MEPBAY was developed as
educational software in the University of Vale at Itajai, Brazil, which can be down-
loaded from http://siaiacad05.univali.br/∼meppe. Wave height information is not
required upon applying MEPBAY; hence, the application procedure is straight-
forward. On the other hand, the parabolic bay shape equation has been integrated
into SMC (Coastal Modeling System), a numeric modeling package developed by
the Ocean and Coastal Research Group (GIOC) in the University of Cantabria
August 6, 2009 14:1 9.75in x 6.5in b684-ch29 FA
for the Spanish Ministry of Environment during the period of 1996–2003. This
model includes wave-current–sediment calculations for short- and long-term beach
and profile changes, and incorporates a beach design module. Comparing with the
MEPBAY, SMC is much sophisticated because calculations of wave-current field
are required while applying it for the design of a headland-bay beach in static equi-
librium. SMC can be downloaded after registration from http://www.smc.unican.es.
Examples of its application to beach design will be given in Sec. 5 later in this
chapter.
Klein et al.21 demonstrated three categories of bay beach stability using
MEPBAY. Their examples are reproduced here to assist the readers in applying
MEPBAY and post assessment of beach stability. The first example is for a bay
beach in static equilibrium at Taquaras–Taquarinhas in the central-north coast
of Santa Catarina, Brazil because the existing shoreline planform is almost iden-
tical to the static bay shape given by the parabolic equation (Fig. 29.2). Second
example depicts the dynamic equilibrium condition at the southern part of Balneário
Camboriú Beach, also in Santa Catarina, Brazil, where the existing shoreline
planform is seaward of that predicted (Fig. 29.3) and its current stability has been
maintained by sediment from updrift and/or riverain source. The third example in
Fig. 29.4 indicates the effect of man-made structure on Porto Beach in Imbituba,
Santa Catarina, Brazil, where downdrift erosion has occurred accompanying by
accretion in the lee of the structure. Nearshore current is responsible for the
transport of sediment from downdrift to the lee, rather than the usual longshore
current. In this case, the existing beach is landward of the static bay shape pre-
dicted by the parabolic model using MEPBAY.
On the other hand, the SMC6,17 software was developed as part of Spanish
Coastal plan, the State Coastal Office and the Coastal and Ocean Research Group
(GIOC) of the University of Cantabria for a new Spanish Beach Nourishment and
Fig. 29.3. Example of a bay beach in dynamic equilibrium at Balneário Camboriú in Santa
Catarina, Brazil.21
Fig. 29.4. Example of a bay beach in unstable condition or natural reshaping at Porto Beach in
Imbituba, Santa Catarina, Brazil.21
Protection Manual (SBM) between 1995 and 2003. The coastal modeling system
(SMC) allows coastal technicians to develop coastal engineering projects based
on a work methodology, a database of littoral morphodynamic information and
numeric tools. The many different numeric models included in the SMC permit for
a specific engineering project, the design and evaluation of coastal alternatives in
different scales for short-term (hours–days), medium-term (seasonal changes), and
long-term (years–decades). The organization logo of the SMC in Spanish is depicted
in Fig. 29.5.
August 6, 2009 14:1 9.75in x 6.5in b684-ch29 FA
Headland-bay beaches in natural condition differ not only in shape and size, but also
with different degrees of sediment availability and exposures to the prevailing waves.
They may appear in the form of single curved embayment with a straight portion
August 6, 2009 14:1 9.75in x 6.5in b684-ch29 FA
Fig. 29.6. Example of natural bay beaches in static equilibrium at Bechino, Tasmania, Australia.
Courtesy of Google Earth imagery.
August 6, 2009 14:1 9.75in x 6.5in b684-ch29 FA
Fig. 29.7. Example of fitting static bay shape to man-made tombolo beaches constructed in the
1980s at Mamzar Beach, Dubai, UAE. Courtesy of Google Earth imagery.
August 6, 2009 14:1 9.75in x 6.5in b684-ch29 FA
Straight groins have been used as inlet jetties or training walls to maintain a nav-
igable channel linking an inland harbor or lagoon with the ocean in many places.
Beach downdrift of these structures may be already curved or could evolve into a
bay periphery, initially in a dynamic condition. One basic principle associated with
the jetty construction is to minimize negative impact on the downdrift beach, by
positioning its tip appropriately with the right length.
The first example is for the Doran Beach on Bodega Bay, about 34 km SW of
Santa Rosa on the Pacific coast of California (Fig. 29.8). Without details of local
bathymetry, wave conditions, and field monitoring, it is not known whether this
beach is in static or dynamic condition. However, a preliminary assessment using
MEPBAY to the satellite imagery indicates that the Doran Beach was in dynamic
condition, as the static bay shape predicted (dotted line marked as A D) is landward
of the existing beach and its equilibrium could be maintained by sufficient sediment
supply from updrift coast and/or riverain source in the back lagoon where the
Bodega harbor is located.26
It is not known when this picture was taken, but with the two inlet jetties in
place, the southern jetty can be applied to test the stability of Doran Beach at the
time after the installation of these jetties. First, application of MEPBAY with the
tip B renders the predicted static bay shape landward of the existing waterline.
This is similar to that before the jetty construction, but this later static bay shape
marked as B D is closer to the existing shoreline than the previous A D; hence, less
potential erosion overall. It can be envisaged that beach may retreat if sediment
supply reduces in the future. However, for a bay beach in dynamic condition, a
Fig. 29.8. Effect of repositioning the tip of an inlet jetty on the stability of Doran Beach in
Bodega Bay, California. Courtesy of Google Earth imagery.
August 6, 2009 14:1 9.75in x 6.5in b684-ch29 FA
Fig. 29.9. Example of mitigating potential downdrift erosion by extending or repositioning the
tip of Zuniga jetty in Coronado Bay, San Diego, California. Courtesy of Google Earth imagery.
August 6, 2009 14:1 9.75in x 6.5in b684-ch29 FA
The parabolic bay shape equation has the potential to improve the stability of an
existing bay beach, and therefore, can help design a static bay shape for shore pro-
tection and recreation. This will create a win–win situation for a local community.
Ideally, a bay beach design requires nourishment to be protected by artificial head-
lands and incorporates with the fill placement with sufficient storm beach buffer to
reproduce a static equilibrium planform. The artificial headland may be a member
of composite groins29 (T -, L-, and Y -shaped; ship anchor or fish-tailed) or detached
breakwater, in addition to the fulfillment of cross-shore profile equilibrium.
Fig. 29.10. An oblique view of a recreational bay beach with T -head groin, submerged detached
breakwater and nourishment at Shirarahama, Wakayama, Japan. Courtesy of Prof. T. Yamashita.
August 6, 2009 14:1 9.75in x 6.5in b684-ch29 FA
Fig. 29.11. Fitting a static bay shape to Shirarahama Beach, Wakayama, Japan. Courtesy of
Dr. T. Mishima.
Fig. 29.12. Artificial bay beach under construction on South Island, Singapore showing predicted
static bay shape at high tide level. Courtesy of Google Earth imagery.
leaking offshore can be minimized as the headland tip position is linked to the
natural equilibrium profile. Moreover, Gonzalez and Medina6 proposed a method
using local wave flux to modify the downdrift control point to the original parabolic
bay shape definition (see Eq. (29.1) and Fig. 29.1). Example of this design concept
is illustrated in Fig. 29.13 for Poniente Beach and San Lorenzo Beach, in the lee of
Gijón Harbor, Spain, aiming to preserve these beaches following the expansion of
harbor facilities.
Another example applying the SMC for bay beach design is available at Dapeng
Bay, Pindong County in southern Taiwan. One of the options put forward to the
authority is shown in Fig. 29.14, which calls for a spur groin of 200-m long butting
out from the tip of the southern inlet jetty to the Dapeng Bay lagoon, with a
300-m long detached breakwater about 800 m to the south of this jetty, as well as
a short detached breakwater about 100 m further southward. To accomplish these
three headland-bay beaches, the total volume of nourishment calculated by the SMC
is about 660,000 m3 . A salient planform is expected in the lee of the 300-m long
detached breakwater in the middle of this coastal sector.
The SMC also outputs the new bathymetry from simultaneous equilibrium in
planform and beach profile, as illustrated in Fig. 29.15. Based on the results of
this expected distribution of bottom contours, the SMC then performs a series of
numeric calculations on waves (for swells and storms) and longshore currents (not
shown in this chapter). With the unique operational system inherited with the SMC
and onscreen modification of bottom contours via the tracking of the mouse on the
screen, graphic outputs are usually available in a matter of minutes for most cases
of bay beach design, once the bathymetry and wave conditions are pre-processed
August 6, 2009 14:1 9.75in x 6.5in b684-ch29 FA
Fig. 29.13. Design of static bay shapes for recreation at Poniente Beach and San Lorenzo Beach,
Gijon, Spain. Courtesy of SMC software.
Fig. 29.14. Design of static bay shapes for recreation on an existing bathymetry at Dapeng Bay
in southern Taiwan.
August 6, 2009 14:1 9.75in x 6.5in b684-ch29 FA
Fig. 29.15. An SMC output of the modified bathymetry following the proposed headland layout
of static bay shapes at Dapeng Bay in southern Taiwan.
and resided in the system. For engineering applications, a feasible option can then
be made after examining a series of trial options.
29.6. Conclusions
The shoreline along the boundary between the land and the sea may be either
straight over a sufficient length or curved in plane separated by headlands, natural
or man-made. The curved beaches are aesthetically beautiful and very stable, if
static equilibrium is reached, compared to their straight counterparts. As more than
70% of the world’s sandy beaches have retreated31 due to natural and man-made
factors, it would seem worthwhile to consider the more permanent geomorphic form
of the bay beach in static equilibrium, for which a parabolic bay shape equation
was developed.15
On shore protection, traditional hard structures and artificial nourishment
without adequate structural protection may not be the best or long-term solution
for an eroding beach, as the diminishing trend of sediment supply has prevailed or
is imminent on almost all the sandy beaches worldwide. A practical and feasible
approach is to emulate a bay beach in static equilibrium in a natural environment
that has remained stable without the need of additional supply in a persistent
swell condition. This concept has been labeled as Headland Control.12,13 For places
receiving storm wave attack from time to time, a bay beach may survive if an ade-
quate storm buffer is provided incorporating headland control methodology.
If a bay beach is already in static equilibrium, it should be spared from any
proposal for large-scale development because an unstable condition would eventuate
with the ongoing erosion problem. However, should a bay beach be in dynamic
equilibrium, it can be converted into static equilibrium by installing an artificial
headland near the tip of the present headland.26
August 6, 2009 14:1 9.75in x 6.5in b684-ch29 FA
Based on the static bay beach concept and with the aid of software MEPBAY21
and SMC6,17,25 demonstrated in this chapter, it should become obvious that the
parabolic bay shape model is beneficial for a wide range of practical applications
in coastal management and shore protection, ranging from the verification of bay
beach stability in natural or man-made condition, for devising a plan of headland
layout to stabilize the sandy beach bounded between them, for mitigating potential
beach erosion downdrift of harbor breakwater and inlet jetty, and for the design of
bay beaches for recreation.
Acknowledgments
The authors wish to acknowledge Prof. R. Medina and Prof. M. Gonzalez, Coastal
and Ocean Research Group (GIOC), University of Cantabria, Spain, for the pro-
vision of SMC implemented in this study. JRCH also thanks the support from Asian
Pacific Ocean Research Center, NSYSU, Taiwan.
References
22. S. K. Tan and Y. M. Chiew, J. Waterw. Port Coast. Ocean Eng. 120, 145 (1944).
23. M. Serizawa, T. Uda, T. Kumada, T. San-Nami and K. Furuika, Ann. J. Coast. Eng.
47, 601 (2000) (in Japanese).
24. K. Sakai, A. Kobayashi, T. Kumada, T. Uda and M. Serizawa, Proc. Coast. Sedi-
ments’03 (2003), p. 1.
25. M. Gonzalez and R. Medina, Proc. 2005 Int. Coast. Plan. Forum, Kaohsiung, Taiwan
(2005), p. 57.
26. J. R.-C. Hsu, L. Benedet, A. H. F. Klein, A. L. A. Raabe, C.-P. Tsai and T.-W. Hsu,
J. Coast. Res. 24, 198 (2008).
27. D. R. Coates, Geomorphology and Engineering, ed. D. R. Coates (George Allen &
Unwin, London, 1980), p. 3.
28. J. R.-C. Hsu, J.-C. Chu, S.-R. Liaw and C.-Y. Lee, Proc. 30th Int. Conf. Coast. Eng.,
ASCE (2006), p. 3762.
29. K. R. Bodge and S. Howard, Proc. 30th Int. Conf. Coast. Eng., ASCE (2006), p. 3756.
30. Y. Tsuchiya, Y. Kawata, T. Yamashita, T. Shibano, M. Kawasaki and H. Habara,
Proc. 23rd Int. Conf. Coast. Eng., ASCE (1971), p. 3426.
31. E. C. F. Bird, Beach Management (John Wiley & Sons, Chichester, 1966).
August 17, 2009 11:4 9.75in x 6.5in b684-ch30 FA
Chapter 30
Beach Nourishment
Robert G. Dean
Department of Civil and Coastal Engineering
University of Florida, Gainesville, FL, USA
dean@coastal.ufl.edu
Julie D. Rosati
U.S. Army Corps of Engineers
Coastal and Hydraulics Laboratory, Mobile, AL, USA
julie.d.rosati@erdc.usace.army.mil
30.1. Introduction
843
August 17, 2009 11:4 9.75in x 6.5in b684-ch30 FA
this deposit will both provide a degree of sheltering and gradually be transported
landward by wave action (Fig. 30.1).
The benefits of beach nourishment can be fairly broad including improving
recreational opportunities,1 providing protection against major storms,2 esthetics
(beaches are more attractive than coastal armoring), and enhancement of ecologic
functions, such as increasing habitat.
The economic viability of beach nourishment depends on a number of factors
including, but not limited to: (1) the value of the upland that is to be protected
including both the structures and the income producing value of a wide recreational
beach (e.g., Ref. 1), (2) the background erosion rate, (3) the availability of relatively
low-cost quality sediment, and (4) the value of ecologic enhancements. Generally,
the beaches most ideally suited for nourishment are those in which the value of the
upland resources to be protected or enhanced are substantial and the background
erosion is low such that with a moderate amount of sediment input, the system can
be “tipped” back to neutrality.
In evaluating an eroded area to determine the appropriateness of beach nour-
ishment, the cause of the erosion should be identified to the degree possible. In
some cases, the cause is anthropogenically induced such as interruption of the long-
shore sediment transport by a jettied inlet and in such cases, it is desirable to
reinstate the transport. In other cases, the cause may be the result of constructing
upland infrastructure too far seawarda and the associated difficulties of maintaining
a beach and in still other cases, the erosion may be completely natural. The most
appropriate response to beach erosion is always more likely to be identified with
a thorough understanding of the cause of the erosion that has led to the need for
nourishment.
This chapter provides a review of the planform and profile processes related to
beach nourishment. The governing equations are introduced to the degree that they
are known. The combination of the linearized longshore sediment transport equation
and the continuity equation results in a second-order partial differential equation
a In some cases, coastal construction has occurred during periods when the beach was at an abnor-
mally seaward position. With the return of more normal conditions, the beach translated landward
and the construction was threatened. Thus, nourishment was considered as an option. In other cases
prior to good coastal management, construction to expand the existing hotel facilities encroached
on the active beach.
August 17, 2009 11:4 9.75in x 6.5in b684-ch30 FA
The need for beach nourishment may occur in a variety of morphologic settings and
the analysis and design challenges differ for each of these. Four different morphologic
settings are considered in this chapter: (1) a pocket beach in which there is little
exchange with the adjacent beaches; (2) a long uninterrupted beach, probably the
most common of the four; (3) a beach downdrift of a partial or complete littoral
barrier; and (4) a barrier island subject to the combined processes of “rollover,”
subsidence, and longshore sediment transport gradients. The unique characteristics
of each of these settings are described briefly below.
Beach nourishment results in a system that is out of equilibrium due to the pro-
tuberance in the planform and a profile that is generally oversteepened. This dis-
equilibrium induces both alongshore and cross-shore sediment flows that tend to
restore equilibrium. Because of the complexity of the sediment transport processes
attending beach nourishment, these processes are usually considered separately
in terms of planform and cross-shore evolution. Usually the timescales of cross-
shore evolution are much shorter than the longshore processes associated with
planform evolution and most designs consider the cross-shore processes to occur
instantaneously. In the following, we consider the two components of sediment
transport separately that are responsible for these two modes of evolution shown
in Fig. 30.2.
Fig. 30.2. Sand transport “losses” and beach profiles associated with a beach nourishment
project.11
height, respectively, and Qx is the total in-place longshore sediment transport rate.
A number of longshore sediment transport equations are available. For purposes
here, we choose the so-called “CERC Equation” expressed as
5/2
Hb g/κ
Qx = K sin 2(β − αb ) (30.2)
16(S − 1)(1 − p)
in which Hb is the effective breaking wave height, β and αb are the azimuths of the
outward normal to the shoreline and the azimuth of the direction from which the
breaking wave arrives, respectively, S is the specific gravity of the sediment, p is
the in-place porosity, g is gravity, κ is the ratio of breaking wave height to water
depth (of order unity), and K is the so-called sediment transport coefficient (of
order unity) and can be considered to vary with sediment size as shown in Fig. 30.3.
These equations are now suitable for calculating evolution of a beach nour-
ishment planform and can be incorporated into a numeric model with the most
direct and simple being an explicit one-dimensional model in which the sediment
transport is applied at the grid lines during one-half time step while the shoreline
positions are held constant and then in the next one-half time step, the transport
is held constant and the shoreline positions are modified in accordance with the
transport equation. This explicit formulation allows internal features and boundary
conditions such as groins or bypassing at inlets to be handled with ease. However,
this solution scheme is subject to the Courant condition which relates the maximum
time step, ∆tmax , grid spacing, ∆x, and, in a broad sense, the energy level of the
1.5
Sediment Transport Coefficient, K
1.0
0.5
0.0
0.0 0.2 0.4 0.6 0.8 1.0
Median Sediment Diameter (mm)
Fig. 30.3. Relationship of sediment transport coefficient with median grain diameter, D50 . From
del Valle et al.,13 published with permission from the American Society of Civil Engineers.
August 17, 2009 11:4 9.75in x 6.5in b684-ch30 FA
∆x2
∆tmax ≤ (30.3)
2G
where the term G is the “longshore diffusivity” given by
5/2
Hb g/κ
G=K (30.4)
8(S − 1)(1 − p)(h∗ + B)
which can be seen to be closely related to the sediment transport that would occur at
a relative wave approach angle, β − αb = 45◦ . With the speed of modern computers,
the Courant condition is usually not restrictive for one-line numeric modeling; in
Florida (which has a rather mild wave climate), we normally use a longshore spacing
of 150 m and a time step of one day. However, it is possible to develop implicit equa-
tions that allow much larger time steps (or smaller grid spacings), but implementing
internal boundary conditions such as discussed above is more difficult.
68.5He2
h∗ = 2.28He − (30.5)
gTe2
b Most nourishment projects are judged by the additional dry beach width and the longevity of
this width.
August 17, 2009 11:4 9.75in x 6.5in b684-ch30 FA
Fig. 30.4. Profile translation, ∆y0 , associated with volume density addition, ∀, with compatible
sands.
∀
∆y0 = (30.6)
(h∗ + B)
where all variables are in consistent units. An early design rule of thumb was
that “the placement of one cubic yard of sediment as nourishment results in the
advancement of the beach by one foot.” Inspection of Eq. (30.6) will show that this
rule of thumb implies that the sum (h∗ + B) = 27 ft. This is a reasonable value for
much of the east coast of the USA; however, the values on the Pacific Coast and
Gulf of Mexico would be greater and less than this value, respectively.
The PC equation will not be derived here; however, the basis is approximating
the term sin 2(β − αb ) by 2(β − αb ) and further approximating β by −(∂y/∂x),
resulting in the PC equation below which is recognized as the linear second-order
heat conduction equation,c
∂y ∂2y
=G 2 (30.7)
∂t ∂x
where G is defined in Eq. (30.4). The solution of this equation requires two boundary
conditions and the specification of an initial condition.
The PC form of the planform evolution equation has significant utilities including
the availability of many known solutions. One of the most useful is the case of an ini-
tially uniform nourishment over a shoreline length, . The solution can be presented
in nondimensional form with the nondimensional time, t , longshore length, x , and
shoreline displacement, y expressed as t = 16Gt/2 , y = y/Y , and x = x/,
respectively. The solution for this case is presented in Fig. 30.5 and the solution is
given by Eq. (30.8).
1 1 1
y (x , t ) = erf √ (2x + 1) − erf √ (2x − 1) (30.8)
2 t t
where the term erf is the so-called “error function” defined as
z
2 2
erf(z) = √ e−u du (30.9)
π 0
in which z is the argument of the error function and u is a dummy variable. It is pos-
sible to integrate Eq. (30.8) to determine the volume remaining in the project area
Fig. 30.5. Nondimensional shoreline evolution for initial rectangular platform. Note that the
initial and subsequent planforms are symmetric about x = 0 and the solutions for x < 0 are not
shown.
√
Fig. 30.6. Proportion of material remaining in region placed M (t) versus the parameter Gt/.
at any future time (assuming that an appropriate value of G is known). The result
of this integration is shown in Fig. 30.6 where the proportional volume remaining
in the project area is denoted by M (t). The volume remaining is a very important
quantity to the design engineer as a key indicator of the project performance.
The PC equation and its associated solutions provide a number of valuable
insights to beach nourishment. Several are illustrated in the following. Available
space does not permit a detailed discussion for which the reader is referred to
Dean.11
Based on Fig. 30.6, a simple rule of thumb can be developed for the volumetric
“half-life” of a project with an
√ initial rectangular planform. We see from Fig. 30.6
that the half-life occurs when Gt50% / ≈ 0.46. Thus, considering reasonable values
for the variables (h∗ + B = 27 ft (or 8.2 m), K = 0.77, κ = 0.78, S = 2.65, and
p = 0.35), we find
2
t50% = K (30.10)
Hb2.5
where K = 8.8 year ft2.5 /mile2 in the English system where the breaking wave
height is in feet, the project length is in miles, and the time is in years, or K = 0.17
year m2.5 /km2 where the breaking wave height is in meters, the project length is
in km, and the time is in years. For example in the English system, for a project
with a length of one mile acted upon by an effective wave height of one foot, one-
half the volume would have spread outside the placement area in a period of 8.8
years. Similarly, in the metric system, for a project of 1 km length acted upon by
an effective wave height of 1 m, one-half the volume would have spread outside the
placement area in a period of 0.17 years.d
Inspection of the PC equation [Eq. (30.7)] demonstrates that it is insensitive
to wave direction. Thus, somewhat counter intuitively, the project evolves indepen-
dently of wave direction. It can also be shown that for projects constructed with
compatible sediments, the project centroid will remain fixed.e If the nourishment
is conducted with sand that is finer or coarser than the native, the centroid of the
planform anomaly moves downdrift or updrift, respectively.
Many projects are renourished when project evolution establishes some
minimum criterion. In the absence of background erosion, it can be shown for this
case that the renourishment intervals increase with renourishment number. However,
the renourishment interval is adversely affected by increasing background erosion
rates such that with large background erosion rates, the resulting renourishment
intervals decrease with increasing renourishment number.
in which h(y) is the water depth at a seaward distance, y, and A is the so-called
“profile scale parameter” which depends on the median sediment size, D or the
sediment fall velocity, w, as presented in Fig. 30.7 and Table 30.1. The profile rep-
resented by Eq. (30.11) is monotonic and concave upward and the profile steepness
increases with sediment size. For noncompatible sediments it is useful for design pur-
poses to consider both the native and nourishment sediments to be represented by
Eq. (30.11). It is thus possible to show that there are three types of nourished pro-
files: (1) intersecting, (2) nonintersecting, and (3) submerged, as shown in Fig. 30.8.
Figure 30.9 illustrates the effects of nourishment with the same volume density and
sediments that are coarser, the same and finer than the native. In panel (a) of
Fig. 30.9 the nourishment sediments are coarser than the native and the additional
dry beach width is 92.4 m. With compatible sediments [panel (b)], the additional
beach width is 45.3 m, approximately half that of panel (a). Panels (c) and (d)
present the results of further decreases in sediment size such that the additional dry
beach width in panel (c) is 15.3 m and zero in panel (d). The relationships between
the initial additional dry beach width, ∆y0 , horizontal distance to the offshore limit
e The PC equation is based on a linearized sediment transport equation. If the full transport
equation is considered, for compatible sediments, there is a very small movement of the nourishment
centroid in the updrift direction!
August 17, 2009 11:4 9.75in x 6.5in b684-ch30 FA
Fig. 30.7. Variation of beach profile scale parameter, A, with sediment size, D, and fall velocity, w.
From Dean,16 published with permission from the American Society of Civil Engineers.
D (mm) 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
0.1 0.063 0.0672 0.0714 0.0756 0.0798 0.084 0.0872 0.0904 0.0936 0.0968
0.2 0.100 0.103 0.106 0.109 0.112 0.115 0.117 0.119 0.121 0.123
0.3 0.125 0.127 0.129 0.131 0.133 0.135 0.137 0.139 0.141 0.143
0.4 0.145 0.1466 0.1482 0.1498 0.1514 0.153 0.1546 0.1562 0.1578 0.1594
0.5 0.161 0.1622 0.1634 0.1646 0.1658 0.167 0.1682 0.1694 0.1706 0.1718
0.6 0.173 0.1742 0.1754 0.1766 0.1778 0.179 0.1802 0.1814 0.1826 0.1838
0.7 0.185 0.1859 0.1868 0.1877 0.1886 0.1895 0.1904 0.1913 0.1922 0.1931
0.8 0.194 0.1948 0.1956 0.1964 0.1972 0.198 0.1988 0.1996 0.2004 0.2012
0.9 0.202 0.2028 0.2036 0.2044 0.2052 0.206 0.2068 0.2076 0.2084 0.2092
1.0 0.210 0.2108 0.2116 0.2124 0.2132 0.2140 0.2148 0.2156 0.2164 0.2172
Notes: 1 The A values above, some to four places, are not intended to suggest that they are known
to that accuracy, but rather are presented for consistency and sensitivity tests of the effects of
variation in grain size.
2 As an example of use of the values in the table, the A value for a median sand size of 0.24 mm is
of the initial equilibrated profile on the native profile, W∗ (= (h∗ /AN )2/3 ),f the berm
height, B, and the “fill” (i.e., nourishment) and “native” sediment scale parameters,
AF and AN , respectively, can be expressed generically as
∆y0 ∀ AF h∗
=f , , (30.12)
W∗ BW∗ AN B
f It
is essential that in the equations to follow, the value of W∗ be based on AN rather than the
AF value.
August 17, 2009 11:4 9.75in x 6.5in b684-ch30 FA
Fig. 30.8. Three generic types of nourished profiles. Adapted from Dean,17 published with
permission from the Journal of Coastal Research.
where the subscripts “F” and “N” refer to “fill” and “native” sand, respectively. It
is noted that these equations must be solved iteratively for ∆y0 . For intersecting
profiles, the volume added is indicated as ∀1 and the equation is
5/3
∀1 ∆y0 3 h∗ ∆y0 1
BW∗
=
W∗
+
5 B W∗
3/2 2/3
(30.13)
AN
1 − AF
Fig. 30.9. Effect of nourishment material scale parameter, AF on width of resulting dry beach.
Four examples of decreasing AF all with the same volume density are shown. From Dean,17
published with permission from the Journal of Coastal Research.
If the nourishment sediment is coarser than the native, for the smaller volumes,
intersecting profiles can result and with increasing volumes, the profiles will become
nonintersecting.
The critical volume separating intersecting/nonintersecting profiles is indicated
by the subscript “c1 ”
3/2
∀ 3 h∗ AN
= 1+ 1− (30.15)
BW∗ c1 5B AF
and applies only for (AF /AN ) > 1. For larger nondimensional volume densities than
given by Eq. (30.15), the profiles are nonintersecting. Also of interest, the critical
August 17, 2009 11:4 9.75in x 6.5in b684-ch30 FA
Fig. 30.10. Additional dry beach width versus volume of sand added per unit beach length.
h∗ = 6 m and B = 2 m.
volume of sand that will just yield a finite shoreline displacement for nonintersecting
profiles (AF /AN < 1), is indicated by the subscript “c2 ” as
3/2
∀ 3 h∗ AN AN
= −1 . (30.16)
BW∗ c2 5B AF AF
Figure 30.10 presents the additional dry beach width as a result of nourishment
with sediments of three different sizes. The upper curve is for the case of nour-
ishment with sediments coarser than the native and shows that for the smaller
volume densities, the profiles are intersecting. At a critical volume density of approx-
imately 570 m3 /m, the profiles become nonintersecting. For compatible sediments,
the additional beach width is approximately linearly related to the volume density.
For sediments finer than the native, a threshold volume is required to achieve a
nonzero beach width. It is noted that the results presented here are for nourishment
with a single sediment size and therefore are not fully representative of actual con-
ditions. Thus, these results should be considered as qualitative guidance rather
than fully appropriate for design. More complete applications of equilibrium beach
profile methodology considering nourishment with a range of sediment sizes docu-
ments that even for nourishment with a median sediment finer than the native, a
finite additional beach width will always occur even for small volume densities of
beach nourishment. Figure 30.11 presents results in the same form as Fig. 30.10
and includes additional results for sediments characterized by a range of sediment
sizes. The quantity σ represents the so-called sorting of the sediments which is the
standard deviation of the sediments with the size represented on a log (phi) scale
(see Ref. 18).
August 17, 2009 11:4 9.75in x 6.5in b684-ch30 FA
Fig. 30.11. Variation of shoreline displacement with volume density and storing characteristics,
h∗ = 6 m and B = 2 m.
Examples will be presented of two beach nourishment projects on open coasts. These
include the Delray Beach, FL project which has been nourished and monitored for
more than 30 years and the Perdido Key, FL project.
Fig. 30.12. Measured and predicted nourishment volumes at Delray Beach, Florida. Project
length = 2.7 miles, Hb = 1.2 ft, background erosion = 0, and 2 ft/year, h∗ + B = 21 ft.
Fig. 30.13. Perdido Key nourishment area with eastern boundary adjacent to Pensacola Pass
showing profile locations.20
consisted of compatible sand. The eastern end of the project is located adjacent to
Pensacola Pass which has been deepened for navigational purposes. Figure 30.13
presents the project location and the beach and profile nourishment locations.
Figure 30.14 presents the measured and predicted proportions of volume
remaining within the eastern and western halves of this nourishment project and
the total for this project. As for the case of predictions of the Delray Beach nour-
ishment project, the predictions here are based on a procedure developed for the
State of Florida in which all parameters are fixed except quantification of the local
August 17, 2009 11:4 9.75in x 6.5in b684-ch30 FA
Fig. 30.14. Comparison of blind-folded predictions of proportional volume changes with moni-
toring data for Perdido Key Beach Nourishment Project. Background erosion = 0.11
berm height. The effects of storms in reducing volume within the project area are
evident. Also, the effects of greater losses from the eastern half of the project which
is adjacent to Pensacola Pass compared to the western half are noted. However,
the differences in measured performances between the eastern and western project
halves are less than predicted.
Fig. 30.15. Assawoman Island, Virginia, cross-section #1 (modified from Ref. 25).
may factor into barrier island evolution. The MCO model calculates erosion, overwash
due to wave runup, overwash due to inundation of the island, subsequent migration,
and time-dependent consolidation of the substrate due to new loading. The theory and
operation of the model are described by Rosati et al.26 The model was validated with
data from the Virginia barrier islands described previously.27
Model applications compared the migration, dune lowering, and consolidation
of a barrier island with sediment restoration using a one-time, large infusion of sed-
iment versus a more traditional approach of smaller volumes of sediment placed
incrementally through time. Both applications totaled the same volume of intro-
duced sediment over the 100-year period; however, the entire volume of the large-
scale restoration was placed at the start of the simulation, whereas the incremental
volume was set at 10% of the total volume, placed every 10 years, for 100 years.
The total simulation length was 200 years with no additional nourishment placed
during the last 100 years. The large nourishment has the advantage of providing
more protection, but it also causes more consolidation, thereby reducing the barrier
dune elevation.
The simulations were conducted with a cross-section similar to barrier islands in
Louisiana: a barrier island with dune height = 1.5 m and width = 1.5 km, average
storm surge (including astronomic tide) equal to the zeroth-moment deepwater wave
height = 1.5 m, and average peak wave period equal to 5.6 s. The total volume of
sediment placed on the barrier was 2,500 m3 /m. For the incremental placement, 10%
of this volume (250 m3 /m) was placed on the barrier island every 10 years for 100
years. Typical nourishment volume densities in the USA are of the order 250 m3 /m
[Ref. 11, p. 23]; thus, this example is representative of a periodic US beach fill
project. The simulations were run for 200 years, as shown in Fig. 30.16.
August 17, 2009 11:4 9.75in x 6.5in b684-ch30 FA
6
C=Consolidation
450
L=Large Fill
I=Incremental Fill D=Dune Elevation
400
5 M=Migration
350
Dune Elevation, m
L,D
Consolidation
4 300
Migration, m
I,M 250
3
200
2 150
I,D
100
1 L,C
I,C 50
L,M
0 0
0 50 100 150 200
Elapsed Time, years
Fig. 30.16. Application of MCO for incremental (250 m3 /m every 10 years) and large-scale
(2,500 m3 /m at t = 0) beach nourishment.
References
1. J. R. Houston, The economic value of beaches — A 2002 update, Shore and Beach
70(1), 9–12 (2002).
2. R. G. Dean, Realistic economic benefits from beach nourishment, Twenty-first Int.
Conf. Coast. Eng., ASCE, Chapter 116, Malaga, Spain, June 1988, pp. 1558–1572.
3. Delft Hydraulics Laboratory, Manual on artificial beach nourishment, Rijkswaterstaat,
The Delft Hydraulics Laboratory Centre for Civil Engineering Research, Codes and
Specifications, Report No. 130 (1987), 195 pp.
4. R. G. Dean and C.-H. Yoo, Beach nourishment performance predictions, ASCE
J. Waterw. Port Coast. Ocean Eng. 118(6), 567–586 (1992).
5. D. K. Stauble and N. C. Kraus (eds.), Special volume on beach nourishment engi-
neering and management considerations, Proc. Coast. Zone Man., American Society
of Civil Engineers (1993).
August 17, 2009 11:4 9.75in x 6.5in b684-ch30 FA
Chapter 31
Nicholas C. Kraus
US Army Engineer Research and Development Center
Coastal and Hydraulics Laboratory, 3909 Halls Ferry Road
Vicksburg, MS 39180, USA
nicholas.c.kraus@.usace.army.mil
Tidal inlets are part of the coastal sediment-sharing system, and an inlet will
modify the nearshore and estuary morphology, as well as the up-drift and down-
drift beaches. Morphologic response to an inlet varies over several time and spatial
scales. This chapter discusses the inlet morphology and its related functional
design considerations that must balance navigation and shore-protection require-
ments. The first half of this chapter reviews the selected material on the mor-
phology of inlets and introduces empiric predictive expressions found useful for
engineering. The second half of the chapter concerns aspects of engineering of
tidal inlets.
31.1. Introduction
A tidal inlet is a short, narrow waterway connecting a bay, estuary, or similar body
of water with a larger water body such as a sea or ocean upon which the astronomic
tide acts. A tidal inlet is distinguished from other possible embayments and coastal
inlets in that the inlet channel is primarily maintained by the tidal current.1 Water
flow through a coastal inlet can be caused by the tide, wind,2 long-period seiching,3
and by river discharge. For water bodies connected to an ocean or large sea by an
inlet, the astronomic tide is typically the major forcing for water movement and
scouring of the inlet channel. In areas where the wind can be strong and tidal range
small, such as the coast of Texas, USA, wind-generated flow through an inlet can
often dominate the tidal signal, as can seasonal variations in water level at inlets
that experience moderate tide range.4 A tidal inlet can be in a natural state, or
its channel can be dredged as needed for navigation, for improving flushing, or
for keeping the inlet open. Navigable inlets are usually stabilized by one or, more
commonly, two jetties.
867
August 18, 2009 17:33 9.75in x 6.5in b684-ch31 FA
868 N. C. Kraus
This section reviews selected features of tidal inlet morphology. Hayes12 and
FitzGerald13 can be consulted for different overviews than presented here and for
many other references to the literature. The seminal article by Bruun and
Gerritsen14 is still relevant in establishing terminology and elucidating sediment
pathways at tidal inlets.
August 18, 2009 17:33 9.75in x 6.5in b684-ch31 FA
Fig. 31.1. Shinnecock Inlet, New York, April 22, 1997. The picture was taken during a time of
small waves, revealing the inlet morphology.
August 18, 2009 17:33 9.75in x 6.5in b684-ch31 FA
870 N. C. Kraus
Fig. 31.2. Definition sketch for tidal inlet morphology (modified from Ref. 15).
through or displaces the ebb shoal along the axis of the inlet. The island on the west
side of the flood shoal likely consists of material placed during early dredging of the
navigation channel. At Shinnecock Inlet, longshore transport is directed strongly to
the west, making the ebb shoal asymmetric.
The main morphologic features of a natural (nonengineered) tidal inlet are
sketched in Fig. 31.2. Of these, the inlet channel, ebb shoal, and flood shoal are typ-
ically of interest to navigation and to the integrity of the adjacent beaches. The ebb
shoal serves as a pathway for sediment bypassing naturally around the inlet, and it is
sometimes mined as a sand source for beach fill. At many inlets, the channel must be
maintained at greater depth through the ebb shoal (sometimes called the outer bar
or entrance bar) than in the inlet and bay because of breaking waves on the shoal,
which cause greater vertical excursion of vessels. The flood shoal often encroaches
on the inlet channel or is located where a waterway passes, and channels around
or through flood shoals must be dredged. The planform of the ebb shoal and flood
shoal can vary greatly according to the relative action of waves (littoral transport)
and tidal range. Coastal geologists refer to the ebb- and flood-tidal shoals as deltas
in analogy to the shape of river deltas. Here, the terminology shoal is employed to
emphasize the typical littoral, as opposed to riverine, provenance of shoals at tidal
inlets, which is also in accord with navigation usage.
Galvin16 classified inlet planform configurations according to relative strength
of longshore sediment transport rates (typically as a volume per year), in which QR
and QL , represent right- and left-directed transport, respectively, as viewed from
August 18, 2009 17:33 9.75in x 6.5in b684-ch31 FA
the inlet and facing the ocean (Fig. 31.3). The configurations were termed as over-
lapping offset, up-drift offset, down-drift offset, and negligible offset. Representative
locations of the main channel and the ebb (E) and flood (F) shoals are indicated.
An overlapping offset inlet forms by spit growth from the direction of strongly
dominant longshore sediment transport (QR in top drawing of Fig. 31.3). The inlet
channel becomes longer, unless the spit growth is contained in some way. Fire Island
Inlet, New York (Fig. 31.4; Ref. 17) is a clear example of an overlapping offset inlet.
Eventually, such a channel becomes so long that friction weakens the tidal flow, and
the spit grows toward shore to close the inlet. The inlet channel may not be fixed
by a jetty; in the case of Fire Island Inlet, the predominant transport was so strong
Fig. 31.3. Planforms of inlets (modified and expanded from Ref. 16). Lengths of arrows denote
relative strength of longshore sediment transport rate.
August 18, 2009 17:33 9.75in x 6.5in b684-ch31 FA
872 N. C. Kraus
Captree
Island
Sexton Island
Cedar Is. Oak Is.
Fig. 31.4. Fire Island Inlet, Long Island, New York. The inlet migrated 8 km east since 1827, and
a jetty placed on Democrat Point in 1939–1941 was impounded and bypassed within 10 years.
Shorelines oriented approximately east-west.
that sediment buried the relatively short jetty and continued spit development down
drift. The channel may cut through the ebb shoal, which tends to orient toward the
down-drift shoreline in a submerged extension of spit elongation.
If there is an unequal but adequate (available) up-drift source of sediment, the
inlet may be offset with only minor spit growth and not tend to exhibit strong spit
growth. Bypassing will occur through a relatively well-connected ebb-tidal shoal.
Shinnecock Inlet, New York, and Ocean City Inlet, Maryland, are examples of up-
drift offset inlets. If there is an inadequate or limited source of sediment from the
predominant direction, then a down-drift offset might occur such that the down-
drift side protrudes, in part because of attachment by or feeding from the ebb shoal
and its bypassing bars, and by local wave and transport reversal through wave
refraction.18 Thus, inlets can have a seaward down-drift offset even without jetties.
This is the mesotidal morphology described by Hayes.19 Finally, for Fig. 31.3, if the
longshore transport rate is balanced or weak, a natural inlet will tend to be wide
and consist of multiple migrating channels and shoals.
Hayes19 and Davis and Hayes20 characterized inlet planform morphology
according to the relative strength of tide and waves, as depicted in Fig. 31.5, with
tidal range serving as a surrogate for tidal prism (volume of water entering or leaving
an inlet in the corresponding half tidal cycle) or tidal current (which moves the
sediment).
Tide-dominated inlets tend to have larger ebb shoals that include channel margin
bars similar to dual jetties. Bypassing at tide-dominated inlets can be through tidal
bypassing, in which the sediment enters the channel on one side at flood tide and
a portion eventually returns to the opposite side at ebb tide (Fig. 31.6). Large
volumes of sand can also be added to the barrier segment on either side of the inlet
by major reorientation of the outer channel that isolates a portion of the ebb shoal
from strong tidal flow.21 Sand bodies can move onshore over the shallower portion
of the ebb shoal exposed to breaking waves. During storms, portions of the channel–
margin bars or other features of the ebb shoal may break off and migrate onshore.
August 18, 2009 17:33 9.75in x 6.5in b684-ch31 FA
tion
5 d ma
a te
w ) r for
in ie
(lo arr
Mean Tidal Range,m
m
Do ) ted fb
– high ina m it o
4 de ( m . li e
Ti Do prox (tid
e – Ap rgy
Tid d Ene ated)
e
3 Mix domin
y
Energ
Mixed inated)
d o m
2 (wave
ted
1 Wave Domina
0
0 1 2
Mean Wave Height, m
Tide Transitional;
Mixed
Wave Shoal
Main transport
pathway
Secondary
pathway
Fig. 31.6. Representative planform morphology and mode of bypassing of natural inlets,
depending on wave or tide dominance.
874 N. C. Kraus
Fig. 31.7. High-resolution survey of ebb-tidal shoal at Ocean City Inlet, Maryland, 2004,
exhibiting a clear concentric shoal characteristic of a wave-dominated inlet. Longshore transport
is from north to south.
Mixed-energy inlets share the features of each of the tide-dominated and wave-
dominated idealized end states (Fig. 31.6; Ref. 22). Anthony and Orford23 investi-
gated mixed-energy inlets on sediment-deficient coasts.
For wave-dominated and mixed-dominance inlets, the semi-circular ebb-tidal
shoal can be subdivided as depicted in Fig. 31.8 into the ebb shoal proper, bypassing
bars to each side of the ebb shoal proper, and attachment bars connecting the
bypassing bars to the shore.24 The ebb shoal was subdivided because the location
and size of the ebb shoal proper is related primarily to the tidal jet. In contrast,
the bypassing bars and attachment bars are controlled by sediment transport pro-
duced primarily by breaking waves. For Shinnecock Inlet (Fig. 31.1), the strong
predominant direction of longshore transport to the west pushes the up-drift
attachment bar toward the beach that is fully impounded directly adjacent to the
east jetty. The down-drift attachment bar is located about 1 km to the west of
the inlet.
August 18, 2009 17:33 9.75in x 6.5in b684-ch31 FA
Updrift
Attachment
Bar
Updrift
Bypassing Bar
it
Sp
in
l
ne
rg
Ebb Shoal
Ma
an
it
Ch
Proper
Sp
b
Eb
in
rg
Ma
od
Fl o
Downdrift
Bypassing Bar
Downdrift
Attachment Bar
Flood
Shoal
Fig. 31.8. Inlet entrance morphology, East Pass, FL, 1990, facing the Gulf of Mexico (adapted
from Ref. 25).
876 N. C. Kraus
Table 31.1. Inlet bypassing and classification of channel area cross-sectional stability (modified
from Bruun and Gerritsen14 and supplemental information).
acts to sweep the inlet sediment as opposed to the volume of sediment brought to the
inlet entrance by longshore transport during a year. Their observations led to the
classification of inlets according to mechanism of sand bypassing and stability of
the channel (Table 31.1). Implications for navigation over the ebb shoal or entrance
bar based on their classification are also noted in Table 31.1.
Another mechanism for natural sand bypassing at inlets is episodic bypassing,
by which a portion of the ebb-tidal shoal or down-drift bypassing bar detaches
from the main body and migrates to the down-drift shore.27–29 Episodic bypassing
initiates after the ebb shoal, typically of a transitional or wave-dominated inlet,
grows large and is disturbed by a storm. A large river flood can also discharge excess
sediment that is not in equilibrium with the prevailing typical tidal-river current
and waves, causing at least a portion of the new material to gradually migrate to
the down-drift shore under wave action. Kraus and Lin30 attribute an increased
rate of migration and tendency for closure of the San Bernard River Mouth, Texas,
to a large volume of sediment discharged from the Brazos River, located 5.6 km up
drift (to the north), during an exceptional flood in early 1992. Hands and Shepsis31
document periodic bifurcation and detachment of a large spit entering Willapa Bay,
Washington, in association with large El Niño that occurs about every seven years.
The detached portion of the spit supplied material to the ebb-shoal complex at this
wide, natural inlet. FitzGerald et al.21 posit nine conceptual mechanisms for natural
sand bypassing at tidal inlets, including those with jetties.
Table 31.2. Example empiric and theoretic equilibrium relationships for tidal inlet morphology.
P : tidal prism; A: area; V : volume; subscripts C, E, and F : channel, ebb shoal, and flood shoal,
respectively; C: empiric or derived coefficient; j, m, n, p, q: empiric or derived power; W : width of
inlet throat; M : gross longshore transport in a year.
empiric relations for tidal inlet morphologic asymmetries with implications for main-
tenance of navigation channels and sediment bypassing to the adjacent beaches.
878 N. C. Kraus
including the third harmonic), the maximum discharge Dm and tidal prism P are
related as:
T /2
2π
P = Dm sin t dt (31.2)
0 T
where T is the tidal period, and t is time. The integration gives:
π
Dm = P. (31.3)
T
Tidal prism can also be calculated as the product of the effective bay surface
area served by the subject inlet and the tidal range. It can also be obtained from
a computation of water discharge, as through a numeric model. By definition of a
discharge, the mean-maximum velocity Vmm is:
Dm
Vmm = (31.4)
AC
where AC is the minimum inlet channel cross-sectional area below mean sea level
(msl).
Although refinements have been made in empiric predictive equations relating
AC and P (see next section), it is convenient to consider the linear relation from
O’Brien34 :
AC = CP (31.5)
where AC is in m2 , P is in m3 , and C = 6.6 × 10−5 m−1 . Then34 :
π
Vmm = . (31.6)
CT
For a pure semi-diurnal inlet, T = 12 h, 25 min = 44,712 s. Equation (31.6) yields
Vmm = 1.06 m/s, in agreement with empirical observations. For a tide that is pri-
marily diurnal, the tidal period is 89,424 s, giving Vmm = 0.53 m/s. The implication
is that an inlet in a diurnal tidal setting may require a smaller mean-maximum
tidal velocity to maintain channel cross-sectional area stability as compared to inlets
in a semi-diurnal setting, which are more common because of the prevalent semi-
diurnal tide.
Jarrett39 compiled annual average maximum ebb velocities for 70 inlets on the
Atlantic coast, 38 inlets on the Gulf coast, and 28 inlets on the Pacific coast of the
USA. The compilation gives mean-maximum velocities of 1.17, 0.75, and 1.06 m/s,
respectively, for the inlet groups. Most inlets on the Gulf coast experience a diurnal
tide and a reduced longshore transport rate (see next section) as compared to the
Atlantic and Pacific coasts. The trend thus agrees qualitatively with the simple
calculation given above.
AC = CP n . (31.7)
Among other observations, Jarrett39 noted that the smaller waves on the Gulf of
Mexico coast relative to those on the Pacific coast and most of the Atlantic Ocean
coast would produce smaller rates of longshore sediment transport. Kraus45 derived
a form of Eq. (31.7) by consideration of a balance of sand transport by the channel-
clearing inlet current and channel infilling by longshore sand transport, resulting in
Table 31.3. Inlet-area and spring or diurnal tidal prism regression values found by Jarrett39 for
AC = CP n (area units of m2 , prism units of m3 ) tidal inlets on US coasts.
Location C n C n C n
All inlets 1.576 × 10−4 0.95 3.797 × 10−5 1.03 7.490 × 10−4 0.86
Atlantic coast 3.039 × 10−5 1.05 2.261 × 10−5 1.07 1.584 × 10−4 0.95
Gulf coast 9.311 × 10−4 0.84 6.992 × 10−4 0.86 Insuff. data Insuff. data
Pacific coast 2.833 × 10−4 0.91 8.950 × 10−6 1.10 1.015 × 10−3 0.85
August 18, 2009 17:33 9.75in x 6.5in b684-ch31 FA
880 N. C. Kraus
n = 0.9, and:
0.3
4/3
απ3 N 2 We
C= (31.8)
Qg T 3
in which α is the transport-related coefficient on order of 0.1–1, N is the Mannings
coefficient, We is the width of the inlet at equilibrium, and Qg is the gross longshore
transport rate arriving at the inlet. The prediction for C indicates that this quantity
increases for wider inlets and for smaller longshore transport rates (sheltered coasts
or coasts with small transport rates in general, such as on the west coast of Florida).
Also, the inverse dependence on T suggests that inlets located on coasts having a
predominantly semi-diurnal tide should be more stable than those experiencing a
diurnal tide, all other conditions being equal. This prediction is contrary to that
given in the preceding section.
The data points for all inlets tabulated by Jarrett39 are plotted in Fig. 31.9. The
equation at the top of the figure is that given by Jarrett, and the one on the bottom
of the figure was computed by the author. There is a small difference, mainly because
several points appear to have been omitted by Jarrett in his correlation calculation.
The solid lines are the predictive equations which overlap, and the dashed lines give
95% confidence limits.
Byrne et al.,54 Riedel and Gourlay,35 and Hume and Herdendorf36 studied the
inlet channel stability on sheltered (protected) coasts and demonstrated that larger
values of the empiric coefficient C (in accord with Ref. 32) and smaller values of
n in Eq. (31.7) apply to coasts with limited littoral transport. The aforementioned
three studies also indicate that the mean-maximum velocity required to maintain
stability of the inlet channel is less (reaching approximately one-third less) than the
typical 1 m/s required to maintain a channel on an exposed coast.
6
10
US, All Inlets
4
10
3
10 -4 0.954
A = 1.528 * 10 *P
R2=0.925
2
10 5 6 7 8 9 10
10 10 10 10 10 10
3
Tidal Prism, m
Fig. 31.9. Data on inlet channel area and spring tidal prisms (after Ref. 39).
