Sunteți pe pagina 1din 141

Transport Formulas for

Graded Sediment


Behaviour of Transport Formulas
and Verification with Data







P. van der Scheer
J.S. Ribberink
A. Blom





Civil Engineering
University of Twente
The Netherlands


ISSN: 1568-4652
Research report 2002R-002





Enschede, April 2002

Abstract
University of Twente III
Abstract
This study is aimed at the behaviour analysis and verification of sediment transport
formulas for graded sediment in equilibrium conditions. Graded sediment is a
mixture of grains with different sizes. This study is a joint project of Rijkswaterstaat
RIZA and the University of Twente. The predictive ability of transport formulas for
graded sediment is investigated since these predictions still have many uncertainties.
Some processes important for transport of graded sediment are given below:
Incipient motion: Larger grains require a higher shear stress at the point of
incipient motion. This means that finer grains are transported more easily than
coarser grains.
Hiding/exposure: Coarse grains protrude more from the bed and, as such, they
are more exposed to the flow than in a situation in which only these large
grains would be present. Small grains can hide behind coarser grains making
them less mobile compared to a situation in which only these small grains
would be present.
In this study, ten transport formulas are analysed for their validity ranges, their
behaviour and their transport predictions. The predicted transport rates and
compositions are verified with measurements from experimental series. Uniform and
fractional versions of these transport formulas result in 17 different formulas:

formula transport type uniform or fractional hiding/exposure
correction
Ackers & White (1973) total uniform
Ackers & White (1973) total fractional Day
Ackers & White (1973) total fractional Proffitt & Sutherland
Parker (1990) bed fractional Parker
Engelund & Hansen total uniform
Engelund & Hansen total fractional
Meyer-Peter & Mller bed uniform
Meyer-Peter & Mller bed fractional Egiazaroff
Meyer-Peter & Mller bed fractional Ashida & Michiue
Van Rijn bed uniform
Van Rijn bed fractional
Hunziker/Meyer-Peter & Mller bed fractional Hunziker
Gladkow & Shngen bed fractional
Wu et al. bed fractional Wu et al.
Wilcock & Crowe bed fractional Wilcock & Crowe
Ribberink bed uniform
Ribberink bed fractional Ashida & Michiue
The validity ranges of the fractional formulas were determined by analysing the data
used for the original development (calibration and verification) of the transport
formula. Three different parameters were chosen to represent the validity of the
formula: the median grain size of the bed material (D
50
), the geometric standard
deviation (!
g
) of the bed material and the Shields parameter ("). The analysis showed
that the fractional formulas of Wu et al., Ackers & White with the hiding/exposure
correction of Day and Ackers & White with the hiding/exposure correction of Proffitt
Abstract
IV University of Twente
& Sutherland have the largest validity ranges. This is caused by the large data sets
that were used for deriving these formulas.
Four series of calculations have been done with hypothetical hydraulic and
sedimentological conditions to investigate the behaviour of the sediment transport
formulas under different circumstances. Both the median grain size (D
50
) and the
geometric standard deviation (!
g
) of the bed material were varied. The grain size was
set at either 2 or 10 mm and the geometric standard deviation was set at either 2 or
5. Combining these sediment properties gives four different sediment mixtures. The
Shields parameter in each series was varied between 0 and 0.6.
The transport rates predicted by the several formulas differ much in conditions
with Shields parameters below 0.3. This is mainly caused by the different ways of
modelling incipient motion since it is difficult to predict sediment transport at these
conditions. The behaviour of the transport rate versus the Shields parameter shows
no changes when the median grain size (D
50
) of the bed material is varied. Varying
the geometric standard deviation (!
g
) of the bed material however, does result in
changes of the behaviour of some transport formulas. The predicted transport rates
of the uniform formulas do not differ much from their fractional counterparts when
the conditions are far from incipient motion.
At low Shields parameters, all formulas predict a transport composition finer
than the bed composition. At higher Shields parameters, the transport composition
predicted by most formulas approximates the composition of the bed material. The
largest differences in predictions of the transport composition occur at low Shields
parameters due to different ways of modelling incipient motion. The behaviour of the
transport composition versus the Shields parameter of the formulas is not very
sensitive to variations in D
50
of the bed material, but does show differences when !
g

is varied.
The transport formulas were verified by comparing transport predictions with
measurements originating from ten sets of laboratory experiments. These
experiments were carried out in flumes with sediment recirculation systems. In this
study, only data from equilibrium phases of the tests have been used. The initial bed
material has been used as input for the transport predictions.
The fractional formula of Wu et al. [2000] gives the best transport rate
predictions, but the formula should be used carefully in conditions with low Shields
parameters and relatively coarse material. The fractional formula of Wilcock & Crowe
[2001] is a good second option. For all formulas, the transport rates are difficult to
predict in Shields parameter ranges below 0.1. The fractional formulas give better
results than their uniform counterparts. Many formulas give good transport rate
predictions for experiments with a median grain size (D
50
) smaller than their validity
range.
The formula of Wu et al. gives the best transport composition predictions. All
formulas have problems predicting a good transport composition at low Shields
parameters. At higher Shields parameters and a median grain size (D
50
) even smaller
than their validity range, the composition predictions are good.
Abstract
University of Twente V
Using the surface layer composition instead of the composition of the initial material
as input for the transport formulas was only possible for six of the ten experimental
series. For most formulas, using the surface layer composition leads to better
predictions. The formula of Wilcock & Crowe gives now the best predictions of the
transport rate followed by the formula of Wu et al. The formula of Wu et al. remains
the best formula to predict the transport composition, but has a much better score
than when using the initial bed composition.
Overall, the formula of Wu et al. [2000] is the best formula to predict transport of
non-uniform sediment. The formula has large validity ranges due to the large
amount of data that was used for the original calibration of the formula. The formula
shows stable behaviour concerning the transport rate and the transport composition.
Even outside the validity ranges of the formula, it gives good predictions. Based on
the present study, it can be concluded that the formula of Wu et al. is the best
formula for sediment transport predictions.
The formula of Meyer-Peter & Mller with the hiding/exposure correction of Ashida
& Michiue [1973] shows good results for the sediment transport predictions if the
Shields parameter is high enough (>0.15). The validity range of this formula is much
smaller than the formula of Wu et al., but the formula of Meyer-Peter & Mller with
Ashida & Michiue shows good predictions outside its validity range as well. The
formula is not suitable for conditions with Shields parameters below 0.1. The
formula predicts no sediment transport in this Shields parameter range while
experiments show that sediment transport occurs at these conditions.