August 18, 2009 17:33 9.75in x 6.5in b684-ch31 FA
7
10
NZ Rivers
Inlets, NZ
6
10
NZ Harbors
5
2
10
MSL Area, m
NZ Bays
4
10
US, All Inlets
US & NE Coast of NZ Northeast Coast of North
3
10 Island, New Zealand
Rivers, New Zealand
Inlets, New Zealand
2
10
Bays, New Zealand
Fig. 31.10. Trends from Jarrett39 for US inlets and New Zealand inlet categories as compiled by
Hume and Herdendorf.36
Trend lines from the data of Jarrett39 and of Hume and Herdendorf36 are plotted
in Fig. 31.10. For the same value of tidal prism, channel cross-sectional areas for
New Zealand entrances in sheltered areas (bays, protected inlets) tend to plot higher
than those for unprotected or unsheltered entrances. Entrances on the (unsheltered)
northeast coast of New Zealand plot on top of the all-inlet trend line for US inlets.
882 N. C. Kraus
lobe of the large ebb shoal at Grays Harbor, Washington, after long jetties were
built in the late 1890s to the early 1900s. The collapsing shoal increased beach
width to the north of the inlet by more than a kilometer. The ebb shoal in front
of the ebb jet migrated seaward to the edge of the continental shelf as a result
of the jetty construction. Collapse of an ebb shoal can give a false impression as
to the net direction of longshore transport. For example, at Grays Harbor the
regional net direction of transport is to the north,57 yet the wide beach to the north
created by the collapsed ebb shoal suggests (incorrectly) that the net transport is to
the south.
Once an ebb shoal develops, if it is not translated too far offshore by the ebb
jet; it provides an efficient pathway for sand to bypass around the inlet during
times of larger waves that can break on the shoal. Therefore, the volume of an ebb
shoal will approach an equilibrium value, after which the sediment transported to
the ebb shoal by the wave-and wind-induced longshore current will be bypassed to
the down-drift beach or transported to the channel and then to the flood shoal.
It was proposed several times, but to the author’s knowledge never executed, to
artificially build an ebb shoal at the location of a newly cut inlet so as to more
quickly reestablish natural sand bypassing.
Walton and Adams40 examined the volumes of ebb shoals as a function of
tidal prism by analyzing 44 inlets on the Atlantic, Gulf, and Pacific coasts of
the USA that were judged to be near or at equilibrium. They estimated the
volume from nautical charts and other surveys by assuming that the ebb shoal
formed on top of parallel-depth contours that can be estimated from those far
from the inlet, a procedure developed by Dean and Walton.58 They further clas-
sified the inlets according to slope of the continental shelf (location) and wave
climate. The volume of the ebb shoal for inlets on the mildly exposed coasts was
larger than those on highly exposed coasts for the same tidal prism. Walton and
Adams40 hypothesized that larger waves would tend to drive the sediment toward
the shore.
Others have confirmed the essential finding of Walton and Adams40 that the
volume of the ebb-tidal shoal is related to the tidal prism. Marino and Mehta59
examined 18 inlets on the east coast of Florida. They found that tidal prism was
a leading parameter controlling ebb-shoal volume, with the ratio of inlet width to
depth being a secondary factor. Hicks and Hume42 conducted a similar analysis for
17 inlets in New Zealand and confirmed the overall results of Walton and Adams.
Hicks and Hume42 also considered the angle between the ebb jet and the shoreline
in their correlations.
For the 44 inlets, Walton and Adams40 found:
VE = CE P 1.23 (31.9)
10
10
Tidal Prism-Outer Bar Storage Relationship
3
9
10
VE = 2.121 * 10-2 P1.1673
8
10
7
10
Exposure
6
10 High
Moderate
Mild
5
10 5 6 7 8 9 10
10 10 10 10 10 10
3
Tidal Prism, m
Fig. 31.11. Volume of ebb-tidal shoal as function of tidal prism, all coasts (after Ref. 40).
884 N. C. Kraus
results pertaining to Florida conditions are typified by mild waves, and small tidal
prisms, and hard limestone bottom at many Atlantic coast inlets. In the following,
the first subscript, F denotes flood shoal, and the second subscripts N , F , and T
denote near field, far field, and total, respectively.
In these equations, volume and prism are expressed in units of m3 . It is possible for
these equations to give a total volume less than the sum or the near field and far
field, demonstrating variability in the data available to the study.
hE = 0.5 hC . (31.12)
The equation was valid irrespective of whether the river flow was or was not trained
by structures. Floyd38 concluded that it is not possible to increase the depth over
the ebb shoal with jetties, and that the greatest depth across a bar can only be
obtained by increasing the depth in the channel, as perhaps by dredging.
Buonaiuto and Kraus48 analyzed the bathymetry of 18 inlets around the coast
of the USA to determine a predictive expression for the minimum depth over crest
hCr of the ebb shoal. It was reasoned that, because both incident waves and the
tidal prism are expected to be controlling independent variables for ebb shoal devel-
opment, a parameter combining both average annual significant wave height HS and
August 18, 2009 17:33 9.75in x 6.5in b684-ch31 FA
12
Best Fit
Grays H.
10 95% Confidence Limit
Columbia R.
8
hc, m (MLLW)
Willapa Bay
St. Marys Ent.
6 Lake Worth Inlet
Charleston Harbor
4 Moriches Inlet
New Pass
1995
2 Shinnecock Inlet
New Pass
1997
0
0 50 100 150 200 250
1/4
(H sP) ,m
Fig. 31.12. Minimum depth over crest of an ebb shoal and product of average annual significant
wave height and tidal prism as (HS P )1/4 .
tidal prism would provide the best predictive capability. The parameter (HS P )1/4 ,
which has units of meters, was devised to represent the combination.
Correlation with the data determined the following predictive relations (metric
units):
Morphologic response to an inlet typically has a long timescale and great spatial
extent. Sediment pathways are shared with the adjacent beaches and estuaries in
a complex hydrodynamic environment. Therefore, consequences of the presence of
inlets, both natural and engineered, can be subtle and far-ranging. It is also difficult
to transfer experience and monitoring results among inlets because of different bal-
ances of kinds and strengths of the acting processes and conditions, of which tidal
range, wave height and direction, wind, river flow if any, bay or estuary surface
August 18, 2009 17:33 9.75in x 6.5in b684-ch31 FA
886 N. C. Kraus
(i) Maintain inlet stability (minimal dredging) while assuring sediment bypassing
continues to the adjacent beaches.
(ii) Provide inlet navigability (strong tidal current to scour the channel) while
assuring navigation reliability and safety (avoiding an excessively strong inlet
current).
The first conflict concerns the design objective of protecting the channel from
excessive sediment shoaling so that navigable depth is maintained for the longest
possible time. Traditionally, this was done by building long jetties to block infil-
tration by sediment moving alongshore. Jetties also promote a stable location and
orientation of the channel, and partially shelter vessels from waves in the surf zone.
However, long jetties will interrupt longshore transport and deprive the down-drift
beach and, perhaps, the estuary of sediment. In the USA, older long jetties have
been deteriorating, and it is a significant question as to whether they should be
rehabilitated.
The second conflict concerns the design requirement of promoting a tidal current
of adequate strength to contribute to maintenance of channel depth by scour. On
the other hand, safe navigation for smaller vessels requires moderate ebb current so
that steep waves that pose a navigation hazard are not created by the wave–current
interaction. Numerous considerations enter in these conflicting requirements, and
some are discussed below.
In construction of a new inlet, the ebb and flood shoals can be assumed to be
formed by sand transported from the adjacent beaches. Although this will be a
long-term process, the volume removed from the beaches should be considered in
the sediment budget.
For a relocated inlet, abandonment of the old ebb shoal can be an advantage
in nourishing the down-drift beach and factored into the sediment budget. Kana
and Mason62 and Kana and McKee63 discuss the benefit for shore preservation of
moving Captain Sams Inlet, South Carolina, a pioneering and successful effort. Vila-
Concejo et al.64 discuss morphologic change for relocation of an inlet in Portugal and
also review the literature. Kraus et al.17 perform a regional sediment management
study and evaluated tradeoffs of relocating Fire Island Inlet, New York (Fig. 31.4).
Cialone et al.65 discuss response of the flood shoal at Barnegat Inlet, New Jersey, to
construction of a south jetty almost parallel to the existing north jetty and dredging
of a shoal that had formed in the entrance of the previous arrowhead configuration
jetties. As a response to this and other engineering actions, the historic growth of
the flood shoal halted.
888 N. C. Kraus
Fig. 31.13. Inferred sediment pathways and chronically eroding, sediment-isolated down-drift
beach (area “3”), Shinnecock inlet (from Ref. 73).
many inlets near-field recession of the shoreline is chronic and requires special mea-
sures of shore protection.73 This erosion may thin barrier islands to the point that
breaching adjacent to the inlet becomes a concern. The far-field shoreline response
can extend many kilometers beyond the inlet. The existence and extent of the
shoreline adjustment depend in great part on (1) length of jetties, (2) placement fre-
quency and location of material dredged from the channel or bypassed mechanically,
(3) balance of net and gross longshore sediment transport rates, and (4) elapsed
time after jetty construction, among many factors. Shoreline-change numeric models
can give an estimate of adjustment of the shoreline to be expected. Such modeling
must include the anticipated configuration of the ebb-tidal shoal in the wave trans-
formation.
Walther and Douglas74 document recovery of the ebb shoal at Boca Raton
Inlet, Florida, which was mined as a source for beach nourishment. Buttolph
et al.75 observationally and numerically investigated offshore migration of the ebb
shoal at Ocean City Inlet, Maryland, that resulted from raising the outer portion
of the south jetty. This study was aided by high-resolution bathymetry surveys
(Fig. 31.7).
890 N. C. Kraus
equilibrium volume VAe , all sand supplied to it by the bar is bypassed to the
beach. The model thus requires values of the input and output rates of transport
from each morphologic feature and their respective equilibrium volumes. Compli-
cated sediment pathways, multiple connections, and time-varying transport can be
represented.79–81
Here, a simple analytic version of the model (closed-form solution) is reviewed.24
In applications, the governing equations are solved numerically and can treat
complex, time-dependent conditions. The continuity equation governing change of
the volume VE of the ebb-tidal shoal is:
dVE
= Qin − (QE )out (31.14)
dt
where t is the time and (QE )out is the rate of sand leaving (going out of) the ebb
shoal. The input transport rate is assumed known; for example, it could be the
right-directed transport QR or the left-directed transport QL .
The remaining unknown is the output transport rate. The reservoir model
assumption is that the output rate is proportional to the input rate times the volume
of sand in the beaker divided by the equilibrium volume. Therefore, rate of sand
leaving or bypassing the ebb shoal (QE )out , is specified as:
VE
(QE )out = Qin (31.15)
VEe
in which Qin is taken to be constant here, although this is not necessary in a numeric
model. For the present situation, Eqs. (31.14) and (31.15) give:
dVE VE
= Qin 1 − . (31.16)
dt VEe
August 18, 2009 17:33 9.75in x 6.5in b684-ch31 FA
892 N. C. Kraus
With initial condition VE (0) = 0, the solution of Eq. (31.16) is exponential growth:
VE = VEe 1 − e−αt (31.17)
in which:
Qin
α= . (31.18)
VEe
The quantity 1/α is a characteristic timescale for growth of the ebb shoal. For
example, if Qin = 1 × 105 m3 /year and VEe = 2 × 106 m3 , which are representative
values for a small inlet on a moderate-wave coast, then 1/α = 20 years. The shoal
would be predicted to reach 50% and 95% of its equilibrium volume after 14 and
60 years, respectively, under the constant imposed transport rate. The parameter α
is essentially the inverse of the r-parameter introduced by Bruun and Gerritsen14
[Eq. (31.1)].
This simple situation for constant input longshore transport magnitude and
direction gives the volume of the bypassing bar as:
Qin VE
VB = VBe 1 − e−βt , β= , t = t − (31.19)
VBe Qin
The coefficients 1/β and 1/γ function similarly to 1/α in representing timescales
for the bypassing bar and attachment bar, respectively. The quantities t and t in
Eqs. (31.19) and (31.20) are lag times that account for a delay in development of
the respective features. After formation of an inlet, a certain time is required for the
bypassing bar to receive a significant amount of sand from the shoal and a longer
time for the attachment bar or beach to receive sand as it moves around the inlet
from the up-drift side (delays).
The following are obtained for the bypassing rate of the bar (QB )out , which
is equal to the input of the attachment (QA )in , and the bypassing rate of the
attachment (QA )out , which is the input to the beach, (Qbeach )in :
VE VB
(QB )out = Qin = (QA )in (31.21)
VEe VBe
VE VB VA
(QA )out = Qin = (Qbeach )in . (31.22)
VEe VBe VAe
The rate (QA )out describes the amount of sand reaching the down-drift beach as a
function of time and is a central quantity entering beach nourishment and shore-
protection design near inlets.
Figure 31.15 illustrates the sediment pathways developed for Sebastian Inlet,
Florida, accounting for seasonality in wave direction and longshore sediment
August 18, 2009 17:33 9.75in x 6.5in b684-ch31 FA
Fig. 31.15. Sediment pathways conceptualized for Sebastian Inlet, FL (E: ebb shoal; B: bypass
bar; A: attachment bar; C: channel; T: sand trap; F: flood shoal; Y: bay; SS : south fillet; IN : north
fillet; IS : south fillet; and O: offshore loss).
transport.81 The inlet opens to the Atlantic Ocean on a north–south trending coast.
Predictions shown in Fig. 31.16 agree with most measurements of flood and ebb
shoal volumes at Sebastian Inlet, but underestimate volumes determined for the late
1980s. The calculations and measurements match for the past decade with respect
to the overall consequence of sand bypassing. The simulation shown in Fig. 31.16
includes sediment volume removed from the sand trap to represent sand-bypassing
projects conducted between 1972 and 1999. Zarillo and Brehin82 give an update on
the status of Sebastian Inlet and its morphologic evolution.
This chapter concerns morphologic change around and engineering of tidal inlets.
The hydrodynamics at a tidal inlet, the water movement that transports sediment
and determines inlet morphologic forms, could not be discussed due to space lim-
itations. Also, advanced numeric modeling of inlets was not covered. Here, some
elements of these subjects are presented for completeness.
The hydrodynamics of inlets is fascinating, for which much mathematic ele-
gance has been devoted. Keulegan83 developed the basic one-dimensional equation
of motion that is used today. Seabergh84 investigated some predictions of the
August 18, 2009 17:33 9.75in x 6.5in b684-ch31 FA
894 N. C. Kraus
Fig. 31.16. Inlet Reservoir Model simulation of sediment volumes at Sebastian Inlet, Florida,
1950–2050. Solid symbols indicate shoal volumes estimated from topographic data and analysis of
aerial images. Solid arrows indicate sand bypass events.
Keulegan approach for inlet channel stability. The reader is directed to Chap. 13 of
Dean and Dalrymple85 and to Seabergh86 for thorough reviews of simple tidal inlet
hydraulics that can be of great aid in understanding and engineering design. Seelig
and Sorensen87 demonstrate the utility of such an approach.
First-order analysis of inlet stability rests on the important “Escoffier stability
curve”37 that depends on a calculation of the current through an inlet. Seabergh and
Kraus88 discuss properties of the Escoffier stability curve and provide a desk-top PC
program to calculate it. The program is based on the analytic solution for an inlet
current given by DiLorenzo,89 which represents overtides — higher harmonics of the
dominant tide component generated through tidal wave shoaling (see Ref. 90). The
Escoffier stability curve also requires a predictive relation for the minimum channel
area for a given tidal prism, as discussed in Sec. 31.2.5.
Sediment pathways and morphology change around an idealized dual-jetty inlet
similar to Shinnecock Inlet were investigated by Militello and Kraus91 with a sophis-
ticated numeric model. This work demonstrates significant differences in sediment
transport depending on the wave climate as either typical or a storm condition.
Recently, Fortunato and Oliveira92 report an interesting two-dimensional numeric
model application investigating inlet stability and minimization of channel dredging.
Consideration of nonlinear processes associated with tidal flats is included. Such
works are among many demonstrating the engineering utility of numeric models of
inlet hydrodynamics and morphology change.
August 18, 2009 17:33 9.75in x 6.5in b684-ch31 FA
In considering options for design or maintenance of a navigable tidal inlet, two con-
tradictory requirements of inlets must be balanced or reconciled in the engineering
design. These are (1) maintaining inlet stability while assuring natural functioning
of sediment bypassing around the inlet, and (2) providing inlet navigability while
promoting safe navigation. Requirement 1 implies inlet morphology retains equi-
librium plan form and depths over which sediment moves in an efficient way, whereas
Requirement 2 implies that inlets must be dredged to a necessary depth in support
of navigation. A channel deeper than the natural channel depth will intercept
more of the sediment moving toward it, which must subsequently be dredged. Sed-
iment may be jetted farther offshore by the constraining jetties, depriving the
beaches of that material or delaying its arrival. If the inlet has a shallow channel,
sediment can cross or bypass easily, but the depth may not be adequate for
navigation.
Williams et al.67 describe the design and functioning of a new inlet, Packery
Channel on the Texas coast, for which monitoring during its first three years has
indicated no need to dredge and no significant negative response of the adjacent
beaches. The inlet was designed with awareness of many of the considerations
described in this chapter,93 in particular that the jetties not intercept all sand
moving alongshore and that the hydraulic efficiency (entrance width to depth ratio)
be less than 100.39
The main issues in sediment management are interruption of the littoral drift
by the inlet jetties; creation, growth, and mining of ebb- and flood-tidal shoals; and
resultant changes in position of the shoreline, which may advance seaward on the
up-drift side and recede on the down-drift side. Optimal placement of beach-quality
material on the adjacent shores that is removed from the channel during new-
work dredging (original dredging) and maintenance dredging operations, as well as
mechanical bypassing of littoral material that is blocked by the up-drift jetty, are
also central elements of a sediment-management plan. Dean61 has discussed such
processes and associated policies. Inlet design and sediment management consid-
erations are therefore linked through interruption of the littoral drift, dredging of
an inlet, increase in tidal current through the dredged channel, and the water and
sediment circulation around the inlet.
Larger tidal inlets can evolve over hundreds of years, indicating that regional
responses must be appreciated or anticipated in engineering design. Kraus et al.17
took a regional sediment management approach in hypothetic relocation of Fire
Island Inlet, New York. Among several aspects, ebb-shoal collapse as a form of beach
nourishment was found to yield a large benefit for a chronically eroding down-drift
beach. Regional applications will typically involve multiple tidal inlets to the same
or connecting bays. Changes in one inlet can cause a response in the others. Batten
et al.94 made a morphologic study documenting the decrease in size of Pass Cavallo,
the natural tidal inlet to Matagorda Bay, Texas. This inlet is estimated to be about
2,600 years old and has been at the same location for the past 200 years. Opening
of the deep-draft Matagorda Ship Channel to the bay in the early 1960s “captured
the tidal prism”, causing a loss of prism through Pass Cavallo and reduction in size
August 18, 2009 17:33 9.75in x 6.5in b684-ch31 FA
896 N. C. Kraus
of the ebb-tidal shoal. Seabergh95 investigated the stability of two inlets serving
the same bay system in Guatemala, taking an engineering approach with Escoffier
stability diagrams.
The science and engineering of tidal inlets are a challenge, and it is hoped that
this chapter will be of some small assistance to those interested in this complex
coastal environment.
Acknowledgments
This work was performed as an activity of the Inlet Morphology and Channels
work unit of the Coastal Inlets Research Program administered by Headquarters,
US Army Corps of Engineers. Permission was granted to publish this information.
I am indebted to Julie Dean Rosati, William Seabergh, and Dr. Gary Zarillo for
thoughtful and helpful reviews.
References
15. R. A. Davis and D. M. FitzGerald, Beaches and Coasts (Blackwell Science, 2004),
p. 419.
16. C. J. Galvin, Jr., Wave climate and coastal processes, Water Environments and Human
Needs, ed. T. Ippen (Parsons Lab. for Water Res. and Hydrodynamics, MIT, 1971),
pp. 28–78.
17. N. C. Kraus, G. A. Zarillo and J. F. Tavolaro, Hypothetical relocation of Fire Island
Inlet, New York, Proc. Coast. Sediments’03, World Sci., CD-ROM (2003), 14 pp.
18. M. O. Hayes, V. Goldsmith and C. H. Hobbs, III, Offset coastal inlets, Proc. 12th
Coast. Eng. Conf., ASCE (1970), pp. 1187–1200.
19. M. O. Hayes, Barrier island morphology as a function of tidal and wave regime, Barrier
Islands, ed. S. P. Leatherman (Academic Press, 1979), pp. 1–28.
20. R. A. Davis, Jr. and M. O. Hayes, What is a wave dominated coast? Mar. Geol. 60,
313–329 (1984).
21. D. M. FitzGerald, N. C. Kraus and E. B. Hands, Natural mechanisms of sediment
bypassing at tidal inlets, ERDC/CHL CHETN-IV-30, U.S. Army Engineer Research
and Development Center, Vicksburg, MS (2001).
22. D. M. FitzGerald, Sediment bypassing at mixed energy tidal inlets, Proc. 18th Coast.
Eng. Conf., ASCE (1982), pp. 1094–1118.
23. E. J. Anthony and J. D. Orford, Between wave- and tide-dominated coasts: The middle
ground revisited, J. Coast. Res. SI36, 8–15 (2002).
24. N. C. Kraus, Reservoir model of ebb-tidal shoal evolution and sand bypassing,
J. Waterw. Port Coast. Ocean Eng. 126(3), 305–313 (2000).
25. E. E. Carr and N. C. Kraus, Morphologic asymmetries at entrances to tidal inlets,
ERDC/CHL CHETN-IV-33, U.S. Army Eng. Res. and Develop. Center, Vicksburg,
MS (2001), http://chl.wes.army.mil/library/publications/chetn.
26. P. Bruun, Stability of Tidal Inlets (North Holland Pub. Co., Amsterdam, 1960), 123 pp.
27. T. W. Kana, M. L. Williams and D. Stevens, Managing shoreline changes in the
presence of nearshore shoal migration and attachment, Proc. Coast. Zone’85, ASCE
(1985), pp. 1277–1294.
28. D. M. FitzGerald, Shoreline erosional-depositional processes associated with tidal
inlets, Hydrodynamics and Sediment Dynamics of Tidal Inlets, eds. D. G. Aubrey and
L. Weishar (Springer, 1988), pp. 186–225.
29. D. J. Gaudiano and T. W. Kana, Shoal bypassing in South Carolina tidal inlets:
Geomorphic variables and empirical predictions for nine mesoscale inlets, J. Coast.
Res. 17(2), 280–291 (2000).
30. N. C. Kraus and L. Lin, Coastal processes study of the San Bernard River Mouth,
Texas: Stability and maintenance of the mouth, TR ERDC/CHL-02-10, U.S. Army
Engineer Research and Develop Center, Coastal and Hydraulics Lab., Vicksburg, MS
(2002).
31. E. B. Hands and V. Shepsis, Cyclic channel movement at the entrance to Willapa
Bay, Washington, USA, Proc. Coast. Sediments’99, ASCE (1999), pp. 1522–1536.
32. L. J. LeConte, Discussion on river and harbor outlets, notes on the improvement
of river and harbor outlets in the United States, Paper No. 1009 by D. A. Watts,
Transactions ASCE 55, 306–308 (1905).
33. M. P. O’Brien, Estuary and tidal prisms related to entrance areas, Civil Eng. 1(8),
738–739 (1931).
34. M. P. O’Brien, Equilibrium flow areas of inlets on sandy coasts, J. Waterw. Port.
Harbor. Div. 95(WW1), 43–52 (1969).
35. H. P. Riedel and M. R. Gourlay, Inlets/estuaries discharging into sheltered waters,
Proc. 17th Coast. Eng. Conf., ASCE (1980), pp. 2550–2562.
August 18, 2009 17:33 9.75in x 6.5in b684-ch31 FA
898 N. C. Kraus
59. J. N. Marino and A. J. Mehta, Inlet ebb tide shoals related to coastal parameters,
Proc. Coast. Sediments 87, ASCE (1987), pp. 1608–1622.
60. E. E. Carr de Betts and A. J. Mehta, An assessment of inlet flood deltas in Florida,
Proc. Coastal Dynamics ’01, ASCE (2001), pp. 252–262.
61. R. G. Dean, Sediment interactions at modified coastal inlets: Processes and policies,
Hydrodynamics and Sediment Dynamics of Tidal Inlets, eds. D. G. Aubrey and
L. Weishar (Springer, 1988), pp. 412–439.
62. T. W. Kana and J. E. Mason, Evolution of an ebb-tidal delta after an inlet relo-
cation, Hydrodynamics and Sediment Dynamics of Tidal Inlets, eds. D. G. Aubrey
and L. Weishar, Lecture Notes on Coastal and Estuarine Studies 29 (Springer, 1988),
pp. 382–411.
63. T. W. Kana and P. A. McKee, Relocation of Captain Sams inlet — 20 years later,
Proc. Coast. Sediments’03, World Sci., CD-ROM (2003), 12 pp.
64. A. Vila-Concejo, O. Ferreira, B. D. Morris, A. Matias and J. M. A. Dias, Lessons from
inlet relocation: Examples from southern Portugal, Coast. Eng. 51, 967–990 (2004).
65. M. A. Cialone, W. C. Seabergh and K. D. Watson, Flood shoal response to inlet
modifications at Barnegat Inlet, New Jersey, Proc. Coast. Sediments’99, ASCE (2003),
pp. 1434–1449.
66. P. D. Komar, J. R. Lizarraga-Arciniega and T. A. Terich, Oregon coast shoreline
changes due to jetties, J. Waterw. Harbor. Coast. Eng. Div. 102(WW1), 13–30 (1976).
67. D. D. Williams, N. C. Kraus and C. M. Anderson, Morphologic response to a new inlet,
Packery Channel, Corpus Christi, Texas, Proc. Coast. Sediments’07, ASCE (2007),
pp. 1529–1542.
68. C. G. Creed, E. J. Olsen and K. R. Bodge, Performance of an interim sand-tightening
measure at an inlet jetty, Proc. 7th Natl. Conf. Beach Preserv. Tech., FSBPA (1994),
pp. 374–388.
69. R. G. Dean, A framework for sediment management practices at jettied inlets, Proc.
Coast. Sediment’03, World Sci., CD-ROM (2003), 13 pp.
70. R. A. Dalrymple, An extended one-line model for jettied inlets, Proc. Coast.
Sediments’03, World Sci., CD-ROM (2003), 12 pp.
71. P. Bruun, The development of downdrift erosion, J. Coast. Res. 11(4), 1242–1257
(1995).
72. P. Bruun, Bypassing at littoral drift barriers, Encyclopedia of Coastal Science,
ed. M. L. Schwartz (Springer, 2005), pp. 210–215.
73. H. Hanson and N. C. Kraus, Chronic beach erosion adjacent to inlets and remedi-
ation by composite (T-head) groins, ERDC/CHL CHETN IV-36, U.S. Army Engineer
Research and Development Center, Vicksburg, MS (2001).
74. M. P. Walther and B. D. Douglas, Ebb shoal borrow area recovery, J. Coast. Res.
SI18, 221–223 (1993).
75. A. M. Buttolph, W. G. Grosskopf, G. P. Bass and N. C. Kraus, Natural sand bypassing
and response of ebb shoal to jetty rehabilitation, Proc. 30th Coast. Eng. Conf., Ocean
City Inlet, Maryland, USA, World Sci. (2007), pp. 3344–3356.
76. W. C. Seabergh and N. C. Kraus, Progress in management of sediment bypassing
at coastal inlets: Natural bypassing, weir jetties, jetty spurs, and engineering aids in
design, Coast. Eng. J. 45(4), 533–563 (2003).
77. M. J. F. Stive, M. Capobianco, Z. B. Wang, P. Ruol and M. C. Buijsman, Morpho-
dynamics of a tidal lagoon and the adjacent coast, Physics of Estuaries and Coastal
Seas, eds. J. Dronkers and M. Scheffers (Balkema, 1998), pp. 397–407.
August 18, 2009 17:33 9.75in x 6.5in b684-ch31 FA
900 N. C. Kraus
78. K. M. Erickson, N. C. Kraus and E. E. Carr, Circulation change and ebb shoal
development following relocation of Mason Inlet, North Carolina, Proc. Coastal
Sediments’03, World Sci., CD-ROM (2003), 13 pp.
79. A. Militello and N. C. Kraus, Shinnecock Inlet, New York, site investigation, Report 4,
evaluation of flood and ebb shoal sediment source alternatives for the west of
Shinnecock interim project, New York, TR CHL TR-98-32, Coastal Inlets
Research Program, U.S. Army Research and Develop. Center, Coastal and Hydraulics
Laboratory, Vicksburg, MS (2001).
80. M. A. Dabees and N. C. Kraus, General methodology for inlet reservoir model analysis
of sand management near tidal inlets, Proc. Coast. Dynamics 05, ASCE, CD-ROM
(2006), 14 pp.
81. G. A. Zarillo, N. C. Kraus and R. K. Hoeke, Mophologic analysis of Sebastian Inlet,
Florida: Enhancements to the tidal inlet reservoir model, Proc. Coast. Sediments’03,
World Sci., CD-ROM (2003), 14 pp.
82. G. A. Zarillo and F. G. A. Brehin, Hydrodynamic and morphologic modeling at
Sebastian Inlet, FL, Proc. Coast. Sediments’07, ASCE (2007), pp. 1297–1310.
83. G. H. Keulegan, Tidal flows in entrances: Water level fluctuations of basins in com-
munication with seas, Committee on Tidal Hydraulics TB-14, U.S. Army Engineer
Waterways Experiment Station, Vicksburg, MS (1967).
84. W. C. Seabergh, Long-term coastal inlet channel area stability, Proc. Coast.
Sediments’03, World Sci., CD-ROM (2003), 8 pp.
85. R. G. Dean and R. A. Dalrymple, Coastal Processes with Engineering Applications
(Cambridge University Press, 2002), 475 pp.
86. W. C. Seabergh, Hydrodynamics of tidal inlets, Coastal Engineering Manual, Part 2,
Chap 6, Engineer Manual 1110-2-1100, U.S. Army Corps of Engineers, Washington,
DC (2002).
87. W. N. Seelig and R. M. Sorensen, Numerical model investigation of selected tidal inlet-
bay system characteristics, Proc. 16th Coastal Eng. Conf., ASCE (1978), pp. 1302–
1319.
88. W. C. Seabergh and N. C. Kraus, PC program for coastal inlet stability analysis using
Escoffier method, CETN-II-11, U.S. Army Engineer Waterways Experiment Station,
Coastal and Hydraulics Laboratory, Vicksburg, MS (1997).
89. J. L. DiLorenzo, The overtide and filtering response of small inlet/bay systems,
Hydrodynamics and Sediment Dynamics of Tidal Inlets, eds. D. G. Aubrey and
L. Weishar (Springer, 1988), pp. 24–53.
90. C. T. Friedrichs and D. G. Aubrey, Non-linear tidal distortion in shallow well-mixed
estuaries: A synthesis, Est. Coast. Shelf Sci. 27(5), 521–545 (1988).
91. A. Militello and N. C. Kraus, Numerical simulation of sediment pathways at an ide-
alized inlet and ebb shoal, Proc. Coast. Sediments’03, World Sci., CD-ROM (2003),
14 pp.
92. A. B. Fortunato and A. Oliveira, Case study: Promoting the stability of the Óbidos
Lagoon inlet, J. Hydraul. Eng. 133(7), 816–824 (2007).
93. N. C. Kraus and D. J. Heilman, Packery Channel feasibility study: Inlet functional
design and sand management, Report 1 of a two-part series. TR TAMU-CC-CBI-96-
06, Conrad Blucher Institute for Surveying and Sci., Texas A&M U.-Corpus Christi,
Corpus Christi, TX (1997), 106 pp.
94. B. K. Batten, N. C. Kraus and L. Lin, Long-term inlet stability of a multiple inlet
system, Proc. Coast. Sediments’07, ASCE, Pass Cavallo, Texas (2007), pp. 1515–1528.
95. W. C. Seabergh, Approaches to understanding multiple-inlet stability, Proc. Coast.
Sediments’07, ASCE (2007), pp. 1391–1404.
August 17, 2009 11:8 9.75in x 6.5in b684-ch32 FA
Chapter 32
Enclosed bays are one of the most productive areas in the sea, and they have
long provided us with uncountable benefits. Since enclosed bays are the boundary
domain between land and open sea, they are also significant for global material
cycles. However, eutrophication has been occurring in many enclosed bays around
the world. To maintain sustainable utilization of such bays’ ecosystems, advanced
interdisciplinary research is needed. This chapter describes the mechanisms of
water quality variation under typical currents and introduces an ecosystem model
as a tool for the integrated management of enclosed bays.
32.1. Introduction
Enclosed bays have always been natural environment next to urban areas, and
provided us with uncountable benefits. Due to their particular geographic feature,
the enclosed bays are stable when compared with open seas, and cannot be easily
influenced by ocean weather conditions. Because of this, many harbors are located
in enclosed bays, and hence many big cities in the world, even today, are located
around enclosed bays.1 Enclosed bays also provide fish and shellfish resources, and
navigation and recreation potentials. In addition, the suitability of coastal zone for
various human activities has led to reclamation of shallow water areas on mildly
sloping bottom.2
However, compared with the long history of our relationship with enclosed bays,
our understanding of these areas is not so deep enough. This may be because they
have an extremely wide variety of environments depending on location and time
since they have so complicated geographic features bordering inland and outer ocean
and hence are affected by both of them.
Recently, deterioration of water quality has been occurring in many enclosed
bays in the world. The damages caused by the events of red tides, harmful
algal bloom, and decrease in dissolved oxygen in the bottom water are occurring
901
August 17, 2009 11:8 9.75in x 6.5in b684-ch32 FA
Diatoms can multiply very rapidly and dominate in nutrient-rich marine waters.
Their distinguishing feature is a hard mineral shell made by polymerized silicic acid.
They float in the water column or attach to surface of sediment. Diatoms are major
contributors to production, especially in spring blooms.
Dinoflagellates have a single cell and are smaller than diatoms. They drift with
the water by using flagella to move. Diatoms do not have swimming capability
and they passively follow the flow. Therefore, dinoflagellates with swimming capa-
bility apparently seem to have an advantage. However, since consuming energy by
swimming and a bodily organs become complicated, dinoflagellates are inferior in
multiplication capability compared with diatoms. So, it cannot be said that dinoflag-
ellate has advantageous for survival compared with diatom. The amount of phy-
toplankton is maintained under a struggle for survival in environmental change.6,7
All higher order biota in the ecosystem can use stably the organic matter which
phytoplankton produces.
Since nutrients, such as nitrogen and phosphorus, are especially abundantly
supplied from a river to bays, much phytoplankton can live and it enables
other various aquatic biota to become abundant through preying on this. For
this reason, enclosed bays serve as remarkable area where the commercial fish
and shell fish production is high, although the area is small compared with an
open ocean.
This formula represents a process which produces organic matters and oxygen
from nutrients (nitrogen and phosphorus), carbon dioxide, water by using the energy
of light.
Figure 32.1 shows (a) the variations of wind speed and its direction, (b) water
temperature at the surface, middle, and bottom, (c) daily solar radiation and
precipitation, (d) chlorophyll a at the surface, (e) nitrogen at the surface, and
(f) phosphorus at the surface.18 The increase in chlorophyll a occurred only at the
time when salinity is low and nitrogen is sufficient. Thus, the increase in the phy-
toplankton in this area was generated when nutrients were supplied from river.
However, nitrogen was exhausted with the increase in phytoplankton and became
undetected after the phytoplankton increase. Phosphorus was not exhausted by phy-
toplankton blooms. Moreover, phosphorus increases after the middle of July when
phytoplankton did not increase. Therefore, it is hard to think that phosphorus is the
limiting factor of phytoplankton. Nitrogen is a limiting factor for a phytoplankton
increase in this area.
Under a sufficient nutrient condition, phytoplanktons produce their biomass with
a nearly constant P:N:C ratio of 1:16:106 (by molecular ratio), where P denotes
phosphorus; N, nitrogen; and C, carbon. This ratio is called the Redfield ratio.19 For
example, when there are 20 mol of nitrogen and 1 mol of phosphorus in sea water,
4 mol of nitrogen will remain after phytoplanktons assimilate all phosphorus. Then,
they cannot increase further. This kind of nutrient is called limiting nutrient and it
changes with location and time.
August 17, 2009 11:8 9.75in x 6.5in b684-ch32 FA
8
Temperature (MJm-2day-1) (m/s)
Solar Radiation Wind
(a) 0
-8
Precipitation
30 0
(mm/day)
20 50
(b) 10 100
0 Surface 150
32 Middle
30 Bottom
28
Water
(°C)
(c) 26
24
22
40
Chlorophyll a
30
( µ g/l)
(d) 20
10
0
30
( µ mol/l)
Nitrogen
PO -P
Phosphate
(( µ mol/l
(f) 1.5 4
1
0.5
0
7/1 7/6 7/11 7/16 7/21 7/26 7/31 8/5 8/10 8/15 8/20 8/25 8/30
Fig. 32.1. (a) Wind vector, (b) solar radiation and precipitation, (c) water temperature,
(d) chlorophyll a, (e) nitrogen, and (f) phosphate from July to August 2001.
Figure 32.2 shows the bathymetry of the Tokyo Bay, Japan and observation
sites. Figure 32.3 shows dissolved nutrients at the two stations and variations of
river discharge.20 Station 1 is located near the river mouth in the west side of the
Tokyo Bay. Station 2 is located at the head of the bay about 15 km from the river
mouth. From these figures, the timing of the high nitrate and phosphate concen-
tration in the surface layer coincides with increases in the river discharge. Phosphate
concentration showed less temporal fluctuations than nitrate and the decrease in
phosphate corresponds with the increase of chlorophyll a. This is more similar at
Stn. 2 than Stn. 1. Since Stn. 1 is closer to the river mouth, it has relatively small
variations in the consumption by phytoplankton compared with Stn. 2.
N–P ratio is estimated from these figures, and it has exceeded 16 in general,
except for the seabed during summer. Hence, a limiting factor of phytoplankton is
phosphorus in this area. For the nutrient limitation, we should consider not only
for the N–P ratio but also the other nutrients and their absolute concentration.
The limiting factor derived from the Redfield ratio suggests a possibility of relative
shortage, and absolute values suggest an overall nutrient restriction.
August 17, 2009 11:8 9.75in x 6.5in b684-ch32 FA
30
20m
Stn.3 N
Kanazawa
20
Yokosuka
10
Kanaya
40m
0
0 10 20 30 40 (km)
Fig. 32.2. Bathymetry of the Tokyo Bay, Japan, and its adjacent coastal area showing the location
of the measurement in the bay.
Fig. 32.3. Isopleths of (a) nitrate (NO3 -N), (b) ammonia (NH4 -N), (c) phosphate (PO4 -P),
(d) silica (SiO2 -Si) at Stn. 1 (left), and Stn. 2 (right). (e) River discharge of Kyu-Edogawa River
(m3 /s) and (f) chlorophyll a at Stn. 2.
30
Solar radiation
25
(MJ/day)
20
15
10
5
0
120 Stn.1
Chlorophyll a
100 Stn.2
( µ g/l)
80
60
40
20
0
Apr/1
Apr/11
Apr/21
May/1
May/11
May/21
May/31
June/10
June/20
June/30
Jul/10
Jul/20
Jul/30
Aug/9
Aug/19
Aug/29
Sep/8
Sep/18
Sep/28
Oct\8
Oct\18
Oct\28
Fig. 32.4. Relation between the solar radiation (top) and chlorophyll a (bottom) at the head of
Tokyo Bay in 1999.
August 17, 2009 11:8 9.75in x 6.5in b684-ch32 FA
of the bay will be stabilized, and, various fish and shellfish will be stabilized and
live abundantly.
In contrast, once nutrient load is increased beyond the level absorbable by phy-
toplankton, then the aquatic environment becomes an eutrophic system, phyto-
plankton growth rate becomes nutrient saturated, and changes in nutrient supply
can only cause relatively slow changes in phytoplankton biomass. As a result, phyto-
plankton biomass varies mainly with solar radiation, temperature, and the thickness
of the mixed layer.21 Phytoplankton production exceeds the grazing capacity of
zooplankton and large fluctuations in phytoplankton biomass may occur. Phyto-
plankton grazers, such as copepods, whose life cycle is in the order of days to
weeks, cannot respond to these quick fluctuations in phytoplankton biomass, and
a portion of the organic matter remains ungrazed.22–24 Subsequently, it sinks from
the euphotic zone into subpycnocline waters where it is rapidly metabolized.25 Most
parts of a bay does not have enough depth for degradation of phytoplankton during
falling on the bottom. Subsequent sedimentation of the bloom also constitutes a
major input to the benthic ecology. Even if only 1% of the organic matter pro-
duced by photosynthesis in the euphotic zone is deposited to sediment in an ocean,
it reaches to 10–50% and 90–99% of the organic matter deposited on the seabed
is decomposed by the microorganisms on the seabed.26 As a result, instability of
plankton population has important ecologic consequences. Such accumulations of
ungrazed organic material often lead to microbial-dominated food webs, which is
characterized by a greater decomposition and oxygen consumption.
Oxygen-depleted water occurs under stratified conditions. Higher temperature
or lower salinity blanket of water overlies the heavier, lower temperature, or higher
salinity water in the bay. This overlying water prevents reaeration of oxygen-
depleted water in the lower layer by suppressing vertical exchange. So, oxygen-
depleted water is induced by both a physical process and a biochemical process.
The effect of oxygen-depleted water to aquatic ecology had been well docu-
mented. Diaz and Rosenberg27 reported that a high mortality of benthos may occur
under 1.1–2.0 mg/l dissolved oxygen. Once oxygen-depleted water occurs, fish tends
to shunt, to avoid it. In contrast, shellfish like oysters and clams cannot escape and
will die because they have small transportation capability. As a result, a nonbenthos
area will be formed in the large portion of inner-bay during the summer in Tokyo
Bay.
Bivalves such as oysters and clams have the important role to decrease the sus-
pended solid. So, once they decrease, transparency will fall and seaweed will become
extinct. Furthermore, when phytoplankton is decomposed on the seabed, nutrients
will be emitted to the sea water again and a further increase of phytoplankton will
be caused. Thus, oxygen-depleted water causes remarkable damage to the ecosystem
inside the bay.
Figure 32.5 shows observation results of oxygen-depleted water in Tokyo Bay
during the summer in 2003. The oxygen-depleted water (oxygen concentration is less
than 3 ml/l) is widely spread at the center part of inner bay. Oxygen-depleted water
is transported by wind-driven currents, and changes the distribution considerably.28
This oxygen-depleted water is in the inner bay about six months from early summer
to early autumn. Figure 32.5 also shows bivalves distribution in 2003 and bivalves
August 17, 2009 11:8 9.75in x 6.5in b684-ch32 FA
Fig. 32.5. Dissolved oxygen distribution at the bottom for August 2003 (left) and benthic bivalve
distribution on the seabed in November 2003 (right).
do not exist at the center part of the bay. This area is almost the same as the
distributions of oxygen-depleted water.29
Water quality is influenced by a flow directly and indirectly. For example, since
phytoplankton drifts underwater, it is influenced strongly by the flow of the bay.
Motion of bay waters also exerts strong influences on the distribution of dissolved
oxygen in bays. These kinds of influences are direct and obvious. In contrast, even
if the flow is very small, it still has some residual current, flowing in one direction.
In this case, nutrient loads are transported to distant locations and, probably, the
water exchange rate of the bay also increases. Then a permissible nutrient loading
which the bay can receive will also alter through the change of water exchange
rate. In this section, we explain the role of physical processes and the relationship
between water motion and nutrient cycles.
2.0
(a)
Water level 1.5
(m) 1.0
(b) Wind
(m/s) 10m/s
(c)
Instantaneous current
surface 20cm/s
(d)
Instantaneous current
bottom
(e)
Residual current
surface
(f)
Residual current
bottom Oct1 Oct2 Oct3 Oct4 Oct5 Oct6 Oct7
(2005)
Fig. 32.6. Results of measurement of (a) water level, (b) wind, (c) instantaneous currents at
surface, (d) bottom layers, (e) residual currents at surface, and (f) bottom layers, respectively,
illustrating the dependency of instantaneous currents on tide and residual currents on wind stress
at Stn. 1 in 2005.
acoustic Doppler current profiler (ADCP). The ADCP measures the velocity profile
by using Doppler shift of ultrasonic waves, and can measure the flow without con-
tacting and disturbing velocity field.30 Its maintenance is easy. Therefore, ADCP
has commonly been used to measure the flow field in recent years. From the top
figure, we can see the oscillations of water level, with a period of approximately
half a day. Instantaneous currents are also oscillating with the semi-diurnal or
diurnal periods at the surface and bottom. This oscillation pattern of current is
called the tidal current and is the most dominant component of a flow in bays.
The water of a bay is moving continuously by the tidal rhythms. So we need to
measure currents continuously to extract the residual component. For a unidirec-
tional flow like a river flow, we can estimate the discharge by using instantaneous
measurement results. However, bay water moves back and forth during ebb and
flood tides. So we cannot determine the average movements of the bay water only
by the instantaneous current. Residual currents which are extracted by filtering out
the tidal component from instantaneous currents are often used for discussion of a
flow pattern, because it is more adequate when we discuss the net transportation of
substances.31
From these figures, the dependency of temporal residual currents on the wind
is also clear. Since the sea surface has not so many obstacles compared with that
of the land and the roughness is also smaller than the land, the sea wind is often
August 17, 2009 11:8 9.75in x 6.5in b684-ch32 FA
twice as strong as that on land. Surface sea water is dragged by such strong winds.
Then the surface water drags further the lower layer, and the flow is generated
in the down-wind direction. Occurrence of wind-driven currents is very incidental
compared with that of tidal currents. Especially, the strong wind which blows at
typhoons or hurricanes mixes bay water and has a strong impact to the water quality
and ecosystems of bays.
Although a flow in the down-wind direction is easy to understand, density
structure and rotation of the earth make it more complicated. To simplify our dis-
cussion, let us consider the rectangle bay which is very small and has a uniform
depth. It is assumed that Coriolis force and stratification effect can be neglected.
When the wind blows over this bay, it generates a shear stress at the sea surface. The
surface water moves in response to the wind-shear stress; the water surface of the
leeward side becomes higher and the windward side becomes lower (see Fig. 32.7).
If the water pressures at the seabed are compared at this time, the water pressure
of the leeward will become high because of the effect of wind stress. As a result, a
flow which returns to the upwind direction occurs at the seabed. Consequently, a
vertical circulation, for which the direction of the flow are opposite at the surface
and bottom, occurs. In this mechanism, the water near the seabed as well as the
surface responds to a wind in comparatively short time, and moves to the upwind
direction.
When the bay water is stratified and has a pycnocline (a layer with a large
density gradient), the transfer of motions is prevented. Therefore, a comparatively
strong circulation is formed only in the upper layer. Moreover, after the wind
stopped, the pycnocline which was inclined by the wind-induced flow may vibrate,
and a strong flow may occur. Consequently, even when constant wind blows on a
certain bay, the resulting water motions depends on the strength of stratification
and its depth.
Fig. 32.7. Wind-induced current profile and a horizontal distribution of water level.
August 17, 2009 11:8 9.75in x 6.5in b684-ch32 FA
One might consider that the surface water would be forced to move in the same
direction as the wind. However, the water normally moves in different direction
similar to the wind when Coriolis force has significant effect. If a wind blows
continuously beyond half a day in bays, the influence of the Coriolis force will
become significant. It is one of the apparent forces and produced by the earth’s
rotation. It acts rightward and perpendicular to movement in the northern hemi-
sphere, and leftward in the southern hemisphere. Therefore, if the surface water is
dragged by the wind, the lower layer water will change the direction to the right-
hand side in the northern hemisphere, so that a spiral may be drawn in the ver-
tical direction as shown in Fig. 32.8. This flow is called Ekman transport and its
vertical distribution of the flow is called Ekman spiral. Consequently, sea water is
transported in the direction different from a wind direction by Ekman transport
which is confined within top 10–100 m of the water column. The direction is 90◦
to the right of the wind in the northern hemisphere and 90◦ to the left in the
southern hemisphere. As shown in Fig. 32.8, the velocity decreases exponentially
in the vertical direction. Since the velocity decays to 1/e times at z = hE , where
hE is called an Ekman layer thickness and is expressed by hE = 2KZ /f , in
which KZ denotes the eddy viscosity and f denotes the Coriolis parameter. Since
the flow direction changes with the wind direction through Ekman transport, as
shown in Fig. 32.9, upwelling occurs for wind parallel to the coastline and with the
coastline on the left-hand side. Conversely, if water is driven toward the coast by
Ekman transport, downwelling occurs at the coast. Once upwelling occurs, bottom
water which has high concentrations of nutrients and deleted with oxygen appears
to the surface layer and it has large influence on the water quality and ecosystem
of the bay.
Fig. 32.8. Ekman current profile for a zonal (eastward) wind in the northern hemisphere. Note
the surface velocity makes the 45◦ angle with the wind vector and the turning of the current to
the right of the stress vector with depth.
August 17, 2009 11:8 9.75in x 6.5in b684-ch32 FA
Fig. 32.9. Upwelling and downwelling situations in the bay: (left) coastal downwelling induced
by a wind parallel to the coast with the coast to its right (in the northern hemisphere) and (right)
coastal upwelling.
Fig. 32.10. Time series of (a) wind speed, (b) daily solar radiation, chlorophyll a [Stn. 1 (c),
Stn. 2 (d), and Stn. 3 (e)], temperature [Stn. 1 (f), Stn. 2 (g), and Stn. 3 (h)] and salinity [Stn. 1 (i),
Stn. 2 (j), Stn. 3 (k)], from April to June 1999. The wind speed is smoothed by a 6-h running mean.
August 17, 2009 11:8 9.75in x 6.5in b684-ch32 FA
Halocline
Entrainment
These nutrients that are supplied from the seabed during summer are trans-
ported to the head of the bay by the estuarine circulation. Furthermore, they are
again uptaken into the upper water, and are used for the increase in the phyto-
planktons in the euphotic zone. Estuarine circulation plays a key role in the nutrient
cycles of stratified estuaries by increasing the retention, recycling the nutrients, and
stabilizing the ecosystem.8
The phenomena in bays are not only physical phenomena but also biologic and
chemical ones, and they are related to each other. So, the ecosystem and water
quality of bays are highly complicated. To deal with these complex systems quanti-
tatively, numeric modeling has been developed since 1960s along with developments
of computers. Modeling is also important for applications such as understanding
and prediction of bays’ water quality and ecosystems. A water quality model in bays
is comprised of a three-dimensional circulation model and an ecosystem model that
describes pelagic and benthic aspect of nutrients cycling.
A physical model solves essentially Navier–Stokes equation with the forcing
(the wind stress, Coriolis force, and buoyancy force) under adequate approxima-
tions. In order to include the density effects, conservation equations for temperature
and salinity are also solved. In order to obtain a realistic prediction for vertical
stratification, turbulent closure model is also employed.38 As a result, the gravita-
tional, wind-driven, and topographically induced flows can be reproduced within
August 17, 2009 11:8 9.75in x 6.5in b684-ch32 FA
Fig. 32.12. (a) Bathymetry of the Chesapeake Bay and its adjacent coastal area. Depths are in
meters. (b) A horizontal curvilinear coordinate system designed for resolving the complex coastlines
and the deep channel in the bay (from Ref. 37).
August 17, 2009 11:8 9.75in x 6.5in b684-ch32 FA
Fig. 32.13. An example of vertical grid system: (a) a Cartesian or z-coordinate vertical grid and
(b) sigma-coordinate vertical grid system.
For vertical coordinate systems (shown in Fig. 32.13), a Cartesian grid (z-
coordinate vertical grid) and a sigma-coordinate grid are widely used. The Cartesian
grid is equivalent to the rectangular horizontal grid. However, it is sometimes more
accurate than the sigma-coordinate, featuring the presence of steep bottom topog-
raphy. The sigma-coordinate system is convenient in that it can introduce an essen-
tially “flatting out” the variable bottom at z = −h(x, y). The sigma-coordinate
system has long been used, in both meteorology and oceanography.44,45
Various kinds of models are also proposed for an ecosystem model46–50 ; it also
solves conservation equations for relevant components with appropriate source and
sink terms. If C is the concentration of some component in a model, the time rate
of variation of C is given by:
dC
= Fin − Fout + R (32.2)
dt
where Fin and Fout are the fluxes into and out of the target control volume which
is calculated by using the physical model results, and R is the net increase or
decrease due to internal production and removal of biogeochemical effect. These
models provide a quantitative description of the influences of physical circulation
on biologic and chemical processes in bays.
Figure 32.14 shows schematic interactions of a lower trophic ecosystem model
which is used in Tokyo Bay.51 This model has 13 state variables: phytoplankton, zoo-
plankton, nutrients (nitrogen, phosphorus, and silica), labile detritus, and refractory
detritus for each nutrients and dissolved oxygen as well as sedimentation process
of particulate organic material. For example, certain labile compounds which are
rapidly degraded, such as sugars and amino acids in the particulate organic matter
deposited on the sediment surface are decomposed readily; others such as cellulose
are more refractory, or resistant to decomposition. Figure 32.15 shows the calcu-
lation results of an annual budget of nitrogen and phosphorus in Tokyo Bay. The
August 17, 2009 11:8 9.75in x 6.5in b684-ch32 FA
Sea surface
Detritus N,P,Si
NO3-N NH4-N PO4-P SiO2-Si •Dissolved
Sea •Labile Particulate DO
N
bed 2 •Refractory
Detritus N,P,Si
NO3-N NH4-N PO4-P SiO2-Si •Dissolved
DO
•Labile
Particulate
•Refractory
Detritus N,P,Si
•Dissolved
NO3-N NH4-N PO4-P SiO2-Si
•Labile
Particulate
Sediment •Refractory
Fig. 32.14. Idealized nutrients cycling in Tokyo Bay’s ecosystem in the model of Koibuchi et al.51
Cycling between the 13 state variables: phytoplanktons, zooplanktons, nutrients (nitrogen, phos-
phorus, and silicate), labile detritus, and refractory detritus for each nutrients and dissolved oxygen
as well as sedimentation processes of particulate organic material.
Fig. 32.15. Summary of fluxes and process rates calculated in Tokyo Bay, January 1999 to January
2000. Units are given in ton per year for each element.
August 17, 2009 11:8 9.75in x 6.5in b684-ch32 FA
32.5. Conclusion
This chapter dealt with the variation mechanisms of water quality and an ecosystem
under typical currents in enclosed bays. Since enclosed bays are the boundary
domain of land and open seas, these are complicated phenomena. In order to under-
stand the recycling of nutrients for integrated management of bays ecosystem,
advanced interdisciplinary research is needed.
Over the last half of the century, constructions of a sewage treatment plant
has been performed to reduce nutrients load for the maintenance of water quality
of bays. However, there is a limitation in nutrient removal in a sewage treatment
plant by the present technology. In order to stop negative spiral of eutrophication
phenomena, an onsite countermeasure also needs to be carried out. Enclosed bays
forming the edge of our world in many ways are next to our life and are not as so
vast as open seas. So, if our understanding of phenomena of bays can be deepened,
we will have a potential to control or redesign them and therefore move toward a
new, more stable, and attractive environment.
References
1. H. J. Walker, The Coastal Zone. In the Earth as Transformed by Human Action. Local
and Regional Changes in the Biosphere over the Past 300 Years, eds. B. L. Turner I,
W. C. Clark and R. W. Kates (Cambridge University Press, Cambridge, 1990).
2. I. Valiela, Marine Ecological Processes (Springer, NewYork, 1995).
3. G. Ærtebjerg, J. H. Andersen and O. S. Hansen, Nutrients and Eutrophication
in Danish Marine Waters (Ministry of the Environment, National Environmental
Research Institute, 2003).
August 17, 2009 11:8 9.75in x 6.5in b684-ch32 FA
45. N. G. Freeman, A. M. Hale and M. B. Danard, J. Geophys. Res. 77(6), 1050 (1972).
46. J. N. Kremer and S. W. Nixon, A Coastal Marine Ecosystem: Simulation and Analysis,
Ecological Studies (Springer-Verlag, Heidelberg, 1978).
47. M. J. R. Fasham, H. W. Ducklow and S. M. Mckelvie, J. Mar. Res. 48, 591 (1990).
48. F. Chai, R. C. Dugdale, T. H. Peng, F. P. Wilkerson and R. T. Barber, Deep-Sea Res.
II49, 2713–2745 (2002).
49. M. J. Kishi, M. Kashiwai, D. M. Ware, B. A. Megrey, D. L. Eslinger, F. E. Werner,
M. Noguchi-Aita, T. Azumaya, M. Fujii, S. Hashimoto, D. Huang, H. Iizumi, Y. Ishida,
S. Kang, G. A. Kantakov, H.-C. Kim, K. Komatsu, V. V. Navrotsky, S. L. Smith,
K. Tadokoro, A. Tsuda, O. Yamamura, Y. Yamanaka, K. Yokouchi, N. Yoshie,
J. Zhang, Y. I. Zenko and V. I. Zvalinsky, Ecol. Model. 202, 12 (2007).
50. B. deYoung, M. Heath, F. E. Werner, F. Chai, B. A. Megrey and P. Monfray, Science
304, 1463–1466 (2004).
51. Y. Koibuchi, J. Sasaki and M. Isobe, Proc. Coast. Eng. 48, 1076 (2001).
August 14, 2009 9:1 9.75in x 6.5in b684-ch33 FA
Chapter 33
Elena Sánchez-Badorrey
Departamento de Mecánica de las Estructuras e Ingenierı́a Hidráulica
University of Granada, Granada, Spain
elenasb@ugr.es
† Corresponding author.
923
August 14, 2009 9:1 9.75in x 6.5in b684-ch33 FA
evaluation of the risk during the ship passage in the harbor entrance channel
is considered. Again, based on risk analysis and decision theory, the problem is
formulated and solved to evaluate the stoppage probability along the entrance
channel to the harbor of Motril of a bulkcarrier during a storm.
33.1. Introduction
Throughout history man’s relationship with the sea has been both intense and
fraught with difficulty. In this sense the sea had been and still is a source of immense
richness and great potential for mankind. It offers maritime routes that from the
beginning of time have fomented commerce and communication between civiliza-
tions. On the less positive side, it has brought illnesses and epidemics, and opened
the door for marauders and invasions. In this way the coast has played an important
role in the history, culture, and economy of countries with coastal boundaries.
Until the middle of the 20th century this relation was symbiotic. The impacts
produced in the majority of cases were either reversible or absorbed by the sea.
In recent decades, anarchic and unsustainable use of the sea and its resources has
becoming the rule rather than the exception, something that is causing the pro-
gressive deterioration of the nearshore system. The most evident consequences of
the increase in activities and uses in such a limited space are: (1) intensive land use,
(2) increase of loads and pressure on the coastal ecologic trace, (3) massive transfers
to the sea, and (4) decrease of the biodiversity in both mediums.
The socioeconomic progress of the Western civilization began its growth after
the Second World War. At that time the slogan was “development must not stick at
nothing.” In the early 70s of the last century, the first alarms arrived with the first
oil crisis. The response was a new doctrine: development in view of the oil (energy)
price. Fifteen years later arrive the next alarm, the environment which halted the
progress: we are killing the planet earth. The introduction of environmental studies
to forecast the environmental damage and, just in case, to develop a new area
in exchange for a damaged area, proportionated a solution to proceed forward.
However, data obtained during the first years of this century confirms that the actual
socioeconomic progress is exhausting the energy, water, and coastal zone resources.
A more serious and rigorous slogan must drive the progress today: (1) socioeco-
nomic and environmental progress must be concomitant, (2) use of basic resources
must be minimized, and (3) operationality and safety of the human works must be
maximized.
These new demands ask for a new coastal and ocean engineering philosophy
based on new design principles, considering the socioeconomic and environmental
impact of the human intervention during its useful life. However, coastal and ocean
engineering must deal with the environmental events, and their random nature.
Thus, the response to the problem has to include the associated uncertainty, among
others, to the occurrence of the atmospheric and maritime agents.
One method of effectively dealing with this situation is through Integral Coastal
Zone and Harbor Management (ICZ&HM), based on a rational decision-making
process in which economic, social, and environmental factors are considered.
August 14, 2009 9:1 9.75in x 6.5in b684-ch33 FA
In this chapter, a summary of some of the new design principles and tools that
can help to match the society demands are presented. It is only the start, but it
will be a long way before the maritime works are optimized and their uncertainty
bounded.
The chapter is organized as follows. First, the problem of an integrated coastal
and harbor management is formulated within a new framework. Next, the morpho-
logic evolution of a stretch of coast is considered,1 and the main tools developed
for the risk analysis are given. The evaluation of the risk during the ship passage
in the harbor entrance channel is analyzed. Again, the probabilistic analysis of the
problem is solved. Two case studies are presented. Firstly, the probable evolution of
the coastline in V years for a stretch of coast in the south of Spain is analyzed. Sec-
ondly, the stoppage probability along the entrance channel to the harbor of Motril
of a bulkcarrier during a storm is evaluated. The chapter concludes with a discussion
and some recommendations.
33.2. Formulation
Natural phenomena are not exempted from uncertainty. In particular, those pro-
cesses in which climatic agents are involved inherit their stochastic character at
different temporal and spatial scales. This is the case for coastal morphodynamics
where changes occurring at short, medium, or large spatial scales are clearly asso-
ciated to the corresponding time scales of the forcing mechanisms.19
In the past few decades, a big effort has been made to model the physical and
geomorphologic processes governing the medium- and long-term natural changes of
the coastal zone and its alteration by human activities.
Most of the tools and concepts (e.g., HUMOR UE Project) deal with simplified
geometries and stationary input conditions that can be considered as “state models”
in the sense that they can only analyze the morphodynamic response to a certain
climatic state defined by a constant energetic level. Inherently, their application
relies in the hypothesis that if the climatic state lasts for enough time, the mor-
phology will arrive to a stable morphodynamic equilibrium that will remain until a
new climatic state moves off it to another morphodynamic state.
Although these models are capable of describing the main morphologic processes
in the coastal zone, coastal morphology does not only respond to average climatic
conditions but also to extreme events (impulses) and, therefore, in a long-term basis,
the littoral morphology is the result of the sequential action of a random number
of storm and nonstorm events.
August 14, 2009 9:1 9.75in x 6.5in b684-ch33 FA
In this framework, deterministic approaches cannot cope with the inherent ran-
domness of the processes and any predictive model for the evolution of the littoral
zone over years or decades has to be based upon statistic tools capable of dealing
with the uncertainty of the prognosis.
To overcome this limitation, a procedure for the assessment of intrinsic uncer-
tainty to be applied for the long-term prediction of the evolution of a certain mor-
phologic feature driven by climatologic agents, was proposed by Payo et al.3 The
methodology is based on the random simulation of climatic states and the sequential
modeling of the morphologic state response and is suitable to be applied, with fore-
casting purposes, to morphodynamic problems evolving in time and space. The
overall procedure is summarized in Fig. 33.1.
where ε is the difffusion coefficient and w(x, t) includes the alongshore variation of
wave breaking height and breaking angle, θb , and a source/sink term, q(x, t)/D,
where D is the depth of closure,
q(x, t) ∂
w(x, t) = − (ε(x)θb (x)). (33.2)
D ∂x
August 14, 2009 9:1 9.75in x 6.5in b684-ch33 FA
The boundary value problem (BVP) consists in the resolution of Eq. (33.1) in the
domain x ∈ [0, l] and t ≥ 0 with the initial condition
yS (x, 0) = f (x), x ∈ [0, l] (33.3)
and the boundary conditions
∂yS (0, t)
c1 yS (0, t) + c2 = g(t), t≥0
∂x
(33.4)
∂y (l, t)
c3 yS (l, t) + c4 S = h(t), t≥0
∂x
g(t) and h(t) describe the behavior of the shoreline at the extremes of the beach,
depending on the physical restrictions imposed. If a groin is partially blocking the
sediment transport, it is assumed that the volume of sand bypassing the obstacle
is proportional to the distance from the tip of the groin to the depth of closure
isobath. A detailed analysis of the boundary conditions is given by Payo et al.4
yi = µ
+ zi1 e1 + · · · + zip ep i = 1, . . . , N (33.6)
In order to show the potential use of the methodology, the distribution function
of two random variables defined in terms of the final shoreline position is calculated
next for p = 2.
August 14, 2009 9:1 9.75in x 6.5in b684-ch33 FA
where
Sαe = (Z1 , Z2 ) | e11 Z1 + Z2 e12 ≤ α1 − µ1 ; · · · ; eM M
1 Z1 + Z2 e2 ≤ αM − µM .
(33.15)
For the particular case in which αi = α, i = 1, . . . , M the probability of the variable
Ymin , minimum advance/retreat of the shoreline can be evaluated:
F (α) = Pr(Ymin ≤ α) = Pr(min{y(x)} ≤ α) (33.16)
= 1 − Pr(min {y(
x)} > α) = f (Z1 , Z2 )dZ1 dZ2 (33.17)
Sα
with
Sα = (Z1 , Z2 ) | e11 Z1 + Z2 e12 > α − µ1 ; · · · ; eM M
1 Z1 + Z2 e2 > α − µM . (33.18)
August 14, 2009 9:1 9.75in x 6.5in b684-ch33 FA
Fig. 33.3. Scheme of the probabilistic model and methodology for the estimation of safety depth
requirements.
then the duration of the time-state scale tst can be defined as:
tst = min Tstsea , Tstwc , Tstmor , Tstship , TC . (33.19)
tions and ship response to the climate agents. Consequently, it can be defined as
ship ship
lst ∼ ship ship
= Lst . Note that Lst = f Tst , V .
Then, in order to calculate the stoppage probability along an entrance channel, the
passage is spatially divided into “subsets” and temporally into “navigation states.”
Taking into account that τ is defined as the ratio between the subset length
ship , we found that, in
and the average velocity of the ship in the subset, τ = lsp /V
general, τ ∼
= tst because lst ∼
= Lst and tst ∼
ship ship
= Tst . Hence, usually, the navigation
state duration is determined by the maneuvering conditions. Thus, each subset can
be divided into a set of navigation states.
Besides the stationary condition, a complete, exhaustive, and mutually exclusive
set of navigation states concurrent with the passage through the subset can
be defined as the union of possible: (1) water column states, (2) sea states,
(3) morphologic states, and (4) maneuvering states during the passage.
Arranging this information, for a passage under known climatic, morphologic,
and maneuvering conditions, a complete set of navigation states in a subset can be
defined by means of:
(1) A water column state characterized by the average local mean water column
level and typical deviation.
(2) A sea state characterized by the state descriptors (i.e., mean wind velocity and
direction, characteristic wave height, and period and direction) and probability
functions of the significant climatic agents.
(3) A set of morphologic states characterized by state descriptors (i.e., mean bed
slope, and mean height of bed morphology), and their spatial probability
distributions.
(4) A set of maneuvering states characterized by state descriptors and/or proba-
bility distributions of the maneuvering conditions (i.e., mean velocity, direction,
trim, path, and human factor) and of the significant ship movements.
SW = wc − ws − SW
0
(33.23)
where wc represents the minimum distance between the entrance channel borders
and the center of mass of the ship hull; ws is the sum of all the random variables and
parameters that may contribute to the variation of the minimum distance between
the ship and the “way-marks” of the channel, and Sw 0
represents an extra security
margin due to model uncertainty.
The model considers two “natural” verification approaches for the stoppage
domain calculation:
(1) A “clearance state approach” in which the terms of the verification equations
are averaged in a characteristic timescale of the subset, tSC .
(2) An “instantaneous clearance approach” in which the average timescale is equal
to a characteristic timescale of the short-time random variables. The instanta-
neous clearance timescale is represented by tIC .
The instantaneous clearance approach is only strictly necessary when the safety
requirements are not fulfilled in the state clearance approach, or when the levels of
use and exploitation of the channel have to be optimized.
where P m is the total stoppage probability in the subset CSm , and M is equal to
the total number of subsets. The probability P m can be expressed as:
P m = PK
M m
+ PW m
− PK∩W (33.25)
m m
where PK is the stoppage probability related to the under-keel stoppage mode; PW
m
is the stoppage probability related to the way-mark stoppage mode; and PK∩W is
the joint probability of both stoppage mechanisms. Any of the above probabilities
(henceforth Pim ) can be calculated as:
Nm
Pim = P Sim ≤ 0N Snm · P N Snm (33.26)
n=1
where P [Sim ≤ 0|N Snm ] represents the stoppage probability of the stoppage mecha-
nism i in the subset CSm conditioned to the occurrence of the navigation state
N Snm ; P [N Snm ] is the probability of the navigation state during the passage through
August 14, 2009 9:1 9.75in x 6.5in b684-ch33 FA
Nm m
the subset CSm . Note that, n=1 P [NS n ] = 1 according to the completeness
requirement of the navigation state definition. Finally, Nm is the total number of
possible navigation states during the passage through the subset.
In the state clearance approach, the probability P [Sim ≤ 0|NS m n ] is calculated
as the probability associated to the stoppage domain of the state clearance approxi-
mation of the verification equation Sim (i.e., Sim ∼ m
= SIC,i ) during the environmental,
morphologic, and maneuvering conditions defined by the navigation state NS nm .
Assuming the statistic independence of both approaches (based on the condition
tIC tSC ), it can be expressed as:
m m m m
P Sim ≤ 0NS m m
n ≈ P SIC,i ≤ 0 SSC,i ≤ 0, NS n · P SSC,i ≤ 0 NS n . (33.27)
where Pmax represents the maximum stoppage probability admissible for the
stoppage mechanism under consideration. The choice of the Pmax value depends:
(1) on the management point of view adopted (e.g., that of harbor pilots, ship
captains, and harbor authorities); and as recommended by Rom 0.0,7 (2) on the
economic, social, and environmental impact of the possible occurrence of an accident
(i.e., on the operational nature of the channel.13
and eventually reparation) of the entrance channel versus probability. The use of
Bayesian or assimilation techniques data during the exploitation of the entrance
channel will help to facilitate the adjustment of parameters of the probabilistic
models and the reduction of the uncertainty.
Fig. 33.5. Definition sketch of the reference frame and initial position.
partially blocking the alongshore sediment transport that is assumed to take place
over the entire beach profile up to the depth of closure (Fig. 33.5).
and gave a value around λ = 9. For every storm event, the duration of the storm,
the maximum value of Hm0 achieved and the corresponding values of the mean zero
upcrossing, Tz , and the mean direction, θ were calculated. Figure 33.6 shows that
over the threshold value, the most probable sea state can be swell with Tz = 12–14 s
or sea with Tz = 8–10 s.
Fig. 33.7. Initial shoreline position (dashed line) and some final shoreline positions after 50 years
of climate forcing.
N = 200 times. Figure 33.7 shows some of the outcomes of the experiment that
were used for the analysis.
Figure 33.8 shows the first four eigenfunctions and Fig. 33.9 represents the
variance explained by retaining the terms up to each of them. It can be appreciated
that with the first two ones it is possible to account for more than 99% of the total
variability.
August 14, 2009 9:1 9.75in x 6.5in b684-ch33 FA
Fig. 33.10. Plot of pairs of values of the scores of the first three eigenfunctions.
Figure 33.10 plots the pairs of scores of every two eigenfunctions for the first
three ones. The variables Z1 and Z2 are strongly correlated and therefore, are not
independent. It is possible, however, to infer their joint density functions from the
samples of values of the scores as shown in Fig. 33.11.
The distribution function of the random variable that measures the increment
of area of the dry beach, is shown in Fig. 33.11.
Figure 33.12 shows the distribution function of Ymin where the integral in
Eq. (33.17) is estimated with a Monte Carlo technique.
August 14, 2009 9:1 9.75in x 6.5in b684-ch33 FA
Fig. 33.12. Probability distribution function of the random variable increment of dry beach
in V years.
Fig. 33.13. Motril harbor: location, current configuration, and NE entrance channel.
Fig. 33.14. Stoppage probability during the passage for the “ideal” reflective breakwater
case (KR = 1): state clearance, instantaneous clearance approaches, and real time solution.
passage passage passage
PSC = 50.3%; PIC = 42.4%; and PRT = 42.7%.
August 14, 2009 9:1 9.75in x 6.5in b684-ch33 FA
included. The old mound breakwater has been characterized by a reflection coef-
ficient KR = 0.2. The coastline has been supposed fully dissipative. To study the
influence of the enlarged vertical breakwater reflectivity on the stoppage proba-
bility along the entrance channel, we have assigned two “ideal” reflectivities to
the vertical breakwater: (1) total reflection with reflection coefficient KR = 1 and
(2) full dissipation with KR = 0.
Mean wind and current velocities are considered negligible during the passage.
Their possible effects on the maneuvering conditions and ship response have been
included as model uncertainties. The nominal value of the local depth along the
channel was obtained from bathymetric data in LWL. The relative error in the
depth measurements is 5%. The morphologic state during the storm event is con-
sidered stationary (i.e., Tstmor TC ). The presence of small-scale bedforms and
wave-induced morphology has been included in the under-keel verification equation
through the Gaussian ηmor term.
As a first choice, the maximum stoppage probability is set equal to Pmax = 10%.
From Fig. 33.14, it is concluded that the passage under consideration cannot be
considered safe for any of the management criteria. Consequently, after checking it
with the harbor pilot, the access should be denied.
August 14, 2009 9:1 9.75in x 6.5in b684-ch33 FA
(1) Decrease of the usage and operativity levels of the entrance channel.
(2) Increase of the depth requirements to obtain the same level of opera-
tivity. Following the procedure shown in Fig. 33.3, the depth requirements
for reflective and mound breakwaters have been calculated. In Fig. 33.16,
the mean and 95% upper confidence bands of the current bathymetry and
the estimated depth requirements for the current passage are compared. The
results show that, for the passage considered, the volume of dredging for
the reflective breakwater case may be up to 42% higher than for the mound
breakwater.
Both aspects, and their impact on the economical and environmental cost of
entrance channels construction and dredging maintenance, should be taken into
account in the entrance channel and breakwater design. The application of present
methodology can help to find the optimum design.
33.7. Discussion
The validity and field of application of the overall methodology and each of the
steps are questioned in this section:
(1) Data processing. The analysis is based on the observation that a significant
change in coastal morphology and ocean engineering is only valuable when
the energy of the sea state exceeds a certain level. This hypothesis allows the
analysis of wave climate to be performed from a POT regime. The value of the
threshold has to be based on expert judgment, and several values were tested.
August 14, 2009 9:1 9.75in x 6.5in b684-ch33 FA
Fig. 33.15. Agitation conditions along the Motril entrance route. (a) Deepwater and entrance
channel wave climate for the mound and reflective breakwater cases. (b) Ratio between local and
incident wave height along the mean entrance path (detail). Sea-state model as monochromatic
wave with Tp = 9 s, Hs = 3.8 m.
(2) Data uncertainty. The simulation of the climatic forcing starts with the joint
distribution functions of the random variables involved in the process which are
themselves based on the surrogated data. To this respect, the validity of the
model results depends on the length and quality of the climatic database.
As previously mentioned, for the climatic temporal variability at the different
timescales to be assimilated into the models, the time series should comprise
a period of time at least equal to the period to be simulated. Fortunately,
hindcast techniques provide long time series, over 40 years, of climatic statistic
descriptors (i.e., HIPOCAS UE project).
August 14, 2009 9:1 9.75in x 6.5in b684-ch33 FA
Fig. 33.16. Safety depth requirements along Motril entrance channel. Mean values and 95% upper
confidence bands of current entrance channel bathymetry and safety depth requirements (for the
passage considered). Reflective breakwater case versus mound breakwater case.
(3) Sample representativity. Additionally, the size of the sample simulation has
to be large enough to guarantee the establishment of the results, particularly
at the tails of the distribution functions. In this context, it would mean that
the joint probability density function of the random vector which components
are the scores of the eigenfunctions, are required to be accurate. To this
end, it would be necessary to repeat the experiment until establishment is
reached.
(4) The proposed methodology consists of four sequential steps that include: data
processing, simulation of the forcing and the response of the system, EOF
analysis, and assessment of the uncertainty of associated events after a given
period of time.
The procedure has a general character in the sense that it can be applied to
many physical processes evolving in time and space. It is also not dependent
on the particular tools employed in each step, provided that all the information
necessary for the following steps can be obtained from the previous ones.
(5) The synthetic database of the forcing mechanism (wave climate) can also be
obtained from any simulation process, like autoregressive models. Moreover,
the data processing and simulation can be replaced by any technique capable
of providing a sample of the different responses of the system. In the case of
shoreline related problems, a sample with the position of the shoreline mea-
sured from video images would also be feasible for a shorter-term analysis.19
In relation to the one-line and ship movement model, they are just two of the
August 14, 2009 9:1 9.75in x 6.5in b684-ch33 FA
available methods for the analysis of shoreline and ship response to a climatic
forcing, but many other tools can be used.
(6) Finally, the EOF technique can be replaced by any of the methods available in
the literature for the analysis of structure in data sets.
References
1. A. Baquerizo and M. A. Losada, Human interaction with large scale coastal morpho-
logical evolution. An assessment of the uncertainty, Coast. Eng. 55, 569–580 (2008),
doi: 10.1016/j.coastaleng.2007.10.004.
2. B. Cicin-Sain and R. W. Knetch, Integrated Coastal and Ocean Management: Concepts
and Practices (Island Press, Washington, D.C., 1989).
3. A. Payo, A. Baquerizo and M. A. Losada, One-line model with time dependent
boundary conditions, ed. J. M. Smith, Proc. 28th Int. Conf. Coast. Eng. ICCE, ASCE,
Vol. 3, World Scientific (2002), pp. 3046–3052.
4. A. Payo, A. Baquerizo and M. A. Losada, Uncertainty assessment of long term
shoreline prediction, ed. J. M. Smith, Proc. 29th Int. Conf. Coast. Eng. ICCE, ASCE,
Vol. 16, World Scientific (2005), pp. 2087–2096.
5. M. Larson, H. Hanson and N. C. Kraus, Analytical solutions of the one-line model of
shoreline change near coastal structures, J. Waterw. Port Coast. Ocean Eng. 123(4),
180–191 (1997).
6. W. O. Gray, J. Waters, A. Blume and C. Landsburg, Channel design and vessel
manoeuvrability — Next steps, International Workshop on Channel Design and Vessel
Manuverability (2001).
7. ROM 0.0, General procedure and requirements in the design of harbor and maritime
structures, Part I (Ministerio de Fomento, Spain, 2001).
8. D. E. Newland, An Introduction to Random Vibrations and Spectral Analysis, 1st edn.
(Longman Group Limited, London, 1975).
9. S. Haver and T. Moan, On some uncertainties related to the short term stochastic
modelling of ocean waves, App. Ocean Res. 5(2), 93–108 (1983).
10. ROM 3.1.99, Proyecto de configuración marı́tima de los puertos; Canales de acceso y
áreas de flotación (Ministerio de Fomento, Spain, 1999) (in Spanish).
11. O. Ditlevsen and H. O. Madsen, Structural Reliability Methods (Wiley, Chichester,
NY, 1996).
12. E. Castillo, M. A. Losada, R. Mı́nguez, C. Castillo and A. Baquerizo, An optimal
engineering design method that combines safety factors and failure probabilities:
Application to rubble mound breakwaters, J. Waterw. Port Coast. Ocean Eng. 130(2),
77–88 (2004).
13. M. A. Losada and M. I. Benedicto, Target deseign levels for maritime structures,
J. Waterw. Port Coast. Ocean Eng. 131(4), 171–180 (2005).
14. H. Günther, S. Hasselman and P. A. E. M. Janssen, Wamodel Cycle 4, Technical
Report No. 4, Deutsches KlimaRechenZentrum, Germany (1991).
15. J. T. Kirby and R. A. Dalrymple, An approximate model for nonlinear dispersion in
monochromatic wave propagation models, Coast. Eng. 9, 545–561 (1986).
16. A. Falqués, On the diffusivity in coastline dynamics, Geophys. Res. Lett. 30(21),
211910.1029/2003GL017760 (2003).
17. A. Baquerizo and M. A. Losada, Longitudinal current induced by oblique waves along
coastal structures, Coast. Eng. 35, 211–230 (1998).
August 14, 2009 9:1 9.75in x 6.5in b684-ch33 FA
Chapter 34
Hwung-Hweng Hwung
Tainan Hydraulics Laboratory
National Cheng Kung University, Tainan 70101, Taiwan
hhhwung@mail.ncku.edu.tw
The history of human utilization of the coastal area can be traced to hundreds
years ago, when they consumed the natural resources through fishing and pas-
turing. In the previous century, human beings built harbors and dikes to prevent
storms and waves. When it came to the past decades, large-scaled reclamation
lands were developed for the purposes of relaxation, aquaculture, and industry.
Consequently, a large part of ecologic environment in the coastal area was seri-
ously sabotaged. Therefore, this chapter attempts to categorize and explain dif-
ferent types of utilization around the world, to thoroughly review the impacts
caused by local constructions and other factors, and to scheme applicable pro-
grams for response and reparation. Last, the monitoring data of land subsidence
and ecologic transformation in the newly developed industrial area in western
Taiwan serves as good lessons to the ocean engineers that they should work with
other specialists of environment, ecology landscape, and management to solve the
related problems. Only by doing this, the abundant resources in the coastal area
could be preserved and utilized in a sustainable way.
34.1. Introduction
The coastal area is the entity of land and water approximate to both terrain and sea.
According to the US Commission on Marine Science, Engineering, and Resources
(1969), the definition of the coastal area is “the part of the land affected by its
proximity to the sea, and that part of the sea affected by its proximity to the land
as the extent to which man’s land-based activities have a measurable influence on
water chemistry and marine ecology.” Many researches on marine biologic resource
have demonstrated that abundant ecologic resources, including planktons, algae,
and nutritional salt, are contained in the coastal area of natural wetland, river
mouth area, and lagoon. These resources not only serve as the food to aquatic
animals and zoobenthos, but also provide functions like water purification, pre-
venting wave and sand drift and are equipped with landscape value. This area is a
very sensitive and crucial part in the process of ecologic evolution of the earth.
953
August 17, 2009 11:9 9.75in x 6.5in b684-ch34 FA
Earlier utilization of the coastal area was out of human instincts and was based
on its natural condition. A natural gulf with perfect shield from violent storms in
the outer sea often served as a port for ships looking for protection. Fishery was also
restricted to the coastal area. Van de Ven1 mentioned in his Man-Made Lowlands:
History of Water Management and Land Reclamation in the Netherlands that in
the centuries after 800, the Dutch became the decisive factor in the formation and
deformation of the land. The peat areas behind the coast were reclaimed by artificial
drainage, and in the newly reclaimed peat areas, both arable farming and animal
husbandry were practiced.
That is the earliest recorded utilization of coastal area by humankind with
engineering. Under their special circumstances, the Dutch have invested enormous
efforts in the development of coastal area and never stopped their fighting with the
nature until this century. Other cases of human utilization of the coastal area can be
found in the History and Heritage of Coastal Engineering, edited by Kraus,2 which
documented the history of coastal engineering and development of 15 industrialized
countries.
Before the mid-20th century, the utilization of the coastal area was restricted to
its natural conditions. However, after the Second World War, coastal engineering
doubled its speed of advancement and evolved into multiple applications. Wiegel
et al.3 reported that it was not until the late 1960s that the environmental awareness
began to develop in the USA and nurtured the constitution of The Federal Coastal
Zone Management Act in 1972, which served as a legal basis of administration on
the utilization and development of the coastal area.
After the Second World War, in the mid-20th century, the world economy accel-
erated, the number of population boosted, and the needs of lands were conse-
quently desperate. Therefore, human beings began to expand their activities and
construction to the shores. The over-exploitation finally led to a vital damage to the
coastal ecosystem. As a consequence, in the 1970s, urbanized countries worldwide
launch a series of regulations to control the development, utilization, and man-
agement of the coastal area. Based on the principle of sustainable access to the
coastal area, this chapter intends to organize the technological engineering and
the experiences that ever been adopted by countries worldwide in the utilization of
the coastal area to work as a reference for the future entree into the coastal area.
and urban drainage within their borderline. As a result, many sewage treatment
plants and associated factories were built along the coast for convenient operation
and management.
The development and utilization of coastal area, because of the engineering oper-
ation, often caused impacts in the region in certain dimensions. Sometimes, the
influence appeared immediately; sometimes, it took long time to be observed.
Among those impacts, the change of natural environment is inevitable and often
covers a wide scope of land and sea, affecting ecologic, hydrologic, and geologic
system and the landscape of the region. In addition, issues derived from coastal
development and utilization, including inland drainage, sewage disposal, traffic flow,
and air pollution control, are crucial problems needed to be solved one by one.
Moreover, some positive and negative social and economic effects will also emerge
at the same time during the process. Because the cause and effect of coastal devel-
opment and utilization cannot be evaluated in a short period, thorough investi-
gation, exhaustive study, and long-term monitoring of the operation are necessary
to every case related to the development and utilization of the coastal area. It is nec-
essary to provide a systematic explanation on what kinds of environmental impact
will be triggered by the development and utilization of the coastal area, under con-
struction or in operation.
of foreign species. This is a cross-field issue and needs consultation from the profes-
sional specialists before the construction really takes place.
create lands. Over usage of the sea sands not only changes the slope of the seafloor,
but also accelerates the scour of the neighboring coastal area.
34.4. Strategy
Phase II: To hold the kiko meeting for the environmental impact evaluation.
At least one year or four seasons complete investigation, research, analysis, and detailed
planning.
Monitor
o to and
a d track
t ac the
t e environmental
e v o e ta impact
pact assessment
assess e t of
o the
t e constructing
co st uct g project.
p oject.
Fig. 34.1. The procedures of environmental impact assessment on coastal development and
utilization.
survey, (v) sediment transport, (vi) nearshore current, (vii) coastal flying sand, (viii)
STC profile, (ix) coastal water quality, (x) ground water and water quality, (xi)
regional water discharge, (xii) geologic investigation, (xiii) ocean ecology investi-
gation, (xiv) zooplankton and phytoplankton, (xv) ecology of benthos, (xvi) heavy
metal in benthos, (xvii) fishing economy, (xviii) air quality, (xix) noise, (xx) traffic
condition, and (xxi) database and GIS establishment.
Complete environmental background information not only serves as the basis
of impact evaluation and preliminary engineering design, but also functions as the
reference to the recovering work in case the ecosystem is sabotaged.
Fig. 34.2. “Formosa Sixth Naphtha Cracking Project” in Yun-lin Offshore Industrial Park.
(a) Before construction. (b) Construction completed (No. 6 Naphtha Cracking Project).
SPT-N (after DC)
Fig. 34.3. SPT-N value (N value of standard penetration test) by dynamic consolidation method.
From the report of Formosa Plastics Group,4 there were two major important
methods applied to the reinforcement of geologic structure during construction,
which were “dynamic consolidation method” and “preloading method.” Test results
after the practice of above two methods, SPT-N value (N value of standard pene-
tration test) in Fig. 34.3 and CPT-Qc value (Qc value of cone penetration test) in
Fig. 34.4, demonstrated remarkable improvement in the weight capacity of reclaimed
land. Even the anti-liquefaction ability of the land was significantly advanced eval-
uated by the standards of Seed et al.5 On September 21, 1999, a 7.3 magnitude
(Richter scale) earthquake happened in central Taiwan; the site of Formosa Sixth
August 17, 2009 11:9 9.75in x 6.5in b684-ch34 FA
Fig. 34.4. CPT-Qc value (Qc value of cone penetration test) by preloaded soil.
Naphtha Cracking Project was only 50 km from the epicenter. However, not a single
building in the area was damaged and no soil-liquefaction happened. The lesson
told us how important the geologic reinforcement is when the coastal area is under
construction. Ocean engineers need to put special emphasis on this work.
Fig. 34.5. Subsidence records of two stations in Miaoli of “Formosa Sixth Naphtha Cracking
Project.”
pipelines, roads, and drainage system; (iii) increasing the amount of reclamation;
and (iv) producing leaks on the dike, bank, and breakwater, spoiling their func-
tions of preventing storm waves from the outer sea, and then block the operation
of the port. Therefore, when launching the utilization and development program
in the coastal area, engineers need to be especially careful about the possible land
subsidence caused by the construction work and need to have backup plans when
such kinds of situation happened so that the damage to the public facilities and
structures could be reduced to the minus.
2632000
2631900
2631800
2631700
2631600
(m)
2631500
2631400
2631300
2631200
2631100
2631000
162000 162100 162200 162300 162400 162500 162600 162700 162800 162900 163000
(m)
2632000
2631900
2631800
2631700
2631600
(m)
2631500
2631400
2631300
2631200 #90
2631100
2631000
162000 162100 162200 162300 162400 162500 162600 162700 162800 162900 163000
(m)
Fig. 34.7. The scouring area in Miaoli area gradually enlarges dated March 2002.
August 17, 2009 11:9 9.75in x 6.5in b684-ch34 FA
an effective and economic way to prevent the sea bottom from scouring. This case
indicates that the construction of coastal structures will inevitably change the flow
field and further generate unexpected impacts on the environment. Therefore, the
success of coastal development activities hinge on the sound investigation plans.
120
(a)
100
cm/sec
80
60
40
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
25
(b)
20
cm/sec
15
10
0
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
360
(c)
270
degree
180
90
0
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
Fig. 34.8. Long-term current velocity in Miaoli Industrial harbor and Sin-Sing at YLCW station:
(a) average amplitude of current velocity, (b) average current velocity, and (c) current direction.
August 17, 2009 11:9 9.75in x 6.5in b684-ch34 FA
Moreover, the direction of current directs between 25◦ and 90◦ no matter if it is
under high tide or low tide, as shown in Fig. 34.8(c). This is a dramatic difference
before and after the construction of west breakwater. Based on the long-term data
analysis, it is clearly shown that the coastal development activity will change the
flow field and even cause the change of ocean characteristics. This change will further
affect the ocean environment and ecology. More attention should be paid on this
issue.
3000
2500 11.13.15
Abundance (ind./M3)
2000
1500
Construction of Sin-Sing
1000
500
0
9 2 4 6 8 11 3 5 10 1 3 5 7 11 2 5 9 11 2 5 8 11 2 5 8 1 3 5 9 11 2
`93 `94 `95 `96 `97 `98 `99 `00 `01
Month
700
3.5.7.9
3.5.7.9
(b) Noctiluca stations
600 11.13.15
11.13.15
Construction
Constructionof
ofMialiao
Mialiao
Abundance (ind./M3)
500
Construction
Construction of
of Sin-Sing
Sin-Sing
400
300
200
100
0
9 2 4 6 8 11 3 5 10 1 3 5 7 11 2 5 9 11 2 5 8 11 2 5 8 11 3 55 99 11
11 22
`93 `94 `95 `96 `97 `98 `99 `00 `01
`01
Month
Fig. 34.9. Monitoring data in Tan Shi area for (a) zooplankton and (b) noctiluca during 1933–
2001.
August 17, 2009 11:9 9.75in x 6.5in b684-ch34 FA
120
(a) Fish eggs and larvae stations 3.5.7.9
11.13.15
100
Abundance (ind./M3)
80
60
40 Construction of Mialiao
Construction of Sin-Sing
20
0
9 2 4 6 8 11 3 5 10 1 3 5 7 11 2 5 9 11 2 5 8 11 2 5 8 1 3 5 9 11 2
`93 `94 `95 `96 `97 `98 `99 `00 `01
Month
200
3.5.7.9
(b) Crab and shrimp stations 11.13.15
160
Abundance (ind./M3)
Construction of Mialiao
120
80 Construction of Sin-Sing
40
0
9 2 4 6 8 11 35 10 1 3 5 7 11 25 9 11 2 5 8 11 2 5 8 1 3 5 9 11 2
`93 `94 `95 `96 `97 `98 `99 `00 `01
Month
Fig. 34.10. Monitoring data in Tan Shi area for (a) fish eggs and larvae and (b) crab and shrimp
during 1933–2001.
34.5. Conclusion
In the 20th century, every country is desperate for a rapid economic development.
As a consequence, more and more high-tech products were produced while fewer
and fewer resources were left. Energy and water resources become needed; even food
has turned out to be scarce. When the resources on the lands have been utilized
to their limits, people turn to look for development in the coastal area. Expansion
of aquaculture industries, construction of harbors, nuclear and steam power plants,
development of industrial park, recreational activities, and the technology of ocean
outfalls are examples of the utilization of the coastal area. However, no matter what
kinds of human activities will inevitably cause impacts to nature. Sometimes, the
damage is forever and can never be retreated.
In the new era, ocean and coastal engineers need to have visionary foresight and
cross-fielded thoughts to maintain sustainable utilization of the coastal resources.
August 17, 2009 11:9 9.75in x 6.5in b684-ch34 FA
In achieving this goal, they must: (a) based on their professional trainings, lower
the degree of compromise that might be caused to the coastal area when con-
struction is in progress; (b) cooperate with specialists of environment, ecology, and
landscape so that the utilization of the coastal area could merge with its natural
surroundings; (c) keep in mind the idea of sustainable management when con-
ducting construction programs in the costal area instead of the traditional engi-
neering methods; and (d) file the development project with ample budget to uphold
the monitoring program in the area after construction so that the transformation
of each stage will be clearly documented as future reference.
In the immense universe, we have only one planet to live; that is our hometown
for nurturing the future generations. We, ocean and coastal engineers, must dedicate
ourselves to maintaining the ecosystem of the coastal area, extenuating the impacts
of construction to the natural environment and sticking to the vision of sustainable
management of the coastal area.
Nevertheless, one proverb we must always keep in mind: there is no such thing
as free lunch.
References
Chapter 35
Paul D. Komar
College of Oceanic and Atmospheric Sciences
Oregon State University
Corvallis, Oregon 97331, USA
pkomar@coas.oregonstate.edu
Jonathan C. Allan
Coastal Field Office
Oregon Department of Geology and Mineral Industries
313 SW 2nd, Newport, Oregon 97365, USA
jonathan.allan@dogami.state.or.us
Peter Ruggiero
Department of Geosciences
Oregon State University
Corvallis, Oregon 97331, USA
ruggierp@geo.oregonstate.edu
Storm intensities and the heights of ocean waves are being affected by Earth’s
changing climate, including the extreme waves generated by both hurricanes and
extratropical storms, attributed to global warming. The increases in wave heights
have been documented by buoy data collected in the North Atlantic and North
Pacific, with the rates of increase being important to engineering design and in
the development of coastal-hazard zones. Reviews are presented about the climate
controls and investigations that have documented the wave-height increases. The
evidence is variously presented as progressive increases in the annual averages of
the measured significant wave heights, and decadal shifts in histograms for the full
ranges of measured waves. A review is also provided of the statistical techniques
available that account for these climate-induced changes in the extreme-value
projections used in engineering and management applications.
971
August 14, 2009 9:7 9.75in x 6.5in b684-ch35 FA
35.1. Introduction
The heights, periods, and energies of storm-generated waves are the dominant envi-
ronmental factors that affect coasts, with their most extreme measured occurrences
having resulted in the erosion of beaches and backshore properties, and damaged
infrastructure such as harbor jetties and breakwaters. Important is the development
of a wave climate, a documentation of the wave conditions experienced during past
decades, with projections of the most extreme waves that could occur in the future,
potentially posing a threat to the coast.
The establishment of wave climates has been of fundamental importance to
coastal engineers and scientists, the former requiring them in the design of ocean
and coastal structures so they can withstand the forces of the most severe storm
waves and continue to function through their desired lifetimes. The requirements
of coastal scientists are similar, for example, in the establishment of coastal-hazard
zones (setback lines) that will maintain homes and other developments safe from
extreme storms. This places demands on the development of the wave climate,
ideally that the wave data span several decades of buoy measurements or storm-
hindcast assessments, so that reasonably confident projections can be made of the
most extreme waves expected during a century or longer (e.g., the 100-year event,
that having a 1% probability of occurrence during any year). These become the
design parameters in applications; their underestimation could lead to the failure
of a structure, whereas an overestimation would result in construction at an unnec-
essarily inflated cost. Similarly, an underestimation of the most extreme waves and
associated beach processes could result in recommended setback distances that are
insufficient, leading to the eventual loss of homes and other developments.
The establishment of wave climates has become more of a challenge due to
Earth’s changing climate, recognized mainly in the context of global warming that
according to the assessments of nearly all climatologists is the result of the release
of greenhouse gases by humans into the atmosphere. Most of the focus on potential
coastal impacts from global warming has been the expected accelerated rates of
sea-level rise as projected by reports of the International Panel on Climate Change.
However, changes in the global climate can also be expected to produce an intensifi-
cation of storms, resulting in higher-generated waves and progressive shifts in wave
climates toward more extreme conditions. This had already been documented in the
northeast Atlantic1,2 and Pacific3–5 for waves generated by extratropical storms,
and for Atlantic hurricanes with increasing wave heights having been measured by
buoys along the US East Coast.6,7 This necessitates that wave climates account for
multi-decadal, progressive increases in storm intensities and their generated waves,
including shifts in the extreme design conditions needed in engineering and coastal-
management applications.
This chapter is an expansion of that by Goda,8 which provided a basic summary
of wave climates and their analysis techniques, the objective here primarily being
to review the effects of Earth’s changing climate and how it can be dealt with in
projections of future extreme wave conditions. This chapter begins with a summary
of the climate related factors that affect storm intensities and wave generation,
including long-term factors such as global warming and accompanying increases in
August 14, 2009 9:7 9.75in x 6.5in b684-ch35 FA
Ocean Wave Climates: Trends and Variations Due to Earth’s Changing Climate 973
ocean water temperatures, but also the existence of interannual events such as a
major El Niño and climate cycles such as the Pacific Decadal Oscillation (PDO).
The presentation then turns to the available sources of wave data, and a review of
the studies that have documented progressive increases in wave heights and their
climate controls. Of particular interest is how these climate-controlled changes have
affected the coastal-wave climates, and how assessments of extreme wave conditions
required in applications can account for those changes.
There has been growing concern regarding Earth’s changing climate and its environ-
mental consequences, by the public at large and by climatologists whose research has
been directed toward the causes and consequences of those changes. The products
of that research are of direct interest in this chapter in so far as the changing climate
has affected the intensities of storms, accounting for trends of increasing heights and
periods of waves they have generated. Here, we provide a summary of the climate
controls, followed in subsequent sections by the resulting trends of measured waves.
The climate change of principal concern is the progressive global warming
that has increased the numbers and intensities of hurricanes, and likely also has
intensified the extratropical storms at higher latitudes. Other climate controls
include variations in water temperatures and atmospheric-pressure differences that
determine the strengths of the storms, and episodes of interannual climate extremes
such as the development of major El Niños in the Pacific that intensify the storms
and alter their tracks, temporarily affecting the wave conditions along the western
shores of North America. Each of these climate controls needs to be recognized for
its potential effects on wave climates, and how they might change in future decades.
The progress of global warming is well established through worldwide mea-
surements of atmospheric temperatures, which on average have increased over the
decades, though some debate remains concerning whether this increase is due to
human activities that have resulted in the release of greenhouse gases into the atmo-
sphere, or represents some long-term natural climate cycle. Of interest to the inten-
sities of storms and heights of the waves they generate are whether ocean-water
temperatures have increased in parallel with the atmosphere. This connection has
a direct consequence to the potential increases in the intensities of tropical cyclones
(hurricanes in the Atlantic and eastern Pacific, typhoons in the northwest Pacific,
and cyclones in the southwest Pacific). It is well recognized that elevated water
temperatures in the tropics serve as the fuel for the heat engine that is responsible
for the generation and intensity of hurricanes.9 As a consequence, global warming
can be expected to result in progressively greater numbers of tropical cyclones and
their average intensities.
It is well established that during the recent decades sea-surface temperatures
over most tropical oceans have increased. In the North Atlantic the most consistent
increase has taken place since about 1970, of interest in that as will be reviewed
below buoy measurements of hurricane-generated waves have documented pro-
gressive increases spanning that period. Mann and Emanuel10 undertook statistical
August 14, 2009 9:7 9.75in x 6.5in b684-ch35 FA
analyses of the North Atlantic water temperatures, finding that there is a close
correspondence with the decadal-smoothed numbers of hurricanes, both having
experienced dramatic increases since 1970. Trenberth and Shea11 analyzed water
temperatures in the tropical Atlantic between June and October 2005, the year with
record numbers of high-category hurricanes, including Katrina and Rita in the Gulf
of Mexico, and found that the average temperature was 0.92◦C above the average
recorded between 1901 and 1970, the 70-year period that predated the recent surge
in worldwide temperatures. Of particular significance is the study by Emanuel,12
who measured the potential destructiveness of hurricanes in the North Atlantic,
quantified as the total power dissipation (TPD) of each storm that represents the
integration of the magnitude of the cube of its surface wind speeds over the surface
area affected by the storm and spanning its lifetime. His graph of the annual varia-
tions in the total TPD from 1930 through 2002 for the annual numbers of tropical
storms and hurricanes closely followed the same pattern as a curve for the sea-surface
temperatures measured during September (the peak of the hurricane season).
Some debate remains among climatologists as to the quality of the measure-
ments documenting the wind speeds and atmospheric pressures within hurricanes,
and how our ability to collect that data has improved over the decades.13 Of par-
ticular importance has been the use of satellites beginning in the 1970s, and there
appears to be consensus among climatologists that the intensities of hurricanes
have increased since the advent of that technology to monitor the storms. However,
uncertainty persists prior to 1970 due to the poorer quality of the data and because
the technology has evolved, which might systematically have affected the storm-
intensity assessments.
The sophisticated global-climate numerical models developed by climatologists
yield a clear indication of the expected increases in hurricane intensities due to
global warming, and also indicate that the intensities of extratropical storms at high
latitudes in the Northern Hemisphere will increase. However, complicating those
progressive trends are modes of climate variability that affect both the atmosphere
and ocean, a variability that can range from years to decades. The greatest source
of interannual variability in the weather and climate is related to the occurrences
of strong El Niños as part of the full range of climate conditions from El Niños to
La Niñas. The primary attention has been directed toward major El Niño events
such as those that occurred during 1982–1983 and 1997–1998, which resulted in
more intense winter storms in the northeast Pacific, whose tracks crossed the coast
of California rather than having followed their normal paths at more northerly
latitudes.14 The result was extreme wave conditions and property erosion along the
shores of southcentral California. Enhanced erosion also occurred at the latitudes
of northern California, Oregon, and Washington, due in large part to a shift in the
wave directions compared with their normal wave climates, the waves arriving more
from the south–southwest than normal, resulting in the northward displacement of
sand on beaches and “hot spot” erosion north of headlands and tidal inlets.15–18
This illustrates that wave directions and their variations can also be an important
component of wave climates. At the same time it has been shown that the occurrence
of a major El Niño in the Pacific results in decreased hurricane activity in the
Atlantic, reducing coastal impacts along the US East and Gulf Coasts19,20 ; this
August 14, 2009 9:7 9.75in x 6.5in b684-ch35 FA
Ocean Wave Climates: Trends and Variations Due to Earth’s Changing Climate 975
decrease is caused by the formation of high-altitude winds that shear the tops off
hurricanes before they can fully develop.