Preface
University of Twente VII
Preface
This report is the result of a continuation of the project Transport formulas for
graded sediment - Behaviour of transport formulas and verification with
experimental data and is a joint research project between Rijkswaterstaat RIZA and
the University of Twente. Rijkswaterstaat RIZA will bring forward this report as a
discussion paper in the study group Morphological Models of the CHR commission
(Commission for the Hydrology of the Rhine basin).
The aim of this project is to gain more insight in predicting the transport of
graded sediment. This was done by a literature study on different sediment transport
formulas. The data on which the different formulas are originally calibrated have
been analysed to determine the validity ranges of the formulas. The behaviour of the
transport formulas has been analysed by making predictions for hypothetical
conditions. Finally, the predictions of the formulas have been verified with the
measured transport rates and compositions from ten sets of flume experiments.




Peter van der Scheer
15 April 2002
Enschede
The Netherlands
Table of contents
University of Twente IX
Table of contents
1 DESCRIPTION OF RESEARCH PROJECT 13
1.1 INTRODUCTION 13
1.2 RESEARCH OBJECTIVES 13
1.3 ACTIVITIES 14
1.3.1 Overview transport formulas 14
1.3.2 Validity ranges of transport formulas 14
1.3.3 Behaviour of transport formulas 15
1.3.4 Verification sediment transport formulas 16
1.3.5 Surface layer as active layer 16
2 GRADED SEDIMENT 19
2.1 INTRODUCTION 19
2.2 SEDIMENT PROPERTIES 19
2.3 INCIPIENT MOTION 20
2.4 HIDING/EXPOSURE 21
3 SEDIMENT TRANSPORT FORMULAS 23
3.1 INTRODUCTION 23
3.2 ACKERS & WHITE 23
3.2.1 General 23
3.2.2 Uniform transport 24
3.2.3 Fractional transport 25
3.2.3.1 Day's hiding/exposure factor 25
3.2.3.2 Proffitt & Sutherland's hiding/exposure factor 26
3.3 PARKER 27
3.3.1 General 27
3.3.2 Surface-based bed load formula 28
3.4 ENGELUND & HANSEN 30
3.4.1 General 30
3.4.2 Uniform transport 30
3.4.3 Fractional transport 30
3.5 MEYER-PETER & MLLER 31
3.5.1 General 31
3.5.2 Uniform transport 31
3.5.3 Fractional transport 32
3.5.3.1 Egiazaroffs hiding/exposure factor 32
3.5.3.2 Ashida & Michiues hiding/exposure factor 32
3.6 VAN RIJN 33
3.6.1 General 33
3.6.2 Uniform bed load formula 33
3.6.3 Fractional bed load formula 35
3.6.4 Suspended load formula 35
3.7 HUNZIKER/MEYER-PETER & MLLER 36
3.7.1 General 36
3.7.2 Fractional bed load formula 37
3.8 GLADKOW & SHNGEN 38
3.8.1 General 38
3.8.2 Fractional bed load formula 39
3.9 WU ET AL. 40
3.9.1 General 40
3.9.2 Fractional bed load formula 40
3.9.3 Fractional suspended load formula 41
3.10 WILCOCK & CROWE 41
3.10.1 Fractional formula 42
Table of contents
X University of Twente
3.11 RIBBERINK 43
3.11.1 Uniform formula 43
3.11.2 Fractional formula 44
3.12 VALIDITY RANGES OF THE FRACTIONAL FORMULAS 44
4 BEHAVIOUR TRANSPORT FORMULAS 55
4.1 INTRODUCTION 55
4.2 DIFFERENCES BETWEEN FORMULAS 56
4.2.1 Transport rate 56
4.2.2 Transport composition 59
4.3 UNIFORM VERSUS FRACTIONAL 62
4.4 FOCUS ON LOW SHIELDS PARAMETERS 64
4.5 CONCLUSIONS 67
4.5.1 Transport rate 67
4.5.2 Transport composition 68
5 EXPERIMENTAL DATA USED FOR VERIFICATION 71
5.1 INTRODUCTION 71
5.2 BLOM & KLEINHANS 72
5.3 BLOM 72
5.4 KLAASSEN 73
5.5 DAY: HRS A 74
5.6 DAY: HRS B 75
5.7 WILCOCK & MCARDELL AND WILCOCK ET AL. 76
6 VERIFICATION OF TRANSPORT RATE 77
6.1 INTRODUCTION 77
6.2 MEASURED TRANSPORT RATE 77
6.3 RANKING OF THE TRANSPORT FORMULAS 78
6.4 ACKERS & WHITE 79
6.4.1 Ackers & White 79
6.4.2 Ackers & White with Day 80
6.4.3 Ackers & White with Proffitt & Sutherland 81
6.5 PARKER 81
6.6 ENGELUND & HANSEN 82
6.6.1 Engelund & Hansen uniform 82
6.6.2 Engelund & Hansen fractional 83
6.7 MEYER-PETER & MLLER 84
6.7.1 Meyer-Peter & Mller uniform 84
6.7.2 Meyer-Peter & Mller with Egiazaroff 84
6.7.3 Meyer-Peter & Mller with Ashida & Michiue 85
6.8 VAN RIJN 86
6.8.1 Van Rijn uniform 86
6.8.2 Van Rijn fractional 87
6.9 HUNZIKER/MEYER-PETER & MLLER 88
6.10 GLADKOW & SHNGEN 89
6.11 WU ET AL. 90
6.12 WILCOCK & CROWE 90
6.13 RIBBERINK 91
6.13.1 Ribberink uniform 91
6.13.2 Ribberink with Ashida & Michiue 92
6.14 CONCLUSIONS 92
7 VERIFICATION OF TRANSPORT COMPOSITION 95
7.1 INTRODUCTION 95
7.2 MEASURED TRANSPORT COMPOSITION 95
7.3 RANKING OF THE TRANSPORT FORMULAS 96
7.4 ACKERS & WHITE 97
7.4.1 Ackers & White with Day 97
7.4.2 Ackers & White with Proffitt & Sutherland 97
Table of contents
University of Twente XI
7.5 PARKER 98
7.6 ENGELUND & HANSEN 99
7.7 MEYER-PETER & MLLER 100
7.7.1 Meyer-Peter & Mller with Egiazaroff 100
7.7.2 Meyer-Peter & Mller with Ashida & Michiue 101
7.8 VAN RIJN 102
7.9 HUNZIKER/MEYER-PETER & MLLER 102
7.10 GLADKOW & SHNGEN 103
7.11 WU ET AL. 104
7.12 WILCOCK & CROWE 105
7.13 RIBBERINK WITH ASHIDA & MICHIUE 105
7.14 CONCLUSIONS 106
8 SURFACE LAYER AS ACTIVE LAYER 107
8.1 INTRODUCTION 107
8.2 MEASUREMENTS 107
8.3 TRANSPORT RATE 108
8.4 TRANSPORT COMPOSITION 110
8.5 CONCLUSIONS 111
9 DISCUSSION 113
9.1 MODIFICATIONS TO VAN RIJN 113
9.2 BED FORM FACTORS 113
10 CONCLUSIONS 115
10.1 VALIDITY RANGES OF FRACTIONAL FORMULAS 115
10.2 BEHAVIOUR ANALYSIS 115
10.2.1 Transport rate 115
10.2.2 Transport composition 116
10.3 VERIFICATION WITH EXPERIMENTAL DATA 116
10.3.1 Initial bed mixture as active layer 117
10.3.2 Surface layer as active layer 118
10.4 FINAL CONCLUSIONS 119
10.5 RECOMMENDATIONS 119
REFERENCES 121
APPENDIX 1: BED ROUGHNESS ANALYSIS A-1
APPENDIX 2: SPLITTING A SEDIMENT MIXTURE INTO N FRACTIONS A-5
APPENDIX 3: TRANSPORT RATE SERIES 1 AND 2 A-7
APPENDIX 4: TRANSPORT COMPOSITION SERIES 1 AND 2 A-9
APPENDIX 5: UNIFORM VERSUS FRACTIONAL FOR SERIES 1 AND 2 A-11
APPENDIX 6: VANONI AND BROOKS A-13
APPENDIX 7: DATA WILCOCK A-15