An El Niño occurs at irregular intervals of approximately 2–7 years, although
the average is once every 3–4 years, and generally each event persists for about
a year. Other years represent “normal” conditions, or range into the development
of a La Niña. This range is characterized by a number of changes in the environ-
mental conditions, in the oceans and atmosphere, one of the most important being
that the eastern equatorial Pacific sea-surface temperatures (SST) are anomalously
warm during an El Niño, whereas during a La Niña the water is unusually cold.
This change is a response to differences in atmospheric pressures in an east–west
direction along the equator through the tropics, and its effect on the Trade Winds,
an atmospheric variation that is described by the Southern Oscillation; this com-
bination of ocean and atmosphere components is commonly referred to as the El
Niño–Southern Oscillation or ENSO climate event.
The range from El Niños to La Niñas can be characterized by environmentally
based indices, the simplest being the SST in the eastern equatorial Pacific, but
more advanced indices incorporate additional environmental responses. Figure 35.1
is a graph of the annual variations in the Multi-variate ENSO Index (MEI), which
includes a number of responding environmental factors21 ; the more positive the
value of MEI the stronger the intensity of the El Niño, with the major 1982–1983
and 1997–1998 El Niños seen to be distinct peaks in the graph, whereas a negative
value denotes the occurrence of a La Niña. It is evident in this diagram that there
has been multi-decadal cycles of alternating dominance by El Niños versus La Niñas,
with more frequent occurrences and greater intensities of El Niños from about 1976
to 1998. In contrast, the period from 1946 to 1976 was dominated by La Niñas. This
Fig. 35.1. Variations in the MEI through the 20th century, a measure of the occurrence of El
Niños versus La Niñas.
August 14, 2009 9:7 9.75in x 6.5in b684-ch35 FA
cycle corresponds to the Pacific Decadal Oscillation (PDO), defined on the basis of
a multi-decadal pattern of monthly SST anomalies in the North Pacific poleward
of 20◦ N, that is at higher latitudes than the environmental factors used to define
the MEI and occurrences of El Niños. The PDO cycle was formulated by Mantua
et al.22 as a climate control on the abundance of salmon at those higher latitudes,
but from the cycles of El Niños seen in Fig. 35.1 it is apparent that the PDO also
describes the variations in that climate event, and therefore the changes in tracks
of storms and the waves they generate in the northeast Pacific.
Of interest in the context of wave climates is that the “warm” PDO phase dom-
inated by frequent and strong El Niños, 1976 to about 1998, corresponds approxi-
mately to the time frame of buoy measurements of waves that have served as the
primary basis in the development of wave climates along the shores of the USA.
From the recent apparent shift to the “cold phase” of the PDO in 1999, with the
likelihood of there now being more frequent La Niñas, fewer strong El Niños, we
can expect that there will be responses in the US West Coast wave climates that
could persist for the next 25 years: for example, reduced extreme wave events along
the coast of California compared with the past 25 years. On the other hand, this
shift in the PDO could become a factor in more frequent and intense hurricanes in
the Atlantic.
Another important index describing the climate of the Pacific, relevant to storm
intensities and wave generation, is the North Pacific Index (NPI), the difference
between the atmospheric pressures within the Aleutian Low and Hawaiian High.
It can be expected that the greater the difference in these pressures, the stronger
the Westerlies blowing at mid-latitude to high latitude across the Pacific and the
stronger the intensities of extratropical storms that follow tracks generally from west
to east, crossing the western shores of the USA and Canada. Thus, one could antici-
pate that the NPI would provide some degree of prediction of the heights of waves
occurring along those coasts, and as will be reviewed below this has been demon-
strated to be the case. However, there is a reasonably strong correlation between
the MEI and the NPI, even though their assessments incorporate environmental
conditions at different latitudes. This correlation confirms the observation that the
occurrence of a major El Niño has climate consequences at all latitudes, not just
near the Equator, and extends beyond the area of the Pacific Ocean, across the
North American continent and into the Atlantic. One significance of this correlation
is that in analyses of the climate controls on storm-generated waves, there may
be dependencies between measured wave heights and multiple indices, requiring an
interpretation as to which has been the primary control.
There is a comparable set of climate indices for the North Atlantic that poten-
tially exert controls on the intensities of storms, both extratropical storms and hur-
ricanes, although those relationships are not as firmly established as in the Pacific
and there is some debate as to their analyses and interpretations. Comparable to
the NPI for the North Pacific, the atmospheric pressure difference between the
Iceland Low and Azores High defines the North Atlantic Index (NAI), with research
having shown that an increase in this index produces higher intensity storms and
the heights of the waves they generate.23 The Atlantic Multi-decadal Oscillation
(AMO) is somewhat analogous to the PDO in the Pacific, based in practical terms
August 14, 2009 9:7 9.75in x 6.5in b684-ch35 FA
Ocean Wave Climates: Trends and Variations Due to Earth’s Changing Climate 977
on the SST variability at high latitudes in the North Atlantic.24,25 Its variations
are attributed to changes in water currents as part of the global ocean circulation,
which carries warm surface water via the Gulf Stream into the northern latitudes of
the Atlantic where it cools and sinks, returning as cold water at depth southward
across the equator. The North Atlantic deepwater formation is difficult to measure
and to track over time, so assessments of the AMO are made in terms of the SST
variability since it is related to the overturning circulation. In its warm phase, the
decades when the temperature is elevated, some researchers have associated the
AMO with increased hurricane activity,26 but others have disputed this correlation
and even the analysis procedures used to define the AMO.
Research by climatologists of extratropical storms in the North Pacific has doc-
umented a progressive increase in their intensities over the decades, possibly as far
back as the 1950s.27 It has been suggested that this increase is the result of global
warming and higher temperatures of the ocean’s surface water, but it also has been
suggested that the increase in storm intensities is affected by particulate pollution
in the atmosphere, originating in southeast Asia. At the same time it is clear that
there is a significant interannual to decadal variability in storm intensities and wave
climates in the northeast Pacific, associated with the range from major El Niños to
La Niñas that affects the intensities of storms, the tracks they follow, and the lat-
itudes at which they cross the western shore of North America. There are similar
trends of increasing intensities and variability of extratropical storms in the North
Atlantic, related to the climate indices reviewed above, but of particular interest
in the Atlantic is the evidence that the numbers and intensities of hurricanes have
increased since the 1970s, posing a major threat to the eastern shores of North
and Central America. As reviewed below, these changes in Earth’s climate and its
storms are reflected in the wave climates for the North Atlantic and Pacific.
The data required for the development of a wave climate are preferably derived
from programs that for decades have collected hourly or daily measurements of wave
spectra, significant wave heights based on the energy content of the spectra, and a
representation of the dominant wave periods such as the peak-energy period within
the wave spectra. Such programs rely mainly on the deployment of moored buoys,
or recording devices attached to a fixed structure such as an offshore oil platform.
Lacking such direct measurements, or to supplement that data if the records span
too few years to establish a satisfactory wave climate, wave-hindcast techniques have
been employed based on wave assessments derived from storm fetch areas, wind
speeds, and the durations of the storms. Beyond these standard sources of wave
data, information on long-term changes in the wave climate may be obtained from
visual observations from ships, although the quality of that data is questionable.
Measurements of waves along the coasts of the USA are derived primarily from
buoys deployed by NOAA’s National Data Buoy Center (NDBC), and their hourly
measurements are commonly used to define wave climates. Furthermore, the 25- to
35-year records available from many NDBC buoys now have sufficient lengths to
August 14, 2009 9:7 9.75in x 6.5in b684-ch35 FA
permit the analyses of trends of increasing or decreasing wave heights, the presence
of climate-controlled cycles, or annual variations due to events such as a major
El Niño. Prior to August 1979 the data were collected at 3-h intervals, while since
that time the measurements have been hourly. Otherwise, the collection methods
and analysis procedures have for the most part remained the same over the decades.
In 1984 NDBC changed their analysis procedures to improve the inclusion of long-
period motions in the wave spectra. The details of that change and how it affected
the spectra and derived significant wave heights has been analyzed by Earle et al.28
The change is important only where there is a strong mix of swell derived from
distant storms and locally-generated sea having a wide spectrum of lower wave
periods, this causing the fixed-hull accelerometers commonly used by NDBC to
introduce spurious energy into the combined spectra. This problem was largely
accounted for by analysis procedures prior to 1984, but it became apparent that
further improvements were needed. Earle et al.28 produced wave spectra analyzed
by both the pre- and post-1984 procedures to illustrate the potential differences,
and their examples indicate that post-1984 the reported significant wave heights
could in the most extreme cases be increased by roughly up to 10%, the actual
increase depending on the relative proportions of sea and swell being measured.
This difference can be expected to be most important in analyses of specific storm-
wave events having broad spectra, but would likely have a small effect on waves
dominated by swell and on monthly to yearly averages needed in wave climates. In
our investigations of climate-controlled trends of increasing wave heights along the
US West and East Coasts, reviewed below, we did not detect any upward “steps”
in the data trends in 1984 that might have resulted from a change in the analysis
procedures.
Few direct measurements of waves by buoys have been collected in the southern
hemisphere, spanning sufficient lengths of time for the development of wave climates.
For this reason wave hindcasts based on meteorologic records have been the primary
source of data for the establishment of wave climates. This is illustrated by the
hindcasts undertaken by Gorman et al.29 for the coasts of New Zealand. Their
analyses employed the wave-generation model WAM (WAve Model) based on data
for the daily winds across the expanse of the South Pacific, the latitudes from
10◦ S near the Equator southward to the coast of Antarctica, and 100◦ E to 220◦E
in longitude, New Zealand being at the center of that area. The hindcasts were
made at 3-h intervals for the 20-year period from 1979 through 1998, with graphical
results presented for eight sites around the New Zealand Coast. The distributions
of hindcast significant wave heights showed good agreement with data derived from
satellites, although the hindcasts tended to underestimate the occurrences of the
largest wave events; it is uncertain whether this difference represents an error in the
satellite measurements or the hindcast results. Gorman et al.29 did not identify a
statistically significant trend of changing wave heights, which might have reflected
a long-term climate control.
The longest records of ocean wave heights are those that have been collected by
visual estimates from ships, for some areas extending back into the 18th century, but
generally only for major shipping routes, with global coverage being available only
since about 1950. While this data is sometimes incorporated into wave climates,
August 14, 2009 9:7 9.75in x 6.5in b684-ch35 FA
Ocean Wave Climates: Trends and Variations Due to Earth’s Changing Climate 979
its quality remains open to question. Nevertheless, these data have proved to be
valuable in studies of climate controls on variations in wave heights spanning
centuries, extending into the past well beyond the analyses based on buoy mea-
surements or wave hindcasts. An example of such analyses is that of Gulev and
Grignorieva,30 of wave heights in the North Atlantic and North Pacific from the
1890s to 2002, relating their long-term variations to the climate indices reviewed in
the previous section.
Wave climates are most easily defined on coasts that are dominated by one type
of storm system; examples are the North Pacific shores of Canada and the USA,
and those of Europe, where the waves are generated by extratropical storms. Other
coasts in effect have two wave climates, the Atlantic shore of the USA being a prime
example, with one climate consisting of waves from extratropical storms (North-
easters) and the second being waves generated by tropical cyclones (tropical storms
and hurricanes). These two types of storm systems are fundamentally different in
their modes of formation, with largely separate seasons of dominance, and can
be expected to have distinctive climate controls on their intensities and generated
waves. They will therefore be considered separately in this section and throughout
the remainder of this chapter whenever dealing with coasts that are impacted by
waves having these different origins.
Fig. 35.2. The increase in annually averaged significant wave heights measured by a wave recorder
offshore from southwestern England.2
the same throughout those 28 years, and were even undertaken by the same indi-
viduals. The charts were prepared by a combination of meteorologic analyses and
direct observations of winds and waves. The resulting charts cover the entire North
Atlantic, but the published analyses focused on two areas, west of Ireland and
northeast of Newfoundland. For the annual averages the regression for the area west
of Ireland yielded a rate of increase of 0.027 m/year, in close agreement with the
measurements by the Seven Stones Light Vessel. The rate of increase in the western
Atlantic near Newfoundland was similar, 0.023 m/year. However, analyses of wave-
buoy data further to the south along the East Coast of the USA did not find an
increase for the winter wave heights, those generated by extratropical storms.7
A review of the storms and changing wave conditions in the North Atlantic
was undertaken in the late 1990s by the European project WASA (Waves and
Storms in the North Atlantic), to examine the hypothesis of worsening storms and
wave climates during the 20th century.32 Their overall conclusion was that while
the wave heights have increased in the recent decades as measured by the Seven
Stones Light Vessel, the storms and waves earlier in the century had undergone
significant cycles on timescales of decades. It was further concluded that the present-
day storm intensities and wave heights are comparable to those at the beginning of
the 20th century. Their conclusions are generally supported by the study of Gulev
and Grignorieva30 involving analyses of the visual assessments of wave heights from
ships collected during the past two centuries. In the northeast Atlantic the annual
averages of the significant wave heights showed a distinct increase after about 1955,
again in agreement with the Seven Stones Light Vessel data, but earlier in the
century there were clear cycles that correlated with variations in the North Atlantic
Oscillation (NAO), with the wave heights at their maximum early in the century
being some 20 cm higher than at present.
August 14, 2009 9:7 9.75in x 6.5in b684-ch35 FA
Ocean Wave Climates: Trends and Variations Due to Earth’s Changing Climate 981
Attempts have been made to account for the measured wave-height increases
in the North Atlantic since the 1960s through comparisons with potential climate
controls. Bacon and Carter23 correlated the Seven Stones Light Ship data with the
NAI (the atmospheric pressure gradient between the Azores High and Iceland Low),
expected to be important to winds in the North Pacific. The measured pressure
gradients were found to have a high degree of interannual variability during the
first half of the 20th century, but beginning in about 1960 there was a distinct
progressive increase with a strong correlation with the annual and monthly averages
of measured wave heights.
Similar increases in wave heights have been found in the northeast Pacific, first
documented by Allan and Komar3–5 in 2000 with data from six NDBC buoys
along the US West Coast from south-central California northward to the coast
of Washington (and one buoy in the Gulf of Alaska). Figure 35.3 graphs the
decadal increases in annual averages of the winter wave heights measured by the
series of buoys, “winter” being taken as the months of October through March,
the dominant season of significant storms. The highest rate of increase is seen
to have occurred on the Washington Coast, slightly less offshore from Oregon,
with northern to central California being a zone of transition having still lower
rates of wave-height increases, until off the coast of south-central California there
has not been a statistically significant change in the wave heights. Figure 35.4
and Table 35.1 provide more detailed assessments of the wave-height increases
off the Washington Coast by NDBC buoy #46005, the series progressively repre-
senting more extreme assessments of the measured significant wave heights. It is
evident in Fig. 35.4 that the more extreme the assessment the greater the rate
of increase, with the averages of the winter waves having increased at a rate
0.032 m/year, while the annual averages of the five largest measured significant
wave heights had increased at a rate of 0.095 m/year. This progression is more
evident in Table 35.1 where additional regressions are included, together with results
from tests of the statistical significance of the linear regressions. Included is an
analysis of the annual averages of the significant wave heights for the entire year,
its regression yielding a rate of increase of 0.018 m/year, nearly identical to that
found in the North Atlantic by the Seven Stones Light Vessel (Fig. 35.2). Also
included in Table 35.1 is a column listing the ratios for the series of rates of wave-
height increases to this 0.018 m/year value representing the entire year. These ratios
are seen to progressively increase from 1.8 for the averages of the winter waves
to 6.0 for the regression of the single maximum measured significant wave height
each year.
It is seen in Table 35.1 that for the series of linear regressions, only those up
through the analysis of the annual averages of the five largest significant wave heights
are statistically significant at the 95% level. The increasing scatter of the data for
still more extreme assessments is evident in Fig. 35.4, for the maximum annual
measured significant wave heights, accounting for the loss of statistical significance
based on linear regressions. However, as will be reviewed in Sec. 35.6, this lack of
statistical significance most likely is due to the nonsymmetric distribution of the
most extreme wave heights, as analyses employing advanced analysis models are
statistically significant even for these annual maximum wave heights.
August 14, 2009 9:7 9.75in x 6.5in b684-ch35 FA
Fig. 35.3. Annual averages of the “winter” significant wave heights measured by buoys along the
US Pacific Coast.5
Allan and Komar5 analyzed the climate controls that might account for the
increasing wave heights in the northeast Pacific, but were unable to find a corre-
lation with the climate indices. This included the NPI, the difference in atmospheric
pressures between the Aleutian Low and Hawaiian High. However, the NPI was
August 14, 2009 9:7 9.75in x 6.5in b684-ch35 FA
Ocean Wave Climates: Trends and Variations Due to Earth’s Changing Climate 983
Fig. 35.4. Regressions for wave heights off the coast of Washington; the more extreme the
assessment, the greater the rate of increase.5
found to explain a significant portion of the annual variability in the data above and
below the regression lines, basically having accounted for part of the data scatter
seen in Figs. 35.3 and 35.4. The MEI equally accounted for this variability, expected
given the correlation between the NPI and MEI discussed earlier. The conclusion
was that a major El Niño, as measured by MEI, tends to result in higher wave con-
ditions all along the US West Coast, with the annual average plotting above the
regression lines representing the net decadal increases. At the same time, it is recog-
nized that a strong El Niño affects the meteorology of storms even at high latitudes
in the Pacific, and this accounts for the correlation with the NPI.
August 14, 2009 9:7 9.75in x 6.5in b684-ch35 FA
Although the focus of this review has been on the changing wave heights over the
decades, Allan and Komar5 also documented similar increases in the wave periods
measured by the US West Coast NDBC buoys, expected from the intensification
of the storms. The significance of this is that the range of nearshore processes on
beaches, such as wave runup elevations, depend on both wave heights and periods,
so increased runup elevations since the 1970s would have resulted from increases in
both parameters, important to coastal erosion and flooding.
Ocean Wave Climates: Trends and Variations Due to Earth’s Changing Climate 985
Fig. 35.5. Annual averages for significant wave heights greater than 3 m generated by hurricanes,
measured by buoys along the central US Atlantic Coast.7
of the annual averages that included only occurrences when the significant wave
heights were greater than 3 m, it having been shown that those higher waves during
the summer can be attributed in almost all cases to specific hurricanes, whereas
the lower waves represent calm intervals between storms. It is seen that there has
been a latitude dependence, with the highest rate of increase having occurred in the
south: 0.059 m/year (1.8 m in 30 years) for the Charleston buoy; 0.024 m/year for
the Hatteras buoy; and 0.017 m/year for Cape May. It will be shown in the following
section that the most extreme waves measured by these buoys during the summer
hurricane seasons have increased from on the order of 7.5-m significant wave heights
early in the buoy records to greater than 10 m during the 1996–2005 decade.
This increase in hurricane-generated wave heights could depend on several
factors, including changes from year to year in the numbers and intensities of storms,
their tracks that determined whether they traveled northward through the Atlantic
where their generated waves could be recorded by these buoys, and how distant
the hurricanes were from the buoys, whether they passed far offshore within the
central Atlantic or approached the US shore and possibly made landfall. Analyses
by Komar and Allan7 indicated that all of these factors have been important to
the observed wave-height increases, but the increased hurricane intensities found by
Emanuel12 in his analyses of measured wind speeds within the storms provided the
best explanation for the progressive increases seen in Fig. 35.5, since the numbers
and tracks of the storms showed considerable variability from year to year.
August 14, 2009 9:7 9.75in x 6.5in b684-ch35 FA
While trends in annual averages of measured wave heights, as graphed in Figs. 35.3–
35.5, document their progressive decadal increases, another meaningful depiction of
the changing wave climate is provided by histograms for the full range of measured
significant wave heights. Histograms in particular permit one to focus on increases
in the highest measured waves, and to relate them to the results from extreme value
analyses.
The traditional depiction of a wave-height histogram is to graph the frequencies
(percentages) of occurrence for individual significant wave heights within the range
of measurements, this approach having the advantage of normalizing the histogram
to the total number of observations, with the summation being 100%. There are
disadvantages, however, as the most extreme but rarely measured wave heights typ-
ically have frequencies less than 0.1%, the significance of which is obscure, and when
graphed using a linear scale these extreme waves disappear on the diagram, unfor-
tunate since they are most important to the wave climate in terms of applications.
In our investigations of climate controls on wave climates we have adopted an alter-
native procedure by graphing the actual numbers of observations, and employing a
log scale that has the effect of emphasizing the rare but extreme wave heights.35 In
this manner we can directly identify the occurrence of even the single most extreme
measured significant wave height measured over the years, which graphs as a 100 = 1
occurrence.
An example is shown in Fig. 35.6 based on the hourly significant wave heights
measured by the NDBC buoy seaward of Cape Hatteras, North Carolina; the
compiled data are for the “winter” months of November through the following
March, the season when the waves have been generated by extratropical storms
(Northeasters). In contrast to the decadal increase in the heights of summer waves
August 14, 2009 9:7 9.75in x 6.5in b684-ch35 FA
Ocean Wave Climates: Trends and Variations Due to Earth’s Changing Climate 987
Fig. 35.6. Histograms for the range of significant wave heights generated by extratropical storms
in the Atlantic, a histogram for the percentages of occurrence and a graph for the actual numbers
of height measurements.35
generated by hurricanes (Fig. 35.5), the winter waves have not experienced a sta-
tistically significant change, so we can compile a histogram for its full record of
measurements to obtain an assessment of the winter-wave climate.6,7,35 Figure 35.6
includes two depictions of the histograms for this wave record, one being the con-
ventional graph as percentages of observations using a linear scale, the second being
our alternate approach of graphing the actual numbers of observations of measured
significant wave heights with a log scale. Comparing the two it is apparent that the
highest measured waves do not show up on the conventional frequency histogram,
their frequencies being less than the thickness of the lines used to prepare the graph.
However, those extreme waves are readily apparent when graphed as the actual
numbers of observations on a log scale, even though they represent only 1 or 2 occur-
rences among the thousands of significant wave heights measured over the years.
It is seen in this example that the highest measured significant wave height during
the 30 years of buoy operation was 11 m, with heights down to about 8.5 m repre-
senting 10 or fewer observations. Nothing from that range of wave heights appears
on the conventional frequency graph, even though they are of greatest relevance in
applications.
Figure 35.6 also illustrates that it is convenient to compare the histogram with
the results derived from an extreme value assessment, in this example ranging
from the 25- to 100-year projected extremes based on the fit to a Weibul distribution
of the annual maximum significant wave heights.35 Having been graphed with a log
scale we can directly examine the actual numbers of measured occurrences of those
extreme value projections during the time frame of data collection. The 25-year
projected extreme is 11 m, and it is seen that two measured significant wave-height
August 14, 2009 9:7 9.75in x 6.5in b684-ch35 FA
Fig. 35.7. Histograms for the hurricane waves measured by the Cape Hatteras buoy, comparing
the numbers of occurrences from early in the record versus the recent decade, 1996–2005.35
August 14, 2009 9:7 9.75in x 6.5in b684-ch35 FA
Ocean Wave Climates: Trends and Variations Due to Earth’s Changing Climate 989
hurricanes that passed in close proximity to the recording buoy, the numbers of
occurrences generally involving fewer than 10 for a specific wave height, often only
1 or 2, with the irregularities in the histograms resulting from the sensitivity on a
log scale to these low numbers of observations.
It is seen in Fig. 35.7 for the Cape Hatteras buoy that in the 1976–1991 histogram
the highest significant wave height generated by hurricanes during the summer had
been 7.8 m, with an increase to 10.2 m in the 1996–2005 histogram. Although this is
consistent with the general increase in hurricane-generated waves measured by this
buoy, these highest values are in the range of the statistical 25- to 50-year extreme
value projections, with the 100-year extreme expected to be slightly higher. With
the buoy record of measurements having been only 30 years, measured significant
wave heights of 8–10 m could potentially have occurred at any time during those
decades, when the track of a fairly intense hurricane came near the recording buoy,
although the probability of that occurring has gradually increased with the average
intensities of the Atlantic hurricanes.
There is the potential for misinterpretations in comparing a pair of histograms
as in Fig. 35.7, in that there appears to be a marked difference in areas beneath the
two curves. The common interpretation (usually mistaken) is that this is due to a
greater number of measurements having been included in the 1996–2005 histogram
compared with that for 1976–1991. However, this difference persists even when the
data are graphed as percentages, normalized to the numbers of observations, with
each graph then having an identical area representing 100%. Furthermore, in some
of our analyses (as seen below in Fig. 35.8) the histogram for the earlier decade had
greater numbers of observations than the more recent decade, but the disparity in
Fig. 35.8. Histograms documenting changes in the ranges of measured significant wave heights
off the coast of Washington.
August 14, 2009 9:7 9.75in x 6.5in b684-ch35 FA
apparent areas beneath the curves remains much the same as seen here in Fig. 35.7.
This apparent difference in areas is produced by the use of a log scale that in
effect warps the true areas beneath the curves, having accentuated the relatively
rare measurements of the most extreme waves so the area beneath that portion of
the curve is expanded.35 This effect is analogous to the commonly used Mercator
projection of the spherical Earth onto a flat map, with the resulting expansion of
areas located at high latitudes; for example, Greenland’s area on a Mercator map is
much too large compared with a land mass close to the Equator. According to that
analogy, the counterpart to the artificial expansion of Greenland is the expansion
in Fig. 35.7 of the areas represented by rare but highest measured wave heights.
With our investigations having focused on regions where the heights of waves have
progressively increased due to Earth’s changing climate, such comparisons of pairs
of histograms invariably result in the shift of the more recent histogram to greater
numbers of high waves, as seen in Fig. 35.7, with a larger apparent area beneath
that curve.
Another example of a pair of histograms comparing their shifts in response to
the changing climate is presented in Fig. 35.8, based again on data from NDBC
buoy #46005 off the Washington Coast in the US Pacific Northwest, used in the
analyses of the decadal increases in Fig. 35.4 and Table 35.1. While the shift to
higher extreme waves is again apparent in this pair of histograms, here the modes
of greatest numbers of occurrences, centered on 3- to 4-m significant wave heights,
have also shifted to higher values and there is even a suggestion that the lowest
waves, those during the summer, have experienced an increase.
From the comparison of the overall shapes of the two histograms in Fig. 35.8
for the Washington data, it is apparent that the skewness has increased with time;
this is also seen in Fig. 35.7 for the East Coast hurricane-generated waves. This
developing skewness accounts for the systematic pattern in decadal trends seen in
Fig. 35.4 and Table 35.1 for the Washington data, where the more extreme the
wave-height assessment the higher the rate of increase according to the regressions.
Taken together the analyses provide guidance as to the magnitudes of increase of
extreme value projections, even though the histograms in Fig. 35.8 individually
represent only about a decade of data, two short to derive meaningful extreme-value
projections beyond the 25-year event.
Formal statistical analysis procedures are being developed for applications to time-
varying changes in data populations, with many of those applications directed
toward the environmental consequences of global warming (e.g., temperatures,
rainfall, and river discharges). Such investigations represent a significant advance
over classic extreme value theory, and have begun to be applied to the seasonal to
decadal variations in ocean-wave climates. Statistical procedures are now available
that involve a direct extension of the conventional extreme value analyses for a
stationary population, an extension that does not assume the wave climate has
August 14, 2009 9:7 9.75in x 6.5in b684-ch35 FA
Ocean Wave Climates: Trends and Variations Due to Earth’s Changing Climate 991
remained the same over the decades, thereby being able to account for progressive
changes in the extreme value assessments.36–38
Figure 35.9 shows the results of such an application to the Washington buoy
(NDBC #46005), with the data analyzed being the five highest significant waves
measured each year (not their averages as analyzed in Fig. 35.4 to determine the
trends in annual averages). This analysis represents an application of the r largest-
order statistical model to extreme value analyses,36,39 with time being modeled as
a covariate via the location parameter of the generalized extreme value (GEV) dis-
tribution. The three lines included in Fig. 35.9 are for the 25-, 50-, and 100-year
projections, each increasing at a rate of approximately 0.09 m/year as determined
by the nonstationary extreme value analysis. The r largest-order statistical model
is employed due to the relatively short time series of annual maxima for this buoy.
A nonstationary application of the annual maximum method (AMM), based on
the single highest measured significant wave height each year, yields much the
same results as in Fig. 35.9, with the rate of increase having been slightly greater,
0.1 m/year. This result for the increase in the extreme value assessment is in good
agreement with the linear regression for the highest measured significant wave
heights (Fig. 35.4 and Table 35.1), but by having modeled the extreme values
this rate of increase is statistically significant whereas that simple least-squares
regression in Table 35.1 was not, most likely due to the fact that extreme values are
not normally distributed so that a least squares analysis is not a reliable method
for identifying the trends.36
Fig. 35.9. Statistical analyses of the increasing extreme value projections with a rate of increase
of approximately 0.09 m/year, based on the five highest significant wave heights measured each
year by the Washington buoy.
August 14, 2009 9:7 9.75in x 6.5in b684-ch35 FA
Fig. 35.10. Statistical analyses including a GEV model that yielded a rate of increase of
0.0.085 m/year for the 25-year projection, and a model that accounts for the monthly variations.38
One of the most advanced statistical models to define the varying wave climate
is illustrated by the analyses of Méndez et al.38 Their model is based on a time-
dependent version of the Peak Over Threshold model, and permits a consideration
of the annual cycle in significant wave heights, long-term trends, and the inclusion of
dependencies on climate indices. Their application was again to the NDBC #46005
buoy off the coast of Washington, with their analysis results shown in Fig. 35.10,
fitted to the data through 2005 and with its projection up to the year 2020 as
a prediction of future wave climates. The data included in the analysis are those
that exceeded the 99.5% quantile in the distribution of significant wave heights,
this threshold significant wave height being 8.1 m. Each of the curves included
in Fig. 35.10 represents the evolution of the 25-year return period of the signif-
icant wave height. The simplest model (GEV1) is comparable to that presented in
Fig. 35.9 in having determined only a long-term trend; the result yielded a rate of
increase of 0.085 m/year, a lower rate because Méndez et al.38 assumed an expo-
nential rate of increase in extreme waves rather than a linear rate of increase as
applied in Fig. 35.9, and due to differences in the levels of high-wave conditions
being analyzed.
The most-advanced analyses by Méndez et al.38 included modeling the annual
cycles (Fig. 35.10), with the results up through 2005 having also included a depen-
dence on the climate indices for those years, but without that dependence for future
projections since those indices are unknowable. Although this advanced model pro-
vides the best overall assessment of the variations in the wave climate, the net
rate of increase is reduced to 0.035 m/year when including the seasonal variability,
which essentially is the value of the rate of increase in the annual averages of the
winter measured significant wave heights found in the regression analyses of Allan
and Komar,5 graphed in Fig. 35.4. While Méndez et al.38 correctly concluded that
their model provides a rigorous analysis of the long-term variability of Washington’s
extreme wave climate, including the controls as predicted by the climate indices, its
potential use depends on the nature of the application. For example, if the appli-
cation is directed toward the extraction of wave energy in a “wave farm” (which
is under consideration for the Washington and neighboring coasts), the advanced
model in Fig. 35.10 would have particular relevance in accounting for the winter
variability in power generation. On the other hand, if the concern is the stability of a
August 14, 2009 9:7 9.75in x 6.5in b684-ch35 FA
Ocean Wave Climates: Trends and Variations Due to Earth’s Changing Climate 993
affect the magnitudes of the ranges of nearshore processes, including the wave runup
elevations on beaches, an important component in hazard-zone assessments. There
can also be climate-controlled variations on the storm locations and therefore on
the directions of wave arrival at the coastal site of interest, this for example being
important along the US Pacific Coast during El Niños, having been a primary
factor in localized “hot spot” beach and property erosion. All of these possibilities
need to be considered in the development of wave climates, due to the multiple
environmental consequences of Earth’s changing climate.
A problem remains in that in some instances, while we can document with
confidence that the ocean wave heights have been progressively increasing, we are
not always certain as to the underlying climate controls. A full understanding of this
can be a challenge, as natural climate cycles may also be present, extending over
decades or longer, an example being the range from El Niños to La Niñas and how
their relative frequency of occurrences is affected by the Pacific Decadal Oscillation
(PDO), the dominance of one over the other being on the order of a 25-year cycle.
That period in the PDO cycle is effectively the span of most records available from
wave buoys, such that in analyses of the wave climates along the Pacific Coast of
the USA we recognized that the wave data we work with had been dominated by
major El Niños that occurred during the past 25 years.5 With the likelihood of
there having been a recent shift in the PDO so there might now be more frequent
La Niñas, it can be expected that this change will affect the wave climates, but we
are uncertain as to how and to what extent. At the same time, although the heights
of buoy-measured waves have progressively increased along the US West Coast, we
are not 100% certain as to its cause. Therefore, uncertainties exist as to projections
of wave climates into the future. In our assessments of coastal-hazard zones along
the US Pacific Northwest, our response to this uncertainty has been to base the
environmental projections on the expected level of development of the site, the more
extensive the proposed development the greater the expanse of the recommended
hazard zone.
The development of wave climates is no longer a routine undertaking, as the
ocean engineer or coastal scientist needs to account for progressive increases in wave
heights, periods, and changing wave directions, as well as the projected increase in
mean sea levels. With the remaining uncertainty in the climate controls, yet with
an intense level of research underway by climatologists, it can be expected that our
approaches to evaluating wave climates will continue to evolve.
References
Ocean Wave Climates: Trends and Variations Due to Earth’s Changing Climate 995
Chapter 36
Lesley Ewing
California Coastal Commission∗
45 Fremont Street, SF, CA 94105 USA
Viterbi School of Engineering∗
University of Southern California, LA, CA 90089 USA
lewing@coastal.ca.gov
Water level variability has been one of the most significant aspects of coastal
factors for coastal engineering and coastal management. Current factors influ-
encing local water level include tides, wind waves, tsunamis, atmospheric forcing,
seiching, storm surge, and local uplift and subsidence. Changes in global sea level
have influenced coastal conditions over geologic time. Changes in eustatic sea level
have been relatively small for the past several thousand years, but there is evi-
dence from global records, trends in sea level and atmospheric-ocean models that
sea level is now rising more quickly that has been experienced in the recent past,
and this trend will continue or increase during the next 100 years. Coastal engi-
neering and coastal management efforts will need to adjust to these changing sea
level trends. Certain situations may be best addressed by individual responses,
such as armoring, beach nourishment, or retreat. In most cases, the use of several
complementary approaches and the application of the major principles of sustain-
ability may be more appropriate.
36.1. Introduction
Water level variability is one of the most significant aspects of the coast that distin-
guishes coastal land from most other lands. Local water level variability influences
∗ This report has not been reviewed by the California Coastal Commission or Viterbi School of
Engineering.
997
August 20, 2009 16:0 9.75in x 6.5in b684-ch36 FA
998 L. Ewing
where someone will place their beach towel for a day of sun and surf; it effects
where different species can exist comfortably; where it is possible to launch a boat
or where it seems safe to erect buildings, roads, and infrastructure. Just as water
level variability is one of the basic features that differentiates coastal land from
other land, it is also one of the features that differentiates coastal engineering and
coastal zone management from their inland counterparts. Water level variability
is also one of the components of the coast that is most difficult to forecast. The
multiple, often independent forces and system dynamics that drive water level at a
specific location contribute to this difficulty. The observed water level at a specific
location is the cumulative result of global and local factors including oceanic con-
ditions, atmospheric forcing, storm surge, waves, tides, land subsidence or uplift,
wind, and river flows.
Recent examination and discussion of global warming have focused attention on
the long-term and large-scale variability of water levels encompassed by the term
global or eustatic sea level change. Although global sea level rise is occasionally
viewed as a new phenomenon, even a cursory examination of historic and geologic
sea level trends shows that global sea level has only rarely been stable or static. In
fact, the most typical characteristic of global sea level over time has been that it does
change, rising and falling in response to varying climatic and tectonic conditions.
The current interest in global warming has directed new attention to the fluctuations
in global sea level, the ensuing changes in local sea level and the likely engineering
and management responses.
Global sea level is but one of the factors that contribute to observed local sea level
and that influence decisions on coastal engineering and coastal zone management.
To understand the role of global sea level change in coastal engineering and man-
agement, it is important to understand the overall dynamics of water level along
the coast from the global to the local scale, how localized water level variability has
traditionally been incorporated into coastal engineering and management, and the
implications of future variability for these traditional and innovative engineering
and management approaches.
Global climate is influenced by changes to both the ocean and the atmosphere.
During the current Quaternary period, large-scale variability in sea level has been
controlled by glacial and climate change. During periods of cooling and glacial
advance, many millions of cubic meters of water are stored on land-based glaciers
and polar ice. Remaining ocean water is highly saline, cold, and dense. During
periods of warming and glacial retreat, melt water and calving from land-based
glaciers expand the volume of ocean water and increasing atmospheric temperatures
transfer heat into the ocean. Sea level rises during interglacial periods due to the
increased water volume.
Sea level during the Quaternary has been strongly influenced by glacial advance
and retreat; however, over much of the geologic record, the prime control on
August 20, 2009 16:0 9.75in x 6.5in b684-ch36 FA
global eustatic sea level has been the overall rate of global plate movements —
specifically changes in the rates of mid-oceanic spreading.1 Spreading rates of the
mid-oceanic ridges range from 2 to 6 cm/year, causing a global rise in sea level
up to 1 cm/1,000 year.2 During periods of rapid spreading, large portions of the
ocean basin are composed of young, hot basaltic crust. This young crust is both
thicker than the older crust and rises higher on the underlying asthenosphere,
reducing the volume of the ocean basin. Shallower, smaller ocean basins displace
ocean water into the continental land masses. Rapid seafloor spreading causes
eustatic high stands of sea level and during periods of slow spreading rates, the
eustatic sea level stands were low. One of the highest identified sea level stands
occurred in the Cretaceous period (130–65 million years BP), during fragmen-
tation of Gondwanaland and opening of the Indian Ocean and Mediterranean Sea,
when much of interior North America, Europe, and Asia was covered by shallow
continental seas.
Reconstructions of tectonics and paleo-climate can only show large trends. Tem-
poral trends in sea level are often plotted as a continuum, but actual sea level
variability most likely occurred as fairly discrete and sudden changes.3 This vari-
ability has been evident in reconstructions of sea level rise following the most recent
interglacial period, the Late Wisconsin Glaciation. This glaciation was followed by
a rapid rise in sea level, increasing by approximately 120 m from the glacial low
level.4 From 18,000 B.P. to about 6,000 B.P. the rate of sea level rise was about
9.0 mm/year. Over the past 6,000 years, sea level rise slowed to a rate of approx-
imately 0.5 mm/year and to a rate of 0.1 to 0.2 mm/year during the past 3,000
years.5 Recently, Yu et al.6 reconstructed the Holocene sea level history for the
Swedish Baltic coast and identified a rapid rise in sea level ca. 7,600 B.P. that
ranged from 3.0 to 7.2 m above present sea level, potentially linked to decay of the
Labrador sector of the Laurentide Ice Sheet. The Yu research and other studies
provide indications that sea level change has been episodic over many different time
scales.
Over the past 100–150 years, tide gages around the world have provided mea-
surements of local changes in sea level. Since 1992, these measurements have been
supplemented by altimeter measurement of the open ocean water level taken by
the TOPEX/Poseidon and Jason satellites. As shown in Fig. 36.1 (IPCC 2007,
Fig. 5.13), sea level has shown a rise of about 1.7 ± 0.5 mm/year when averaged
over the 20th century and 1.8 ± 0.5 mm/year when averaged over the more recent
time period from 1961 to 2003. Much of the sea level change examination by the
IPCC Fourth Assessment Report (2007) attempted to bridge the gap between the
geologic record of sea level change and the more recent tide gage and satellite
data. Archeologic sources, well data and information from fish pond elevations
from the Mediterranean area, have established that sea level was at a near still
stand from about 3,000 to 2,000 year B.P. until the 19th century. The current
rise in sea level seems to coincide with the industrial revolution, with a noticeable
rise in sea level occurred in the time period between 1850 and 1950. Thus, the
observed increase in sea level of 1.7–1.8 mm/year seems to be only a recent global
condition and represents an acceleration of 1.3 ± 0.3 mm/year/year for the past
150 years.7
August 20, 2009 16:0 9.75in x 6.5in b684-ch36 FA
1000 L. Ewing
Fig. 36.1. Annual averages of the global mean sea level (mm). The thin black line with standard
deviations is the reconstructed sea level since 1870 (from Ref. 7); the shaded gray curve shows
coastal tide gage measurements since 1950 (from Ref. 8) and the solid black line is based on
satellite altimetry (from Ref. 9). The reconstructed sea level field and the tide gage measurements
are deviations from their averages for 1961 to 1990; the satellite altimetry is the deviation from the
average of the reconstructed sea level for the period 1993 to 2001. Error bars show 90% confidence.
(From Ref. 10; Fig. 5.13).
Short-term localized sea level conditions can deviate greatly from average con-
ditions and from global conditions. Localized observed sea level is dependent on
climatic events such as the occurrence of El Niños, frequency of storm surge or atmo-
spheric forcing, pulses of river water and short-term thermal warming or cooling.
Satellite altimetry data for the 1990s (a time period that included the influences of
1997/1998 El Niño) showed a 3 mm/year rise in sea level in the open ocean for the
decade of recorded observations and the global coastline had a sea level rise rate
of 4 mm/year9 for global coastal waters. Some are tempted to see these short-term
rates as an indication that the anthropogenic increases in atmospheric temperature
linked to greenhouse gas emissions have led to acceleration in sea level rise; however,
there are many limitations in the extrapolation of short-term sea level rates for
long-term trends.
sea level as the only component. In most situations, changes in mean sea level will
provide a good first-order approximation for changes in the other tidal constituents.
Furthermore, the range provided in most projections can cover the variation in the
individual constituents.
Over the past century, global temperature has increased about 0.6◦ C and mean
sea level has risen by 15–18 cm globally.11 State-of-the-art climate models predict
that global temperature may rise by 2–9◦ F (1–5◦ C) over the next 100 years.12 There
is a large uncertainty in these estimates of future temperature change and even
greater uncertainty about the sea level response to this warming; however, there is
strong certainty that a future rise in global temperature will be accompanied by a
rise in sea level. The uncertainty is only for the rate and amount of rise that could
occur by any specific time in the future. Future changes in global sea level will
depend, among other factors, upon:
The two main factors contributing to global sea level (eustacy) remain thermal
expansion (steric eustacy) and the addition of water to the oceans from glacial
melting. Both of these factors can cause spatial variability in observed global water
levels. Satellite and tide gage information have found spatial variability in the
recent changes in sea level and large-scale atmospheric–oceanic models have been in
general agreement with these observations. However, observations and model data
sets available at the time of the IPCC Third Assessment Report13 were not able to
fully match observed water level changes with thermal expansion or glacial melting.
The IPCC Fourth Assessment Report found a better matching of the various
data sets and observations and concluded that thermal expansion and melting of
land-based glaciers each accounted for about half of the sea level rise observed
from 1993 to 2003. These agreements provide confidence in modeling of future
trends in sea level attributable to changes in atmospheric conditions and global
warming.
The thermal capacity of the oceans is about 1,000 times that of the atmo-
sphere, with the transfer of heat between the atmosphere and the ocean providing
a buffering mechanism to minimize extreme ranges of atmospheric temperature.13
Levitus et al.14 found that the net heat uptake by the oceans since 1955 has been
approximately 20 times more than the atmosphere. Most of the heat has been stored
in the upper ocean layers. Estimated thermal energy of that layer has increased over
the time period from 1955 to 2003, by 10.9 ± 3.1 × 1022 J or 0.14 ± 0.04 Wm−2 . Such
an increase in energy corresponds to an increase in average temperature of 0.1◦ C
and rise in sea level due to thermal expansion of approximately 0.42±0.12 mm/year.
These trends have not been spatially uniform; approximately half the warming has
occurred in the Atlantic Ocean. Some shallow equatorial areas and high-latitude
areas have experienced cooling over the same period from 1955 to 2003.15 Thermal
August 20, 2009 16:0 9.75in x 6.5in b684-ch36 FA
1002 L. Ewing
expansion for the past decade (1993–2003) alone has led to an average rise in sea
level of 1.6 ± 0.5 mm/year.
The future rate and volume of glacial change is the largest uncertainty for esti-
mating future sea level. Prior to satellite imaging, it had been difficult to get detailed
data on the extent and mass of glaciers. The IPCC Fourth Assessment Report pro-
vides estimates that glaciers (glaciers and ice caps, Greenland Ice Sheet and the
Antarctic Ice Sheet) contributed approximately 1.2 mm/year to sea level rise from
1993 to 2003. The observed thermal expansion and glacial melt for 1961 to 2003
could not be combined to account for most of the observed rise in sea level, leaving
+0.7 ± 0.7 mm/year as an unaccounted difference between observed rise in sea level
and contributions.
Human activities have contributed to changes in the available water and water
sequestration. Table 36.1 shows the general distribution of water. The ocean con-
tains 96.5% of the total water reserves, and fresh water is only 2.5% of the water
reserves. Of the available fresh water, approximately 69% is locked up in glaciers;
approximately 30% is available as groundwater or soil moisture, leaving only 1% of
the fresh water available as surface water or in the atmosphere. Gornitz16 has esti-
mated that deforestation, wetland losses, and fossil fuel combustion release the sea
rise equivalent of 0.4–0.9 mm/year, and sequestration in reservoirs and infiltration
below reservoirs has provided the equivalent of a sea level drop of 1.3–1.8 mm/year.
This overall reduction in potential sea level by 0.9 ± 0.5 mm/year has been a signif-
icant influence on current sea level, especially when compared with the general rise
in sea level of 1.7–1.8 mm/year.
Accelerated melting of the glaciers and ice sheets could lead to substantial
increases in sea level. Melting of the ice sheets on Greenland could result in up
to 7 m of additional sea level rise and melting of the glaciers and ice sheets in
Antarctica could result in an additional 20 m of sea level rise.17 Research and obser-
vations confirm that polar ice caps and glaciers are melting at rates far higher than
previously; however, there is little long-term information on glacial change to put
the current conditions into context. Few researchers have been able to provide any
time periods for these occurrences and accelerated loss of polar ice caps and glaciers
has not been included in any of the current IPCC predictions for future sea level.
Figure 36.2 and Table 36.2 provide the most recent IPCC scenarios for future sea
level rise.18
Fig. 36.2. Time series of global mean sea level (deviation from the 1980 to 1999 mean) in the
past and as projected for the future. For the period before 1870, global measurements of sea level
were not available. The estimates of the past show the uncertainty in the estimated long-term
rate of sea level change; the instrumental record is the reconstructed global mean sea level from
tide gages and for the period from 1991 to present from satellite altimetry, with gray shading to
denote the range of variations from a smooth curve. The projections of the future sea level are
model predictions for the IPCC scenario A1B, relative to the 1980 to 1999 mean and it has been
calculated independent of observations. Beyond 2100, the projections are increasingly dependent
on future greenhouse gas emissions scenarios. Over many centuries or millennia, sea level could
rise by several meters (IPCC, FAQ 5.1, Fig. 1).
Table 36.2. Projected globally averaged surface warming and sea level rise 2090
to 2099 relative to 1980–1999 (Ref. 13, Table SPM-3).
*These estimates are assessed from a hierarchy of models that encompass a simple
climate model, several Earth Models of Intermediate Complexity (EMICs) and a
large number of Atmosphere Ocean Global Circulation Models (AOGCMs).
**Excludes future rapid dynamic changes in ice flow.
***Year 2000 constant composition is derived from AOGCMs only.
August 20, 2009 16:0 9.75in x 6.5in b684-ch36 FA
1004 L. Ewing
The Emission Scenarios of the IPCC Special Report on Emission Scenarios (SRES)
A1: The A1 scenarios describe a future world of very rapid economic growth, global popu-
lation that peaks mid-century and declines thereafter, and a rapid introduction of new and
more efficient technologies. Major underlying themes are convergence among regions, capacity
building and increased cultural and social intersections, with a substantial reduction in regional
differences in per capita income. The three A1 scenarios are distinguished by their techno-
logical emphasis: (A1F1) non-fossil energy sources; (A1T) a balance across all sources; A1B
where balance is defined a not relying too heavily on one particular energy source and similar
improvement rates apply to all energy sources and end use technologies.
A2: The A2 scenarios describe a very heterogeneous world. The underlying theme is self-reliance
and preservation of local identities. Fertility patterns across the region converge very slowly,
resulting in continuously increasing population. Economic development is primarily regionally
oriented and per-capita economic growth and technological change is more fragmented than
other scenarios.
B1: The B1 scenarios describe a convergent world with the same global population as in the A1
scenarios which peaks in mid-century and declines thereafter, but with rapid change in economic
structure toward a service and information economy, with reductions in material intensity and
the introduction of clean and resource efficient technologies. The emphasis is on global solutions
to economic, social and environmental sustainability, including improved equity but without
additional climate incentives.
B2: The B2 scenarios describe a world in which the emphasis is on local solutions to economic,
social and environmental sustainability. It is a world with continuously increasing global popu-
lation, at a rate lower than A2, intermediate levels of economic development, and less rapid and
more diverse technological change than in the B1 or A1 scenarios. While the scenario is also ori-
ented towards environmental protection and social equity, it focuses on local and regional levels.
All scenarios should be considered equally sound. The SRES scenarios do not include additional
climate incentives which means that no scenarios are included that explicitly assume implemen-
tation of the United Nations Framework Convention on Climate Change or the emissions targets
of the Kyoto Protocol.
Most coastal engineering and management decisions need to consider the anticipated
rise in future global sea level. This global rise in sea level will not be an isolated
component of a design or management decision; it must be considered in association
with other changes in water level. Coastal residents, visitors, and observers are aware
of many of the changing water levels that occur throughout the day, from season to
season, over a period of years, or at irregular intervals. The main factors influencing
these changes are astronomic tides; El Niño/Southern Oscillation events (ENSO);
atmospheric pressure variability; wind and storm surge; waves; tsunamis; river dis-
charge; land subsidence and land uplift, in addition to the historic eustatic sea level
change. These factors vary in both spatial and temporal significance. Table 36.3
provides an overview of these factors, many of which are discussed below.
36.3.1. Tides
Tides and tidal currents are caused by the gravitational forces of the moon and
the sun, the centrifugal force from the rotation of the earth and interaction or
August 20, 2009 16:0 9.75in x 6.5in b684-ch36 FA
interference of land masses with these deepwater tides. People recognized a con-
nection between the moon, the sun, and the tides several thousand years ago. One
of the first documented descriptions of tides was presented in the 17 books of Geog-
raphy of Strabo (64 BC to 21 AD) in which he cites subsequently lost writings by
Posidonius of Rhodes:
“Now he (Posidonius) asserts that the motion of the sea corresponds with the revo-
lution of the heavenly bodies and experiences a diurnal, monthly and annual change,
in strict accordance with the motion of the moon.”19
These early writings detailed the observed connections between the daily,
monthly, and annual tidal variations; however, it was sixteen centuries, and many
tidal observations and tidal theories later, that Newton expressed the gravita-
tional forces between astronomic bodies with the equation, F = (Gm1 m2 /r2 ). The
detailed examination and expansion of tidal harmonics continues to this day.
Tidal fluctuations are often the main changes to water level that are considered
in coastal engineering and design. Tidal fluctuations have been included routinely
in wetland restoration projects, where the habitat ranges of wetland vegetation are
determined in large part by the amount of time the plants will be submerged. Veg-
etation zonations tend to follow the tidal zonations, where the intertidal zone, the
subtidal zone, and the supra tidal zone provide habitat for different plant palettes.
Beach construction and shoreline access is often planned for low tides to max-
imize available dry beach area. The Port of Oakland, in Oakland, California,
installed two new cranes that were shipped across the Pacific Ocean, through the
Golden Gate and under the Bay Bridge. Delivery was timed for low tide to provide
adequate clearance for the cranes to fit safely under the bridge. Navigation charts
reference all water elevations to a low tide datum in an attempt to prevent ships
from running aground during low tide periods.
High tide too can be important for coastal engineering and coastal management.
Flick20 found that during the 1982/1983 El Niño storms, most of the storm events
that caused the greatest damage were those storms that occurred during high tide.
Wood21 looked through the 341-year record of coastal flooding throughout North
America and found more than 100 incidents when flooding occurred due to storms
occurring at the same time as the perigean spring tides. Likewise, some of the worst
flooding from Hurricane Katrina occurred with the combination of high storm surge
August 20, 2009 16:0 9.75in x 6.5in b684-ch36 FA
1006 L. Ewing
and high tide.22 Tides are significant to many aspects of coastal engineering, design,
and management.
Changes to eustatic still water levels can be expected to alter tides in several
ways. As inundation areas move inland with a rise in sea level, the area of land
exposed to tidal influence will shift inland; however, there will not be a consistent
shift of the tide range or of the tidal constituents. Flick et al.23 analyzed data from
long-term stations in the US tide gage network and found that most open ocean
areas have experienced a change in tidal range and the rate of change of the tidal
constituents does not parallel the rate of change of the mean. For example, the mean
and diurnal tide ranges at San Francisco increased at a rate of 60 mm/century and
mean high tide and mean higher high tide constituents increased 19% faster than
mean sea level. Changes in the tidal range and tidal constituents will alter the coastal
elevations exposed to tidal influence. An increase in the high tide components will
expose more areas to periodic inundation. Increases in tidal range will increase tidal
currents and the intertidal zone. The dynamics of tidal variability with sea level
have not been well defined, but, future tidal variability can be expected to occur
with global climate change.
The areas of future storm surge will depend on changes in sea level insofar as sea
level will change inundation areas and water depth. However, since surge is primarily
influenced by wind conditions, sea level rise would be expected to cause only minor
changes to the water level variability that is associated with atmospheric pressure or
storm surge. However, there are some indications that global warming could change
both the frequency and intensity of storms. These changes would have significant
influence on the resulting water levels and such changes would greatly alter coastal
engineering and coastal management options in areas subject to large water level
variability due to meteorologic forcing from atmospheric pressure or surge.
1008 L. Ewing
watts per meter of shoreline.27 The energy at the coast from a 2-m wave would
be approximately 5,000 N-m/m2 , transferring four times the energy of a 1-m wave.
In addition, Kanoglu and Synolakis28 have shown that small increases in water
elevation, such as those from sea level rise, may have substantial, nonlinear effects
on runup and impacts on vertical walls.
Small changes in water level can cause significant changes in wave energy and the
potential for shoreline damage from wave forces. Any rise in sea level, the 0.2–0.6 m
rise over the 21st century modeled by the IPCC scenarios would cause significant
changes in nearshore wave energy. And, all IPCC scenarios exclude future dynamic
changes in glaciers and polar ice caps, so the modeled changes in sea level would be
low estimates if glacial and polar melting accelerates. Wave energy is sensitive to
water depth and large rises in sea level will greatly influence coastal conditions; a
sea level rise of 6 or 7 m would transform the coast in far greater ways than a rise
of 0.2–0.6 m.
There is some evidence that global warming and climate change will contribute
to changes in the intensity and frequency of storms. Graham and Diaz29 have found
an intensification of storm activity in the northeast Pacific that may be related to
or indications of changing climatic conditions. A rise in global temperature may
increase not only sea level and the transformation of wave energy, but could change
the characteristics of available wave energy.
clearly showed the life safety benefits of vertical evacuation and the importance of
considering potential tsunami forces for the design of critical facilities.34 NOAA’s
National Tsunami Hazard Mitigation Program has initiated development of design
guidelines for vertical evacuation structures. And, for the development of a new
community in a tsunami-prone area of Humboldt Bay, CA,35 tsunami evacuation
routes are being included in the circulation patterns for the community and potential
tsunami inundation levels are being used to site habitable structures and critical
facilities. Tsunami inundation and runup will be altered by rising sea level. For
most situations, there is little detailed guidance for how to include tsunamis effects
into coastal engineering and management. Engineering and management efforts for
tsunamis clearly require a long-term focus, and this long-term perspective should
consider whether the combined influence of tsunamis and sea level rise will be
important.
36.3.6. Seiches
Seiches are oscillating standing waves that develop in harbors, lakes, or partially
enclosed water bodies. The main characteristic of a seiche is that the wave con-
tinuing pendulum fashion after the initial force has stopped. The initial forcing can
come from winds blowing across the surface of the water body, waves, tides, and
earthquakes.
Seiches often set up in lakes or enclosed harbor basins where the oscillations
depend on the natural periods of the harbor and the energy from the forcing event.36
Most lakes develop seiches in response to wind, tides, or seismic disturbances. Lake
Erie can experience a 4.5-m change in water level due to wind-induced seiching.37
Often the vertical displacement has a small amplitude and the main evidence of
seiching is the occurrence of the large horizontal displacements that can be quite
damaging to moored ships. The lowest natural frequency for harbor seiching is
proportional to the square root of water depth. Seiching is sensitive to changes in
water depth and will be sensitive to changes in local sea level, with harbor responses
and resonance being an area of significant concern.
Seismic-induced seiching is one of the more intriguing seiching conditions. The
1964 Alaska Earthquake caused oscillations in water bodies throughout North
America and the world. Many locations in the Gulf of Mexico area recorded seismic
seiching, but seiching was also recorded as far away as Australia. McGarr and
Vorhis38 analyzed data on all seiches known to have been caused by the 1964
Alaskan Earthquake. Maximum recorded amplitudes were fairly small, all under
0.5 m, but in enclosed reservoirs, oscillations lasted more than 2 h. McGarr and
Vorhis concluded that the distribution of seiches was more dependent upon geo-
logic and seismic factors than hydrodynamic ones. More recently, Barberopoulou
et al.39 examined long-period effects from the 2002 Denali Earthquake on water
bodies throughout the Puget Sound, where vertical oscillations reached 0.8 m near
Seattle. As with harbor seiching, the most damaging aspect of most seismic seiching
comes from horizontal forces; nevertheless, the vertical oscillations are short-term,
potentially significant changes in water level.
August 20, 2009 16:0 9.75in x 6.5in b684-ch36 FA
1010 L. Ewing
to the holds of ships. Loading and unloading may have to be scheduled for low tide
periods to allow greatest access into the ship, or else mooring and cargo handling
facilities will need to be elevated.
If breakwaters or jetties protect the harbor, these structures will become less
efficient as water levels increase. The breakwaters and jetties will need to be enlarged
and heightened to keep up with the rise in sea level, or the harbor will have to
accept a higher level of overtopping and storm surge, and a higher probability of
storm damage.
The increase in water level could also increase the tidal prism of the harbor,
resulting in increased scour at the foundations of any structures in the harbor.
So, it may also be necessary to reinforce the base of the breakwater or jetty to
insure stability. Changes in water depth can also alter harbor seiching/harbor res-
onance characteristics. Harbors with resonance problems may develop different or
new seiching problems; harbors without recorded seiching problems may develop
them. Benefits that could occur from a rise in sea level would be the opportunity
for harbors to accommodate deeper draught ships and a decrease in dredging to
maintain necessary channel depths.
August 20, 2009 16:0 9.75in x 6.5in b684-ch36 FA
1012 L. Ewing
36.4.1.3. Wetlands
Coastal wetlands will be greatly modified by changes in sea level; however, the con-
sequences will vary with the different wetland areas. Overall, there will be greater
areas of inundation. The change in the intertidal area will depend on local topog-
raphy, the future change in tidal range, and the ability of the wetland to migrate
both up and inland. Historically, many wetlands have accommodated the rise in sea
level by increasing the base elevation. Sediment collects in the roots and vegetative
mass of the wetland and provides a substrate for new growth. If the rate of sediment
entrapment equals the rate of sea level rise, the wetland will remain fairly constant. If
the rate of sedimentation exceeds the rate of sea level rise, the wetland will convert to
a wet meadow or other system with more supratidal vegetation. If the sediment rate
is less than the rise in sea level, the wetland will transition to intertidal and subtidal
habitat.
Wetland changes also will be affected by inland development. Historically,
wetland areas migrated both upward and landward as they were inundated. If the
inland area has a slope and soil composition that can support a wetland and is not
already developed, then inland migration may be possible. If there is a steep bluff or
some type of fixed development such as a highway or bulkhead inland of a wetland,
inland migration will not be possible and the wetland area will diminish over time.
Another physical change to wetlands in response to a rise in sea level is an
increase in the tidal currents, with the potential for increased scour. For estuarine
systems there will be a shift in the location of the saltwater–freshwater interface,
and an inland movement of the zone of brackish water. Wetlands have survived for
millennia by moving up and inland. If existing development or other barriers restrict
inland migration, as is often the case, wetland areas will narrow and eventually be
converted to subtidal and submerged habitat.
Various climate change models have attempted to examine changes in run-off
and sedimentation from global warming. Two global circulation models predict
both increases in the number of storm events entering the US from the Pacific
and an increase in run-off state-wide. Along coasts such as California, where rivers
are a major source of coastal sediment supplies, the effects from increased run-
off and sediment loadings will greatly modify coastal conditions. There is con-
siderable uncertainty about the effects of climate change to hydrologic cycles,
particularly on the regional and watershed level and the changes to coastal sed-
iment supplies may change greatly from current levels. For example, some models
August 20, 2009 16:0 9.75in x 6.5in b684-ch36 FA
1014 L. Ewing
will be beneficial for many reasons and not just to accommodate a single prediction
of future sea level rise.44
Hard engineering
• Heighten existing structures or add new layers of armor
• Build more seawalls/bulkheads/revetments/breakwaters/groins/jetties
Soft engineering
• Nourish beaches with sand (restore historic sources or find new sources)
• Perched beaches
Accommodation/adaptation
• Elevate structures
• Switch to salt tolerant crops
Retreat
Planning and regulatory responses to sea level rise
• Setbacks for new development
• Engineering design of new structures on the coast
• Wetland buffers
• Notify property owners of risks with building on the coast
• Prohibit future shoreline protection for new development
• Monitor for subsidence from fluid withdrawal projects
• Studies of shoreline change
• Encourage beach nourishment and opportunitistic nourishment
• Coastal land acquisition
• Public education programs
August 20, 2009 16:0 9.75in x 6.5in b684-ch36 FA
36.4.2.3. Accommodation/adaptation
Another approach to varying sea level is to accommodate or adapt to the changes.
Coastal systems have used adaptation for millennia as a response to climate change.
Adaptive capacity refers to the ability of a system, individual, organization, or
community to adjust to changing conditions, take advantage of opportunities to
cope with the consequences, with the result that damages from the climate change
can be moderated. If structures are at risk, they can be raised so they will not
be inundated. If agricultural lands are at risk, there may be options to switch to
salt tolerant vegetation. Islands or spits could be elevated to keep pace with sea
level rise.
Some of the more accepted adaptation techniques include the use of elevation,
setbacks and buffers to provide physical separation between development and either
hazard or a resource that needs to be protected from development impacts. Rolling
easements have been suggested as a way to site development in a hazardous
area and then move the development back from the hazard when the separation
decreases. Approaches for adaptation or accommodation often require more “hands-
on” involvement than most hard engineering approaches. Accommodative/adaptive
responses are rarely one-time actions; to be most effective they require continued
flexibility and modification to meet the ongoing conditions.
August 20, 2009 16:0 9.75in x 6.5in b684-ch36 FA
1016 L. Ewing
36.4.2.4. Retreat
Historically, one of the natural responses to rising sea level has been retreat.
Wetlands have demonstrated this as an effective response to sea level rise. Retreat
has also been used in some situations to protect development or movable property.
During the 1982/1983 El Niño storms, a mobile home subdivision in Pacifica, CA
lost the seaward most row of lots. The response was to pull the facilities landward
and reduce the number of available sites. During the storms of 1997/1998, a dif-
ferent part of Pacifica, CA was attacked by waves and 10 homes were eventually
removed from the bluff top. In southern Big Sur, portions of Pacific Coast Highway
have been threatened by bluff retreat. Small amounts of revetment have been built
to protect the most threatened portions, but the overall plan is to relocate several
miles of the road to a safer, more inland location. Such examples of retreat may
be more common in the future, as the threat from coastal erosion becomes more
pervasive and the long-term costs of in-place protection increase.
Coastal zone management and land use planning are important tools for addressing
sea level rise. New development can be located far enough away from erosion hazards
or bluff retreat, through either elevation or setbacks, to assure site stability for
foreseeable future conditions. Individual property owners can undertake many of
these actions voluntarily, but in many situations, safe setback and elevation design
conditions are required through the regulatory process. Regulatory efforts also can
establish wetland buffers to allow future inland or upland migration of wetlands.
Deed restrictions concerning hazardous conditions can be used to notify current and
future property owners that the conditions at a site may worsen with time. Along
coastal areas where hard engineering can reduce the quality or quantity of public
beach access and recreation, planning agencies can use the idea of “assumption of
risk” to put property owners on notice that they will need to address the future
consequences of sea level rise without actions such as seawalls or revetments that
would diminish the public beach.
Few states have passed special regulations to address sea level rise. Instead,
most states have coastal programs that address sea level rise as an ongoing coastal
condition and they modify or adapt current regulatory mechanisms to cover the
effects of sea level rise. Texas has a rolling easement program that relocates the
public land boundary to the current line of vegetation. States like South and North
Carolina and Massachusetts have prohibited the construction of any hard shoreline
armoring. This limits the responses that can be used to address sea level rise to
soft engineering, accommodation, and retreat. Maine has regulations that prohibit
rebuilding structures that have been damaged by storms if the new structure could
August 20, 2009 16:0 9.75in x 6.5in b684-ch36 FA
reasonably be expected to be damaged within the next 100 years. This regulation
covers damage from sea level rise or other hazards.
The US Army Corps of Engineers is now incorporating adaptive planning for new
projects, requiring consideration of a range of sea level conditions in the feasibility
analysis of all coastal and estuarine projects. The planning guidance requires,
“A sensitivity analysis should be conducted to determine what effect (if any) changes
in sea level would have on plan evaluation and selection. The analysis should be based,
as a minimum, on the extrapolation of the local, historical record of relative sea level
rise as the low level and Curve 3 from the NRC report [a rise of 1.5 m by 2100] as the
high level.”46
Sustainability and integrated coastal management are part of a new trend in
coastal management and they offer opportunities to address many future coastal
concerns. The European Union has adopted eight principles of best practice for
integrated coastal zone management:47
(1) A broad “holistic” perspective (thematic and geographic)
(2) A long-term perspective
(3) Adaptive management during a gradual process
(4) Reflect local specificity
(5) Work with natural processes
(6) Participatory planning
(7) Support and involvement of all relevant administrative bodies
(8) Use of a combination of instruments.
During times of rapidly changing sea level, the coastal zone is the area that
experiences the most dynamic changes from both rising water levels and adjust-
ments from the inland areas. There is great uncertainty in the potential changes to
water level from global warming and this uncertainty will be combined with ongoing
uncertainties concerning water level changes from tides, waves, seiching, river flows,
atmospheric forcing, uplift, and subsidence. Engineering or planning responses that
are based upon any single fixed water level will have great difficulty maintaining
effectiveness during a period of rising sea level. The principles of sustainability and
reliance upon multiple adaptive responses to the array of changing coastal conditions
can provide many useful approaches for dealing with future changes in sea level.
36.5. Conclusions
Global sea level has been rising since the end of the last ice age about 18,000 years
ago. The rate of rise has varied and for short periods of time, such as the Little
Ice Age, sea level has even dropped. However, based on current climate models and
projections of greenhouse gas emissions, it is expected that future sea level will rise
at a greater rate than it has over the past hundred years.
The continued rise in sea level will increase inundation of low coastal areas.
Nearshore wave heights and wave energy will increase, increasing the potential for
storm damage, beach erosion, and bluff retreat. Ports and harbors will have reduced
cargo transfer capability as ships ride higher along the dock. Wetlands may be
August 20, 2009 16:0 9.75in x 6.5in b684-ch36 FA
1018 L. Ewing
inundated if they are not able to migrate either upward or landward. Almost all
coastal systems will be affected; even groundwater aquifers will be a greater risk
from saltwater intrusion.
Sea level rise is not a new phenomenon; it has been a major component of
coastal change throughout time. While there are a number of possible responses to
future changes in sea level, in many situations, response flexibility will be important.
Reliance on a single response for a certain water level may be useful while that
water level occurs. But, as either the water level continues to rise or the coastal
land subsides, the fixed response will become less efficient and be less appropriate
to the new conditions.
Along most coastal areas, it is likely that a combination of hard engineering, soft
engineering, accommodation/adaptation, and retreat responses will be considered to
address sea level rise. There are situations where each response may be appropriate
and well suited. In all coastal projects, it is important to recognize and accept
that there will be changes in sea level and in other coastal processes over time.
Careful review, siting and permitting of new projects on the coast can increase the
likelihood that these projects will be able to adapt and change to accommodate
future coastal hazards. Public education and efforts to alert property owners to the
dynamic nature of the coast will be important.
The recent interest in coastal sustainability offers planning opportunities for
addressing the consequences of sea level rise. Sustainability principles espouse the
need for integrated planning and the incorporation of a holistic approach to the
coast, with strong consideration of local conditions and the need for adaptive
planning to adjust to changing conditions.
Acknowledgments
The author would like to thank Meg Caldwell and Costas Synolakis for conversations
past and future about the engineering and management concerns relating to a world
of changing sea level. Thanks to Eli Davidian and Susan Hansch for their input
throughout, Mark Johnsson for his geologic input and generously sharing material
from his library, and Dr. Kim for starting this project.
Portions of this paper have been developed from earlier work: (1) L. Ewing,
J. Michaels, and R. McCarthy (1990) Draft Report: Planning for an Accelerated
Sea Level Rise along the California Coast, California Coastal Commission, San
Francisco, CA and (2) California Coastal Commission Staff (2001) Overview of Sea
Level Rise and Some Implications for Coastal California.
References
1020 L. Ewing
22. American Society of Civil Engineers Hurricane Katrina External Review Panel, The
New Orleans Hurricane Protection System: What Went Wrong and Why (ASCE
Publications, 2007).
23. R. E. Flick, J. Murray and L. C. Ewing, Trends in United States tidal datum statistics
and tide range, J. Waterw. Port Coast. Ocean Eng., ASCE, 129(4), 155–164 (2003).
24. R. S. Nerem and G. T. Mitchum, Observation of sea level change from satellite
altimetry, Sea Level Rise: History and Consequences, Vol. 75, eds. B. C. Douglas,
M. S. Kearney and S. P. Leatherman (Academic Press, San Diego, International Geo-
physics Series, 2001).
25. NOAA National Weather Service, Service Assessment, Hurricane Katrina August
23–31, 2005, June 2006, http://www.weather.gov/os/assessments/pdfs/Katrina.pdf.
26. American Society of Civil Engineers Hurricane Katrina External Review Panel, op. cit.
27. R. G. Dean and R. A. Dalrymple, Coastal Processes with Engineering Applications
(Cambridge University Press, 2002).
28. U. Kanoglu and C. E. Synolakis, Long wave runup on piecewise linear topographies,
J. Fluid Mech. 374, 1–28 (1998).
29. N. E. Graham and H. F. Diaz, Evidence for intensification of North Pacific winter
cyclones since 1948, Bull. Am. Meteorol. Soc. 82, 11869–11892 (2001).
30. R. L. Wiegel, Oceanographic Engineering (Prentice-Hall, Englewood Cliffs, NJ, 1964).
31. K. Iida, Magnitude, energy, and generation of tsunamis, and catalogue of earthquakes
associated with tsunamis, Proc. Tsunami Mtgs. Assoc, with the 10th Pac. Sci. Cong.,
Monograph No. 24, IUGG, 1963 (cited in R. L. Wiegel, op. cit.).
32. C. E. Synolakis, J.-P. Bardet, J. C. Borrero, H. L. Davies, E. A. Okal, E. A. Silver,
S. Sweet and D. R. Tappin, The slump origin of the 1998 Papua New Guinea tsunami,
Proc. Royal Soc. (London), Ser. A 458, 763–789 (2002).
33. C. E. Synolakis, J. Borrero and R. Eisner, Developing inundation maps for Southern
California, Proc. Solutions to Coastal Disasters, ASCE (2002).
34. C. E. Synolakis and L. Kong, Runup measurements of the December 2004 Indian
ocean tsunami, EERI Spectra 22(S3), S67–S91 (2006).
35. Humboldt County Community Development Services, Samoa Town Master Plan,
Recirculation Draft 2: Tsunami Vulnerability Evaluation (2007), http://co.humboldt.
ca.us/planning/samoa/docs/.
36. R. L. Wiegel, Oceanographic Engineering (Prentice-Hall, Englewood Cliffs, NJ, 1964).
37. S. J. Bolsenga and C. E. Herdendorf (eds.), Lake Erie and Lake St. Clair Handbook
(Wayne State University Press, Detroit, 1993).
38. A. McGarr and R. C. Vorhis, Seismic Seiches from the March 1964 Alaska Earthquake,
U.S. Geol. Surv. Profess. Pap. 544E, E1–E43 (1968).
39. A. Barberopoulou, A. Qamar, T. L. Pratt and W. P. Steele, Long-period effects of the
Denali Earthquake on water bodies in the puget lowland: Observations and modeling,
Bull. Seism. Soc. Am. 96(2), 519–534 (2006).
40. R. O. Castle, M. R. Elliott, J. P. Church and S. H. Wood, The evolution of the
Southern California uplift, 1955 through 1976, U.S. Geol. Prof. Paper #1362 (1984).
41. U. Kanoglu and C. E. Synolakis, op. cit.
42. D. Scavia et al., Climate change impacts on US coastal and marine ecosystems,
Estuaries and Coasts 25(2), 149–164 (2002).
43. B. D. Collins, R. Kayen and N. Sitar, Process-based empirical prediction of land-
slides in weakly Lithified coastal cliffs, San Francisco, California, USA, Landslides and
Climate Change — Proceedings of the International Conference on Landslides and
Climate Change, eds. R. McInnes, J. Jakeways, H. Fairbank and E. Mathie, Isle of
Wight, United Kingdom, May 2007, Taylor & Francis, London, pp. 175–184.
August 20, 2009 16:0 9.75in x 6.5in b684-ch36 FA
44. D. M. Mageean, A. Constable and M. D. Van Arsdol, Jr., Impacts of rising sea level on
coastal populations in California and Maine, ISA Research Committee #24, Montreal,
Canada (1998).
45. D. Cayan, A. L. Luers, M. Hanemann, G. Franco and B. Croes, Scenarios of climate
change in California: An overview, CEC-500-2005-186-SF (2006).
46. US Army Corps of Engineers, ER 1105-2-100 Planning Guidance Notebook,
Appendix E, Section IV.E-24.k, April 2000.
47. Commission of the European Communities, Communication from the Commission to
the Council and the European Parliament on Integrated Coastal Zone Management:
A Strategy for Europe; Annex 1 (COM(2000) 547 final, 2000; http://ec.europa.eu/
environment/iczm/comm2000.htm).
August 20, 2009 16:0 9.75in x 6.5in b684-ch36 FA
Chapter 37
IPCC projections indicate that the rate of sea level rise (SLR) during the 21st
century may be about an order of magnitude greater than the 20th century rate
of 1–2 mm/year. This accelerated SLR will in turn result in much faster coastline
retreat with particularly severe impacts on low-lying areas. The socioeconomic
impact of such accelerated coastline retreat could be massive due to the rapid
growth of coastal communities and infrastructure over the past five or six decades.
The method most commonly used to estimate coastline retreat due to SLR is
the simple two-dimensional mass conservation principle known as the Bruun rule.
However, in view of the high level of predictive accuracy that is clearly needed to
facilitate informed planning decisions for the future, can we continue to depend on
the Bruun rule? This chapter discusses the evidence for and against the Bruun rule
and suggests alternative methods that may be more suitable for the 21st century.
37.1. Introduction
The IPCC1 projections for 21st century sea level rise (SLR) range from
0.18 to 0.79 m by 2090–2099 relative to 1980–1999, including an allowance of 0.2 m
for uncertainty associated with ice sheet flow. Very recent research also suggests that
the measured SLR over the last decade is under predicted by the IPCC models,2
and that a maximum SLR of 1.4 m by 2100 (relative to 1990 levels) is not unlikely.3
1023
August 25, 2009 14:7 9.75in x 6.5in b684-ch37 FA
Compared to the 1–2 mm/year rate of SLR that was experienced in the last century,
these 21st century projections constitute an order-of-magnitude increase in the rate
of SLR.
It has been long known that, in the absence of other compensatory mechanisms,
any rise in the mean sea level will generally result in the retreat of unprotected
coastlines.4 Fortunately, the still-stand (slow) conditions of SLR during the last
century has resulted in slow and mostly manageable coastline retreat (recession).
However, the potential order-of-magnitude increase in the rate of SLR in the 21st
century is likely to result in much faster coastline recession. The socioeconomic
impact of such accelerated coastal recession could be massive due to the unprece-
dented growth of coastal communities over the past 50 years or so which has led to
billions of dollars worth of developments and infrastructure within the coastal zone.
To ensure the safety of growing coastal communities and to avoid massive economic
losses in the future, it is now imperative that any predictions of coastal recession
due to SLR be highly accurate.
The IPCC has successfully raised awareness on political and societal levels, which
in many nations has resulted in including SLR scenarios in new designs of shore
protection works, both hard (structures) and soft (nourishment) or combinations
thereof.5 In this context, it is crucial that the impact of SLR in low-lying coastal
areas be quantified accurately. The inundation of such low-lying areas will result in
significant coastline retreat, the magnitude of which is governed by the local coastal
slope. As coastal slopes in such areas may be as mild as 1 in 1,000, the resulting
coastline retreat could be three orders of magnitude greater than the rate of SLR.
Quantifying the impact of SLR on dune and barrier coasts is less straightforward.
This is because the response of dune and barrier coasts to SLR is a complex mor-
phodynamic issue. The most commonly used method to quantify the recession due
to SLR at such coastal locations is the simple two-dimensional mass conservation
principle known as the Bruun rule, which predicts a landward and upward dis-
placement of the cross-shore profile in response to a rise in the mean sea level. For
many of the world’s coastlines, where the nearshore beach slope is about 0.01–0.02,
the Bruun rule predicts a coastline retreat between 50 × SLR and 100 × SLR, which
are proportionalities that are commonly used as a rule of thumb.
Although, coastal scientists and engineers have been routinely using the Bruun
rule for almost five decades, mainly due its simplicity and the lack of any other easy-
to-use alternative methods, it has been receiving some heavy criticism in the recent
past.6,7 While the many attempts to verify the Bruun rule against field and labo-
ratory data over the last four decades (e.g., Refs. 8–12) have qualitatively confirmed
the basic concept of the Bruun rule, not many have resulted in convincing quan-
titative comparisons between measurements and Bruun rule predictions.13 Zhang
et al.14 who undertook a large-scale study of a 220-km stretch of the US East coast
is the only study which resulted in reasonably good comparisons between measured
shoreline recession and Bruun rule predictions.
However, does even a good comparison between measured and predicted values
under present still-stand SLR conditions (1–2 mm/year SLR) mean that the Bruun
rule is conclusively validated? Based on the observations made in the USA, the
Netherlands, the Mediterranean, and Australia, Stive15 showed that the net natural
August 25, 2009 14:7 9.75in x 6.5in b684-ch37 FA
shoreline change due to cross-shore processes is ∼1 m/year, whereas the net natural
shoreline change due to longshore processes is ∼0.1–1 m/year. However, the Bruun
rule predicts a recession rate between 0.1 and 0.2 m/year for the current long-term
SLR rate of about 2 mm/year (using the rule-of-thumb approach). Thus, the Bruun
effect (i.e., the coastal recession due to eustatic SLR alone) is, at best, an order
of magnitude less than the observed net natural shoreline changes. This means
that the Bruun effect are likely to be overridden by other coastal processes under
the present still-stand SLR conditions. Therefore, any good comparisons between
shoreline recessions measured in the last century and the corresponding Bruun rule
predictions are likely to be fortuitous and cannot be considered as conclusive verifi-
cation of the Bruun rule. However, it should be noted that the quantitative accuracy
of the Bruun rule has been validated for contemporary systems that have undergone
a rapid relative SLR due to subsidence.9,16 Holocene coastal evolution modeling
using the Bruun concept under considerable rates of SLR and fall also lends con-
vincing support to the concept.17,18 It is also noteworthy that large- and small-scale
laboratory tests of dune erosion under high storm surge levels (e.g., Refs. 19 and
20) suggest a Bruun-type response at the much shorter timescales of storm duration
(hours-day).
Although the available evidence suggests that the Bruun rule is most likely
conceptually correct, arguably it addresses only one potentially important effect of
a range of effects. If no other sediment sources or sinks are present or if no other
sediment transport gradients in cross-shore and longshore directions prevail, the
Bruun effect is the only operational effect. However, this idealized situation is the
exception rather than the rule, and there are other effects of SLR on the coastal
sediment budget, which are generally much larger, or at least of the same order
of magnitude as the Bruun effect. In the following section we present the available
evidence pointing to the presence of mechanisms other than the Bruun effect, which
may play an equal or more crucial role in governing coastline retreat/advance.
A general point of view, triggered by Bird,21 is that since 70% of the world’s sandy
beaches are in a state of erosion, global SLR has to be the most probable cause (cf.
Ref. 22). However, there are numerous coastal systems that have been accretive in
the Holocene, even though sea level was rising. A few examples are the Australian
coast,23 deltaic coasts (Mississippi, Ebro, Po, Yangtze, and many other deltas at
earlier stages of the Holocene) and composite coasts such as the US Northwest
Washington coast and the Dutch coast.17 This implies that there must be a number
of other processes that can override the Bruun effect, which is generally erosive,
to such an extent that the resultant response is coastline advance. In contrast,
many other coasts experience larger erosion than is explained by the Bruun effect.
This implies that there must be coastal processes other than the Bruun effect that
contribute to coastline retreat/advance. The important question then is, what are
these other processes and are they likely to be affected by the accelerated SLR that
is likely to occur in the 21st century?
August 25, 2009 14:7 9.75in x 6.5in b684-ch37 FA
The other processes that may govern coastline retreat/advance may be collec-
tively referred to as “sediment availability,” which is implicitly included in earlier
first approximation kinematic models of long-term (millennia) coastal change.24
Swift25 extended Curray’s ideas into a general framework for long-term coastal
change entailing transgression (landward retreat) and regression (seaward advance)
of the shoreline due to SLR and fall, respectively, with corresponding tendencies
toward retrogradation and progradation due to net sediment losses or gains from
alongshore gradients in sediment transport.
Cowell et al.26 show how Swift’s concepts can be quantified and related back to
the Bruun rule, when upper shoreface sediment balance is considered. Cowell et al.
assume that, to a first approximation, the upper shoreface is form invariant relative
to mean sea level over time periods for which profile closure occurs (1 year).27
The upper shoreface is represented by an arbitrary, but usually concave-up, profile
h(x) to a depth h∗ (a morphologically active depth) and a length L∗ , in which x is
the distance from the shore.28 Assuming that the cross-shore profile shape remains
constant over time, sediment-volume conservation for profile kinematics requires (for
a Cartesian coordinate system with seaward and upward directions positive) that
∂h ∂h
+ cp =0 (37.1)
∂t ∂x
∂zb ∂zb ∂M SL
+ cp = (37.2)
∂t ∂x ∂t
where cp is the horizontal translation rate of the shoreline position. The sediment-
transport balance equation for a fixed spatial control volume is
where qx and qy are the cross-shore and alongshore sediment transports, and s is a
local source or sink. These equations may be combined to yield
−1 −1 −1
∂M SL ∂h ∂qx ∂qy ∂h ∂h
cp = − − − −s (37.4)
∂t ∂x ∂h ∂y ∂x ∂x
∂M SL ∂Qy
cp h∗ = L∗ − (qx,sea − qx,dune ) − −s (37.5)
∂t ∂y
transgression transgression
m3 m -1a -1
sea-level rise
-10
virtual sink
and
due to
7200
retrogradation
E A T
6000
T R
R E
5200
75n progradation
325s 75s 3350
*
4500
325n
sink 750n -20 1500n 750s 1500s 20 source
m3 m -1a -1 m 3 m -1a -1
retrogradation
V A N C E
A D
virtual source
sea-level fall
m 3 m -1a -1
due to
progradation
and
regression 10 regression
Fig. 37.1. Evolution of the central Netherlands coast (Hoek van Holland to Den Helder) as a time
trajectory in sediment-supply/accommodation phase space (abscissa and ordinate, respectively,
scaled in cubic meters per year per meter of shoreline). Numbers along the trajectory indicate
time (years BP); suffixes n and s denote north and south of Haarlem, respectively (after Ref. 26).
Table 37.1. Typical cross-shore losses (negative) and gains (positive) and associated shoreline
changes (negative = retreat).15
∗ The net ephemeral loss will be zero, as these losses will be recovered in time.
What can we conclude from these cross-shore process quantifications? Under the
present SLR conditions the Bruun effect is at least an order-of-magnitude smaller
than contributions from other processes and is therefore negligible. Obviously, if the
rate of SLR increases five- or ten-fold, then the losses due to the Bruun effect will be
of the same order-of-magnitude as the gains due to chronic accretionary processes
(e.g., wave asymmetry and boundary layer flow), which may even reverse the net
shoreline change from advance to retreat.
Table 37.2. Typical longshore sediment transport rates (integrated across the surf zone)
along low-energy coasts (e.g., the Mediterranean coast) and associated shoreline changes
(negative: retreat).15
Table 37.3. Typical longshore sediment transport rates (integrated across the surf zone) along
high-energy coasts (e.g., Holland coast, eastern US coast) and associated shoreline changes
(negative: retreat).15
Table 37.4. Typical longshore sediment transport rates (integrated across the surf zone) along
a high-energy barrier coast, e.g., the Frisian Wadden coast or the eastern US coast and asso-
ciated shoreline changes (negative: retreat).15
The effect of backbarrier sources and sinks are represented by the last term on
the right-hand side in Eq. (37.5). This term could play a significant role when the
backbarrier consists of a river, estuary, or tidal lagoon. In case of a river or an
estuary there may be a natural supply of sediment to the coastal system that can
compensate for all cross-shore and longshore losses mentioned above in Secs. 37.3
and 37.4. In this case, a delta will form and evolve with time. The delta evolutionary
characteristics will depend on the relative role of waves, tides, and river flow. In
the Holocene, many deltas have been outbuilding as a result of abundant sediment
supply due to erosion of the catchment basin. Over the past five decades many
deltas have started to disintegrate due to human intervention in the form of dam-
regulated river discharge that decimates downstream sediment supply.
In case of a tidal lagoon or an estuary with low freshwater inputs, the backbarrier
tidal basin area may act as a source or a sink for the coastal sediment budget.
Classic examples of sink behavior are the Frisian Wadden basins along the Dutch
and West German North Sea coast. Dronkers34 analyzed the net sediment transport
behavior of these basins and showed that these basins are generally flood dominant,
i.e., there is a tendency to accumulate sediment within the basin as SLR, restoring
dynamic equilibrium geometry. Stive and Wang35 further analyzed this response
and showed that, in this case, the Bruun rule can be extended as follows to express
the impact of SLR on inlet-influenced coasts:
∂M SL L∗ ∂M SL Ab
cp = + (37.7)
∂t h∗ ∂t h∗ Lac
where Ab is the tidal basin area and Lac is the length of the adjacent coast impacted.
In the above equation the first term on the right-hand side expresses the Bruun
effect and the second term expresses the basin accommodation effect. The Bruun
effect is exceeded by the basin effect when:
Typical orders of magnitude for L∗ and Lac are 1 and 10 km, respectively, meaning
that the direct impact of basin areas larger than O (10 km2 ) on coastline retreat
overrides the Bruun effect.
Friedrichs and Aubrey36 presented a similar analysis for a large number of
schematized tidal basins in eastern USA. They showed that depending on basin hyp-
sometry, tidal basins could be either flood or ebb dominant. This implies that SLR
may lead to both importing and exporting basins. Figure 4 of Zhang et al.14 indi-
cates that long stretches of coastline, which they denote as inlet-influenced, expe-
rience stronger recession rates than the noninlet-influenced stretches of coastline
where the Bruun rule was used. This gives rise to the hypothesis that backbarrier
basins along that coastline are flood dominant. However, when the basins are
August 25, 2009 14:7 9.75in x 6.5in b684-ch37 FA
ebb dominant, SLR may cause an export of sediment. This will decrease or even
compensate for the Bruun effect to such an extent that shoreline advance may occur.
If SLR forces positive accommodation in flood-dominant basins, surf zone gene-
rated sediment transport will be diverted into the tidal basin by flood currents.
These sediments will be trapped in the flood delta. Typical barrier length scales
are O (10 km). Therefore, typical values for shoreline changes in the vicinity of
backbarrier sinks can be estimated as shown in Table 37.4.
Table 37.4 indicates that along inlet interrupted coastlines the tidal basins linked
to the inlets have a considerable influence on coastline change. Interestingly, the
magnitude of tidal basin influenced coastline change (as indicated above) is signi-
ficantly larger than the coastline changes estimated by Zhang et al.14 along their
noninlet-influenced coastline (cf. Fig. 37.3 of Ref. 14).
In essence, therefore, while the underlying concepts of the Bruun rule have been
verified, its quantitative accuracy remains unverified; largely due to the presence of
other coastal processes that easily override the Bruun effect under present still-stand
SLR conditions. Furthermore, the numerous restrictive assumptions associated with
the Bruun rule preclude its application in most natural environments, and especially
in the vicinity of inlets linked to tidal basins larger than 10 km2 . However, what is
needed now, in view of the massive socioeconomic impacts along the world’s highly
developed coastal margins that are likely to result from accelerated SLR, is a robust
and widely applicable method that can be confidently applied to obtain accurate
predictions of coastline retreat due to SLR. From the above discussion, it is apparent
that the Bruun rule is of limited use in this context. Therefore, alternative methods
to address this critical issue must now be considered. In this regard, a promising
alternative philosophy is the “coastal tract” concept presented by Cowell et al.26,37
and to the lower shoreface. The coupling mechanisms govern systematic lateral
displacements of the shoreface, and therefore determine trends in shoreline advance
and retreat. These changes manifest as the most fundamental modes of coastal
evolution upon which higher-order (shorter-term, i.e., subdecadal scale) changes are
superimposed.
Prediction of shoreline change adopts different approaches, depending on
the space and timescale over which predictions are required. For short-term
(subdecadal) coastal change (event and synoptic-scale changes occurring over hours
through seasons to years), the focus is generally on the local sediment dynamics.
These affect the shoreline plan form and the cross-shore profile (e.g., shoreline and
profile models) in response to fluctuations in environmental conditions (i.e., the
wave climate, littoral sediment budgets, sea level, and the effects of anthropogenic
activities). Theoretical and empirical approaches to these subdecadal timescales
generally focus on changes to the upper shoreface (defined loosely as the active
zone; cf. Ref. 38), which correlate with shoreline movements. These changes are
moderated by littoral sediment budgets and by sediment “production” via shoreline
erosion cutting into onshore sand reserves (e.g., eroding dunes or cliffs), or through
artificial nourishment of beaches.
The practical imperative for long-term prediction (decades or longer) requires
an expanded scope as included in the coastal tract concept that includes the lower
shoreface and the interaction between the shoreface and backshore environments
(Fig. 37.2). The upper shoreface has cross-shore length scales that are typically
two to three orders of magnitude less than for the lower shoreface (depicted in
Fig. 37.2). This scale difference means that changes on the lower shoreface are
associated with disproportionately larger changes on the upper shoreface, due to
mass continuity for sediment exchanges between the two zones.39,40 The upper
shoreface is subject to a similar interaction with the backshore, which comprises a
morphologically active zone located between the upper shoreface (ocean beach) and
Fig. 37.2. Physical morphology encompassed by the coastal tract (after Ref. 37, see text for
explanation).
August 25, 2009 14:7 9.75in x 6.5in b684-ch37 FA
the mainland. This zone may variously include dunes, wash over surfaces, flood-tide
deltas, lagoon basins, tidal flats (Fig. 37.2A), mainland beaches (Fig. 37.2B), and
fluvial deltas (Fig. 37.2C). Each of these may be present or absent, depending on
local conditions, especially the regional substrate slope.17
The sediment exchanges depicted by the arrows in Fig. 37.2 occur in principle
during any average year and on all longer timescales. These exchanges are summa-
rized schematically in Fig. 37.3 that differentiates sediment fluxes into sand and mud
fractions. For coastal change on any scale, antecedent morphology, sea-level change,
fluvial discharges, and littoral sediment budgets can be regarded as boundary con-
ditions for the coastal area of interest.
For subdecadal prediction of horizontal movements in the upper shoreface, sand
exchanges with the lower shoreface (Fig. 37.3B) are usually ignored because these
fluxes are so small that resulting morphologic change is negligible: i.e., the annual
closure-depth concept.27,41 The fluxes of fine sediments (Fig. 37.3C and 37.3D) are
not directly relevant to the upper-shoreface sediment budget because mud depo-
sition there is negligible. For long-term predictions, like on the scale of climate
change, however, none of the internal sediment exchanges depicted in Fig. 37.3
can be ignored. This is because systematic residual fluxes, that are small on the
subdecadal timescale, eventually cumulate through time enough to produce nonneg-
ligible (i.e., measurable) morphologic changes. Moreover, the changes in morphology
of the backbarrier, lower shoreface, and upper shoreface cause these three zones to
interact dynamically: i.e., the sediment exchanges themselves become influenced by
the morphologic changes.
Fig. 37.3. Schematic representation of mechanisms steering the location of the upper shoreface
(after Ref. 37).
August 25, 2009 14:7 9.75in x 6.5in b684-ch37 FA
37.7. Conclusions
The characteristics of coastline response to SLR suggested by the Bruun rule seem
conceptually correct. However, under present near-still-stand SLR conditions, the
Bruun effect can easily be subordinate to a host of other processes. Therefore,
any good comparisons between shoreline recessions measured in the last century
under near-still-stand SLR and corresponding Bruun rule predictions are likely to
be fortuitous and cannot be considered as conclusive verification of the Bruun rule.
Furthermore, due to the many restrictive assumptions associated with the Bruun
rule, it is not applicable in most natural environments, as exemplified by the
exclusion by Zhang et al.14 of almost 70% of the study area from their analysis. The
applicability of the Bruun rule is particularly compromised in the vicinity of tidal
basins associated with inlets as the Bruun effect can be overridden when the basin
area is larger than 10 km2 .15
Clearly, the Bruun rule is not suitable to obtain exact and site-specific predic-
tions of coastal recession due to SLR, particularly in view of the accelerated SLR
projected for the 21st century. At best, any predictions obtained via the Bruun rule
should be considered only as broadly indicative, order-of-magnitude estimates that
are not suitable for direct use in making planning decisions. The only robust solution
to the problem lies in comprehensive bottom-up (small-scale, process-based) and
top-down (large-scale, behavior-based) numerical modeling that adopts the coastal
tract philosophy as a conceptual template. Once comprehensively validated by field
data, such numerical models can be strategically applied to determine quantitative
forcing-response relationships of complex, nonlinear coastal processes. These rela-
tionships can then be aggregated and/or parameterized and embedded into a robust
and easy-to-use numerical model, which accounts for at least the primary physical
processes governing coastline response to SLR. Such a process-based approach would
constitute a significant step forward from the Bruun rule and is likely to provide
more scientifically robust and reliable predictions of coastline retreat due to future
SLR.
In closure we note that here we have only discussed the impact of a change
of the rate of SLR on coastline evolution and not the impacts associated with
other climate change driven impacts such changes in regional hurricane or typhoon
climates, changes in dominant wave direction, storm frequency and intensity,42 and
ENSO, NAOs, and SOI oscillations (cf. the chapter by Komar et al. and Ref. 43),
which may have local and/or regional impacts as important as or even exceeding
the impact due to SLR alone.
Acknowledgments
M.J.F.S. was supported by the project “Sustainable Development of North Sea and
Coast” (DC-05.20) of the Delft Cluster research project dealing with sustainable
use and development of low-lying deltaic areas in general and the translation of
specialist knowledge to end users in particular.
August 25, 2009 14:7 9.75in x 6.5in b684-ch37 FA
References
1. IPCC, Contribution of Working Group I to the Fourth Assessment Report of the Inter-
governmental Panel on Climate Change. Summary for Policymakers (2007), 18 pp.
2. S. Rahmstorf, A. Cazanave, J. Church, J. Hansen, R. Keeling, D. Parker and
R. Somerville, Science 316, 709 (2007).
3. S. Rahmstorf, Science 315, 368 (2007).
4. P. Bruun, Sea-level rise as a cause of shore erosion, J. Waterw. Harbor. Div., ASCE,
88, 117–130 (1962).
5. L. Hamm, M. Capobianco, H. H. Dette, A. Lechuga, R. Spanhoff and M. J. F.
Stive, A summary of European experience with shore nourishment, Coast. Eng. 47(2),
237–264 (2002).
6. O. H. Pilkey and J. A. G. Cooper, Society and sea level rise, Science 303, 1781–1782
(2004).
7. J. A. G. Cooper and O. H. Pilkey, Sea level rise and shoreline retreat: Time to abandon
the Bruun rule, Global Planet. Change 43, 157–171 (2004).
8. P. A. Rosen, A regional test of the Bruun rule on shoreline erosion, Marine Geol. 26,
M7–M16 (1978).
9. E. B. Hands, The great lakes as a test model for profile responses to sea level changes,
Handbook of Coastal Processes and Erosion Boca Raton, ed. P. D. Komar (CRC Press,
1983), pp. 176–189.
10. C. H. Everts, Sea level rise effects on shoreline position, J. Waterw. Port Coast. Ocean
Eng. 111(6), 985–999 (1985).
11. O. H. Pilkey and T. W. Davis, An analysis of coastal recession models: North Carolina
coast, Sea-level Fluctuation and Coastal Evolution, SEPM, eds. D. Nummedal,
O. H. Pilkey and J. D. Howard (Society for Sedimentary Geology) Special Publication
No. 41 (Tulsa, Oklahoma, 1987), pp. 59–68.
12. R. G. Dean, Beach response to sea level change, The Sea, Vol. 9 (Wiley, 1990),
pp. 869–887.
13. SCOR Working Group 89, J. Coast. Res. 7(3), 895 (1991).
14. K. Zhang, B. C. Douglas and S. P. Leatherman, Climat. Change 64, 41 (2004).
15. M. J. F. Stive, Climat. Change Editorial Comment 64, 27 (2004).
16. N. Mimura and H. Nobuoka, Verification of Bruun rule for the estimate of shoreline
retreat caused by sea-level rise, Coastal Dynamics 95, eds. W. R. Dally and
R. B. Zeidler (American Society of Civil Engineers, New York, 1995), pp. 607–616.
17. P. J. Cowell, P. S. Roy and R. A. Jones, Simulation of LSCB using a morphological
behaviour model, Marine Geol. 126, 45–61 (1995).
18. J. E. A. Storms, Event-based stratigraphic simulation of wave-dominated shallow-
marine environments, Marine Geol. 199(3–4), 83–100 (2003).
19. H. J. Steetzel, Cross-shore transport during storm surges, PhD. thesis, Delft University
of Technology (1993).
20. M. R. A. van Gent, E. M. Coeveld, D. J. Walstra, J. van de Graaff, H. J. Steetzel and
M. Boers, Dune erosion tests to study the influence of wave periods on dune erosion,
Proc. Int. Conf. Coast. Eng., ASCE, San Diego (2006).
21. E. C. F. Bird, Coastline Changes (Wiley & Sons, New York, 1985), p. 219.
22. S. P. Leatherman, K. Zhang and B. C. Douglas, Sea level rise shown to drive coastal
erosion, EOS Trans. 81(6), 55–57 (2000).
23. A. D. Short, A survey of Australian beaches, Keynote Lecture, Coastal Sediments
2003, Clearwater, Florida (2003).
24. J. R. Curray, Transgressions and regressions, Papers in Marine Geology, ed.
R. C. Miller (McMillan, New York, 1964), pp. 175–203.
August 25, 2009 14:7 9.75in x 6.5in b684-ch37 FA
25. D. J. P. Swift, Continental shelf sedimentation, Marine Sediment Transport and Envi-
ronmental Management, eds. D. J. Stanley and D. J. P. Swift (Wiley, New York, 1976),
pp. 311–350.
26. P. J. Cowell, M. J. F. Stive, A. W. Niedoroda, D. J. P. Swift, H. J. de Vriend,
M. C. Buijsman, R. J. Nicholls, P. S. Roy, G. M. Kaminsky, J. Cleveringa, C. W. Reed
and P. L. De Boer, The Coastal-Tract (Part 2): Applications of aggregated modelling
to lower-order coastal change, J. Coast. Res. 19(4), 828–848 (2003).
27. R. J. Nicholls, W. A. Birkemeier and G.-H. Lee, Evaluation of depth of closure using
data from Duck, NC, USA, Marine Geol. 148, 179–201 (1998).
28. R. G. Dean, Equilibrium beach profiles: Characteristics and applications, J. Coast.
Res. 7, 53–84 (1991).
29. R. G. Dean and E. M. Maurmeyer, Models of beach profile response, CRC Handbook
of Coastal Processes and Erosion, eds. P. Komar and J. Moore (CRC Press, Boca
Raton, 1983), pp. 151–165.
30. D. J. Beets, L. van der Valk and M. J. F. Stive, Holocene evolution of the coast of
Holland, Marine Geol. 103, 423–443 (1992).
31. P. J. Cowell, M. J. F. Stive, P. S. Roy, G. M. Kaminsky, M. C. Buijsman, B. G. Thom
and L. D. Wright, Shoreface sand supply to beaches, Proc. 27th Int. Conf. Coast.
Eng., ASCE (2001), pp. 2495–2508.
32. A. J. Bowen, Simple models of nearshore sedimentation, beach profiles and longshore
bars, The Coastline of Canada, ed. S. B. McCann, Geological Survey of Canada Paper,
80-10 (1980), pp. 1–11.
33. J. H. List and A. S. Farris, Large-scale shoreline response to storms and fair weather,
Proc. Coast. Sediments ’99, ASCE, Reston, VA (1999), pp. 1324–1338.
34. J. Dronkers, Morphodynamics of the Dutch delta, Physics of Estuaries and Coastal
Seas, eds. J. Dronkers and M. Scheffers (Balkema, Rotterdam, 1998), pp. 297–304.
35. M. J. F. Stive and Z. B. Wang, Morphodynamic modeling of tidal basins and coastal
inlets, Advances in Coastal Modeling, ed. C. Lakhan (Elsevier, 2003), Chapter 13,
pp. 367–392.
36. C. T. Friedrichs and D. G. Aubrey, Non-linear tidal distortion in shallow well-mixed
estuaries: A synthesis, Est. Coast. Shelf Sci. 27, 521–545 (1988).
37. P. J. Cowell, M. J. F. Stive, A. W. Niedoroda, H. J. De Vriend, D. J. P. Swift,
G. M. Kaminsky and M. Capobianco, The Coastal-Tract (Part 1): A conceptual
approach to aggregated modelling of low-order coastal change, J. Coast. Res. 19(4),
812–827 (2003).