Description of research project
University of Twente 13
1 Description of research project
1.1 Introduction
Many formulas have been developed the last five decades to predict the sediment
transport. Formulas for both uniform sediment and for graded sediment exist. All
formulas were derived in a different way and calibrated with different data. This can
lead to large differences in the predictions of the sediment transport between
different formulas. In this project, we will give a clearer picture about which formula
gives good sediment transport predictions, and which does not. The present project
is a continuation of a joint research project between Rijkswaterstaat RIZA and the
University of Twente (Van der Scheer et al., 2001). Rijkswaterstaat RIZA will bring
forward this report as a discussion paper in the study group "Morphological Models"
of the CHR commission (Commission for the Hydrology of the Rhine basin).
The present project deals with the behaviour and the validity of several
sediment transport formulas for graded sediment. Research conducted in the
framework of the M.Sc. thesis of P. van der Scheer [Van der Scheer, 2000] is an
important basis of this project. During this thesis, software written in the software
language MATLAB 5.3 was developed for making predictions by several sediment
transport formulas for both uniform and graded sediment. These sediment transport
formulas were verified with data of flume experiments, which were performed in
1999/2000 for the Ph.D. project of A. Blom [Blom, 2000].
Within this project, the behaviour of the transport formulas is investigated
further. All transport formulas are calibrated with experimental and/or field data.
The validity range of the formulas is limited to the specific hydraulic and
sedimentological ranges of the calibration data. The influence of important
parameters in sediment transport processes like the grain size distribution and the
Shields parameter will be investigated.
1.2 Research objectives
There are three research objectives in this project. These are:
1. For several fractional sediment transport formulas we determine the validity
ranges. These validity ranges express for which sediment mixtures and which
hydraulic conditions the formula is calibrated and should give good transport
predictions.
2. We investigate the behaviour of the transport formulas by predicting transport
rates and compositions for hypothetical cases. The predicted transport rates
and compositions are compared with each other, which indicates how large is
the difference between predictions of sediment transport formulas under
different conditions. An important aspect is to gain knowledge on the
circumstances when selective transport processes start playing a role and
fractional sediment transport formulas should be used.
Description of research project
14 University of Twente
3. We study the performance of the transport formulas. How well a transport
formula predicts the transport rate and composition is investigated by verifying
predictions of these formulas with measurements in ten sets of experiments.
1.3 Activities
1.3.1 Overview transport formulas
Table 1.1 shows a list of all formulas considered within this project. For this study,
the sediment transport formulas of Gladkow & Shngen, Wu et al., Wilcock & Crowe
and Ribberink are included in the list.
Table 1.1: List of sediment transport formulas considered in present study.
formula transport type uniform or fractional hiding/exposure
correction
Ackers & White (1973) total uniform
Ackers & White (1973) total fractional Day
Ackers & White (1973) total fractional Proffitt & Sutherland
Parker (1990) bed fractional Parker
Engelund & Hansen total uniform
Engelund & Hansen total fractional
Meyer-Peter & Mller bed uniform
Meyer-Peter & Mller bed fractional Egiazaroff
Meyer-Peter & Mller bed fractional Ashida & Michiue
Van Rijn bed uniform
Van Rijn bed fractional
Hunziker/Meyer-Peter & Mller bed fractional Hunziker
Gladkow & Shngen bed fractional Shngen
Wu et al. bed fractional Wu et al.
Wilcock & Crowe bed fractional Wilcock & Crowe
Ribberink bed uniform
Ribberink bed fractional Ashida & Michiue
1.3.2 Validity ranges of transport formulas
Van der Scheer [2000] found that the hydraulic and sedimentological ranges for
which the formulas were originally calibrated are often not clearly given in the
scientific papers describing these transport formulas. In this study effort will be put
into determining the validity ranges of some fractional transport formulas by means
of an extensive literature study. The Shields parameter range will be used to show
the hydraulic range of the calibration data. The bed material will be analysed for its
median grain size and gradation.
Another important aspect in the literature research will be the definition of the
active layer used to derive a representative grain size for transport predictions. The
active layer is defined as the layer that determines the rate and the composition of
the transported sediment [Van der Scheer, 2000]. Van der Scheers research showed
that most formulas do not state which part of the bed is considered to be the active
layer and how its composition should be determined. Van der Scheer showed that the
composition of the active layer has a large influence on the predictions by the
transport formulas. Thus, it is important to present a good method for determining
the composition of the active layer to determine the correct composition of the bed
material, which is used for sediment transport predictions. If, in the specific papers,
Description of research project
University of Twente 15
no definition is found of the active layer, the initial bed composition will be used as
the composition of the active layer.
A third aspect that will be studied is the type of experiments that are used for
calibration of the model. Flume experiments can roughly be divided into two groups:
experiments with a sediment recirculation system or with a sediment feed system.
Erosion experiments are sediment feed experiments without sediment input at the
upstream end of the flume. This leads to erosion of the bed and eventually to
armouring. Experiments with a sediment recirculation system are often used to
maintain equilibrium during the experiment and measure the sediment transport in
that period. Both types of experiments have been used to develop sediment
transport formulas and since different processes occur in the two types of
experiments, it is important to distinguish them.
1.3.3 Behaviour of transport formulas
Here we deal with the changes in the sediment transport rate and composition when
varying hydraulic and sedimentological parameters. Calculations with pre-set
hydraulic and sedimentological conditions will be done to get insight in these
changes.
The following parameters will be varied: Shields parameter ("
50
), grain size (D
50
)
and the geometric standard deviation of the grain size (!
g
). The variation of these
parameters will be within realistic ranges occurring in rivers in The Netherlands and
in laboratories. The predicted transport rates and compositions will be plotted versus
the Shields parameter. The sediment mixture will be divided into ten fractions
according to Ribberinks method [1987]. Ribberink assumes that the grain size
distribution is log-normal. The grain size and the probability of each fraction are
calculated by giving the values for D
50
and !
g
.
The range of the Shields parameter is within 0 to 0.6. The Shields parameter
will be changed by changing the water depth. The water surface slope and the Chzy
value will be held constant at the values 0.0005 and 40 m

/s, respectively.
Calculations will be done for different Shields parameters and four different bed
compositions. The different bed compositions are characterised by their
representative grain size (D
50
) and their geometric standard deviation of the grain
size (!
g
) (see Table 1.2).
Table 1.2: Bed material properties of the four series
!
g
(-)
D
50
(mm)
2 5
2 Series 1 Series 2
10 Series 3 Series 4
For the four series, we studied the influence of the Shields parameter on both the
sediment transport rate, and the geometric mean grain size of the transported
material.
Description of research project
16 University of Twente
1.3.4 Verification sediment transport formulas
The predictions given by the transport formulas listed in Table 1.1 will be verified
with the results of laboratory experiments. A data-file will be created of the
experimental data. The following experimental series are examined in this study:
Blom and Kleinhans, [1999], Non-uniform sediment in morphological
equilibrium situations - Data report Sand Flume experiments 97/98, Research
report CiT 99R-002/MICS-001, Civil Engineering and Management, University
of Twente.
Experimental Series 1 was conducted in the Sand Flume Facility of WL | delft
hydraulics in 1997/1998. Five tests were performed with a sediment mixture
from the river Rhine.
Blom, [2000], Flume experiments with a trimodal sediment mixture Data
report Sand Flume experiments 99/00, Research report CiT: 2000R-004/MICS-
013, Civil Engineering & Management, University of Twente.
Experimental Series 2 was conducted in the Sand Flume Facility of WL | delft
hydraulics in 1999/2000. Four tests were performed with a tri-modal mixture.
Klaassen, [1991], Experiments on the effect of gradation and vertical sorting on
sediment transport phenomena in the dune phase, Proc. Grain Sorting Seminar,
Ascona, Switzerland, pp 127-145.
Six tests with graded sediment were conducted in the Sand Flume Facility of WL
| delft hydraulics in 1990.
Day, [1980], A study of the transport of graded sediments, HRS Wallingford,
Report No. IT 190.
HRS A, experiments conducted by Day. 11 tests with graded sediment were
conducted in the flume of HR Wallingford.
HRS B, experiments conducted by Day. Nine tests with graded sediment were
conducted in the flume of HR Wallingford.
Wilcock and McArdell, [1993], Surface-Based Fractional Transport Rates:
Mobilization Thresholds and Partial Transport of a Sand-Gravel Sediment, Water
Resources Research, Vol. 29, No. 4, pp. 1297-1312.
Bed Of Many Colours (BOMC): Ten tests with a sand gravel mixture. The tests
were carried out at Johns Hopkins University.
Wilcock et al., [2001], Experimental Study of the Transport of Mixed Sand and
Gravel, Water Resources Res., Vol. 37, No. 12, pp. 3349-3358.
Four experimental series with different sand gravel mixtures were conducted at
Johns Hopkins University. 37 runs are used for verification in this report.
1.3.5 Surface layer as active layer
The active layer is defined as that part of the bed material, which determines the
transport rate and composition. The bed surface layer (the material present in the
surface layer of the moving bed forms or of a flat bed) seems the most logical
representation of the active layer. Namely, the water flow directly interacts with the
Description of research project
University of Twente 17
bed surface and only grains at the bed surface can be directly transported.
Measurements of the bed surface composition are only available from Wilcock &
McArdell [1993], Wilcock et al. [2001], and Blom [2000]. However, for most
experiments only the composition of the initial bed mixture was known. For this
reason, initially, the composition of the initial bed mixture is used as input for the
sediment transport predictions. Later the composition of the surface layer is used as
input to check if the predictions of the formulas become better.