38. M. J. F. Stive and H. J. De Vriend, Modelling shoreface profile evolution, Marine
Geol. 126, 235–248 (1995).
39. P. S. Roy, P. J. Cowell, M. A. Ferland and B. G. Thom, Wave dominated coasts,
Coastal Evolution: Late Quaternary Shoreline Morphodynamics, eds. R. W. G. Carter
and C. D. Woodroffe (Cambridge University Press, Cambridge, 1994), pp. 121–186.
40. P. J. Cowell, D. J. Hanslow and J. F. Meleo, The shoreface, Handbook of Beach and
Shoreface Morphodynamics, ed. A. D. Short (Wiley, Chichester, 1999), pp. 37–71.
41. R. J. Hallermeier, A profile zonation for seasonal sand beaches from wave climate,
Coast. Eng. 4, 253–277 (1981).
42. R. Ranasinghe, D. Lord, D. Hanslow and K. McInnes, Climate change impacts on
NSW coasts and estuaries, Proc. Coasts and Ports ‘07, Melbourne, VIC, Australia,
CD ROM published by Engineers Australia (2007).
43. R. Ranasinghe, R. McLoughlin, A. D. Short and G. Symonds, The southern oscillation
index, wave climate, and beach rotation, Marine Geol. 204, 273–287 (2004).
August 25, 2009 14:7 9.75in x 6.5in b684-ch37 FA
Chapter 38
This chapter deals with flood risk analysis and assessment. The conceptual model
source pathway receptor consequence for flood risk analysis is presented and
its components are analyzed. The methodology to extract the predicted proba-
bility of coastal flooding from risk sources and pathways, as well as the expected
damages from risk receptors are introduced and examined. Reliability analysis of a
coastal system is also briefly discussed. Quantitative methods to define acceptable
flooding probabilities on the level of the protected area are presented. Tools such
as cost–benefit analysis, utility models, and the life quality index are introduced
to define the “tolerable” risk of flooding.
38.1. Introduction
A flood is a temporary covering by water of land normally not covered by water. This
definition includes floods from rivers, mountain torrents, Mediterranean ephemeral
water courses, and floods from the sea in coastal areas and may exclude floods
from the sewerage systems. Worldwide, over 65% of the large cities with more than
2.5 million inhabitants are located in coastal areas, deltas, and estuaries and almost
40% of the population of our planet is living within a 100-km-wide coastal strip. In
the past, flooding disasters have struck human society all over the world and in the
future they may be expected to do so again. Flooding still remains one of the most
widely distributed natural hazards in the whole world.
Flood risk is the combination of the probability of a flood event and of the
potential adverse consequences to human health, the environment, and economic
activity associated with the flood event. For coastal areas, the hazards that con-
tribute to the risk of flooding are related to the water surface elevation process.
1039
August 25, 2009 14:14 9.75in x 6.5in b684-ch38 FA
The surface elevation can be described by the water level elevation and a number of
parameters for individual waves, e.g., the wave height, the wave period, the wave-
length, and the wave steepness.
Risk can be defined in a number of different ways.1 Some definitions of risk
are the following: (1) risk is the probability of an unwanted event; (2) risk is the
consequence of an unwanted event; (3) risk is the function of the probability and
the effect; and (4) risk is defined as risk = probability × consequence.
The first two definitions are not well applicable because the risks with small prob-
abilities and very large consequences and those with great probabilities and small
consequences cannot be described properly. The third definition is more general and
helps assigning a weight to the consequence of an unwanted event, which depends
on the seriousness of the consequence. This is especially important for small prob-
abilities and great consequences and can include matters such as risk aversion in
the assessment of risk. The definition of risk as product of probabilities and con-
sequences (fourth definition) is simpler and offers a good basis for comparisons of
risks.
Following the fourth definition, risk is a combination of the chance of a par-
ticular event, with the impact that the event would cause if it occurred. Risk has
two components: the chance or probability of an event occurring and the impact or
consequence associated with the event. The consequence can be desirable or unde-
sirable; however, risk is typically concerned with the likelihood of an undesirable
consequence.
The probability component of the risk consists of1 : (a) the probability that
the unwanted event occurs, (b) the probability that the unwanted event leads to
a possible effect, and (c) the probability that this effect leads to the considered
consequence. For example, the probability of flooding at a defined area consists of
the probability that a dike slips, the probability that water flows into the polder,
when the water level is high and the probability of damage and casualties due to
this effect. The total probability component of risk amounts to: P (E1 ) · P (E2 | E1 ) ·
P (G | E1 ∩ E2 ), where E1 is the unwanted event (e.g., collapse of dike), E2 is the
effect (e.g., inundation of polder), and G is the consequence (e.g., number of dead
people by drowning). The probability P (E1 ) · P (E2 | E1 ) is usually the probability
of failure Pf .
The consequence of flooding is an impact such as economic, social, or environ-
mental damage/improvement. It can be expressed quantitatively (e.g., monetary
value), by category (e.g., high, medium, and low), or descriptively. Typical descrip-
tions of consequence are: (a) economic damage, (b) number of people/properties
affected, (c) occurrence of specified event, and (d) degree of harm to an individual
(injury, stress, etc.). The consequence of an unwanted event can be both deter-
ministic and statistic. If a statistic effect is involved, a probability density function
can be defined for the risk, for which the expected value can be determined.
Failure to recognize the full spatial extent of the consequences will bias the
decision-making process and could lead to suboptimal decisions. Equally, the tem-
poral scale of the consequences must be explicitly considered. For example, the
impact of flooding should not only be considered with reference to the physical time
for the floods to recede, but also the time taken to reestablish community businesses
August 25, 2009 14:14 9.75in x 6.5in b684-ch38 FA
as well as stress- and health-related issues that may persist in flooded communities.2
Only with proper consideration of the variety and temporal and spatial scale of the
consequences can appropriate decisions be made.
To consider the significance of risk, reference must be made not only to the
numeric value of the product of probability and consequence, but also to how it
will be perceived by society or the individual. Also, in some cases it is important to
understand the nature of risk, distinguishing between rare, catastrophic events and
more frequent less severe events. For example, risk methods adopted to support the
targeting and management of flood warning represent risk in terms of probability
and consequence, but low probability/high consequence events are treated very
differently to high probability/low consequence events.
Flood risk management deals with the analysis and governance of the flood
hazards, the flood vulnerability (the resilience of a particular group of risk receptors)
and the resulting flood risk. It considers all natural and societal processes related
with the flood issues of a coastal cell. Based on such a risk analysis, the tolerability
of risk has to be assessed and risk reduction options to be defined. Decision-making,
implementation, and control of actions are also parts of flood risk management.
Flood risk management includes the analysis and assessment of risks and the
formulation and implementation of risk reduction options. Risk analysis includes
hazard as well as vulnerability determination and estimates the integral risk. Risk
assessment deals with the evaluation and weighing of costs and benefits to derive a
tolerable level of risk, taking risk perception into account. Risk reduction deals with
planned and realized measures and instruments supposed to decrease risks before
(pre-flood), during the event and after an event (post-flood). Figure 38.1 shows the
three tasks included in flood risk management.
Risk analysis is divided into hazard, vulnerability, and risk determination and
constitutes the main subject of interest of the present chapter. It deals with extreme
Hazard Vulnerability Risk Risk Risk Pre-Flood risk Flood event Post-flood risk
determination determination determination perception weighing reduction risk reduction reduction
Probability and Potential Probability of Overall view of Agreement on Physical Physical Physical
characteristics social, certain social, risk held by a tolerability of measures, measures, measures,
(e.g. water economic, and economic, and person or risk, weighing regulatory, regulatory regulatory,
depth, flow ecological ecological group benefits, and financial, and instruments, financial, and
velocity, damage damage to a depending on costs communicative and communicative
duration) of depending on certain hazard cultural and depending on instruments to communicative instruments to
flood events value and personal individual or reduce the risk activities to deal with
susceptibility values, collective by prevention reduce the risk existing flood
to a certain experiences, perception and and/or of an ongoing damages
type of hazard and feelings interest preparedness event
coastal events, their characteristics, and their probability of occurrence and the
potential damage (social, ecologic, and economic) that they may cause if they
occur, depending on the susceptibility of the society to a certain type of hazard.
Risk analysis encompasses investigations on the impact of climate and societal
change. Risk determination is the main outcome of this component of the flood risk
management process.
In comparison to the “objective” analysis, risk assessment includes individual
and collective perception and weighing of risk and will also be further analyzed in the
present chapter. There are psychologic–cognitive and sociologic–cultural approaches
of explicating risk perception, the overall view of risk depending on cultural and
personal values, experiences, and feelings. Risk assessment also covers the evalu-
ation and weighing of risk. Evaluation is dedicated to determine effectiveness, effi-
ciency, robustness, flexibility, and other criteria, whereas weighing means to balance
probable costs of risk and risk reduction measures with expected benefits by using
flood-prone areas.3
Finally, risk reduction includes pre-flood, flood event, and post-flood risk
reduction, namely physical measures, regulatory, financial, and communicative
instruments to reduce the risk by prevention and/or preparedness, by coping with
the risk of an ongoing event and by dealing with existing flood damages, respec-
tively. As far as the overall principle of risk reduction is concerned, flood risk man-
agement is based on a shift from resistance against floods toward a resilience of the
entire flood risk system. The system-based risk analysis provides the prerequisites
for a holistic ex ante analysis of the outcomes of risk reduction.3
Since flood risk management has been indicated as a societal task, the decision
making and development process of responsible actors come into the foreground.
Decision-making is fundamentally about making choices. The societal decision
process concerning the protection against flooding disaster shows an increasing
demand on the performance of the flood defence system over time. Society is a
continuous process of decision making on the protection against flooding. Develop-
ments in science and technology have led to a demand for a quantitative method for
judging flooding risk. The decision process in a risk-based approach relies upon a
variety of decision criteria and must be able to distinguish the merits and demerits
of an action over another (i.e., it must be able to compare the risk and performance
of one option with another).
The conventional engineering distinction between failure and nonfailure has for
many years been translated to the design of flood defences via the concept of a
“design load,” almost always expressed in terms of a return period. In the current
design practice three approaches exist2 : (a) a deterministic, where design is on the
whole based on the concept of design loads, be they wave heights or water levels for
coastal areas and where precise, single values are used for all variables; (b) a deter-
ministic including sensitivity testing, where deterministic outcomes are tested by
systematically varying input values; and (c) a probabilistic, where the variability of
input values is taken into account to provide a probabilistic result. In cases (a) and
(b), the design equations or constraints are written in terms of safety factors. Safety
factors have the advantage of being easily interpreted in terms of their physical
or engineering meaning, but do not give clear information on the reliability of the
August 25, 2009 14:14 9.75in x 6.5in b684-ch38 FA
structure. However, the deterministic methods are aimed at preventing the worst,
without taking the likelihood of the various circumstances into account. The main
advantage of the probabilistic-based design [approach (c)] is that the reliability of
the structure can be evaluated if the statistics are well defined. However, it has
the shortcoming of being very sensitive to tail assumptions, and in some cases,
as, for example, rubble-mound stability, runup, overtopping, and geotechnical sta-
bility, parametric dependencies and statistics are difficult to define. Applications of
approach (c) in coastal engineering have been published among others by Bakker
and Vrijling,4 Burcharth et al.,5 and Voortman et al.6 In the last few years all
methods have been improved by applying optimization techniques.7 The reliability
design method with partial safety factors is superior to the conventional methods,
in the sense that it takes into account the uncertainties involved in the load and
resistance parameters. Nevertheless, it cannot predict the amount of damage when
it occurs, just like the conventional methods.8
The benefit of a risk-based approach, compared to the other approaches to
design or decision making, is that it deals with outcomes. A risk-based approach
enables informed choices to be made based on comparison of the expected out-
comes and costs of alternative courses of action. This is distinct from, for example,
a standards-based approach that focuses on the severity of the load that a par-
ticular flood defence is expected to withstand. Using a risk-based approach, risk
assessment often involves more complete representation of the flood/erosion system.
This more holistic approach will be able to account for: (a) complex physical mech-
anisms, spatial, and temporal variations in natural hazards (e.g., natural vari-
ability in wind, wave, rainfall, and water level conditions); (b) descriptions based
on sparse/incomplete data; and (c) multiple stakeholders with differing, often con-
flicting, values and objectives.2
The decision process in a risk-based approach is split into three levels.1 Level A
incorporates decision making on the appropriate probability of flooding on the level
of protected areas. On level A, the cost of protection on the level of the full system
and the consequences of flooding in the protected area are considered. Level B con-
sists of the decision making on the failure probability per individual structure in the
protection system for a given value of the probability of flooding. The minimization
of the construction costs of the full system is utilized for the allocation of a given
probability of flooding to individual sections of a system. Finally, level C includes
decision making on the geometry of an individual flood defence structure for a given
probability of failure. Considering the cost of elements of the cross-section and a set
of failure modes, the construction costs of an individual section of the protection
system are minimized. Risk-based design is performed only on level A. On levels B
and C, reliability-based design is applied.
Section 2 deals with flood risk analysis. The conceptual model source–pathway–
receptor–consequence for flood risk analysis is presented and its components are
analyzed further. The methodology to extract the predicted probability of coastal
flooding from risk sources and pathways, as well as the expected damages from risk
receptors are introduced and examined. Reliability analysis of a coastal system is
also briefly discussed in the subsection dealing with the pathways of risk. Section 3
concentrates on risk assessment. Quantitative methods to define acceptable flooding
August 25, 2009 14:14 9.75in x 6.5in b684-ch38 FA
probabilities on the level of the protected area are presented. Tools such as cost
benefit analysis, utility models, and the life quality index are introduced to define
the “tolerable” risk of flooding. Section 4 presents the main ideas and conclusions
of the chapter, recommending an approach to risk management that embraces all
sources, pathways, and receptors of risk and considers combinations of structural
and nonstructural solutions.
Risk Sources
Modeling of loading
and resistance of each
Hydrological/ hydraulic boundary conditions, including
defence structure
uncertainties and focusing on extremes and joint
probabilities
Pfc
Fig. 38.3. The link between risk sources and the predicted probability of failure.
larger amount of data in the inference, which is expected to reduce uncertainty. This
approach, referred to as the (POT), is widely used in the field of hydrology.17 The
choice of the threshold should be based on physical and meteorologic information.
For POT methods, independence requires a suitable combination of both threshold
and minimum separation time between events.
There have been many proposals for the distribution functions of extreme wave
heights. The current consensus is to choose a function best fitting to the extreme
wave data among several candidates, which include the Fisher–Tippett type I
(FT-I), Fisher–Tippett type II (FT-II), and Weibull distributions. A difficulty in
selecting a distribution function best fitting to a data set of storm wave heights is
the lack of information on the population distribution of extreme events.
Goda8 (Fig. 38.4) exhibits the difference between the distribution functions of
the partial duration and annual maximum series data in terms of return period for a
given wave height. The 50-year wave height is set at 8.0 m. Three distributions FT-II
and two Weibulls are shown with their converted annual maximum distributions.
Because of the definition of annual maxima, its return period must be greater than
1.0 year. Except for the zone of short return period, the distribution for the annual
maximum series exhibits little difference from that for the partial duration series.
When an extreme value model (GEV or POT) is fitted to the data of waves,
storm surges or any other variable of interest, the parameters of the applied model
are estimated using different methods. Three of them are the maximum likelihood
estimation (MLE), the L-moments (LM), and the Bayesian estimation procedure.
Fig. 38.4. Comparison of the distribution functions of partial duration and annual maximum
series data for the 50-year wave height.8
August 25, 2009 14:14 9.75in x 6.5in b684-ch38 FA
Fig. 38.5. Wave height estimates using ML, Bayesian, and LM estimation procedures.18
Galiatsatou18 (Fig. 38.5) estimates median and 95% confidence intervals of return
level estimates for wave heights, when extreme value parameters are estimated with
(a) maximum likelihood (ML), (b) Bayesian with noninformative prior distributions,
and (c) L-moments (LM) estimation procedures.
The choice of the water level distribution is another subject which has received
considerable interest. The observed water level is the sum of a deterministic astro-
nomic tidal component and a stochastic meteorologically induced component, the
surge component. Surges are generated by a combination of air pressure and the
wind field over the surrounding region. The observed water level is modeled either
after an analysis into its components or as a single variable. Galiatsatou and
Prinos19 show estimates of the median and 95% confidence intervals of surge (raw
and log-transformed data) return levels, using the Bayesian approach and the MLE
procedure to estimate the parameters of the extreme value models fitted to two
stations off the Dutch coast. Figure 38.6 shows return level plots for the station
Eld (Eierlandse Gat) at the northern part of the Dutch coast in the North Sea.
Figure 38.6(a) presents estimates of the surge level for raw data, when noninfor-
mative prior distributions for the parameters in the Bayesian framework are utilized
and Fig. 38.6(b) gives corresponding estimates when information from a neighboring
station is used.
Spatial dependence of a single process (e.g., storm surge) has been recently
included in the estimation of return levels, using extreme value methodologies.
Galiatsatou and Prinos20 use a “spatial linkage” assumption to extract results of
surge (return) levels for a process of daily storm surges, taking account of the degree
of spatial dependence among nine sites of the Dutch coast in the North Sea. Return
levels and their standard errors are extracted under the assumption of ξ-linkage
and compared to the ones when no parameter linkage is considered. Variance and
August 25, 2009 14:14 9.75in x 6.5in b684-ch38 FA
(a) (b)
Fig. 38.6. Surge levels against return period for the Eld station based on ML-estimates (dashed
line) and the posterior Bayesian distributions (solid line) for the raw data, using (a) noninformative
prior distributions and (b) information from the neighboring station Son.19
efficiency gains in the model parameters are estimated and discussed further in
the paper.
In Voortman,21 two different possibilities of modeling the distribution of water
levels are examined: (a) by direct statistic analysis of the observed water levels and
(b) by combining the probability distribution of wind speed with a relevant physical
model. Realizing that physical models are often imperfect, in option (b) an estimate
of the model uncertainty is necessary.
Where the source consists of one or more variables (e.g., coastal flooding caused
by extreme wave heights and water levels), it is necessary to consider their joint
probability. Joint exceedance probability refers to the chance of two or more par-
tially related variables occurring simultaneously. Joint exceedance combinations of
wave heights and sea levels with a given chance of occurrence are defined in terms
of sea conditions in which a given wave height is exceeded at the same time as a
given water level (or its surge component) being exceeded. There are different levels
of complexity for joint probability methods but all require some assessment of the
dependence between variables. A relatively simple method uses the marginal distri-
butions of the variables involved and an assessment of their dependence. The joint
return period is expressed in terms of combinations of the marginal return periods.
The main disadvantage of the method is that the joint return period of the vari-
ables (e.g., wave height and water level) is not the same as the return period of the
response variable. A more sophisticated approach involves fitting a probability dis-
tribution to the joint distribution of variables and extrapolating the joint density
function. The benefit of this approach is that the return period of the response
function can be easily determined. The main disadvantage is that this approach
needs a large amount of concurrent data of the variables involved. For the second
approach Multivariate EVT is used to extract joint exceedance probabilities and
joint return periods.
Multivariate Extreme Value Theory (MVE) is used to describe the joint dis-
tribution of two or more variables, and appropriate methodology has only been
August 25, 2009 14:14 9.75in x 6.5in b684-ch38 FA
developed in recent years. Coles and Tawn,22,23 Joe et al.,24 Zachary et al.,25 and
Schlather and Tawn26 give the main aspects of MVE with applications to oceano-
graphic and other sort of datasets. Fang and Hogben,27 Haver,28 Athanassoulis
et al.,29 Morton and Bowers,30 De Haan and De Ronde,31 Repko et al.,32 and
Ferreira and Soares33 describe the joint probability distribution function of long-
term hydraulic conditions. Especially when the main interest is in the design
of flood defence structures, the extreme conditions are important, which implies
that the dependence between hydraulic conditions needs to be accounted for. The
joint probability analysis of extreme waves and water levels thus is significant in
order to estimate more accurately the extreme environmental loading on a coastal
structure. Because wind setup (storm surge) and wave conditions depend on the
same driving force, a strong dependence between them is observed under extreme
conditions.
Galiatsatou18 compares different pairs of bivariate observations of extreme waves
and surges with reference to joint exceedance probabilities, in order to find the
most severe sea state caused by the two variables. A bivariate logistic model and a
sequential estimation procedure are used for this case to extract joint exceedance
probabilities. The parameters of the margins of the bivariate distribution are defined
by maximum likelihood (ML), Bayesian with flat prior distributions, and L-moments
(LM) estimation procedures. Figure 38.7 presents the effect of different marginal
parameters to the joint return levels of surges and wave heights for return periods
of 1, 10, 50, and 100 years, for the most severe sea state.
Voortman6 describes a model for the hydraulic boundary conditions, where
the JPDF (joint probability density function) of hydraulic conditions nearshore is
written as a function of properties of the wind field, geometry of the North Sea
basin, astronomic tide, and nearshore bathymetry. A description of the boundary
conditions on the basis of parametric models found on physical concepts is favored
in the thesis over pure statistic methods.
EVT is progressively utilized to compute limit state equations, which corre-
spond to different failure modes of coastal and offshore structures. The input param-
eters of the limit state function are the stochastic load and strength parameters
Fig. 38.7. Bivariate return levels for surges and wave heights with three methods of estimating
marginal parameters (ML, Bayesian, LM) (a) Bayesian and ML estimators and (b) ML and LM
estimators.18
August 25, 2009 14:14 9.75in x 6.5in b684-ch38 FA
corresponding with the failure mode, like wave heights and wave periods. It should
be noted that oceanographic data like wave heights and periods experience depen-
dence, which leads to the use of joint probability analysis to describe in a better
way the wave climate and make probabilistic design and risk-based optimization
possible. Galiatsatou and Prinos34 implement four bivariate models (bivariate log-
normal, bivariate log-normal with correction for wave skewness, a parametric distri-
bution based on transformations to normality, and bivariate logistic distribution) to
the data of significant wave height and wave period from the field station Eld off
the Dutch coast and compare them mainly on their ability to correctly describe the
data and on their behavior when extrapolation is of interest.
In practice such multivariate problems are often simplified to a univariate
setting, focusing on the relevant variable for the structure under consideration
(structure variable approach). Brunn and Tawn35 compare the method which uses
the joint distribution of wave heights and water levels and the structure variable
approach in the general framework of estimating the risk of coastal flooding in
∆1 ∆2 ∆3
northwest Europe. The problem is structured as: W −→ X(O) −→ X(N) −→ Z,
where W is the wind field over the northwest European continental shelf; X(O) and
X(N) are the sea state variables offshore and nearshore, respectively; Z is the impact
variable on the flood defence system; and ∆1 , ∆2 , and ∆3 is a series of complex
hydrodynamic numeric models. The vector X(O) = (X1 , X2 ) is defined as the vector
of significant wave height and water level. The structure variable Z is analyzed with
the nested numeric model ∆3 (∆2 ) approximated by: Z = 0.3X1 + X2 − υ, where υ
is defined as the height of the dike and flooding is assumed if Z > 0. In case of using
the joint distribution of waves and water levels, the dependence model of Ledford
and Tawn36 characterizes the asymptotic structure of the joint survivor function,
which enables the identification of a whole class of models, for which the limiting
dependence of the componentwise maxima is independence. Figure 38.8 shows (a)
estimates of the design parameter υp for the failure region 0.3X1 + X2 − υp esti-
mated using the joint probabilities method and the structure variable approach and
(b) the failure regions {(x1 , x2 ) : 0.3x1 + x2 > v} with v = 2.5, 3.5, 4.5, and 5.5:
the regions are shown after transformation to logged unit Fréchet marginal scales
[log10 (Y1 ), log10 (Y2 )]. In both (a) and (b), (X1 , X2 ) are the Dutch offshore data
and the thresholds used correspond to the 80% empiric quantiles of the variables
involved.
The analysis of offshore variables (wave heights, water levels, and wave periods)
and the extraction of high enough return levels for them, seems to be a suitable
compromise in the analysis of extreme events. Contrarily to the associated nearshore
variables, offshore data are regionally homogeneous (e.g., on the 50-km scale) as they
correspond to observations at a depth at which complex local interactions between
the processes do not occur.35 One statistic analysis is required, from which estimates
at different sites can be obtained by suitably changing the complex hydrodynamic
and hydraulic numeric models which propagate wave heights and water levels near
the shore. The assessment of the wave and water level conditions near the coast is
essential for the estimation of flood characteristics. Using the output of the analysis
as input to nearshore wave and water level models, the wave climate next to the
coastline is produced.
August 25, 2009 14:14 9.75in x 6.5in b684-ch38 FA
Fig. 38.8. (a) Estimated design parameter vp for the failure region 0.3X1 + X2 − vp : estimates for
the JPM (—) and SVM (. . . . . . .) and (b) failure regions {(x1 , x2 ) : 0.3x1 + x2 > v} with υ = 2.5,
3.5, 4.5, and 5.5.35
For nearshore water level predictions (tide and surge), the first-generation
models are 2DH models providing results of tide and surge components across a
given area. They solve the nonlinear shallow water equations and use inputs of
wind fields and atmospheric pressure over the modeled area. More advanced models
include the effects of breaking waves, causing setup of water levels in nearshore
areas. MIKE21 HD/NHD is a representative model from this category. The second-
generation models are 3D models that include the effects of temperature and salinity,
in addition to the characteristics of first-generation models. POLCOMS, TELEMAC
2D, and FINEL 3D are second-generation nearshore water level models.
For nearshore wave prediction there are phase-resolving and phase-averaging
models. The first-generation models of the former category (e.g., “mild slope”
models) are 2DH models which provide instantaneous surface elevations over a
given area. They include a linear representation of refraction, mild shoaling, and
an approximate representation of diffraction. First-generation models of the latter
category are 2DH wave-tracing models that provide results at a point or in an area
and they have a linear representation of refraction and shoaling (e.g., COSMOS).
The second-generation models (Boussinesq models) are 2DH models which include
nonlinear representation of diffraction, refraction, and mild shoaling. The second-
generation models are 2DH models which provide averaged results of tide and
surge components across a given area. Examples are the MIKE21 NSW and the
STWAVE models. The third-generation models are like the second-generation ones,
but include an explicit representation of the nonlinear transfer of energy resulting
from the primary wave–wave interaction frequencies (e.g., SWAN).
Response probability refers to the occurrence of a particular response (such as
overtopping or failure), which in turn depends on the joint occurrence of water levels
August 25, 2009 14:14 9.75in x 6.5in b684-ch38 FA
and waves. From a very large number of combinations of wave heights and water
levels with the same joint return period, only one will be the worst case for each
response. The probability of occurrence of the response function (e.g., overtopping
or force) calculated from the worst-case combination of wave heights and water levels
will be higher than the joint exceedance probability. In practice, a small margin of
safety is added to the joint exceedance probability predictions to try to offset this
discrepancy with the return period of the response.37 A more detailed analysis of
the pathways of risk and estimation of the predicted probability of flooding is given
in the following sections.
A first approach to determine the probability of extreme events and their impli-
cation for flooding would be to analyze the joint probability of strong winds and
high water levels, or large waves and high water levels. However, since flooding often
is a result of the combined effect of the mean water level and the wave properties
through the runup, it is also worthwhile to analyze the runup level, which is the
sum of the mean water level and the runup height.
Runup height is a critical quantity to estimate when assessing the probability of
flooding since it defines the highest elevation to which the waves might reach. The
runup height is normally referenced to the still water level and it includes the wave
setup.38 The earliest formula for wave runup height R was developed by Hunt39 :
R tan β
= (38.1)
Ho Ho /Lo
where β is the beach slope, L the wavelength, and o denotes deepwater conditions.
The runup height calculated for every value in the calculated wave time series can be
added to the applicable still-water level given by the water level time series. These
generated data may constitute a base for determining the risk of flooding including
overtopping of defence structures at various alongshore locations in the study area.
Hanson and Larson38 used the equation of Hunt to determine runup heights
in the southern Baltic Sea for the period 1982–2004, from which the runup levels
were derived by taking into account the water levels. Empiric distribution functions
were then fitted to the data to extrapolate the total water level (runup level) to
high return periods. Figure 38.9 illustrates the empiric distribution function for the
annual maximum runup level plotted with Gringorten’s formula together with a
fitted Gumbel distribution.
Fig. 38.9. Annual maximum runup level plotted against the reduced value from the Gumbel
distribution using the Gringorten plotting position formula together with a linear fit.38
Risk Pathways
Joint probability
distributions from
flood risk sources
Flood
Breaching initiation and growth inundation
Reliability analysis: Pf
E(D)
Performance of entire defence system and its components,
including breach growth as a key issue to provide initial
conditions for the assessment of flood wave propagation Pfc
and inundation
Fig. 38.10. The link between risk pathways and the predicted probability of failure.9
the area is estimated, taking into account the existing morphology. Although this
is a static approach, it is possible to introduce some kind of dynamics by including
the existing knowledge in local coastal response (e.g., observed variability of beach
profiles protecting the hinterland). Then, the beach response to the selected storm is
August 25, 2009 14:14 9.75in x 6.5in b684-ch38 FA
Fig. 38.11. Storm-induced profile changes: eroded volume ∆V (m3 /m) and beach retreat ∆X (m).
estimated and the inundation of the area after considering the beach modification,
(e.g., beach lowering and flattening and coastal retreat) is evaluated.40 The main
variables used to characterize storm-induced profile changes: eroded volume ∆V
(m3 /m) in the inner part of the beach and, beach retreat, ∆X (m) are shown in
Fig. 38.11.
Beach evolution models describe cross-shore and longshore transport and con-
sequently, the change of coastal morphology. One-line models (e.g., Ref. 41), and
coastal area models (e.g., Ref. 42) tend to be based on representations of physical
processes and typically include forcing by waves and/or currents, a response by
sediments to this energy input, and a morphologic updating module.43 COSMOS,
DELFT 2D/3D, LITCROSS SBEACH are some cross-shore beach evolution models,
while GENESIS is a well-known longshore model of the same category.44 Walkden
and Hall45 have recently developed a long-term model of the erosion and profile
development of soft rock shores, called SCAPE. This models the development of
the shore platform, beach, talus, and cliff at a series of representative cross-shore
profiles, each of which is represented by a column of elements. Cowell et al.46 sim-
ulated large-scale coastal change using a morphologic behavior model. The method
involved extensive parameterization of processes into simplistic behavior-oriented
models, avoiding the need for process-based determination, but reliant on obser-
vation over greater periods of time. Behavior-oriented models have been successfully
applied along coastlines where data spanning many decades were available. There
are also hybrid methods being applied which combine, for example, deterministic
process based models with parameterized behavior.43
As part of a full systems approach to flood forecasting, it will be important to
recognize the influence of defences. Beach morphology is often a crucial component
of a sea defence. The assessment of sea defence assets seldom takes account of the
August 25, 2009 14:14 9.75in x 6.5in b684-ch38 FA
morphologic variability of the bed levels and slopes in front of the defence. Changes
in bed morphology can have a significant influence on the wave conditions reaching
the defence and thus, enhance flooding risks. Numeric models that are used to
predict the evolution of the shape of the shoreline thus play a key role in assessing
the safety of coastal developments.
Wave-driven overtopping of beaches or seawalls is a highly variable process, sen-
sitive to defence structure shape and composition, to water levels and wave con-
ditions. AMAZON-CC, NEWMOTICS, and SKYLLA are some examples of this
category of models. Empiric formulae for the calculation of overtopping are also
known. Apart from defence overtopping, the understanding of the likelihood of a
breach is critical if reliable forecasts are to be made regarding the flood inundation
depths and hence risk to life. Predicting breach growth and maximum size is prob-
lematic and at present beyond the capabilities of most numeric tools. However,
breach events represent the most significant of flood scenarios and are of consid-
erable importance in determining flood risk, with two issues of primary importance
to the forecaster: (a) breach probability and, (b) breach size and invert level.44
HR BREACH and NWS BREACH belong to the category of shoreline response-
breaching models. Some advanced third-generation flood inundation models (e.g.,
FINEL 2D–3D nested) include the breaching process.
Flood inundation models, which combined with digital terrain models, describe
the processes taking place in the flood plain, have shown significant progress in
the latest years.44 The basis for flood inundation models is a digital terrain map
(DTM) of the flood plain area. DTMs are often produced based on laser scan data
and from joining survey and laser scan data. Empiric methods are often described
as pure mapping. No physical lows are involved in the simulations performed. They
are rather simple methods compared to the others, with low cost, but they provide
poor estimates of flood risk in large low lying or extensive areas where flows through
a breach may be critical in determining the flood extent. First-generation models
are essentially 1D models used with a 2DH grid. These models calculate water level
in each flood cell at given output steps and therefore enable the duration of flood
to be estimated. In cases where the flood plain is extensive, such models can give
poor results, because they do not consider the propagation of floodwater within
each cell. Infoworks RS, Infoworks CS, and ISIS are typical examples of this cat-
egory of models. Second-generation models are 1D/2DH hybrid models and fully
2D. The models of this category use the St Venant equations to model channel flow;
however, a 2D continuity equation is used to approximate flow over the flood plain
area. HYDROF and LISFLOOD-FP are second-generation models of method 2D− ,
which means that they are 2D excluding the law of conservation of momentum for
the floodplain flow. MIKE21 and TELEMAC 2D are 2D models, which solve the
2D shallow wave equations. TELEMAC 3D is an available finite element 3D model
included in the category of second-generation models. Third-generation models
simulate breaching in 3D with the flood inundation in 2D. They provide better
simulations of the flood inundation as the flow velocities at the boundary are
accurately simulated. FINEL 2D is a 2DH finite element model of medium cost
and runtime and high accuracy. Third-generation models such as CFX, FLUENT,
August 25, 2009 14:14 9.75in x 6.5in b684-ch38 FA
and PHOENIX are usually applied for local predictions of 3D velocity fields in
floodplains.
In a lot of cases, flooding occurs when there is failure of a defence. This can
arise from a functional failure (the conditions exceed those for which the defence
is designed) or a structural failure (some element or component does not perform
as intended under the design conditions). Structural failures are more dangerous
because they are unexpected. Excessive overtopping leading to erosion of the back
of the crest of the defence, or damage to the toe layers, and toe failure where
erosion of the foreshore of the defence occurs to such an extent that the structure is
undermined and collapses.43 In structural design, the failure probability is important
to evaluate the reliability of the structure.
Vulnerability
(Risk Receptors)
Flood wave
propagation and
inundation
Guidelines and methods for socioeconomic flood damage
evaluation
E(D)
Fig. 38.12. The link between risk receptors and the expected damage of coastal flooding.
of the direct damage. A further categorization of flood damage refers to the fact
whether the damages can be specified easily in monetary terms (tangible damage) or
whether the damage effects are better quantified by nonmonetary measures (intan-
gible damage), like number of lives lost or square meters of ecosystems affected by
pollution.48 Regarding loss of life (direct intangible damage), some empiric models
were developed by Jonkman et al.49 and Penning-Rowsell et al.50 for the context
of their nations. Indirect tangible damage is really difficult to model and thus,
it is rarely considered in practical vulnerability analyses. Intangible and indirect
damages have been rarely considered to date, due to methodologic difficulties. The
damage due to flooding is influenced by two factors: the size of the flooded area and
the water level in the area. These two factors can be extracted using one of the flood
inundation models described in the section dealing with pathways (see Fig. 38.10).
From an economic methodologic point of view, direct tangible flood damages are
relatively easy to measure and to model ex ante because the damage can be related
to market prices. In every flood damage evaluation study, at first, an appropriate
approach based on the scale, the study objective, the availability of resources, and
the availability of pre-existing data, has to be found. Then, an appropriate damage
category is determined (e.g., buildings and inventories). It is common place to use
the most important damage categories to reduce the effort of the study. The major
tasks of flood damage evaluation, the estimation and calculation of economic values
of potentially damaged tangible goods, and the gathering of sound information are
then carried out. The intensity of different flood types by means of various inun-
dation and flood characteristic indicators (e.g., depth and area of inundation, time
of occurrence, and velocity), the gathering of land use data for the area at risk and
information on the value of assets at risk belong to this stage of damage evaluation.
Finally, all information gathered previously is brought together and flood damages
August 25, 2009 14:14 9.75in x 6.5in b684-ch38 FA
to be expected from a certain flood event are calculated. Depending on the choice
to work with relative or absolute damage functions the percentage of property value
approach needs to be distinguished from the absolute of direct damage estimate
approach, which is more precise but much more resource and data intensive.48
The damage potential of a specific area represents the maximum possible amount
of damage which may occur if the area becomes inundated. To estimate the con-
sequences of coastal flooding, vulnerability aspects must be considered to calculate
the proportion of damage potential which will finally materialize (i.e., to determine
expected damages). In many aspects a vulnerability factor is derived for the most
important vulnerability indicators having a substantial impact on the degree of
damage produced during a flood event. In some vulnerability analyses, such a factor
is derived from expert knowledge and empiric data on flood damages and then
expressed on a scale between 0 (no loss at all) and 1 (total loss) to quantify the
expected damage reduction for certain categories of elements at risk.47 The most
important vulnerability indicator in current flood damage analysis is the exposure
indicator “inundation depth.”
The distribution of flooding damage can generally be described by a probability
density function with two or more peaks. The conditional distribution of the flooding
damage can be found by defining damage factors as a function of flooding properties.
The damage factor21 is defined as:
D
c= (38.2)
Dmax
where D is the flooding damage and Dmax is the maximum possible flooding damage
(damage potential). The maximum possible flooding damage is generally taken
equal to the value of the full inventory of the area. To find the distribution of
the consequences of flooding, it is necessary to find the distribution of the damage
factor for every type of flooding damage. If the damage factor is explicitly written
as a function of water level, flooding speed (speed of water level increase), current
velocity, wind speed, and wind direction, the distribution of the damage factor can
be found from the joint probability distribution of the flooding properties by appli-
cation of probabilistic methods. Delft hydraulics provides a function for flooding
damage to dwellings which is a function of the water level only (Fig. 38.13). Com-
bination of the damage function with the distribution of the water level inside the
area directly leads to the distribution of the damage factor.
The casualty factor is estimated to be in agreement with the general damage
factor:
Nd
ccas = (38.3)
Np
where Nd is the total number of casualties and Np is the population size in the
affected area. The methodology described above can be used to establish the prob-
ability distribution of the casualty factor. An alternative methodology for the risk
of loss of life is the loss of life expectancy at birth caused by flooding. This measure
of loss of life is used in the model of Nathwani et al.51 The problem of quanti-
fying the effects of flooding risk on the life expectancy at birth is limited to finding
August 25, 2009 14:14 9.75in x 6.5in b684-ch38 FA
Economic consequences consist also partly of direct economic damage and partly
of indirect economic damage. Damage factors, introduced above, can be used to
quantify this kind of direct damage. Indirect economic damage is defined as the
damage caused by the primary damage due to relations that exist between the
flooded area and the rest of the economy. The estimates of economic consequences
of flooding depend on the choice of the scope of the study and on the definition of
cost. It may be clear that quantification of the macro-economic effects of flooding
requires extensive modeling.
Within the European Research Program FLOODsite (Integrated Flood Risk
Analysis and Management Methodologies) methods of flood damage evaluation
in four European countries: England, the Netherlands, the Czech Republic, and
Germany, which feature very different histories of flood protection policy and dif-
ferent institutional settings, were reviewed. It can be noted that the major differ-
ences relate to the damage categories considered, the degree of detail, the scale of
analysis, the application of basic evaluation principles (e.g., replacement cost versus
depreciated cost), and the application or nonapplication of results in benefit–cost
and risk analyses.52
In England and Wales, there is a long history of developing and applying methods
of flood damage evaluation. Different methods for flood damage evaluation are
used at different spatial scales (national, regional, intermediate, and local level).
In almost all levels standard damage data developed by the Flood Hazard Research
Center (FHRC, Middlesex University) are used for flood damage evaluation.
August 25, 2009 14:14 9.75in x 6.5in b684-ch38 FA
The newest and updated version is the multi-colored manual (MCM).53 The MCM
provides methods for the quantification of many damage categories in monetary
terms, not only for direct tangible damages, but also for indirect and some intan-
gible losses, particularly regarding the damages to residential and nonresidential
properties.
Regarding residential properties, at first a classification is carried out after type,
age, and social status. After that a typical inventory is compiled for each of these
altogether 100 types. The depreciated value of the complete building structure
inclusive its inventories are then determined according to replacement costs and
market prices. On the basis of the assessment of loss adjusters, the susceptibility
of these assets to inundation depth is then assessed and finally absolute depth-
damage functions are derived. Depth-damage functions show the total damage of
the valuable property (e.g., buildings, cars, and roads) or its relatively damaged
share as a function of inundation depth. Moreover, long and short durations are dis-
tinguished. Figure 38.14 shows the damage functions for different types of houses.
This standard depth-damage data is used for damage evaluation on all spatial levels.
At all spatial levels there are also examples for the use of the Social Flood
Vulnerability Index (SFVI) as a measure for the coping capacity of the flood
affected population. The index was developed by FHRC (Flood Hazard Research
Center) including indicators for vulnerable groups and persons like elderly people,
lone parents as well as persons with preexisting health and financial deprivation
problems. FHRC also developed a FORTRAN-based software for the calculation of
damages called ESTDAM, which is not GIS based.
In the Netherlands, especially inundation depth is needed for the damage eval-
uation. In case of residential buildings, impacts of velocity and waves are also con-
sidered. Regarding casualties, three different inundation characteristics are taken
into account: velocity, rise rate, and inundation depth. The amount of tangible
damage for each grid cell is calculated by the formula:
n
S= ai ni Si (38.5)
i=1
Fig. 38.14. Absolute damage functions for different residential house types (sector mean).53
August 25, 2009 14:14 9.75in x 6.5in b684-ch38 FA
where ni is the number of units of category i, Si is the maximum damage per unit
in category i, and ai is the damage factor category i. The three essential parts of
the damage evaluation are: the gathering of land use information, the estimation of
maximum damage amounts per unit of each category, and the derivation of (depth-)
damage functions. Altogether 11 damage functions are derived from a study by
Vrouwenvelder.52,54 The damage functions are mostly depth-damage functions.
In the Czech Republic, the methods of flood damage evaluation focus also on the
estimation of direct, tangible damages. For buildings, the formula for the estimation
of damages to an individual building floor underwater is:
DAMAGE = H · C · %p · A (38.6)
Tolerability refers to willingness to live with a risk to secure certain benefits and in
the confidence that it is being properly controlled. To tolerate a risk means that we
do not regard it as negligible, or something we might ignore, but rather as something
we need to keep under review and reduce still further if and as we can.2
In the past, three criteria for risk acceptance were developed: (a) the personal
acceptance of risks, (b) the social acceptance, and (c) the economic criterion. The
personally accepted risk level is defined as the frequency of suffering a certain degree
of injury as a result of an event, accepted by an individual. The social risk concerns
the risks for the total population. Society looks at the total consequence of an
event, including the number of casualties, material and economic damage, and the
loss of immaterial [matters]. Generally, the consideration of social consequences in
the case of safety problems is limited to the number of casualties as a result of an
event. More often, the social consequence is considered the total material damage.
This definition is more suitable for an economic optimization of the risk level to be
August 25, 2009 14:14 9.75in x 6.5in b684-ch38 FA
βi · 10−4
Pfi ≤ (38.7)
Pd|fi
casualties can be found from the flooding probability and the damage factor for
casualties.
Quantitative methods to define acceptable flooding probabilities on the level
of the protected area have been obtained from literature. Three commonly used
methods are21 : cost–benefit analysis, utility analysis, and life quality analysis.
In cost–benefit analysis, a decision is deemed acceptable if the benefits exceed
the cost of protection. Costs and benefits are expressed completely in monetary
terms. The cost benefit model can be expressed as:
T
T
Pflood (cd cb b(t) + cd d(t))
b(t)
Bref (Pflood , T ) = −I(Pflood ) + t
− (38.9)
t=0
(1 + r) t=0
(1 + r)t
where Pflood denotes the flooding probability, I is the function of the cost of pro-
tection, T is the reference period chosen in the design stage of the protection
system, b(t) are the benefits at time t, d(t) is the maximum flooding damage at
time t, cd is a factor defining the part of the indirect economic damage caused by
the direct economic damage, cb is a factor describing the macroeconomic effects
of flooding, and r is the market interest rate, including inflation. Rules have to
be established that indicate whether a given value of the flooding probability
is acceptable in view of resulting value of cost–benefit model and/or constraint
on the flooding probability itself. Figure 38.15 shows an example based on the
work of van Dantzig21 of the optimization of the flooding probability (acceptable
flooding probability), combining the cost–benefit model with an acceptance rule
which considers the sum of costs and benefits over the reference period to be
nonnegative.
Fig. 38.15. Analysis of acceptable flooding probability by requiring a nonnegative value of the
sum of costs and benefits.21
August 25, 2009 14:14 9.75in x 6.5in b684-ch38 FA
Fig. 38.16. Total expected utility over the reference period as a function of flooding probability. 21
August 25, 2009 14:14 9.75in x 6.5in b684-ch38 FA
Fig. 38.17. Application of life quality index to a case study after van Dantzig.21
at birth and the gross domestic product per capita. The summed life quality over
the reference period is defined as:
T
L(Pflood , ccas , T ) = 1(t, Pflood , ccas ) (38.11a)
t=0
where ccas is the damage factor for casualties, g is the gross domestic product per
capita, e is the life expectancy at birth, and w is the relative lifetime spent working.
Figure 38.17 shows the application of life quality index to a case study after van
Dantzig.
38.4. Conclusions
Risk is a combination of the chance of a particular event, with the impact that the
event would cause if it occurred. Risk has two components: the chance or probability
of an event occurring and the impact or consequence associated with the event. For
a risk to arise there must be a hazard that consists of a “source” or initiator event
(i.e., high wave height and storm surge), a “receptor” (e.g., people and property),
and a pathway between the source and the receptor (i.e., flood routes including
defences).
Flood risk management includes the analysis and assessment of risks and the
formulation and implementation of risk reduction options. A measure of the flood
August 25, 2009 14:14 9.75in x 6.5in b684-ch38 FA
risk level which is appropriate for the decision making under consideration can be
formulated by calculating the difference Rfc − Rft of the predicted and the acceptable
risk of coastal flooding, the remaining risk.
The EVT is used to estimate return levels and joint probabilities of the sources
of flooding risk. It is important to find good statistic models that give accurate
parameter predictions and corresponding uncertainty estimates, and is therefore
of interest to find both marginal probability distributions as well as joint proba-
bility distributions for combined parameters. Runup height is a critical quantity
to estimate when assessing the probability of flooding since it defines the highest
elevation to which the waves might reach.
Risk pathways include loading and failure modes of flood defences, morphologic
changes, breaching initiation and breach growth, reliability analysis, and flood inun-
dation. All these processes help to understand the performance of the entire flood
defence system and its components, and therefore contribute significantly to obtain
an estimate of the overall probability of flooding. At the beginning, the probability
of flooding of the area is estimated, taking into account the existing morphology.
Then, the beach response to the selected storm is estimated and the inundation
of the area after considering the beach modification (e.g., beach lowering and flat-
tening and coastal retreat) is evaluated. Beach evolution, shoreline response, and
flood inundation models are utilized for this purpose.
Knowledge gaps in the science of coastal morphology lead to uncertainty in
the solutions derived by engineering methods. This is because the methods are
based on an incomplete understanding of the physical processes that govern the
morphodynamics of dunes and beaches. Filling these knowledge gaps is expected
to reduce the uncertainty in our understanding and improve our predictions of
morphology and flood risk.
In risk analysis, a good analysis of the functioning of the system and the sub-
system is of great importance. The reliability of a system is the extent to which a
system meets the requirements and is determined by the reliability of the elements
and by the relations between those elements. The reliability of an element depends
on the margin between the resistance to failure and the loads. Reliability-based
design is defined as a design approach where the probability of failure is used as
a measure of the performance of the structure. A maximum failure probability is
defined and the structure should meet the requirement.
Vulnerability refers to the resilience of a particular group of risk receptors,
e.g., people, property, and the environment and their ability to respond to haz-
ardous conditions. The damage due to flooding is influenced by two factors: the
size of the flooded area and the water level in the area. Damage factors are used
to estimate the conditional distribution of the flooding damage as a function of
flooding properties. The most important vulnerability indicator because of coastal
flooding in current flood damage analysis is the exposure indicator “inundation
depth.” Thus, the calculation of the possible damages in most European countries
is mainly based on the so-called depth-damage functions. These functions describe
either the monetary damage or the percentage of the total value of the object related
to inundation depth.
August 25, 2009 14:14 9.75in x 6.5in b684-ch38 FA
Three criteria were developed to define acceptable risk: the personal acceptance
of risks, the social acceptance, and the economic criterion. Maximum allowable
probabilities of exceedance for all possible consequences and all risk categories
are given in the literature. Three commonly used quantitative methods to define
acceptable flooding probabilities on the level of the protected area are cost–benefit
analysis, utility analysis, and life quality analysis.
A general purpose of the flood risk management framework is providing a basis
for rational decision making. Recording different variants with associated risks,
costs, and benefits, in a matrix or decision tree serves as an aid for making decisions.
Once the predicted flood risk (Rfc ) and the acceptable flood risk (Rft ) are
obtained, a measure of the flood risk level that is appropriate for the decision making
under consideration can be formulated as a function of cost and further intangible
losses. For instance, a risk scale G = Rfc − Rft /Rft = 0 shows an optimum risk level.
Negative G-values mean overdesign, while positive G-values mean underdesign. If
the flood risk management framework is used in the design of processes and objects,
the steps are often repeated several times with adjusted system specifications to
obtain an optimal design. An economic optimization calculates the costs of the
process or object and the risk with every iteration step. The design is optimal when
the costs are minimal.
A risk-based approach offers significant advantages compared to the current
practice. A flood-risk approach allows the choice of the safety levels to be further
rationalized if the consequences of flooding and the costs of protection are made
explicit. Risk-based approaches exist also in other fields where safety levels have to
be defined, so that a risk-based approach to flooding safety opens the probability
of comparison of risk levels.
The lack of adequate data to specify parameters probabilistically and the diffi-
culty in communicating uncertain results to the public and other professionals are
the main drawbacks of a risk-based approach. However, the skepticism as to the
ignorance of the techniques can be combated by routine application of risk tech-
niques. In a risk analysis framework uncertainty is understood and attempts are
made to handle it transparently.
Acknowledgment
Disclaimer
This chapter reflects the authors’ views and not those of the European Community.
Neither the European Community nor any member of the FLOODsite Consortium
is liable for any use of the information in this chapter.
August 25, 2009 14:14 9.75in x 6.5in b684-ch38 FA
References
21. H. G. Voortman, Risk-based design of large scale flood defence systems, PhD. thesis,
Delft University of Technology, the Netherlands (2003).
22. S. G. Coles and J. A. Tawn, Modelling extreme multivariate events, J. Roy. Stat. Soc.
B 53, 377–392 (1991).
23. S. G. Coles and J. A. Tawn, Statistical methods for multivariate extremes: An appli-
cation to structural design, Appl. Stat. 43, 1–48 (1994).
24. H. Joe, R. L. Smith and I. Weissman, Bivariate threshold methods for extremes,
J. Roy. Stat. Soc. B 54, 171–183 (1992).
25. S. Zachary, G. Feld, G. Ward and J. Wolfram, Multivariate extrapolation in the off-
shore environment, Appl. Ocean Res. 20, 273–295 (1998).