Graded sediment
University of Twente 19
2 Graded sediment
2.1 Introduction
In experiments, we often use uniform sediment to increase our insight in sediment
transport processes. However, river sediment is mostly graded, i.e. the sediment
consists of a range of grain sizes. The gradation of the bed material shows great
variation among various rivers and it influences the processes that are involved in
sediment transport. This chapter describes several sediment properties and the
influence of the gradation of mixtures on some important transport processes.
2.2 Sediment properties
Several sediment properties have to be known to describe the sediment transport.
These include size, density, gradation and cohesion. In this report, only non-
cohesive material, like sand and gravel, is considered. The sediment transport
formulas analysed in the present study are not suitable for clay, silt and other
cohesive sediments.
The grain sizes of graded sediment are usually given in weight percentages.
The grain sizes are noted as D
x
, wherein the subscript x denotes the weight
percentage of the sediment mixture with a grain size smaller than D, e.g. D
50
= 2.5
mm indicates that 50% of the sediment mixture has a grain size smaller than 2.5
mm.
The density of natural sediments (#
s
) is approximately equal to 2650 kg/m
3
, as
they mostly consist of small quartz particles. The sediment density is generally used
in sediment transport formulas related to the density of water (#) in the form of
specific gravity:
#
#
s
s = (-)
Or in the form of relative density:
1 s
s
$ =
$
= %
#
# #
(-)
In literature several definitions of the geometric grain size exist. The most commonly
used definition is:
&
' =
i
i i m
D p D (m) (2.1)
in which p
i
is the probability of fraction i in the active layer, which determines the
transport rate and composition. D
i
is the grain size of fraction i. However, Parker
[1990] uses a different method to calculate the geometric mean grain size. It is
widely accepted that grain size distributions of sediment mixtures are log-normal.
Graded sediment
20 University of Twente
Calculating the geometric mean grain size according to this distribution leads to the
following expression for D
m
:
i
2
i
D log =$ ( grain size on phi-scale (-)
&
' =
i
i i
p ( ( mean grain size on phi-scale (-)
( $
= 2 D
m
geometric mean grain size (m) (2.2)
In this report, definition (2.1) will be used unless mentioned otherwise.
The geometric standard deviation gives a good indication of the gradation of the
sediment:
)
)
*
+
,
,
-
.
+ =
16
50
50
84
g
D
D
D
D
5 0. ! (-)
Engelund & Hansen [1967] stated that some sediment properties, like shape and
gradation, received too little attention. Since then, much research has been done on
the gradation of the sediment, but knowledge on the influence of shape is still
relatively small.
2.3 Incipient motion
The flow of water over the riverbed causes a stress known as the bed shear stress
(/
b
). The bed shear stress can be divided into a part related to the grains and a part
related to the bed forms:
/ / / 0 0 + 0 =
b

in which / notes the shear stress caused by the grains and / notes the shear stress
caused by the bed forms.