26. M. Schlather and J. A. Tawn, A dependence measure for multivariate and spatial
extreme values: Properties and inference, Biometrika 90(1), 139–156 (2003).
27. Z. S. Fang and N. Hogben, Analysis and prediction of long-term probability distribu-
tions of wave height and periods, Technical Report of the National Maritime Institute,
London (1982).
28. S. Haver, Wave climate off northern Norway, Appl. Ocean Res. 7(2), 85–92 (1985).
29. G. A. Athanassoulis, E. K. Skarsoulis and K. A. Belibassakis, Bivariate distributions
with given marginals with an application to wave climate description, Appl. Ocean
Res. 16, 1–17 (1994).
30. I. D. Morton and J. Bowers, Extreme value analysis in a multivariate offshore envi-
ronment, Appl. Ocean Res. 8, 303–317 (1996).
31. L. De Haan and J. De Ronde, Sea and wind: Multivariate extremes at work, Extremes
1, 7–45 (1998).
32. A. Repko, P. H. A. J. M. van Gelder, H. G. Voortman and J. K. Vrijling, Bivariate
description of offshore wave conditions with physics-based extreme value statistics,
Appl. Ocean Res. 26, 162–170 (2004).
33. J. A. Ferreira and C. G. Soares, Modelling bivariate distributions of significant wave
height and mean period, Appl. Ocean Res. 24, 31–45 (2002).
34. P. Galiatsatou and P. Prinos, Bivariate analysis and joint exceedance probabilities
of extreme wave heights and periods, Proc. 31st Int. Conf. Coastal Engineering 2008
(2008), pp. 4121–4133, doi: 10.1142/9789814277426 0342.
35. J. Brunn and J. Tawn, Comparison of approaches for estimating the probability of
coastal flooding, Appl. Stat. 47(3), 405–423 (1998).
36. A. W. Ledford and J. A. Tawn, Modelling dependence within joint tail regions, J. Roy.
Stat. Soc. B 59, 475–499 (1997).
37. DEFRA/Environment Agency Flood and Coastal Defence R&D Programme — Joint
Probability: Dependence Mapping and Best Practice: Technical Report on Dependence
Mapping, R&D Technical Report FD2308/TR1 (2005).
38. H. Hanson and M. Larson, Extreme waves and water levels in the southern Baltic
Sea: Implications for flooding at present and future conditions, J. Hyd. Res. 46, Extra
Issue 2, 292–302 (2008).
39. I. A. Hunt, Design of seawalls and breakwaters, J. Waterw. Harbor. Div. 85(WW3),
123–152 (1959).
40. D. Alvarado-Aguilar, J. A. Jimenez and A. Sanchez-Arcilla, Evaluation of risk and
vulnerability to floods in the coastal zone, Poster, Flood risk management research —
From extreme events to citizens involvement, Proc. Eur. Symp. Flood Risk Manag.
Res. (EFRM 2007), Dresden (DE), IOER (2007), p. 205.
41. H. Ozasa and A. H. Brampton, Mathematical modelling of beaches backed by seawalls,
Coast. Eng. 4, 47–63 (1980).
August 25, 2009 14:14 9.75in x 6.5in b684-ch38 FA
Chapter 39
Michael J. Briggs
Coastal and Hydraulics Laboratory
US Army Engineer Research and Development Center
3909 Halls Ferry Rd., Vicksburg, MS 39180-6199, USA
michael.j.briggs@erdc.usace.army.mil
This chapter summarizes the physical modeling of tsunami waves. The first section
contains an overview of historic studies. The next section describes equipment
and procedures used in laboratory studies for generating, measuring, and ana-
lyzing tsunami waves. The third section summarizes some early National Science
Foundation-funded laboratory studies of runup on vertical walls, plane beaches,
and circular islands that were conducted at the Coastal and Hydraulics Labo-
ratory. The next section describes the latest Network for Earthquake Engineering
Simulation-funded, the state-of-the-art tsunami wave-making facility at Oregon
State University, and several recent studies. Finally, the last section is a summary
and conclusion of laboratory modeling of tsunami waves.
39.1. Introduction
Tsunamis are long waves relative to the ocean (on the order of 20 times ocean
depth) as their wave periods are much longer than typical gravity waves. In the
deep ocean, one barely notices the tsunami as it passes underneath at the speed
of a jet plane. As the tsunami approaches shallower water, it begins to “feel” (i.e.,
react to) the bottom as the entire water column is moving. It undergoes refraction,
diffraction, reflection, and breaking. The tsunami increases in height and steepness
with complicated currents and multiple wave trains. Edge waves may even develop
depending on beach slope and bathymetry, coastline irregularity, and incident wave
1073
August 25, 2009 14:33 9.75in x 6.5in b684-ch39 FA
direction. Reflections from adjacent shorelines may affect the number of tsunami
waves and their amplitudes. The wave moves ashore as a fast-rising tide or as a
bore (if broken) flooding everything in its path and producing massive destruction
due to the movement of large objects, such as trees, cars, and trains.
Historically, people have worried about far-field tsunamis caused by the tec-
tonic events or earthquakes on the ocean floor. They are known to “export death
and destruction.”1 The sudden movement and displacement of the Earth’s plates
deforms the seafloor and produces tsunami waves on the ocean surface that travel
great distances to distant coastlines. Although volcanic eruptions and asteroid
impacts can trigger tsunamis, these are far less likely. Recent research indicates
that many tsunamis can be generated from co-seismic landslides, triggered by the
violent shaking of the Earth’s crust.2 Concern has recently focused not only on
the landslide-generated tsunami, but also on potential near-field or local tectonic
sources. Both these tsunami scenarios have the characteristic of far less travel time
(and reduced warning time for evacuation) from the generation area of the tsunami
to the coastal inundation site.
The Indian Ocean Tsunami of December 26, 2004 was a “wakeup call” about the
dangers of tsunamis.3 In the past 400 years, it is estimated that more than a million
people have died by tsunamis, with billions of dollars in damages. Prior to the Indian
Ocean Tsunami, Synolakis4 reported that 12 major tsunamis had occurred around
the Pacific Rim in the past 10 years, causing more than 3,000 deaths and $1B in
damages.
Wave runup is the most devastating hazard associated with tsunamis, yet it is the
least understood. The Intergovernmental Oceanographic Commission of UNESCO
(IOC) Tsunami Glossary defines runup as the difference between the elevation of
maximum tsunami penetration (inundation line) and the sea level at the time of the
tsunami.5 It further describes runup as the elevation reached by seawater measured
relative to a reference datum. Vertical elevation can be as high as 30 m and usually
is on the order of 10 m.6 A good understanding of tsunami wave runup is essential in
calculating forces on coastal structures and predicting inundation zones for coastal
flooding, especially as this impacts civil defense evacuation plans and routes.7
Physical models are one of the best tools for understanding the complex physics
governing tsunami wave transformation and runup. Numeric models are constantly
being improved, but require good laboratory or field data to validate the latest
enhancements. Field data are becoming more available for a variety of scenarios,
but collecting and verifying data quality is difficult.8 Laboratory experiments have
the advantage of controlled input parameters and conditions. Especially with larger-
scale models (closer to prototype), scale and boundary effects become less of a
problem. Also, laboratory models have the advantage of enabling people to see and
hear the processes as they occur.
In 1992, the National Science Foundation (NSF) funded a three-year research
program to better understand three-dimensional (3D) tsunami wave runup9 based in
part on recommendations from the International Workshop on Long Wave Runup.10
Goals of this project were to (a) advance the theoretic and numeric modeling tech-
niques for simulation of 3D free surface flows, (b) generate small- and large-scale
experimental databases for verification and modification of numeric models, and
August 25, 2009 14:33 9.75in x 6.5in b684-ch39 FA
(c) improve the predictive models used for tsunami hazard mitigation by including
improved reproduction of the important physical parameters in tsunami runup
models. This joint research study included participants from Cornell University,
Harvard University, University of Washington, University of Southern California,
and the Coastal and Hydraulics Laboratory (CHL) of the US Army Engineer
Research and Development Center (ERDC). This tsunami research was the first to
be systematically designed to include multiple institutions with multi-disciplinary
expertise.
Due in part to the large number of destructive tsunamis in the 1990s, the NSF
funded the Network for Earthquake Engineering Simulation (NEES) for earthquake
and tsunami research in the academic community in 2003. Oregon State University
(OSU) is one of the engineering schools in the USA selected for tsunami research,
especially for laboratory modeling.
This chapter summarizes physical modeling of the hydrodynamics of tsunamis.
The first section contains an overview of historic laboratory studies, with particular
emphasis on early pioneering research. The next section describes the physical
model design, equipment, and procedures for generating, measuring, and analyzing
tsunami waves. The third section summarizes the NSF-funded, large-scale labo-
ratory studies conducted at CHL to better understand the runup on vertical walls,
plane beaches, and circular islands. The fourth section describes the latest state-
of-the-art tsunami wave-making facilities at Oregon State University that are part
of NEES. Finally, the last section summarizes physical modeling and what we can
expect in future studies. It is, of course, not possible in this short chapter to describe
all the results and provide comprehensive data lists. Interested readers may contact
any of the authors for additional information and data.
The first systematic laboratory experiments relevant to the study of tsunamis were
introduced by Russell11 after he accidentally observed the generation of a solitary
wave in August 1834 while investigating ship waves. His historic remark is as
follows:
of one or two miles I lost it in the windings of the channel. Such, in the
month of August 1834, was my first chance interview with that singular
and beautiful phenomenon which I have called the Wave of Translation,
a name which it now very generally bears; when I have since found to be
an important element in almost every case of fluid resistance, and ascer-
tained to be the type of that great moving elevation of the sea, which,
with the regularity of a planet, ascends our rivers and rolls along our
shore.
This discovery of solitary waves motivated Russell to build a water tank for
generating waves of translation by the addition of a volume (7.29 P) by dropping
weights at one end of the water tank. The breadth of the tank was 30.5 cm and
the still water depth was 13 cm. The original sketch of his experiments is shown
confirmed that the propagation speed c could be predicted
in Fig. 39.1. Russell
accurately by c = g(h + a), where h is the quiescent water depth and a is the
height of the wave crest above the water level. While he measured the characteristics
of a solitary wave for its permanent and stable form, Russell also performed the
experiments by the removal of a volume by lifting weights as shown in Fig. 39.2.
His experiments generated a radiating oscillatory wave train, and the propagation
speed was slower than the equivalent wave of translation (solitary wave).
In 1972, more than 100 years later, Hammack12 performed his experiments at
Caltech generating waves by volume displacement, just like the experiments by
Russell. Hammack’s experiments were specifically designed for tsunamis, generating
Fig. 39.1. Scott Russell flume experiment for generating solitary waves.
waves by displacing vertically a small end portion (61 cm long) of the tank floor.
The tank was 38.6 cm wide and 31.7 m long. Unlike Russell’s experiments, the
tsunami generation was precisely controlled by a hydraulic servo-system, and the
displacement of water at the floor was a closer mimic of the real seismic seafloor
displacement. Hammack’s experiments were the first to use the KdV-soliton theory
in a real fluid environment.13 The theory predicts that an initial water displacement
will evolve in the farfield to generate a finite number of solitary waves — the larger
one leads the smaller ones — followed by oscillatory waves. Hammack and Segur14
and Tadepalli and Synolakis15 later demonstrated by their scaling argument that
such theoretic prediction may not be realizable in the real world. They noted that
their flume was not physically long enough for the solitons to accurately sort into
separate waves. The Earth does not have sufficient size for a real tsunami, typically
generated by co-seismic faulting, to evolve into the separation of solitons. The real
tsunami is often too linear and too nondispersive.
Tsunami generation in a 3D basin was first performed by Takahashi and
Hatori.16 Tsunamis were generated in a 25 m × 40 m × 0.6 m deep water tank by
rapid dilation of an elliptic-shaped, rubber membrane driven by compressed air.
Wave heights and propagation speeds were measured in their experiments.
Besides co-seismic fault displacement, tsunamis can also be triggered by
submarine and subaerial landslides. Perhaps, the first systematic and careful exper-
iments were those performed by Wiegel.17 In his experiments, rigid bodies of several
shapes, sizes, and weights were used to slide down the submerged sloping surface.
Wiegel found that the generated wave has what is now referred to as an N -wave
profile; the leading wave has an elevation followed by a depression wave of amplitude
greater than the leading elevation.18 Note that his focus on tsunamis propagating
offshore was evidently motivated by the 1946 Aleutian tsunami. Wiegel’s keen
foresight to examine a potential submarine landslide associated with this event is
surprising — nearly 50 years later, several scientists discussed a conjecture of a
submarine landslide scenario for the 1946 Aleutian event (e.g., Refs. 18 and 19).
Similar experiments in an even smaller laboratory apparatus were conducted 50
years later by Watts.20 Raichlen and Synolakis21 conducted other experiments in a
similar manner — sliding a rigid body on an inclined surface. They used a relatively
wide tank so that marginal 3D runup motions along the beach could be examined,
although the alongshore propagation distance was too short to examine their evo-
lution. Fritz et al.22 studied subaerial landslides by sliding granular material instead
of a rigid body. Detailed velocity fields were measured with the particle imagery
velocimetry (PIV) technique.
Runup of tsunami onto a sloping beach is often modeled by imposing an incident
solitary-wave form for convenience, taking advantage of a solitary wave being a
permanent and stable wave form. Although earlier attempts had been made to
study this problem (e.g., Refs. 23 and 24), Synolakis25 was the first to demon-
strate careful validation of the theory with his experiments. The key to generating
a “clean” solitary wave is the precise control of wave-paddle motion using “non-
linear” wavemaker theory; such an algorithm was given by Goring and Raichlen.26
This algorithm was later generalized for the creation of arbitrary long waves by
Synolakis.27
August 25, 2009 14:33 9.75in x 6.5in b684-ch39 FA
Another incident-wave model often used for studying tsunami runup is a bore.
Miller28 may be the first to systematically examine the runup of a bore that was gen-
erated by a flat vertical piston. This type of bore generation may generate unwanted
simple depression waves when the piston stops. To circumvent this contamination,
Yeh et al.29 generated a single bore by lifting a vertical gate that initially sepa-
rated the quiescent water on the beach from the deeper water behind the gate. The
laboratory observations of the maximum runup locations were always significantly
lower than the predictions made by the nonlinear shallow water theory.
Aiming at evaluating tsunami forces on onshore structures, Cross30 measured the
forces caused by the incident surge impact on a vertical wall. He generated a single
bore using the dam-break method. Arnason31 (also see Ref. 32) investigated bore-
impinging forces on vertical columns as well as the detailed flow fields surrounding
the column with the combination of laser Doppler velocimetry (LDV) and PIV.
There are many benchmark experiments for harbor resonance. Following the
pioneering work by Ippen and Raichlen,33 Ippen and Goda34 and Lee35 performed
careful laboratory experiments on harbor resonance induced by waves that can be
similar to the tsunami problem.
In general, historic experiments were performed primarily to explore character-
istics and behaviors of tsunamis. On the other hand, a majority of recent experi-
ments aims at providing adequate benchmark data sets for validation of numeric
models. For example, benchmarking exercises for numeric models were conducted
at the 1995 Friday Harbor Workshop36 and the 2004 Catalina Island Workshop.37
Objectives of laboratory experiments have evolved together with advances in
measuring instruments.
inconsistency (i.e., viscous scale effect) caused by the Reynolds number discrepancy,
but introduces complex interpretation of data due to the violation of geometric
similitude. The scale factor affects the design and cost of the experiment including
the selection of flume or basin, model size, wavemaker, and instrumentation.
To achieve a dynamically similar flow condition in a scaled model, the Froude
number F , the Reynolds number R, and the Euler number E should match those
of the prototype. When the flow is transient, the Strouhal S number should be
matched as well. These parameters are defined as:
V ρV 2pf
F =√ ; R= ; E= ; S= (39.1)
G µ ρV 2 Vt
where V is the characteristic velocity, g is gravitational acceleration, is a repre-
sentative geometric distance or length, ρ is fluid density, µ is dynamic viscosity,
pf is the characteristic fluid pressure, and t is the characteristic time. For a scaled
model in the tsunami tank, it is impractical to use a different fluid because of its
basin size. All the dynamic similitude parameters can usually be matched except
the Reynolds number. For experiments under Froude law, wave period and velocity
scale as the square root of the length scale factor Lr .
Although R cannot normally be matched in scaled experiments with that of
the prototype, it is customary to assume that, if the model R is large enough to
generate fully developed turbulent flows, the model results are relatively insensitive
to R. The prototype R is typically on the order of 107 , using flow depth as the
characteristic length . In the case of tsunami basin experiments of 50 scale, the R
would be about 3 × 104 . Hence, it is anticipated that scale effects associated with
R should not be significant.
(c) update control signals, and (d) collect and evaluate wave heights from active
wave absorption systems. Active wave absorption (although important in wind wave
studies) is not required for tsunami studies as usually only one wave is created
and the experiment is completed by the time the reflected wave returns to the
wavemaker.
2d √
L= arccosh 20 (39.3)
γ
so that it is equal to the distance between two points in the symmetric surface
profile (i.e., one near the front and one near the tail) where the height is only 5%
of the height at the crest H.
The N -wave is a special type of tsunami wave that has been shown to represent
nearshore or local tsunamis better than the solitary wave.40 At generation most
tsunamis are either leading-elevation or leading-depression N -waves. When close to
shore, they retain their leading-elevation or leading-depression shape.
Changes in runup magnitude and in the shape of the runup tongue can be inves-
tigated in the 3D basin by varying the number of paddles used in each experiment
and the eccentricity of the source. The number of paddles is referred to as the source
length because this is the physical parameter most closely represented. The source
width is modeled by the wave height, since smaller solitary waves have larger wave-
lengths for a given depth. When generating other long waves, the source width is
represented by varying the stroke. Eccentricity is a parameter used to relate the
cross-shore and longshore distances between the tsunami source and the coastal
location.
39.3.3. Instrumentation
39.3.3.1. Wave gages
Capacitance wave gages are typically used to measure surface wave elevations
since they are very reliable. They are more robust and less temperature sen-
sitive than resistance gages. The number of gages should be sufficient to ade-
quately determine incident and transformed wave conditions without having to
repeat runs. A grid layout can be used to optimize the placement of the gages.
Incident gages (i.e., measure incident wave properties) are usually located par-
allel and normal to the wavemaker at a distance that may vary for each wave
case since there is a certain amount of evolution in the solitary wave form as
it travels. Prior to each run, the incident wave gages are often moved seaward
from the toe of the island to a distance equivalent to half-a-wavelength (i.e.,
L/2) of the wave to be generated. This procedure ensures that the tsunami wave
is always measured at the same relative stage of evolution. Other gages can be
located in a cross-shore array to measure wave transformation normal to the
wavemaker.
These wave gages should be calibrated as necessary to ensure an acceptable
level of accuracy. These multiple gage calibrations are necessary to rule out any
drift in the gages due to temperature and humidity (especially in some laboratory
environments) and error from floating debris in the water. The length of the mea-
surement rods on the gages is usually determined by the water depth and wave
height. The wave rods can be manually or automatically raised three or more heights
or steps in the water column to obtain a reading. At CHL automatic Jordan con-
trollers and stepper motors are used to raise and lower the rods through 22 steps:
11 up and 11 down. A linear least-squares averaging technique is used to obtain
calibration coefficients. In the event that the calibration coefficients do not meet a
minimum threshold for a particular gage(s), the calibration is selectively repeated
for them.
At CHL, data are sampled for a minimum of 10 samples per wavelength. Typ-
ically, a sampling rate of 10 Hz is adequate, especially for longer-period tsunami
waves; because even at model scale many more than 10 points are collected for
each wave. Higher sampling rates (i.e., 20–60 Hz) may give more definition of the
recorded wave profile, but not necessarily more accurate wave parameters. The main
disadvantage is the collection and storage of more data, although this is not much
of a problem with modern computer technology.
It is a good practice to conduct a short “calibration” phase prior to the
“production” phase to make sure the waves were created correctly and are
being simulated accurately. It is also a good practice to repeat the individual
wave cases, if possible, to ensure that the recorded values are correct and
give some confidence in the repeatability of measured values from a statistic
standpoint.
Analysis for wave data typically consists of plots of the measured surface ele-
vation and calculations of wave height and period. Tsunamis are usually charac-
terized by only a few wave forms, so it is relatively easy to measure the main or
largest wave in the profile. These calculations are usually done in the time domain
by estimating the average wave height in the record.
August 25, 2009 14:33 9.75in x 6.5in b684-ch39 FA
runup gages showed great promise as a tool for reducing the manual measurement
of tsunami runup, a considerable time and cost savings.
The CHL has been conducting tsunami research since the 1960s, with more than
27 publications documenting physical, numeric, and analytic modeling of far-field
tsunami hazards in the Pacific Ocean. Two early physical modeling studies involved
the work of Oswalt and Boyd45 and Senter.46 Oswalt and Boyd45 conducted a lab-
oratory study of Hilo Harbor, HI, to evaluate steady flow stability for a tsunami
barrier. Senter46 studied the effect of tsunami waves on a proposed Crescent City,
CA, harbor design. In the 1970s and the 1980s, most of the tsunami research
involved pioneering numeric models for tsunami flood predictions in the Pacific
Ocean. The work of Houston et al.47 and Houston48 is still referenced today
as a deterministic method for predicting flooding due to tsunami inundation,1
although many numeric models provide both deterministic and risk-based
predictions.
In the early 1990s, a joint tsunami study was funded by the NSF to identify
important physical parameters involved in 3D tsunami runup. Over the course of this
study, four physical models of a vertical wall (flume), plane beach (flume and basin),
and a circular island (basin) were investigated.7 The following sections summarize
the important issues for these experiments.
August 25, 2009 14:33 9.75in x 6.5in b684-ch39 FA
Fig. 39.7. Wavemaker control signals for three Cases A, B, and C.36
Fig. 39.8. Plume of water shooting up against vertical wall at end of flume.
Fig. 39.9. Test setup in flume for study of runup on a plane beach.43
Fig. 39.10. Test setup in basin for study of runup on a plane beach.43
August 25, 2009 14:33 9.75in x 6.5in b684-ch39 FA
depth region of the model was 32 cm, which is the same as the flume experiments
described earlier. The toe of the slope was located 12.4 m in front of the wavemaker.
The CHL’s directional spectral wave generator (DSWG) was used to generate
solitary waves. The electronically controlled DSWG was 27.4 m long and consisted
of 60 paddles, 46 cm wide and 76 cm high. Briggs and Hampton52 provide additional
details of the DSWG. The rear of the basin (behind the DSWG) was lined with
wave-absorbing material backed by a concrete wall. The basin sides were lined with
absorber material and the right side (looking in the positive x-direction) was open
to the adjacent basin.
The global coordinate system was aligned with the X-axis (X) perpendicular
and the Y -axis (Y ) parallel to the DSWG. The origin was located at the mid-point
of the DSWG, in line with the front surface of all paddles at their rest positions.
A 12-m by 20-m grid was established between X = 12–24 m and Y = −10 to 10 m.
Spacing between grid points in both x and y directions was 1 m, except for 50 cm
spacing in the x-direction for grid points between X = 20 and 24 m. Also, grid
points with 1-m spacing were laid out along the centerline (Y = 0) between the
face of the DSWG and the toe of the slope. The smaller spacing was to allow more
accurate measurements of runup on the beach.
Thirty capacitance wave gages were used to measure surface wave elevations
within this grid. The first three gages were located at X = 3, 6, and 9 m along the
centerline in the constant depth region to measure incident wave conditions. The
same procedure of moving one of them prior to each run was followed to ensure
they always recorded the wave at the same relative stage of evolution. The other
two gages remained in their fixed positions. Twenty-seven gages were positioned in
three cross-shore transects in an 8-m by 6-m measurement area between the toe
of the slope and the still water level (SWL) to measure wave evolution. Spacing
between transects in the longshore direction was 3 m, with the center transect over
the centerline and the other two on either side of center. Spacing between the nine
gages in each transect was 1 m.
Eight target (nondimensional) wave heights from H = 0.01 to 0.20 were sim-
ulated. Measured wave heights were approximately 70–85% of these target values.
The decrease in measured wave height from the target was due to losses in the
mechanical generation of the solitary waves because of gaps between the floor and
the wavemaker. The reduction in measured wave height is not critical to the study
as even in the most carefully controlled experiments, the wave height is usually
less than target values. What is important is that the solitary wave profile is not
adversely affected. All waves were nonbreaking until final stages of transformation
near the shoreline (where gentle spilling occurred), except for H = 0.20 waves,
which broke nearshore.
Changes in runup magnitude and in the shape of the runup tongue were inves-
tigated for selected cases in the basin by varying the number of paddles used in
each experiment and the eccentricity of the source. The maximum vertical runup
along the sloping beach was measured at each gridline above the SWL. These hori-
zontal measurements parallel to the slope of the beach were converted to maximum
vertical runup using the bathymetric survey data or a surveyor’s rod and transit.
August 25, 2009 14:33 9.75in x 6.5in b684-ch39 FA
39.4.2.3. Results
39.4.2.3.1. Comparison of results
Normalized maximum vertical runup R was compared between the flume and the
basin results as a function of normalized wave height H . Synolakis25 had postulated
a runup law given by
R = 2.831H 1.25 cot β (39.4)
where β is the beach slope. His data represented nonbreaking waves between 0.005 ≤
H ≤ 0.04, whereas the flume and basin data included this range and breaking waves
from 0.04 ≤ H ≤ 0.4. Breaking occurred near the shoreline in both the flume and
the basin for a measured H > 0.04. The runup law slightly overpredicted wave
runup, since it is based on previous laboratory results, which did not fully cover
the range of these experiments. Agreement between the flume and basin tests was
not as good as expected, especially for the nonbreaking cases. Possible explanations
included the fact that the source lengths and cross-shore distances between the
wavemaker and the toe of the slope were different between flume and basin, and
the flume was more reflective than the basin.
with a correlation coefficient r 2 = 0.99. Later, all 10 values of H were included and
a linear relationship was still produced. A nonlinear relationship could also have
been produced, but the high r2 indicated that the simplest relationship would work
just as well. Of course, this equation should predict zero runup for zero velocity,
but the small intercept is due to the curve fitting. Friction losses amounting up to
20% of the total energy during runup were suggested as the mechanism accounting
for the differences in the measured and predicted runup.
August 25, 2009 14:33 9.75in x 6.5in b684-ch39 FA
39.4.2.3.3. 3D effects
Several 3D effects on runup were investigated in the basin including source length,
eccentricity, and amplitude evolution. The source length was varied from 1/4 to 4
DSWG modules, all centered on the DSWG centerline. Runup heights measured in
the longshore direction along the beach showed very good uniformity for different
source lengths.
The effect of eccentricity of the source on runup was studied by varying the
offset of the source from the measurement points on the beach. The source length
was varied from 1/4 to 1 module and offset to the right of the center of the beach.
Runup values were largest directly opposite the center of the source and decreased
linearly with longshore distance due to diffraction. Runup showed a strong linear
trend with source length, increasing as the source length increased.
The final results illustrated the evolution of maximum amplitude with cross-
shore distance in the basin. Test results showed that dimensionless wave height
ηmax = ηmax /d increases as source length S increases and water depth h decreases,
in agreement with earlier findings of Synolakis39 relative to Green’s law. This may
be the first instance where it has been proven that this evolution law is valid for
3D waves.43
39.4.3.1. Setup
The model island was constructed in the center of the same basin as previously
used for the plane beach experiments (Fig. 39.11). The island had the shape of a
truncated, right circular cone with diameters of 7.2 m at the toe and 2.2 m at the
crest. The vertical height of the island was approximately 62.5 cm, with a 1 vertical
on 4 horizontal beach slope. The water depth was set at 32 cm in the basin. The
surface of the island and basin were constructed with smooth concrete. The basin
was lined with wave absorbers to minimize unwanted reflections.
Twenty-seven capacitance wave gages were used to measure surface wave ele-
vations (Fig. 39.12). The first four gages were located parallel to the wavemaker
to measure incident wave conditions. The same procedure for locating the incident
gages as described for previous experiments was followed. A measurement grid of
August 25, 2009 14:33 9.75in x 6.5in b684-ch39 FA
Fig. 39.11. Circular island in DSWG basin with trapped edge wave wrapping around lee side.54
six concentric circles covered the island to a distance 2.5 m beyond the toe. Mea-
surement points were located at the intersection of these concentric circles and
the 90-degree radial lines. The spacing between grid points was a function of the
water depth.
of transformation near the shoreline (where gentle spilling occurred) except for the
Case C wave, which broke nearshore.
39.4.3.3. Results
Maximum vertical runup R was measured at 20 locations around the perimeter
of the island. Sixteen locations were evenly spaced every 22.5 degrees around the
perimeter. Four radial transects with uneven spacing were located on the backside
of the island to improve the resolution. Changes in runup shape and magnitude
were investigated by varying the water depth, wave height, source length (number
of modules), and eccentricity of the source. Figure 39.13 is a polar plot of maximum
vertical runup around the island for Case C. Waves approach the island from the
bottom or 270 degrees. The island crest, waterline, and toe are shown for reference.
Two runs are overlain, demonstrating excellent repeatability.
When the tsunami wave hit the front side of the island, the wave was split into
two components. The first component wave propagated along one side of the island,
whereas the second propagated along the other side. Refraction and diffraction
caused the waves to bend around the island as edge waves with their maximum wave
amplitude along the shoreline when the crest was perpendicular to the shoreline.
This wave trapping only occurs when the tsunami wavelength is comparable to the
Fig. 39.13. Polar plot of maximum vertical runup around the island for Case C.51
August 25, 2009 14:33 9.75in x 6.5in b684-ch39 FA
island diameter. Shorter waves are only refracted and dissipated by breaking on the
beach. Longer waves pass around the island without trapping. When these two waves
collided on the lee side of the island (i.e., the shadow zone), substantial magnification
occurred near the shore, resulting in significant flow runup onto the beach. The
runup was slightly larger on the front of the island, but the flow velocities were
larger on the lee side due to this focusing effect. The physical model and numeric
simulations verified these mechanisms. This phenomenon is very interesting since
most people would feel “safe” on the backside of an island, which is not normally
exposed to wind waves.
Table 39.2. Short-, medium-, and long-term goals for tsunami research.
Term Description
Given limited funding, an addendum to the report indicated that “it is more
important to the research needs of numerical model validation that an adequate
three-dimensional facility be developed.” In 2003, the National Research Council
released a report “Preventing Earthquake Disasters: The Grand Challenge in Earth-
quake Engineering.” Table 39.2 lists the short-, medium-, and long-term goals for
tsunami research.
39.5.3. Projects
Since the start of the NEES Tsunami Research Program, the facility has conducted
five major research projects as of publication of this handbook:
(1) Runup and rundown generated by 3D sliding masses by Profs. P. Liu, T.-R.
Wu, F. Raichlen, C. Synolakis, and J. Borrero (Cornell University, Caltech, and
University of Southern California), 2005.
(2) Tsunami-structure interactions by Prof. P. Liu (Cornell University).
(3) 3D tsunami evolution with landslides by Prof. H. Fritz (Georgia Institute of
Technology, Savannah).
(4) Performance-based tsunami engineering by Prof. R. Riggs (University of
Hawaii).
(5) Multi-scale tsunami model by Profs. P. Lynett and P. Liu (Texas A&M Uni-
versity and Cornell University).
Most of these projects are new and work is ongoing. In most cases, the mea-
surements are still being analyzed. The results will be archived and documented for
each project for future tsunami wave studies. Two example summaries of landslide
studies are presented to illustrate the level of research in the NEES projects and to
highlight the differences between landslide and distant tsunamis.
Fig. 39.16. Model landslide mass sliding (right to left) into NEES TWB (from Project (1) on
p. 1100).
The runup and rundown generated by the sliding mass was used to validate a
new numeric model that is based on the large-eddy simulation (LES) approach.
The Smagorinsky subgrid scale-model was used to provide turbulence dissipation
and the volume of fluid (VOF) method to track the free surface and shoreline
movements. A numeric algorithm for describing the motion of the sliding mass was
also implemented.
Very good agreement was shown between the numeric predictions and laboratory
measurements for the time histories of runup and generated waves from the wedge.
Details of the complex 3D flow patterns and free surface and shoreline deformations
were illustrated by the numeric models. The maximum runup heights were discussed
as a function of the initial slide elevation and specific weight. Finally, effects of the
TWB on the maximum runup was discussed.
TWB at OSU (see Fig. 39.4), the researchers simulated the impact of landslides
that occur both above and below the water’s surface. Fritz’s team constructed the
landslide tsunami generator as “an open box” that is mounted on a steel slide and
filled with up to 1,350 kg of gravel. The box accelerates down the slide by means
of four pneumatic pistons. The granular mass is accelerated inside the box and
released while the sled is slowed down pneumatically. The box is 2.1 m by 1.2 m by
0.3 m with subdivisions to adjust initial slide length and thickness, and is placed
on a slide that can vary in length. The box itself is able to travel approximately
2 m before the gravel is released down the 2H:1V slope at initial velocities up to
5 m/s. The researchers have tested two types of landslides: those that fall into the
water and those that occur beneath the surface. Sensors were placed on the simu-
lator to measure the velocity of the gravel. Using cameras placed above and within
the water, the researchers measured the shape, length, and thickness of the gravel
masses while they were in motion. Wave gages were placed to measure the size and
shape of the waves that were generated, including the lateral onshore runup.
The recorded wave profiles were extremely directional, unsteady, nonlinear, and
located mostly in the intermediate water depth wave regime. Among the principal
differences between a tectonic-generated tsunami and a landslide-generated tsunami
is that the latter has a strong directional component that can be devastating to
the immediate area. Because it has a shorter wavelength, however, it dissipates
quickly over a short distance. Landslide tsunamis exhibit a more dispersive and
strongly directional propagation than tectonic tsunamis. Planar PIV was applied to
the tsunami surface and revealed the fully 3D tsunami generation. Currently, more
than 60 successful runs have been completed and the main tsunamigenic parameters
identified that will serve as key benchmarks for numeric models.
This chapter gave a brief overview of physical modeling of tsunami waves. Back-
ground information on historic laboratory studies was presented. Equipment and
procedures used in physical models for generating, measuring, and analyzing
tsunami waves were described. Physical model studies conducted at CHL in the
early 1990s to better understand runup on vertical walls, plane beaches, and circular
islands were also described. The NEES state-of-the-art tsunami facility at OSU was
described in the last section. Since its inception, the NEES has hosted five major
research projects. In general, the data and conclusions from these studies are still
being processed and will be available to the tsunami research community in future
years. Although numeric models have come a long way, the future of physical models
for tsunami wave studies seems assured as one will always want to see the many com-
plicated processes that occur as the waves interact with coastlines and structures.
Acknowledgments
for authorizing publication of this chapter. Special thanks to Profs. Fred Raichlen
(Caltech Emeritus), Costas Synolakis (University of Southern California), Phil Liu
(Cornell University), and Hermann Fritz (Georgia Tech, Savannah), Dr. Steven
Hughes and Alex Sanchez (CHL), for assistance in the review of this chapter.
References
1. C. E. Synolakis and E. N. Bernard, Tsunami science before and beyond boxing day
2004, Proc. Roy. Soc. Lond. A 364(1845), 2231–2265 (2006).
2. C. E. Synolakis, A. C. Yalciner, J. C. Borrero and G. Plafker, Modeling of the
November 3, 1994 Skagway, Alaska tsunami, Solutions to Coastal Disasters,
eds. L. Wallendorf and L. Ewing, ASCE (2002), pp. 915–927.
3. P. L.-F. Liu, P. Lynett, H. Fernando, B. E. Jaffe, H. Fritz, B. Higman, R. Morton,
J. Goff and C. Synolakis, Observations by the international tsunami survey team in
Sri Lanka, Science 308, 1595 (2005).
4. C. E. Synolakis, Tsunami and seiche, Earthquake Engineering Handbook, eds. W.-F.
Chen and C. Scawthorn (CRC Press, 2003), pp. 9-1–9-90.
5. International Tsunami Information Center (ITIC), IOC Tsunami Glossary, 3. Surveys
and Measurements (2007).
6. J. C. Borrero, Field data and satellite imagery of tsunami effects in Banda Aceh,
Science 308(5728), 1596 (2005).
7. M. J. Briggs, J. Borrero and C. Synolakis. Tsunami disaster mitigation research in
the United States, Int. Symp. Tsunami Disas. Mitigat. Fut., Kobe, Japan, January
2005, pp. 17–18.
8. C. E. Synolakis and E. A. Okal, 1992–2002: Perspective on a decade of post-tsunami
surveys, Tsunamis: Case Studies and Recent Developments, ed. K. Satake, Adv. Natur.
Technol. Hazards 23, 1–30 (2005).
9. M. J. Briggs, Joint Tsunami Runup Study, The CERCular, U.S. Army Engineer
Waterways Experiment Station, Vicksburg, MS, CERC-92-3 (1992), pp. 7–8.
10. P. L.-F. Liu, C. E. Synolakis and H. H. Yeh, Report on the international workshop on
long-wave runup, J. Fluid Mech. 229, 675–688 (1991).
11. J. S. Russell, Report on Waves, Report of the 14th Meeting of the British Associ-
ation for the Advancement of Science, held in York in September 1844, 311–391 plus
11 plates (1844).
12. J. L. Hammack, A note on tsunamis: Their generation and propagation in an ocean
of uniform depth, J. Fluid Mech. 60, 769–799 (1973).
13. M. J. Ablowitz and H. Segur, Solitons and the inverse scattering transform, SIAM,
Philadelphia (1981).
14. J. L. Hammack and H. Segur, The Korteweg–de Vries equation and water waves.
Part 2. Comparison with experiments, J. Fluid Mech. 65, 289–314 (1974).
15. S. Tadepalli and C. E. Synolakis, Model for the leading waves of tsunamis, Phys. Rev.
Lett. 77(10), 2141 (1996).
16. R. Takahashi and T. Hatori, A model experiment on the tsunami generation from a
bottom deformation area of elliptic shape, Bull. Earthq. Res. Inst. 40, 873–883 (1962).
17. R. L. Wiegel, Laboratory studies of gravity waves generated by the movement of a
submerged body, Trans. Am. Geophys. Union 36, 759–774 (1955).
18. J. M. Johnson and K. Satake, Estimation of seismic moment and slip distribution of
the April 1, 1946, Aleutian tsunami earthquake, J. Geophys. Res. 102, 11765–11774
(1997).
August 25, 2009 14:33 9.75in x 6.5in b684-ch39 FA
42. C. E. Synolakis, The runup of long waves, PhD. thesis, California Institute of Tech-
nology, Pasadena, CA (1986).
43. M. J. Briggs, C. E. Synolakis, G. S. Harkins and S. A. Hughes, Large scale three-
dimensional laboratory measurements of tsunami inundation, tsunami: Progress,
Prediction, Disaster Prevention and Warning, Recent Developments in Tsunami
Research, Advances in Natural and Technological Hazards Research, eds. Y. Tsuchiya
and N. Shuto (Kluwer Academic Publishers, 1995), pp. 129–149.
44. M. J. Briggs, C. E. Synolakis and G. S. Harkins, Tsunami runup on a conical
island, Int. Symp. Waves — Physical and Numerical Modeling, IAHR, Vancouver, BC,
Canada (1994), pp. 446–455.
45. N. R. Oswalt and M. B. Boyd, Steady-flow stability tests of navigation opening struc-
tures, Hilo Harbor tsunami barrier, Hilo, Hawaii, U.S. Army Engineer Waterways
Experiment Station, Technical Report No. 2-742, Vicksburg, MS (1966).
46. P. K. Senter, Design of proposed Crescent City Harbor, California, tsunami model:
Hydraulic model investigation, U.S. Army Engineer Waterways Experiment Station,
Technical Report H-71-2, Vicksburg, MS (1971), pp. 1–33.
47. J. R. Houston, R. D. Carver and D. G. Markle, Tsunami-wave elevation frequency of
occurrence for the Hawaiian islands, WES Technical Report No. H-77-16, U.S. Army
Engineer Waterways Experiment Station, Vicksburg, MS, August 1977.
48. J. R. Houston, Tsunami elevation predictions for American Samoa, WES Technical
Report No. HL-80-16, U.S. Army Engineer Waterways Experiment Station, Vicksburg,
MS, September 1980.
49. U. Kanoglu and C. E. Synolakis, Long wave runup on piecewise linear topographies,
J. Fluid Mech. 374, 1–28 (1998).
50. M. J. Briggs, C. E. Synolakis and S. A. Hughes, Laboratory measurements of tsunami
run-up, IUGG/IOC Int. Tsunami Symp., Wakayama, Japan (1993), pp. 585–598.
51. M. J. Briggs, C. E. Synolakis, G. S. Harkins and D. R. Green, Laboratory experiments
of tsunami runup on a circular island, PAGEOPH 144(3/4), 569–593 (1995).
52. M. J. Briggs and M. L. Hampton, Directional Spectral Wave Generator Basin
Response to Monochromatic Waves, WES Technical Report CERC-87-6, U.S. Army
Engineer Waterways Experiment Station, Vicksburg, MS (1987), pp. 1–90.
53. H. Yeh, F. Imamura, C. Synolakis, Y. Tsuji, P. Liu and S. Shi, The Flores Island
tsunamis, EOS Trans. 74(33), 369–373 (1993).
54. H. Yeh, P. L.-F. Liu, M. J. Briggs and C. E. Synolakis, Propagation and amplification
of tsunamis at coastal boundaries, Nature 372(24), 353–355 (1994).
55. P. L.-F. Liu, Y.-S. Cho, M. J. Briggs and C. E. Synolakis, Runup of solitary waves on
a circular island, J. Fluid Mech. 302, 259–285 (1995).
56. K. Fujima, M. J. Briggs and D. Yuliadi, Runup of tsunamis with transient wave
profiles incident on a conical island, Coast. Eng. J. 42(2), 175–195 (2000).
57. P. L.-F. Liu, T.-R. Wu, F. Raichlen, C. E. Synolakis and J. Borrero, Runup and
rundown generated by three-dimensional sliding masses, J. Fluid Mech. 536, 107–144
(2005).
58. H. M. Fritz, Physical modeling of landslide generated tsunami, Caribbean Tsunami
Hazard, eds. A. Mercado-Irizarry and P. L.-F. Liu (World Scientific, Singapore, 2006),
pp. 308–324.
August 25, 2009 14:33 9.75in x 6.5in b684-ch39 FA
Chapter 40
During the past two decades, a large number of multidirectional wave facilities had
been built around the world. In parallel, wave generation and analysis techniques
have also advanced so that it is now possible to ensure realistic wave conditions
that mimic the nature in laboratory basins and flumes. With these capabilities,
testing of coastal and offshore structures can be carried out with greater accuracy
ensuring thereby optimal designs for structures in terms of cost and safety. This
chapter provides a brief review of the wave generation and analysis techniques
that are commonly used for simulating uni- and multidirectional waves and also
shares the experience gained at the Canadian Hydraulics Centre of the National
Research Council Canada, in this field.
40.1. Introduction
Physical modeling is still the best means of design optimization of coastal and
offshore structures, in spite of the fact that it was speculated in the 1980s that
the numerical modeling will take over, making physical models redundant. The
numerical models have indeed improved over the years that it is now possible to
use them to solve a large range of practical problems. However, with the trend of
locating marine structures in deeper waters and in more severe environment the
numerical models do not seem to cope with some of the environmental factors and/or
their associated complexities in a reliable fashion. Hence, it is expected that the
physical modeling techniques will continue to be relevant. Furthermore, the use of a
combination of both numerical and physical modeling techniques in large projects is
becoming more common. In such situations, numerical models are used to study the
simpler components of the problem while physical models are employed for solving
the complex components. There are also some situations where physical models are
used to validate the numerical modeling results on some test cases, before applying
the numerical tools to find solutions for the entire project.
∗ Retired staff of the Canadian Hydraulics Centre.
1107
August 22, 2009 9:56 9.75in x 6.5in b684-ch40 FA
Waves being the most important input to any physical model study, correct
simulation of waves in laboratory basins is crucial. This chapter briefly reviews
the wave generation techniques commonly used in major laboratories and shares
the experience of the Canadian Hydraulics Centre (CHC) of the National Research
Council Canada (NRC) in this field.
The flowchart (Fig. 40.1) lists the main steps involved in laboratory simulation of
waves. They include:
given variance spectral density with a phase spectrum created by a random number
generator and then obtaining a time series of desired length by inverse Fourier
transform. Since random phases are used in this procedure, time domain charac-
teristics such as wave grouping, cannot be controlled by this synthesis technique.
However, the spectral characteristics are well reproduced, as it uses the amplitude
spectrum from the target spectral density.
Another method that uses the random phases, but has no complete control
over the spectral characteristics, is called the random complex spectrum method.
Its synthesis procedure is as follows. A Gaussian distributed white noise complex
spectrum, with a standard deviation of 1, is first generated and then filtered using
the amplitude spectrum derived from the target spectral density. Subsequent inverse
Fourier transform results in a time series of desired record length. Unlike the pre-
vious method, the waves produced by this technique will not match exactly the
desired spectral density. Obviously, this method does not also exercise control on
the time domain characteristics. However, both these synthesis methods based
on random phases have their own proponents. Funke and Mansard3 describe the
rationale associated with each of these methods.
One of the requirements during the synthesis of a wave train from a spectral
density is the choice of the length of wave train to be created. Since the Gaussian
distribution is used to describe the probability distribution of prototype water
surface elevations and the Rayleigh distribution is used to describe the distri-
bution of wave heights, the longer the wave train, the better is the fit with the
above functions. However, long wave trains can increase the aggregate testing time
and thereby the cost of a physical model study. Hence, a suitable compromise is
required.
The length of the wave train is also characterized by the number of waves con-
tained in that train and choices may range between 200 and 1,000 waves in a wave
train. Often smaller lengths of wave trains (or smaller number of waves) are chosen
to conduct sensitivity studies, such as penetration of wave energy in a harbor basin.
Longer records are used where it is critical to ensure an appropriate distribution of
wave heights. For instance, in breakwater stability studies, simulation of appropriate
values of H1/10 and Hmax . are important. In general, the longer is the wave record,
the larger would be the wave heights, such as H1/10 and Hmax . Studies using rel-
atively short records (i.e., 200 waves) in shallow or breaking wave situations could
potentially result in wrong interpretation of the test results (see Ref. 4). In CHC,
wave records containing at least 1,000 waves are used in most studies.
an appropriate wave generator control signal. This conversion requires static and
dynamic calibration of the wave machine and its associated components such as
servo controller and digital-to-analog filter. The hydrodynamic transfer function
that computes the displacement required to produce a desired wave height is based
on the well-known Biesel relationship. However, the dynamic response of the wave
machine components can vary to a large extent from one facility to another and
needs to be compensated for, to ensure a faithful reproduction of the desired wave
train inside the laboratory basin. NRC uses a procedure called dynamic calibration
to estimate the compensations required for correcting the dynamic response of the
wave machine. This procedure is described below.
The static calibration of the wave machine is first established by driving the
wave machine with known and slowly varying inputs of the voltage and then mea-
suring the response in terms of paddle displacement. Then, a time series composed
of wavelets of different frequencies is fed into the system and the corresponding
displacement of the paddle is monitored and analyzed.
Figure 40.3 presents an example of the drive signal used for dynamic calibration,
while Fig. 40.4 illustrates an expanded version of a portion of this time series (i.e.,
contained between 180 and 210 s).
The cross-correlation of the input wave signal and the measured displacements
of the paddle provides a complex transfer function that can be used to generate a
wave board control signal that can ensure the desired waves.
The phase and the amplitude response of this transfer function are shown in
Fig. 40.5. The decreasing amplitude response and increasing phase difference, gen-
erally found in such systems when the frequency increases, can be seen clearly.
The real-time control and data acquisition then ensure the appropriate control
of the machine and the sampling of data by sensors such as wave gages.
target value can be attributed to loss of energy during propagation along the basin
and/or to some breaking.
In studies where the influence of group-induced long waves is important, the
long energy period energy (i.e., subharmonics) found in this figure will be analyzed
in more detail.
Note that in shallow water situations where substantive sub- and superharmonics
are expected to be present, the second-order wave generation is recommended to
be used to ensure correct reproduction of these harmonics. The techniques used for
this purpose is beyond the scope of this paper. Readers are requested to refer to
Barthel et al.10 and Sand and Mansard11 for these techniques.
An increasing number of laboratory wave flumes and basins are now equipped with
wave machines that can function both as wave generators and absorbers. Such
systems are called active wave absorbers to distinguish them from the passive wave
absorbers that are also used in basins. Active absorbers are used mainly to improve
wave quality by preventing waves reflected by the structure being tested from
being re-reflected back toward the structure by the wave machine paddle. Thus,
active absorption allows the desired incident wave field to be maintained at the test
structure, while preventing spurious wave energy from building up in the section of
the flume between the structure and the wave generator. Active absorption can also
be used to greatly reduce the stilling time between tests in a flume by removing low-
frequency waves, which would otherwise persist for some time. Similarly, it can also
prevent spurious resonant oscillations in a basin or flume during long duration tests.
August 22, 2009 9:56 9.75in x 6.5in b684-ch40 FA
There are two main types of active wave absorption systems. The first method
uses an array of wave elevation probes located a few meters in front of the wave
machine. Real-time reflection analysis of the wave probe signals is used to separate
the incoming and outgoing wave trains and the incoming wave component at the
position of the wave machine is then computed by linear dispersion theory. The cor-
responding paddle motion to absorb the incoming wave is then computed by using
the linear hydrodynamic transfer function with compensation for the amplitude and
phase response of the wave machine’s servo control system. The main advantage
of this method is that it can be easily installed on existing wave machines without
designing a new servo control system. The propagation delay between the wave
probe array and the paddle gives the wave machine time to respond, so existing
analog servo controllers with relatively large phase lags can still be used. No correc-
tions are required for evanescent waves because the wave probes are located some
distance from the paddle. One disadvantage of this method is that it relies on linear
dispersion theory so nonlinear shallow water waves will introduce phase errors that
reduce the absorption capability.
The second method uses a wave elevation probe mounted on the face of the
paddle and moving with it. This is the method most commonly used on segmented
wave machines with one wave probe on each segment. The expected wave elevation
at the paddle is subtracted from the total measured wave elevation and the residual
wave elevation is assumed to be due to the incoming wave to be absorbed. The
commanded paddle motion required to absorb the incoming wave is then computed
by linear wave theory. The digital servo control system for this method must be
very carefully designed and tuned so that there is very little phase lag over the full
frequency range because the paddle motion must respond immediately to the mea-
sured incoming wave signal. The control system must also compensate for the
amplitude and phase of the evanescent waves since the wave probe is mounted on
the paddle. This is especially important at higher frequencies. Careful tuning may
also be required to ensure a stable control system since phase lag must be kept to
a minimum.
Depending on the wave frequency and the distance between the wave machine
and the reflecting structure being tested, the paddle motion required for absorption
may be in phase with the motion required for wave generation. Consequently, the
paddle motion required for simultaneous wave generation and absorption is larger
than that required for wave generation alone although no additional power is needed.
When designing new wave machines, the paddle stroke should typically be 30%
larger than that required for wave generation alone if active absorption is to be
used. The wave machine controller must also ensure that the paddle motions do not
exceed the mechanical stroke limits and the percentage of active absorption control
may have to be reduced when generating large waves to stay within the limits of
the machine.
The two main active wave absorption methods that have been used on segmented
wave machines are known as quasi-3D and fully-3D. In the quasi-3D method, the
direction of the incident waves approaching the wave machine is not measured but
is set a priori based on the configuration of the basin and the structure being
tested. The resulting paddle motions will have the correct phase when absorbing
August 22, 2009 9:56 9.75in x 6.5in b684-ch40 FA
an oblique wave but the amplitude may be in error. However, since the amplitude
varies as cos(θ), the quasi-3D method still works very well in most cases since
the amplitude error is small when the incident waves are within 20 degrees of the
estimated direction. In fully-3D systems, the angle of the incident waves is measured
directly so that the correct amplitude can be used. This can either be done by using
a wave probe array in the basin or by using signals from wave probes mounted
on several adjacent segments to determine the incident wave angle. Schäffer and
Klopman18 provide a good review of the various active absorption techniques that
have been used for segmented wave machines.
The Canadian Hydraulics Centre of the NRC developed a 2D active wave
absorption system for wave flumes in 1999 that uses an array of three wave probes
to measure the incoming and outgoing waves. This system provides a convenient
and inexpensive way to add active absorption to an existing wave machine since
the original servo controller can be used. CHC has recently also developed a new
digital active wave absorption control system for segmented wave machines that
uses a wave elevation probe mounted on each paddle. This system uses two drive
signals per segment, which define the paddle motion for wave generation and the
expected wave elevation at the probe including the evanescent component. The con-
troller subtracts the expected wave elevation from the measured wave elevation to
obtain the elevation of the incoming wave to be absorbed. Both of these methods
have been tested experimentally using a special wave flume equipped with two wave
generators as shown in Fig. 40.7. In wave absorption tests, the waves are generated
by wave machine A and absorbed by wave machine B. Standard reflection analysis
of data from a wave probe array is then used to measure the reflection coefficient
of the active wave absorber. In other tests, wave machine B performs simultaneous
generation and absorption while wave machine A remains stationary to provide a
reflecting boundary.
Some regular wave test results for the two types of active absorbers are shown
in Fig. 40.8. The three-probe array method has good performance with an average
Fig. 40.8. Active wave absorption test results for regular waves.
Fig. 40.9. Irregular wave test results for a JONSWAP spectrum with Tp = 2.0 s.
reflection coefficient of about 10% over the main range of wave periods from 1.0
to 3.0 s. However, the reflection coefficient increases at longer wave periods due
to phase errors associated with nonlinear shallow water waves. The new digital
controller has very good performance over the full range of wave periods with an
average reflection coefficient of 5%.
Some irregular wave test results for the new digital controller with a JONSWAP
spectrum with a peak period of 2 s are shown in Fig. 40.9. It can be seen that the
system also has excellent performance in irregular waves with an average reflection
coefficient of 4%.
The results of a test of simultaneous generation and absorption are shown in
Fig. 40.10. In this case, waves were generated at a resonant frequency of the flume.
It can be seen that the wave height steadily increases when active absorption is
turned off but a stable standing wave is quickly established and maintained when
active absorption is on.
August 22, 2009 9:56 9.75in x 6.5in b684-ch40 FA
another 260 s. The two measured wave trains are almost identical for this first 100 s,
which means that the AWA has accurately identified that there are virtually no
incoming waves to absorb during that period. The periods from 100 to 160 s consists
mainly of the first reflection of waves by the generator A (see Fig. 40.7). During
this period, the AWA on wave height is approximately half as large as the AWA
off, indicating the good absorption when AWA is on. Finally, a comparison of the
results for the period 160–360 s shows that the AWA system has very effectively
reduced the residual wave energy in the flume.
The importance of testing under multidirectional seas has been described in many
publications such as Funke and Mansard,19 Mansard,20 Hoklie et al.,21 Franco
et al.,22 and Stansberg et al.,23 and many multidirectional wave facilities have been
built around the world during the past 15–20 years. In parallel, numerous techniques
have been developed to simulate multidirectional waves for laboratory model inves-
tigations. The next sections will provide a brief review of the techniques used for the
generation and analysis of multidirectional seas. They will also describe a numerical
model that is used at NRC for designing experiments in an existing multidirectional
wave basin or for designing new facilities.
The flowchart given earlier in Fig. 40.1 also provided a sketch of the different
steps involved in the generation of multidirectional seas. As it can be seen from that
figure, the synthesis part of the target wave train is different for generation of uni-
directional and multidirectional seas, while the generation component through the
wave machine is similar. Wave analysis will of course be different for the two cases.
where S(f ) is a spectral density and D(f, θ) is the directional spreading function
satisfying the relationship:
2π
D(f, θ) dθ = 1. (40.2)
0
This function can either be the same for all frequencies or the parameters θ0 and s
may vary with frequency:
Γ(s + 1) π
D(f, θ) = √ cos2s (θ − θ0 ) for |θ − θ0 | < . (40.3)
πΓ(s + 2 )
1 2
be used directly in the wave synthesis, since the values of u(t) and v(t) describe
adequately the directional characteristics of the waves.
The various synthesis techniques available at NRC for multidirectional wave
generation can be summarized as follows:
• Single summation method with equally spaced wave angles for specified wave
spectrum and spreading function.
• Single summation method with random wave angles for specified wave spectrum
and spreading function.
• Single summation method with random wave angles for specified wave train and
spreading function.
• Double summation method with equally spaced wave angles for specified wave
spectrum and spreading function.
• Discrete FFT wave synthesis for specified time series of wave elevation η(t) and
horizontal velocities u(t) and v(t).
The above methods perform not only the synthesis of water surface elevation,
but also compute the required paddle motions based on the snake principle method
of Sand and Mynett.27 These paddle motions are then compensated for dynamic and
static transfer functions as in the case of unidirectional waves. The single summation
method with random wave angles is most commonly used since it provides the
most accurate frequency spectrum but equally spaced angle methods provide a
more accurate spreading function at the expense of some spectral distortion. Test
durations corresponding to 1-h full-scale are adequate when the spreading function
does not vary with frequency. However, test durations of 5 h or more full-scale are
August 22, 2009 9:56 9.75in x 6.5in b684-ch40 FA
Among these methods, the MLM and MEM are those that are commonly used.
The instrumentation that best corresponds to these methods is either a wave probe
array or an η−u−v sensor.
The experience gained at NRC, which used these two popular methods, is
described below.
Both the MLM and the MEM are based on cross-spectral analysis of the various
sensor signals. The MLM assumes that D(f, θ) can be expressed as a linear combi-
nation of the cross-spectra, whereas the MEM estimates D(f, θ) by maximizing an
entropy function subject to cross-spectra constraints. However, the MLM, which in
fact is easier to implement, tends to provide an estimate of the spreading function
wider than the target function.
The NRC technique developed by Nwogu et al.,30 and based on MEM technique
using η−u−v data, was found by Benoit et al.29 to be superior to other methods.
This technique was later adapted to work with data from an array of wave
gages, partly because of the fact that the current meters could be more expensive
than wave gages and also susceptible to errors caused by contamination of the
current-turbulent fluctuations produced at the same frequencies as the wave-induced
kinematics.
The wave probe array consisted of five probes arranged in a trapezoidal fashion,
and the water surface elevation data from the probe array were used to resolve the
directional characteristics of the sea state.
More recently, this method has been modified to use the wave slopes derived from
the water surface elevation measured by these gages in the MEM analysis rather
than using directly the water surface elevations. This modification is triggered by
the fact that the previous method was efficient only over a limited frequency range
where the energy contained in the spectrum was substantial. Computational effort
August 22, 2009 9:56 9.75in x 6.5in b684-ch40 FA
in terms of convergence of solution was also relatively high. It was also sensitive to
some extent, to the spacing between gages.
The new method is analogous to the η−u−v method since it uses the water
surface elevation η, and the orthogonal surface slopes dη/dx and dη/dy. Figure 40.14
shows a definition sketch of the five-probe array.31
Gages A, B, C, and D are located on the circumference of a circle having a
radius R and gage E is located at the center of the circle. The water surface ele-
vation is derived from gage E, while the orthogonal slopes dη/dx and dη/dy are
derived from the wave elevation differences between gages A and C and B and D,
respectively.
Numerical simulations were undertaken to test this technique (called 5-η MEM
at NRC) on synthesized multidirectional waves with a JONSWAP spectrum and
a cos2s spreading function. The traditional well-proven technique, called η−u−v
MEM that uses η−u−v data was used for comparison purposes. Figure 40.15(a)
shows the results of this comparison. The mean direction, the standard deviation
of the spreading function, and the spreading function at the peak frequency esti-
mated by these two techniques (5-η MEM and η−u−v MEM) are compared with
the corresponding target (or imposed) function. The two analysis techniques ensure
a good agreement with the target spreading function. In fact, the 5-η MEM result
even comes closer to the target. Similarly, the values of standard deviation of the
spreading function and the mean values of the direction resolved by these two tech-
niques are nearly indistinguishable and also agree well with the target values.
Following the good performance of 5-η MEM, additional investigations were
undertaken to investigate the accuracy of the analysis if only four gages were used
instead of five. In this case, the water surface elevations from gages A, B, C, and D
will be averaged rather than using the information from gage E.
August 22, 2009 9:56 9.75in x 6.5in b684-ch40 FA
Fig. 40.15. Numerical validation of wave slope method using data from four- and five-probe
arrays.
The results of these investigations suggested that a value of R/Lp ≈ 0.02 would
be adequate to yield reliable results. More details on these investigations can be
found in Cornett et al.31
(a) (b)
Fig. 40.16. Directional wave analysis results for a set of wave basin data.31
August 22, 2009 9:56 9.75in x 6.5in b684-ch40 FA
lesser accuracy of the analysis results. Furthermore, in this case, as it can be clearly
seen from the variance spectral density measured at 10-m water depths, waves were
highly nonlinear, whereas the entire analysis technique is based on linear theory.
Given below are the conclusions that were drawn from basin studies conducted
at NRC by Cornett et al.31 :
The surface slopes MEM method (i.e., 4-η MEM) performs about as well as the
η−u−v MEM over a wide range of wave conditions and water depths. Tests conducted
with bimodal seas, nonlinear shallow water waves, and even breaking waves, indicate
that both methods perform reasonably well, even under these challenging conditions.
The surface slopes MEM offers a number of operational advantages, such as cali-
bration compared with alternative methods, and is a viable alternative for accurate
and reliable directional wave analysis in the laboratory. However, the optimal array
radius being dependent on the dominant wavelength could be perceived as a disad-
vantage of the surface slopes method.
Readers are requested to refer to Cornett et al.31 for more detailed information
on the validation tests.
Numerical models have been developed for evaluating the design options for new
multidirectional wave basins but they can also be very useful tools for designing
experiments in existing wave basins. They provide an effective way to select the
best wave generation method and basin configuration to optimize the quality of
the generated waves in a designated region of the basin for a particular type of
model test. One such numerical model named WAGEN was originally developed by
Isaacson and Qu32 for the case of oblique unidirectional regular waves generated by
one or more segmented wave machines in a basin of constant depth. It is a linear
diffraction model, which uses a large number of point wave sources to represent
the action of the wave machine segments and the other fully reflecting or partially
reflecting boundaries of the basin. Hiraishi et al.33 provides some results on the
experimental validation of this model in a multidirectional wave basin.
The WAGEN model was subsequently extended by CHC to cover the case of
multidirectional irregular waves and wave machines with active wave absorption. It
uses an iterative technique to compute the primary wave field and the secondary
wave fields produced by partial reflection of the primary waves from the passive
wave absorbers in the basin as well as the re-reflection of any incident waves from
the segmented wave generators.
In the case of regular waves, WAGEN computes the wave height and the hori-
zontal u- and v-wave velocity components over a specified x–y grid in the basin for
a target wave train defined by period, wave height, and wave propagation angle.
In the case of multidirectional irregular waves, the target wave field is typically
synthesized by approximately 2,000 wave components with individual frequencies,
amplitudes, directions, and phases. The amplitudes and phases are computed by
the random phase method for a specified wave spectrum and the wave directions
are selected at random based on the cumulative distribution defined by a specified
August 22, 2009 9:56 9.75in x 6.5in b684-ch40 FA
directional spreading function. The velocity potential Φ(x, y) is first computed for
each of these wave components and the final velocity potential for the total wave
field is obtained by linear superposition. WAGEN then computes the significant
wave height Hs , the mean wave direction θm , and the directional spreading width
σθ as functions of x and y on the basin grid. Time series of wave elevation η(t) and
velocities u(t) and v(t) can also be computed at specified points in the basin so that
directional wave spectra can be computed by the MEM.
Most multidirectional wave basins have a segmented wave machine on one side
and passive wave absorbers on the other three sides. The passive absorbers are
essential to prevent reflected waves from propagating back to the model test site
but they also cause substantial variations in wave height due to wave diffraction.
There are also wave height variations due to reflection from the passive absorbers.
WAGEN can be used to compute the wave height variation over the basin so that
the location and orientation of the structure being tested can be chosen to minimize
the effects of diffraction and reflection. In some cases, partial length guide walls
are used to reduce diffraction and WAGEN is also very useful for selecting the best
length and position for such devices. For example, diffraction effects can be reduced
by using intentional reflection off a short guide wall at one end of a segmented
wave generator (i.e., extending from X = 0 to 5 m in this case) for the case of
unidirectional regular waves propagating at an angle of 30 degrees relative to the
x-axis of the basin with a target wave height of 0.2 m (see Fig. 40.17). The wave
generator is located on the left side of the basin. It can be seen that the corner
reflection method has significantly reduced the size of the diffraction zone on the
lower right side of the basin. The increased wave height near the origin due to the
corner reflection technique is also evident.
Fig. 40.17. Contour plots of regular wave height with and without corner reflection.
August 22, 2009 9:56 9.75in x 6.5in b684-ch40 FA
Fig. 40.18. Normalized significant wave height for irregular multidirectional waves.
Fig. 40.19. Normalized significant wave height for a basin with two wave generators.
Fig. 40.20. Mean wave direction for a basin with two wave generators.
August 22, 2009 9:56 9.75in x 6.5in b684-ch40 FA
40.7. Conclusions
Wave simulation technology has made great strides in the past two decades that it
is now possible to ensure realistic ocean waves inside laboratory environments for
testing purposes. Sophisticated analysis techniques ensure that the desired charac-
teristics of the waves are faithfully reproduced. Advanced tools are also available
for properly designing experiments in laboratory basins and for ensuring improved
wave quality by preventing waves reflected by structures being re-reflected back to
the structure by the wave generator. Further research is required to develop tech-
niques that can analyze nonlinear multidirectional waves, separate the incident and
reflected components from coexisting multidirectional waves and also to perform
true 3D active wave absorption.
References
13. E. P. D. Mansard and E. R. Funke, The measurement of incident and reflected spectra
using a least squares method, Proc. 17th Int. Conf. Coast. Eng., Sydney, Australia
(1980), pp. 154–172.
14. E. P. D. Mansard, S. E. Sand and E. R. R. Funke, Reflection analysis of non-linear
regular waves, National Research Council Hydraulics Laboratory Technical Report
TR-HY-0ll (1985).
15. E. P. D. Mansard and E. R. R. Funke, On the reflection analysis of irregular
waves, National Research Council Hydraulics Laboratory Technical Report TR-HY-
017 (1987).
16. E. P. D. Mansard, On the estimation of the incident wave train from reflection
analysis, Proc. 9th Cong. Asian-Pacific Div. Int. Assoc. Hydraul. Res., Singapore,
24–26 August 1994.
17. M. Isaacson, D. Papps and E. P. D. Mansard, Oblique reflection characteristics of
rubble mound structures, Proc. 12th Canadian Hydrotech. Conf., Ottawa, Canada
1, 445–454 (1995).
18. H. Schäffer and G. Klopman, Review of multidirectional active wave absorption
methods, IAHR Sem. Multidirectional Waves, 27th IAHR Cong., San Francisco
(1997), pp. 159–182.
19. E. R. R. Funke and E. P. D. Mansard, On the testing of models in multidirectional
seas, Presented at the 23rd Int. Conf. Coast. Eng., Venice, Italy (1992).
20. E. P. D. Mansard, Directional wave generation and its application, Proc. Symp. Real
Sea ’98, Taejon, Korea (1998).
21. M. Hoklie, C. T. Stansberg and P. Werenskiold, Model tests with a single
point mooring system in short-crested seas, Paper 4644, 15th Offshore Technology
Conference, Texas, USA (1983).
22. C. Franco, J. W. van der Meer and L. Franco, Multidirectional wave loads on vertical
breakwaters, Proc. 25th Int. Conf. Coast. Eng., Orlando, Florida, USA (1996).
23. C. T. Stansberg, J. R. Krokstad and F. G. Nielsen, Model testing of the slow drift
motion of a moored semi submersible in multidirectional waves, Proc. IAHR Sem.
Multidirectional Waves Their Interaction with Structures, 27th IAHR Cong., ed.
E. Mansard, San Francisco, USA (1997).
24. M. D. Miles and E. R. Funke, A comparison of methods for synthesis of directional
seas, ASME J. Offshore Mech. Eng. 111(1), 43–48 (1989).
25. M. D. Miles, A note on directional random wave synthesis by the single summation
method, Proc. 23rd IAHR Cong., Vol. C, Ottawa, Ont., August 1989, pp. 243–250.
26. E. R. Jefferys, Directional seas should be ergodic, Appl. Ocean Res. 9(4), 186–191
(1987).
27. S. E. Sand and A. E. Mynett, Directional wave generation and analysis, Proc. IAHR
Sem. Wave Anal. Generation Lab. Basins, 22nd IAHR Cong., Lausanne, Switzerland
(1987).
28. A. Cornett and M. Miles, Simulation of hurricane seas in a multidirectional wave
basin, OMAE J. 113, 219–227 (1991).
29. M. Benoit, P. Frigaard and H. Schäffer, Analysing multidirectional wave spectra: A
tentative classification of available methods, Proc. IAHR Seminar on Multidirectional
Waves and Their Interaction with Structures, 27th IAHR Congress, ed. E. Mansard,
San Francisco, USA (1997).
30. O. Nwogu, E. P. D. Mansard, M. D. Miles and M. St. de Q. Isaacson, Estimation of
directional waves spectra by the maximum entropy method, Proc. Sem. Wave Anal.
Generation, 22nd IAHR Conf., Lausanne, Switzerland (1987), pp. 363–376.
August 22, 2009 9:56 9.75in x 6.5in b684-ch40 FA
Chapter 41
J. William Kamphuis
Department of Civil Engineering
Queen’s University, Kingston, ON, Canada K7L 3N6
kamphuis@civil.queensu.ca
41.1. Introduction
a The opinions are mainly based on North American experience, but the author’s extensive interna-
tional experience, both in universities and in practice, indicates that other jurisdictions have similar
problems and that both academia and engineering practice look up to and emulate American
“success stories.”
b The terms coastal engineering, civil engineering, and engineering are all used in this chapter,
denoting if a point refers specifically to coastal engineering or is more general and applies to civil
engineering or engineering.
1135
August 19, 2009 13:34 9.75in x 6.5in b684-ch41 FA
1136 J. W. Kamphuis
of the database. Hence this is a work in progress, which should really involve the
thoughts and opinions of us all. This chapter extends Kamphuis,1 which presents
a broad background to practice, education, and research in coastal engineering and
management, tracing their origins and sociologic settings from historic times.
The American Society of Civil Engineers (ASCE) is making very valuable contri-
butions to address the mismatch between civil engineering practice and education.
In fact, the ASCE views itself as being a leader in bringing about change in both
practice and education of civil engineers. Since it is the civil engineering organization
with the largest (global) exposure this is a natural step and one to be carefully fol-
lowed. The ASCE has published a number of valuable documents by various authors
from the civil engineering community. In addition, it is defining a required body of
knowledge for civil engineers.2,3 In a recent document titled “The Vision for Civil
Engineering in 2025”,4 they define the required knowledge, skills, and attributes.
This is a very helpful publication and will be quoted later. In that vein, Patricia
Galloway, a recent past president of ASCE has just completed a book titled: “The
21st Century Engineer — A Proposal for Engineering Education Reform.”5 This
very worthwhile and hard-hitting book makes many of the points made in this
chapter. It has, however, a different but complementary focus. Galloway writes as
CEO of a large international engineering company and emphasizes different subjects
from this chapter that is written from the point of view of an academic/practicing
engineer.
Engineering projects and their designs must fit within the existing socioeconomic
system. A very important aspect of this is the global market and a short discussion
of its impact will be given in this section.
Global market forces affect all aspects of engineering practice and education,
which in turn involve research and teaching in the engineering schools and univer-
sities. Global market force calculations simplistically break large complex systems
into connected series of commodities, which are essentially represented by simple
$cost units.c
The market operates by maximizing economic $ benefit and minimizing $ cost
through simplistic calculations such as:
• $ profit = ($ benefits) −
($ cost units)
• $ BCR = ($ benefits)/ ($ cost units).
c We use “$” as a shorthand symbol to denote money (in any denomination). $ cost calculations
normally do not consider other cost definitions such as lifecycle cost. The focus is on $ return to the
organizations and their shareholders. Therefore, we can describe economic cost simply as “$ cost.”
All of us know that considering only $ costs and $ benefits is dangerous, simplistic, and myopic.
It generates and has generated much disaster, dislocation, and grief. But this is the common
denominator of all transactions and, therefore, it is the stage on which engineering projects and
education must play.
August 19, 2009 13:34 9.75in x 6.5in b684-ch41 FA
Engineering services are also computed into project cost via $ cost units. They
are essentially commodities that are bought and sold. The market is global because
many of the actors and many projects are international; impacts are transnational,
and all purchases, also engineering services, are made globally.
Social values are also global. Impact of global warming, consequences of emis-
sions, and fallout of wars over water or unlimited access to fossil fuel have no national
borders.
Important ethical values such as safety, health, quality of life, and quality of the
environment are normally introduced via regulations, and in contemporary juris-
dictions via the public through stakeholder representation and regulation hearings.
Such ethical/social aspects are also reduced to simple $ cost units in the project
cost calculations.
Universities are also conglomerates of $ economic units. They are normally
funded through government grants, student’s tuition fees, and income from endow-
ments, fund drives, and research. Universities work very hard to maximize the
incoming $ funding on all these fronts and minimize their $ costs. They must. No
one pays universities anymore simply to think and to provide quality education. The
lofty ideals — a community of scholars with primary purposes of education of the
young and the betterment of society — have long disappeared, except in mission
statements. Unfortunately, the largest $ cost items for universities are . . . creating
a reflective, thoughtful environment, and education of the students — the raisons
d’être of universities.
University income is maximized by accepting and graduating the maximum
number of students. The $ costs to the universities of teaching students are min-
imized by increasing class sizes, student/teacher ratios, and student/space ratios.
These calculations are forcing the universities away from being places of learning
and quiet reflection. They have become factories that produce products (grad-
uates and research papers) for a global market on a delicately balanced budget.
The student/space ratio, in particular, severely restricts engineering education,
which needs rather extensive laboratory space. Thus, coastal laboratories with their
requirements for large spaces that often stand empty do not show up well on the
student/space scale. Biological, geological, and other field stations are also perceived
as a liability.
All this is purported not to affect the quality of education. Even if this were so,
the drive to maximize $ research income at minimum $ cost does affect the education
of engineering students. Research has really become just one type of machine within
the university factory. Its primary purpose is to provide a product (research papers
in this case) at minimum cost. The cost is kept low by paying graduate students
and post-doctoral employees as little as possible and by experimenting as much as
possible on computers, rather than in the field or in the physical laboratory. The
students and post-docs are considered first as cost units and only secondly as young
people who came to the university for higher education. The implications of this are
described in greater detail below.
Clearly, the above statements are generalizations. They are made to sketch
the background against which decisions are made. Fortunately, there are many
exceptions.
August 19, 2009 13:34 9.75in x 6.5in b684-ch41 FA
1138 J. W. Kamphuis
1140 J. W. Kamphuis
with rock. There were many problems with such designs and failures were common.
No one had told the inlet not to move and therefore there were unexpected and
unplanned breakthroughs and blockages, but eventually (by the 1960s) coastal engi-
neers became quite proficient at designing such inlets. They knew about stability
of rock in and through the breaker zone. They were becoming familiar with the
hydrodynamics (flow) through inlets and the morphodynamics (the movement of
sand and the change of shape) in and around such inlets.
What began as a small, natural tidal inlet is now a small port city behind a
“stable” tidal inlet. But there are many “new” problems, such as:
• The original ingenious design had not correctly taken into account the movement
of sediment along the shore (littoral drift), resulting in the collection of sand
into a beautiful beach (accretion) on one side (the updrift side of the inlet), but
because the sand can no longer move past the inlet, or is moved offshore by the
inlet structures, there is erosion of the shore downdrift of the inlet.
• Much of the habitat for fish and shellfish, which was the original reason for the
development, has been destroyed through overfishing and degradation of the water
quality from the industry and the ships.f
f For example, an interesting history about the world-renowned New York oyster industry may be
found in Kurlanski.6
August 19, 2009 13:34 9.75in x 6.5in b684-ch41 FA
professors know best because that is what they do in their research. At the same
time, students who were quite erudite communicators in elementary and secondary
school (show and tell, science projects, compulsory language training, etc.) become
less proficient at both written and verbal communication during their engineering
studies because it is apparently not required in their education and in a subsequent
engineering career. As a result, most practicing engineers, because of their lack
of training in communication and because they were educated to view their real
work as design and application of mathematical expressions, are not interested in
the communication part of the task. Yet the communication about a design must
be done by the engineers. Any other type of “communication expert” would be
severely handicapped to perform this communication task. They are not trained
in the engineering and design concepts, which are not exactly easy to understand
without adequate training.
Reviewing the tasks of the contemporary coastal engineer of our example in more
detail, she must first be able to do the functional design of the inlet stabilization.
This involves the application of knowledge about, for example:
• Hydrodynamics: Tides, tidal exchange, wind generated waves, wave climate, long
waves such as tsunamis and storm surge.
• Sediment transport and morphology: Accretion and erosion along the sea shore,
scour and deposition in the channel.
• Design and analysis of inlet structures, scour protection around bridge founda-
tions, channel revetments, etc.
• Risk analysis about flooding, storm surge, tsunamis, closing of the inlet, damage
to fisheries.
• Environment: Impact, water quality, habitat.
Because the engineer does not have detailed knowledge of all the above design
aspects, she needs to communicate with experts in all the above areas where she
has insufficient expertise: climatologists, wave modelers, geologists, biologists, con-
tractors, etc. This involves both multidisciplinary and interdisciplinaryg communi-
cation with the other scientists and engineers.
Further, as part of contemporary functional design she must have an under-
standing and awareness of:
• The social impact and limitations to the design, such as posed by land use,
relevant laws and regulations.
• Economic impacts: project costs, availability of people, construction materials
and equipment, disruption of the existing economy, etc.
But functional design is only the first part of the task of the engineer. The second
task is to persuade the public and the various stakeholder groups to buy into
the design and cooperate in bringing the design to fruition. That could involve
g Multidisciplinary is defined as interacting with related (science and engineering) disciplines, using
each other’s expertise and understanding, while interdisciplinary is defined as actually working
together with related disciplines to solve a problem.
August 19, 2009 13:34 9.75in x 6.5in b684-ch41 FA
1142 J. W. Kamphuis
among others:
h Transdisciplinary communication for engineers is defined as the interaction with disciplines who
are not trained in sciences or engineering. From another vantage point: multidisciplinary and inter-
disciplinary (for engineers and scientists) means solving well-defined problems; transdisciplinary
involves solving ill-defined problems, involving both the physical/environmental system and the
socioeconomic system. Stabilizing an inlet is itself a well-defined problem. Integration of the design
into the socio-political context of regulators, politicians, and stakeholders surrounding the project
makes it an ill-defined problem.7
August 19, 2009 13:34 9.75in x 6.5in b684-ch41 FA
Coastal Issue
Decision Makers
Coastal
Engineers
Project Coastal
Formulation Scientists
Project
Design
Implementation
Desk Study
Preliminary Post-
Design Modeling Design Implementation Implementation
Design
Desk Study
Preliminary Post-
Design Modeling Design Implementation Implementation
Design
Approval
integral and central part of the process, as shown by the arrows. Even information
exchange at the time of the preliminary design is crucial to success of the project,
as will be seen later.
A sample of the decision making processes used at the present time is shown in
Fig. 41.4. The heavy boxes show the two new players who have central roles in the
August 19, 2009 13:34 9.75in x 6.5in b684-ch41 FA
1144 J. W. Kamphuis
Coastal Issue
Decision Makers
Government
Coastal
Manager
Law
Coastal
Engineers Physics
Coastal
Scientists
Chemistry
Project
Formulation
Theoretical and Biology
Empirical
Alternatives Relationships
Geology
Others
Modeling
(uncertainties)
Judgment
Solution
Governments
Socioeconomic
Input from Nongov’t Orgs
Stakeholders
Approvals
Interest Groups
Implementation Citizens
Monitoring
process: the stakeholders and the coastal manager. The stakeholders have already
been discussed. The coastal manager is a central figure between the decision makers,
the laws and regulations, and the design process.
The coastal manager is ideally someone who communicates well with all the
players and who can translate the information to the other parties so everyone
understands each other. This position should be filled by someone with a background
in science and engineering, since much of the translation involves the communi-
cation of socioeconomic facts to the technically trained engineers and translating
technical details of the design to the socioeconomic partners, such as stakeholders,
regulators, and lawyers. Yet, in practice, most coastal managers have little formal
August 19, 2009 13:34 9.75in x 6.5in b684-ch41 FA
Stakeholders
Different backgrounds, biases,
risk perceptions and aversions, Public involvement,
poor understanding of Meetings, Press, etc
uncertainties
Pre-Engineering
Communication
Decision Makers
Resilient System
technical training. Therefore, unless the engineers themselves are proficient in trans-
lating their design into common terminology, while at the same time being able to
understand the boundary conditions imposed by the socioeconomic framework, the
project is headed for problems.
The contemporary design process has become very complex. Figure 41.5 shows
the complexity involved. It is clear that stakeholder and public involvement are
integral parts of the contemporary design process, but it is very important that
the stakeholders and public are part of the design process from the very beginning.
They cannot be simply used as sounding boards because they must take ownership
of the project and the later they appear on the scene, the more likely it is that they
oppose the process and the project.
It is essential to involve the public and the stakeholders even at the preliminary
design stage, so that they feel like partners. This necessitates “pre-engineering.” No
design has taken place at that time. But at the same time the engineers must under-
stand the project well enough to be able to present the project outlines and central
thrusts to the stakeholders. Then, with the subsequent input from the decision
makers and stakeholders, the engineers work on a conceptual design.
Recently, it has been realized that failures do occur for various reasons such
as poor infrastructure design, rampant real estate development in dangerous areas,
land subsidence, sea level rise, and global warming. Sometimes the failures can
be disastrous and they often involve the coastal zone and coastal designs. Recent
examples are the Indian Ocean Tsunami in 2004, the flooding of New Orleans as
a result of Hurricane Katrina in 2005, and the frequent flooding of Bangladesh by
monsoon runoff and impact from cyclones.
August 19, 2009 13:34 9.75in x 6.5in b684-ch41 FA
1146 J. W. Kamphuis
Even the inlet design in our example is subject to increased probability of failure
because of:
Failure means that the system must be resilient. Resilience may be broadly
defined as: how does a system recover after failure? But the exact definition of
resilience as it applies to a project depends on the socioeconomic context of the
project, which very much involves the reactions and flexibility of the local govern-
ments and populations. The next important design step after pre-engineering is,
therefore, to define resilience as it pertains to the particular project and to explore
the requirements as well as the opportunities to introduce resilience into the project.
A detailed discussion of resilience is beyond the scope of this chapter.
The conceptual combination of the PES and SES for coastal engineering projects
is shown in Fig. 41.6. The PES must be supported by the SES. In the simpler
designs SES influence is small and limited essentially to the permitting process. In
more complex designs, for example, the reconstruction of the New Orleans flood
protection, SES involvement is very large.
The public and stakeholders must be involved in the discussions on the definition,
requirements, and opportunities for resilience and they must be involved in the
resilience design process itself.
The system design sequence (the five middle boxes in Fig. 41.5) requires
great communication skills (also listening skills) on the part of the engineers.
The engineers must also take on the difficult task of simultaneously being leaders
(in the design process) and servants of the public (in meeting regulations and
designing a resilient system). It is quite clear that most engineers are not capable
of doing this and, in fact, engineers are generally not interested in the communi-
cation/leader/servant tasks. This must change! Engineers must begin to view these
expanded tasks as challenges and opportunities, instead of as threats and onerous
burdens.
The above section has shown that communication and leadership have become very
large elements of the design process. But experience indicates that engineers’ facility
with tasks not directly related to design calculations is insufficient. To improve the
training of engineers we must look at the universities and engineering schools. In
this section we examine the engineering school/university systems to see what is
feasible and how we can bring about improvement.
An engineering professor’s task within the university is a combination of
teaching, research, and administration. These three elements compete within the
university and vie for each professor’s time and interest. University administration
is not discussed in detail here, but engineers in high university administrative posi-
tions should be able to effect important changes to the engineering curriculum, if
they understand what changes are needed. But university administrations and also
the engineers within administrations have not made many changes. Changes that
would improve the quality and usefulness of the engineering graduates are quite
distant concerns as the universities struggle to stay financially viable within a com-
petitive global market.
The other two (very closely related) aspects of a professor’s task — research and
teaching — are discussed in more detail below.
1148 J. W. Kamphuis
1 1 1
2 5 2
3 24 9
4 203 39
Generation Understanding
1 Broad
3
Narrower and Deeper
Research in the sciences follows a moderni trajectory, which is based on the fol-
lowing social dynamic. At any one time there is a body of knowledge that is accepted
to be true by consensus within the scientific community. Research consists of the
search for new scientific principles by starting out from this scientific consensus.
Once new hypotheses are proven, the research results are submitted to be published.
They are essentially subjected to an initial process of truth-finding to see if a new
scientific consensus can be built, supposedly on a higher plane of knowledge. This
initial truth-finding step is normally entrusted to peer review, a process used exten-
sively by the editorial boards of the scientific journals. Peer review determines what
new articles will be published for the scientific community at large as a first step in
achieving a new general consensus on a topic. Thus, the “scientific method” builds
up a body of scientific knowledge through research and consensus, which in turn
is based on earlier scientific consensus. And the modern view is that this scientific
process will eventually lead to truth.
Engineering research at universities uses the above science research dynamic as
a model; it follows a similar modern trajectory. The search for new engineering prin-
ciples also attempts to move from one level of consensus to a next higher level of
consensus.j In fact, it is difficult to distinguish engineering research from science
research, and much of what is called engineering research is essentially science
research in that it focuses on analysis, rather than on design (synthesis).
Engineering research in the universities is similar to science research because
they are judged and funded by the same criteria. Funding for research is achieved
through research proposals, which are judged and funded on the basis of what is
purported to be another peer review process. But grants committees normally use
the number of journal publications as the prime indicator of academic prowess. Only
in that way can a committee simply crunch some (assumed to be hard) numbers to
come up with a simple evaluation of the merits of the proposal. Any other criteria,
such as outstanding engineering work performed, are normally considered to be too
subjective and too time consuming to evaluate. This process also tacitly assumes
that the quality of the publications is controlled by the journal editors and that
all the authors listed in the chapter headers contributed equally. Is this really a
thorough peer review, or does this simply transfer the value judgments from one
review process to another?
The workers on the research grants committees are indeed peers of the appli-
cants. But that is only because these positions are poorly paid or not remunerated
at all. Therefore, serving on these committees is only of interest to academic peers
i Modern, a sociological term. A modern research path is based on the belief that with proper
scientific method, given enough time and funding, we will eventually understand “the laws of
nature” and be able to solve all the world’s problems. The postmodern reaction to this recognizes
that the modern methods will not be able to reach these goals and that the modern premises
are false: for example, that there are different viewpoints, that there is no single voice of truth
(pluralism), and that uncertainty will continue to play a large role, even a key role, in our process
of discovery.1
j In view of the narrowing of research fields the “community” around each research subject becomes
quite small and hence “consensus” is between a few individuals. This leaves much opportunity for
shortcuts in the truth-finding process and for cronyism among the individuals involved.
August 19, 2009 13:34 9.75in x 6.5in b684-ch41 FA
1150 J. W. Kamphuis
who can themselves benefit from the system or have benefited from it.k And cer-
tainly it is difficult to find peers from engineering practice, who are not rewarded
by the process and who base their commitments on service to clients.
Promotion and salary committees at universities normally function the same
way. Although they have more time to use more subjective quality assessment
methods, they also use the number of journal publications as the basic indicator of
merit for academic appointment, tenure, promotion, and salary increase. Thus, there
is essentially only one review process — the review process to determine if a paper
should be published. And it was shown to have shortcomings. The other reviews
simply use the results of the publication review process. The researcher with the
most publications is almost automatically considered to be the most valued, is pro-
moted, and receives the largest salary increases based on “merit.” This method of
assessment for research grants, promotion, and salary increases puts faculty under
great pressure to publish as many papers as possible. That is as true for engineering
faculty as for the other academics. Thus, engineering professors are rewarded for pro-
ducing papers, rather than for solving engineering problems. Such a system deepens
understanding of engineering or scientific principles, the analysis side of engineering,
but it does little to advance knowledge about design (synthesis).
There is some encouraging movement in the right direction and there are excep-
tions. In Canada, for example, the national granting agency (NSERC) asks the
applicant for:
• The five most significant (lifetime) contributions to the profession — not papers,
but all work within the profession, the academic world, or social contexts of the
engineering.
• Other evidence of impact — engineering designs, reports, patents, etc.
• Contributions to the training of highly qualified people — not only graduate stu-
dents, post-docs, and undergraduate students, but also technicians and technical
experts, other academics, and engineers from practice.
• It pays to work in a research niche where one can produce papers quickly. This
encourages, for example, papers based on numerical modeling, rather than field
work and physical testing.
• It pays to split a piece of work up into as many papers as possible, published
serially and in different journals and conference proceedings, rather than pro-
ducing one valuable definitive paper at the end of a research program. Surely,
this seriously diminishes the real value of a publication.
• It pays to attach as many authors’ names as possible to each paper, so that
as many people as possible can legitimately claim authorship and add the
paper to their list of publications. Surely, this seriously diminishes the value of
authorship.
• It discourages venturing into really new, but uncharted areas because really new
ideas do not fit well into an accepted consensus pattern and because they may
lead into blind alleys and produce “useless” research.
None of the above practices are conducive to good research and, in fact, they
are an expensive waste of everyone’s time.
b. There are problems with research funding itself. For example, when large
research grants support large research projects or research centers of excellence,
they generate a feeling of ownership of the problem. The research group feels that
no other research groups should be given much funding, since that would in their
opinion dilute their effort (and it might lead to competition). This results in political
moves within in the research community to protect turf. Doidge8 describes this
internecine warfare in brain research; Venter9 describes similar protectionist politics
in connection with the mapping of the genome. These are examples from other
disciplines, but surely this also occurs within the coastal research community.
c. There are few papers describing practical applications, designs, and case
studies in engineering journals. Practicing engineers have no incentive to write
papers for engineering journals. It does not influence their funding, salaries, and
promotions, as it does for academics. Their advancement is based on the quality of
work they perform for clients and on hours of billable time. Since practicing engi-
neers also have little incentive to join editorial boards and peer review exercises,
there are few members from engineering practice on most peer review committees,
which means there is little support for practical papers on such committees.l
d. With the extreme pressure to publish results, the GEN 3 academics must
participate in the positive feedback loop of publications → research funding → more
publications → more research funding, eventually leading to tenure or promotion.
Hence, they cannot afford much time or interest to study and work on aspects
of engineering that will provide a broader and more valuable education to their
students, since this does not directly contribute to their research output. Nor can
they afford to spend time on research of practical engineering applications, which
were GEN 1’s strength.
e. Engineering education, research, and innovation has been forced into a science
type of peer review straightjacket so that the very meaning of engineering (ingenuity)
has been immobilized and innovative designs and their impacts are only discussed
in largely hidden and often proprietary design reports, not in the peer reviewed
literature, where it could benefit other engineers.
This is, of course, a description of central trends. Many encouraging changes are
being made to this restrictive system. Many individuals refuse to follow the system
slavishly. Many academics are dedicated to producing good and valuable research
and to working very hard and equitably on peer review boards. And even more
l Hence, research journals receive few practical papers to review, so they publish few practical
papers. Therefore, few practicing engineers read the research literature and practice and research
drift further apart.
August 19, 2009 13:34 9.75in x 6.5in b684-ch41 FA
1152 J. W. Kamphuis
encouraging, many academics dedicate their lives to educate the students to make
them competent engineers.
1154 J. W. Kamphuis
design report was expected, which was sometimes also presented orally. The periodic
presentations to peers and the presentations at the beginning and the end were
very good for the students, in that they functioned as students, teachers, and critics
at different times. The topics treated in the intermediate weeks were wave theory,
wave statistics, wave calculations, long waves and water levels, wave transformation,
design of coastal structures, coastal management, sediment transport, and physical
and numerical modeling. The tutorials complemented the lectures. Two tutorials
were two laboratory sessions. In one session the students used a large wave flume
to learn about waves, wave dynamics, laboratory methodology and equipment, and
analysis of model results. In the second laboratory session they built a cross-section
of their breakwater design and tested it to destruction.
Further details about this course and the course material may be found in
Kamphuis,1 Introduction to Coastal Engineering and Management. This basic
coastal textbook is essentially an edited and somewhat extended version of the
course notes. The textbook was written at the repeated request of past students
and tutors. The book has been adopted in coastal engineering courses around the
world and thus some of the above course philosophy is trickling into other coastal
courses to add to similar innovation taking place in many different ways in many
schools.
The approach in this course attempted to correct some of the misalignments
of education and practice of that time. The success in meeting the course objec-
tives varied from year to year, but the objectives always formed the framework in
which the course was presented. It exposed the students to the working environment
like they would encounter once they graduate and enter engineering practice. Some
examples: they worked in groups, they were forced to work with the same col-
leagues and with their strengths and weaknesses, they needed to negotiate with
these colleagues, they communicated their work through reports and presentations
to a critical audience (their professors and their peers) and they had to make deci-
sions based on incomplete information. All these involved listening. The students
listened to the other members of their company, to their peers and professors in
the critiques of their presentations, and to their professors and tutors in lectures,
laboratory sessions, and tutorials.
Not everyone worked well in this environment. Some students in their exit com-
ments noted that this was the best course (or even the only course) that in their
opinion came close to real engineering. Others flatly stated that it was the worst
course they had ever attended — the professor taught them nothing. If it were not for
their own hard work and leaning on their colleagues, they would have never passed.
Such a course has obvious advantages in that it teaches communication, lead-
ership, and teamwork. But it also has two negative aspects: time and evaluation.
There is no doubt that such a course requires much more time from everyone — the
professors, the tutors, and the students. In an age when efficiency counts, it is dif-
ficult to convince all the actors that the extra effort is worth it, or even to find the
time required to teach and learn in this way. Professors are forced by the system
to be very engaged in research as was seen in the previous section. Since research
publication is the motor for promotion, etc., the professors prefer to spend less time
in the classroom. The students have to work for other courses, which probably also
August 19, 2009 13:34 9.75in x 6.5in b684-ch41 FA
call for more time. And the students are less prepared to spend long hours on their
coursework than students 30 years ago.
1156 J. W. Kamphuis
Although all the points are important, some are traditionally viewed as primary
(e.g., scientific foundations and design) and some are considered secondary (e.g.,
risk/uncertainty, sustainability, and ethics). The earlier sections of this chapter show
that in a contemporary engineering environment we do this at our peril. Knowledge
about risk/uncertainty and sustainability is essential, as is knowledge of public
policy, business and the social sciences, ethical behavior, health, and safety.n The
above list must not be viewed as a shopping list. It must represent different facets
of an integral knowledge diamond.
Civil engineers also need skills beyond the basic technical education. ASCE4 also
discusses the skill set needed by civil engineers:
The civil engineer is skillful. He or she knows how to:
Unfortunately, such skills are often poorly taught and mostly considered as add-
ons to the real engineering education. Thus, we normally teach and learn real engi-
neering in a core curriculum (often consisting of the professors’ favorite subjects)
and then we teach and learn skills in separate courses, such as:
— elective courses;
— masters courses in management and business administration;
— periodic professional courses after graduation;
— professional in-house training after graduation.
n Risk, uncertainty, and sustainability concerns are actually rapidly rising.1
August 19, 2009 13:34 9.75in x 6.5in b684-ch41 FA
But it is very important that the skills be learned as part of the daily routine,
both while doing the coursework as students and while working as engineers-in-
training (and as practicing engineers). The skills must be seen as more facets on the
same knowledge diamond. Only if the skills are taught integrally with the knowledge
courses will skills become part of the engineering toolbox and not something that
was taught in isolation and is viewed as belonging in a separate compartment.
This integration of skills and knowledge is difficult to accomplish in engineering
classes where the professor and the students are driven by a love for innovation,
implementation of exciting ideas, and use of mathematics.o
The above analysis only covers strict minimum knowledge and skills require-
ments. We also need to return to the early 20th century model, where engineers
were trained in the arts and humanities for a well-rounded education. Today’s engi-
neers need a basic knowledge about, for example, history, philosophy, psychology,
language and culture, music, and the arts. This arts and humanities education
provides a greater balance. Students will have been exposed to important con-
cepts such as truth, belief, and reality, and they will better understand the social
context of their projects. They will also relate better to the stakeholders and their
concerns.
With respect to skills education: skills are often taught subjectively. For example,
communication courses are often feel-good courses in which students must be
encouraged. But skills courses should clearly also meet standards. For example,
communication in university or engineering practice is more than show and tell.
Typical student presentations are: “this is what we did; we used all the correct
formulae and here are the fantastic results” or “we did not achieve our goals
but we did the best we could.” Engineering students (as well as practicing engi-
neers) cannot simply assume that their audience is in agreement because their
reasoning includes the best technical knowledge available. They need to know the
alternatives as well as their solution. They need to be able to explain that there
are no ultimate solutions, only optimum solutions. They must be able to identify
clearly what are the relevant parameters and boundary conditions within which
they made the decisions. They must learn to argue their case for various audi-
ences and make their presentations fitting and convincing. Such communication
should use, or at least refer to contemporary communication tools such as the
Internet, numerical models, and computer games. It should introduce approaches
to negotiation and development of win–win strategies, where necessary. Evalu-
ation of presentations and reports needs to account for all these details, rather
than simply giving an averaged feel-good, encouraging assessment. And communi-
cation is just one example of a whole suite of necessary skills to be so taught and
evaluated.
The above discussion indicates two basic scarcities: time and people.
a. Time. A breakwater design calculation can be made in hours. Preparation
of a presentation of relevant design parameters can take days. Preparation of case
o The author noted consistently over many years that everyone was interested and paid attention
when formulae were presented, but eyes quickly glazed over when design philosophy, coastal man-
agement issues, environmental impacts, etc. were discussed.
August 19, 2009 13:34 9.75in x 6.5in b684-ch41 FA
1158 J. W. Kamphuis
studies needs weeks. Where will this additional time come from? To make time:
b. People. Such reformed education needs special people. For example, it needs
teachers with experience in two-way communication of information as it pertains to
engineering projects. Where do these people come from? They must come, among
others, from the engineers who have experience with this.
41.5.2. Cooperation
The above sections show that much greater cooperation is needed between uni-
versity and engineering practice. Who pays for this? Section 41.4 showed that the
existing (mostly market-driven) emphasis by universities on production of papers
discourages production of well-rounded engineers. Yet the universities have few
alternatives. On the other side, engineering employers (engineering firms, govern-
ments at all levels, industry, innovation networks, etc.) compete in a global market,
where engineering contracts are normally based primarily on cost and only sec-
ondarily on quality. To keep engineering costs down, the engineering staff must
work efficiently and many decisions revolve around “billable time.” Thus, engi-
neering practice and education are both tied down by market requirements. Unless
a win–win scenario can be produced, there will be no progress on ensuring better
engineering education.
Whether or not we agree with this domination by market parameters, it is a
fact of life. Yet, in spite of the huge influence of market forces on our decisions, we
must work toward a social dynamic within this market system that values attributes
beyond simple $ costs. We must assert, for example, that producing good engi-
neers takes priority over eking the last $ out of the system through maximizing bil-
lable time, or maximizing university income through building more efficient research
machines.
This means it is time for engineers to take back the management of engineering
from masters of business administration, accountants, and lawyers. On the academic
side, engineers need to take back engineering education from the grip of university
August 19, 2009 13:34 9.75in x 6.5in b684-ch41 FA
41.5.2.2. Evaluation
It is essential that the present systems of evaluation be modified and broadened.
Evaluations to be reviewed include evaluation of:
p This is parallel to the struggle to convince business that green (concern for the environment) is
better for the bottom line than only black (singular concern for $ profits). That battle has been
quite successful. Recently, business converts to a newly-minted green attitude are marketing their
new green image with vigor because it is now perceived that it is worth $$.
August 19, 2009 13:34 9.75in x 6.5in b684-ch41 FA
1160 J. W. Kamphuis
This task is very difficult, but time has come for this type of renewal in both
engineering practice and education.
Practicing engineers must also regularly spend time to attend and give lec-
tures, seminars, and workshops at the universities, and to be involved in curriculum
updating. Such educational leaves will:
• Expose the students to real life problems, points of view, and solutions.
• Expose engineers to today’s students and their environment.
• Establish direct contacts between the students and industry. This would facilitate
the job marketplace — job postings and hiring, arrangement of student work
terms, etc.
• Strengthen ties between individual academics and engineers.
q And of course we need to encourage sabbaticals and leaves in educational systems and engineering
1162 J. W. Kamphuis
Not much has changed in the basic concept. But the mandate has broadened,
mainly through concern and involvement of a much stronger socioeconomic presence
in the design process. And there are large implications in this updated mandate:
41.6. Conclusion
We live and work in challenging times. The contemporary tasks facing engineers are
very much more interesting than ingeniously putting together building materials in
designs of structures. We must design PES that fit integrally and function effectively
and sustainably within a very complex SES. Many designs are also concerned with
infrastructure renewal. This is much more difficult than starting with a clean slate
to design new PES. What challenges and how interesting! But engineers must be
motivated to see these very complex tasks as challenges and opportunities, rather
than as extraneous drudgery. And engineers must be educated appropriately to be
able to execute these very complex tasks. That involves major reorganizations of
civil engineering practice and education, within very stringent global market con-
straints and within a short time frame. This can be done only through close coop-
eration and mutual respect between civil engineering practice and civil engineering
education. The future? The future is ours to make.
References
1. J. W. Kamphuis, Coastal engineering — Quo vadis?, Coastal Eng. 53, 133–140 (2006).
2. ASCE, Civil Engineering Body of Knowledge for the 21st Century (American Society
of Civil Engineers Publications, Reston, VA, USA, 2004).
3. ASCE, Preparing the civil engineer for tomorrow by raising the bar, Civil Engineering
(American Society of Civil Engineers Publications, Reston, VA, USA, 2007), pp. 64–71.
4. ASCE, The Vision for Civil Engineering in 2025 (American Society of Civil Engineers
Publications, Reston, VA, USA, 2007).
5. P. D. Galloway, The 21st Century Engineer — A Proposal for Engineering Education
Reform (American Society of Civil Engineers Publications, Reston, VA, USA, 2007).
6. M. Kurlanski, The Big Oyster (Ballantine Books, USA, 2006).
7. F. Mercks, Dutch coastal defense research, summary and conclusion, Internal Report,
Rathenau Instituut, The Netherlands (2007).
8. N. Doidge, The Brain that Changes Itself (Viking Press, Penguin, USA, 2007).
9. J. C. Venter, A Life Decoded (Viking Press, Penguin, USA, 2007).