lower regime
dunes
ripples
plane
bed
upper regime
anti -
dunes
t
r
a
n
s
i
t
i
o
n

p
l
a
n
e

b
e
d

a
n
d

s
t
a
n
d
i
n
g

w
a
v
e
s

/
/
/b
u

Figure 2.1: Bed shear stress versus flow velocity (from Engelund & Hansen [1967]).
Graded sediment
University of Twente 21
Figure 2.1 shows the bed shear stress versus the flow velocity. It can be seen that, in
the lower flow regime, the bed form roughness mainly determines the bed shear
stress, while in the upper regime, the grain roughness is the most important factor
for the bed shear stress.
In an open channel the bed shear stress can be calculated with:
w b
i h g ' ' ' = # / bed shear stress (N/m
2
)
in which g is gravity, h is mean depth, and i
w
is the water surface slope. The Shields
parameter (") gives a dimensionless bed shear stress related to the grain diameter
(D) and is given by:
D
i h
D g
b
' %
'
=
' % ' '
=
#
/
" Shields parameter (-)
Each grain needs a certain shear stress to start moving. This stress is called the
critical bed shear stress (/
b,cr
). The grain is moved when the bed shear stress exceeds
the critical bed shear stress. The critical Shields parameter ("
cr
) represents the
dimensionless critical bed shear stress. Shields [1936] determined the relation
between the Reynolds number and the critical Shields parameter. Van Rijn [1984a]
gives a mathematical approximation of this relation. In Section 3.6, this
approximation is given as part of the sediment transport formula of Van Rijn.
Both " and "
cr
are related to the grain diameter. This means that grains with different
sizes start moving at different shear stresses. This can be simply explained in the
case of spherical particles. The weight of a particle is the main force that withholds it
from moving. The drag force, which is proportional to the bed shear stress, only
applies to the surface area of the particle. The weight is a function of the volume of
the particle. When the diameter of a particle is increased, its volume increases faster
than its surface area. This means that a higher bed shear stress is needed to move
larger particles.
2.4 Hiding/exposure
In sediment mixtures, there is a difference in exposure to the shear stress for
different grain sizes. Larger grains protrude more out of the bed and thus are more
exposed to the flow. This is called exposure. Smaller grains can lie in the shade of
the larger grains. This hides them from the flow. The combined effect is called
hiding/exposure. Figure 2.2 illustrates the hiding/exposure effect.

Exposure Hiding

Figure 2.2: Hiding/exposure effect.
Graded sediment
22 University of Twente
The hiding/exposure effects result in a smaller critical bed shear stress for larger
grains and a higher critical bed shear stress for smaller grains. Egiazaroff [1965] was
the first who estimated a hiding/exposure factor for the critical bed shear stress.
Most hiding /exposure factors are derived for a specific sediment transport formula.
They are mostly based on experimental results. This means that these empirical
hiding/exposure factors do not only correct for the critical bed shear stress but may
take other effects of gradation into account as well.

Sediment transport formulas
University of Twente 23
3 Sediment transport formulas
3.1 Introduction
Throughout the last fifty years, many researchers have proposed formulas to predict
sediment transport rates and compositions. Most of these formulas have an
empirical basis. They are mostly based on experiments with uniform bed material.
However, natural sediment mostly consists of non-uniform material. A certain grain
size is present in the bed material in a certain percentage, its probability. A simple
method to determine the transport per size fraction is to multiply the transport
calculated for uniform sediment of this size fraction by the probability of this size
fraction in the bed:
u i i n i
q p q
, ,
' =
In which:
q
i,n
volumetric sediment transport of fraction i in non-uniform sediment;
p
i
probability (volume fraction) of size fraction i being present in the transport
layer;
q
i,u
volumetric sediment transport of fraction i in uniform sediment with similar
hydraulic conditions.
In general, this method is considered to be too simple because it does not take
hiding and exposure effects into account. The following sections give an overview of
the ten transport formulas, listed in Table 1.1, and of available hiding/exposure
corrections. The data on which the several fractional transport formulas were
originally calibrated are analysed to determine their validity ranges.
3.2 Ackers & White
3.2.1 General
Ackers & White [1973] proposed a formula to estimate the total load transport. No
distinction was made between bed load and suspended load. This empirical formula
is based on 925 sets of data of flume experiments with a grain size ranging from
0.04 mm to 4 mm. The water depth was mainly below 0.4 m. The analysis of the
data showed that the transport of fine material (smaller than 0.04 mm) could be best
determined using the shear velocity (u
*
). The mean velocity ( u ) appeared more
suitable for coarser grains (larger than 2.5 mm). The method cannot be used for
grain sizes smaller than 0.04 mm because of their cohesive properties. Ackers &
White excluded data with Froude numbers exceeding 0.8. The method proved not to
be sensitive to bed forms. The following sections present the formulas for uniform
sediment and sediment mixtures.
Sediment transport formulas
24 University of Twente
3.2.2 Uniform transport
Ackers & White define the dimensionless transport rate as:
n
gr
u
u
D s
h X
G
)
)
*
+
,
,
-
.
'
'
'
=
*
dimensionless sediment transport rate (-) (3.1)
m
cr gr
gr
gr
1
F
F
K G
)
)
*
+
,
,
-
.
$ ' =
,
dimensionless sediment transport rate (-) (3.2)
in which:
#
#
s
s =
#
#
'
'
=
q
q
X
s t
mass flux per unit mass flow rate (-)
Combining Equations 3.1 and 3.2 gives the following equation, with which the
sediment transport volume without pore volume can be calculated:
m
cr gr
gr
n
t
1
F
F
u
u
D u K q
)
)
*
+
,
,
-
.
$ '
)
)
*
+
,
,
-
.
' ' ' =
, *
total sediment transport (m
2
/s) (3.3)
in which:
u depth-averaged velocity (m/s)
w
i h g u ' ' =
*
shear velocity (m/s)
( )
n 1
n
gr
D
h
10 32
u
D g
u
F
$
)
)
)
*
+
,
,
,
-
.
' '
'
' % '
=
log
*
sediment mobility number (-)
#
# # $
= %
s
relative density of sediment (-)
#
s
sediment density (kg/m
3
)
# water density (kg/m
3
)
D diameter of bed material (m)
3 1
2
g
D D
/
*
)
*
+
,
-
. ' %
' =
1
dimensionless grain size (-)
6 2
10 15 Te 00068 0 15 Te 031 0 14 1
$
' $ + $ $ = ] ) ( . ) ( . . [ 1 kinematic viscosity (m
2
/s)
Te temperature (C)
According to Ackers & White [1973], the best representation of the bed material is
the D
35
grain size. The coefficients n, m, K and F
gr,cr
(critical sediment mobility
number) are dimensionless. These coefficients depend on the dimensionless particle
size. Ackers & White make a distinction between particles with 1 < D
*
< 60 and
particles with D
*
2 60. Later revisions were made for the K and m coefficients [HR
Sediment transport formulas
University of Twente 25
Wallingford, 1990], because there were uncertainties in the original formula in the
sediment transport for relatively fine and coarse sediments. Van der Scheer [2000]
showed that the results of the Ackers & White formula with the modified parameters,
described in HR Wallingford [1990], are slightly worse than the predictions of the
original formula. This is the reason that the modified formula will not be analysed in
this report. Table 3.1 shows the original coefficients for the Ackers & White formula.
Table 3.1: Coefficients Ackers & White formula
coefficients 1 < D
*
< 60 D
*
2 60
n
) log( .
*
D 56 0 1 ' $
0
m
34 1
D
66 9
.
.
*
+
1.5

K
2
D D 86 2 53 3
10
) (log ) log( . .
* *
$ + $
0.025
F
gr,cr
14 0
D
23 0
.
.
*
+
0.17

3.2.3 Fractional transport
In this section the hiding/exposure corrections for the Ackers & White transport
formula by Day [1980] and Proffitt & Sutherland [1983] are explained. Both
correction factors (3
i
) given in these sections are applied to the critical sediment
mobility number (F
gr,cr
). The fractional sediment transport for fraction i yields:
i i
m
i i cr gr
i gr
n
i i i i t
1
F
F
u
u
D u K p q
)
)
*
+
,
,
-
.
$
'
'
)
)
*
+
,
,
-
.
' ' ' ' =
3
, ,
,
*
,
fractional transport formula (m
2
/s) (3.4)
3.2.3.1 Day's hiding/exposure factor
Day [1980] proposes a hiding/exposure factor based on a large number of
experiments with sediment mixtures (Table 3.2). Unfortunately, Day does not explain
which definition he used for the active layer of for which part of the bed the D
50
was
representative.
Table 3.2: Validity ranges of experimental data used by Day
Experiments D
50

(mm)
!
g

(-)
"
50
range
(-)
Remarks
USWES 1 [1935] 0.42 1.82 0.043 - 0.191 -
USWES 2 [1935] 0.44 1.52 0.062 - 0.260 -
USWES 9 [1935] 4.10 1.46 0.033 - 0.072 -
Gibbs & Neill [1972, 1973] 4.75 2.34 0.042 - 0.108 constant flow, tilting flume
Cecen & Bayazit [1973] 13.9 1.73 0.043 - 0.061 removal of armour layer
Day: HRS A [1980] 1.75 4.21 0.034 - 0.125 sediment recirculating system
Day: HRS B [1980] 1.55 3.50 0.028 - 0.123 sediment recirculating system

Day's correction is a correction factor for the critical sediment mobility number. The
hiding/exposure factor is given by:
Sediment transport formulas
26 University of Twente
6 0
D
D
4 0
i
A
i
. . + ' = 3 hiding/exposure factor (-) (3.5)
in which D
A
denotes the grain size in the sediment mixture that does not experience
any hiding/exposure effects:
28 0
84
16
50
A
D
D
6 1
D
D
.
.
)
)
*
+
,
,
-
.
' = (-)
3.2.3.2 Proffitt & Sutherland's hiding/exposure factor
Proffitt & Sutherland [1983] also propose a hiding/exposure factor for the Ackers &
White formula. This factor can be seen as an extension of Days work, since the latter
was more aimed at threshold conditions, whereas Proffitt & Sutherland include also
the higher transport rates. The hiding/exposure factor is based on laboratory
experiments done by Proffitt [1980] and verified with a large number of experiments
(Table 3.3). Also, Proffitt & Sutherland do not explain what they defined as the active
layer.
Originally, Proffitt & Sutherland's hiding/exposure factor was applied to the
sediment mobility number, which is in contrast to Days factor. The latter is applied
to the critical sediment mobility number. The inverse function of the original Proffitt
& Sutherland factor is given to allow the use of the two hiding/exposure factors in
the same form in formula 3.4:
Table 3.3: Validity ranges of experimental data used by Proffitt & Sutherland
Experiments D
50
range
(mm)
!
g

(-)
"
50
range
(-)
Remarks
Proffitt: run 1 [1980] 2.90 - 6.85 2.26 0.037 - 0.090 armouring experiments
Proffitt: run 2 [1980] 3.25 - 11.7 3.24 0.030 - 0.099 armouring experiments
Proffitt: run 3 [1980] 3.07 - 11.7 2.78 0.027 - 0.098 armouring experiments
Proffitt: run 4 [1980] 4.20 - 5.45 1.95 0.041 - 0.072 armouring experiments
Davies [1974] 2.8 - 4.2 ca. 3 0.050 - 0.087 armouring experiments
Day: HRS A [1980] 1.75 4.21 0.038 - 0.125 sediment recirculating system
Day: HRS B [1980] 1.55 3.50 0.040 - 0.123 sediment recirculating system
Gessler: 1-5 [1967] 1.0 2.78 0.074
Little and Mayer [1972] 1.0 2.05 0.057
East Fork river [1980] 1.3 6.71 0.150 - 0.326
Tanana river [1978] 13 18.9 N.A.
1

Snake river [1980] 30 6.87 0.050 - 0.094 river with paved bed
Clearwater river [1980] 18 10.9 N.A. river with paved bed

4
4
4
5
4
4
4
6
7
8
< < )
*
+
,
-
.
+
)
*
+
,
-
.
'
8
=
$
075 0
D
D
for 5 2
7 3
D
D
075 0 for 1
D
D
53 0
D
D
7 3 for 769 0
u
i
u
i
1
u
i
u
i
i
. .
. . log .
. .
3 hiding/exposure factor (3.6)

1
N.A.: Not Available
Sediment transport formulas
University of Twente 27
D
u
denotes the grain size that requires no correction. D
u
is related to the effective
Shields parameter, which denotes:
50
b
50
D g ' ' % '
=
#
/
" (-)
Proffitt & Sutherland give a relation between D
u
and "
50
that is shown in Figure 3.1.

(a) ACKERS & WHITE
1.5
1.0
0
0.02 0.06 0.10 0.14
" "" "
50
d
u
/d
50

Figure 3.1: Relation between "
50
and D
u
given by Proffitt & Sutherland [1983]
As seen in Figure 3.1, the function consists of four straight lines. The authors of this
report made the following approximation of the function:
4
4
5
4
4
6
7
<
8 < + ' $
8 < + ' $
8
=
50
50 50
50 50
50
50
u
097 0 for 456 0
097 0 046 0 for 40 1 73 9
046 0 040 0 for 944 1 6 21
04 0 for 08 1
D
D
"
" "
" "
"
. .
. . . .
. . . .
. .

3.3 Parker
3.3.1 General
Parker et al. [1982] proposed a bed load transport formula for paved gravel bed
streams. In paved rivers, the surface layer (pavement) is significantly coarser than the
substrate. Parker et al. distinguish a paved bed from an armoured bed in that a
paved bed is transported during peak discharges, in contradiction to armoured beds,
which never move. The formula uses the grain distribution of the substrate and is
further referred to as "substrate-based formula". Parker [1990] gives a final version
of this formula. In the same article, Parker transforms the substrate-based formula
into a surface-based formula that uses the grain distribution of the surface layer
instead of the grain distribution of the substrate. A description of the surface-based
formula is given in the next section. In this project, the substrate-based formula of
Sediment transport formulas
28 University of Twente
Parker is not taken into account since Van der Scheer [2000] showed that the formula
gives poor results for the sediment transport predictions.
3.3.2 Surface-based bed load formula
Parker [1990] transformed the substrate-based formula into a surface-based
formula to predict pavement as well as selective transport. The surface-based
formula includes a hiding/exposure factor. He defined the active layer as the surface
layer of the plane bed, with a thickness equal to D
90
. The transport formula is based
solely on field data (Table 3.4).
Table 3.4: Validity ranges of field data used by Parker
Stream D
50

(mm)
!
g

(-)
"
50
range
(-)
Remarks
Oak Creek 20 5.52 0.091 - 0.147 river with paved bed
Elbow River [1968] 28 2.79 0.103 - 0.139 river with paved bed
Snake River [1974/76] 27 7.09 0.097 - 0.160 river with paved bed
Clearwater River [1974/76] 18 9.19 0.058 - 0.134 river with paved bed
Vedder River [1974] 19 5.93 0.083 - 0.103 river with paved bed
The dimensionless surface-based transport formula denotes:
)] ( [
* *
i 0 0 sg rs i
g G W W 9 : ; ' ' ' = transport parameter (-) (3.7)
The fractional bed load without pore volume relates to the dimensionless transport
parameter as follows:
i
2
3
b
i bs
i
F
g q
W
'
)
*
+
,
-
.
% ' '
=
#
/
, *
(3.8)
The variables in the equations above are described by:
4
4
4
5
4
4
4
6
7
<
8 8 $ ' $ $ '
<
)
)
*
+
,
,
-
.
$
=
1 for
59 1 1 for 1 28 9 1 2 14
59 1 for
853 0
1 5474
G
2 14
2
5 4
: :
: : :
:
:
:
.
.
. ] ) ( . ) ( . exp[
.
.
) ( (-)
sg
i
i
D
D
= 9 (-)
&
' = ) ln exp(
i i sg
D F D surface layer geometric mean diameter (m)
*
rsg
sg
0 sg
/
"
: = (-)
sg
b
sg
D g ' ' % '
=
#
/
" Shields parameter D
sg
of surface layer (-)
The hiding/exposure factor for the surface-based formula is given by:
Sediment transport formulas
University of Twente 29
<
9
$
)
)
*
+
,
,
-
.
=
sg
i
i 0
D
D
g ) ( reduced hiding function (-)
The surface-based formula uses a generalised straining factor to better agree with
the equal mobility concept. The straining function yields:
) ( 1 1
0
0
$ ' + = ;
!
!
;
:
:
straining function (-)
&
'
=
=
>
?
@
@
A
B
=
i
2
sg i 2
F
2
D D
) ln(
) ln(
:
! arithmetic standard deviation (-)
F
i
denotes the volume fraction of fraction i in the surface layer. The parameter <
equals 0.0951. The reference values
*
rs
W and
*
rsg
/ equal to 0.0025 and 0.0386,
respectively. The values ;
0
and !
:0
depend on :
sg0
, as shown in Figure 3.2. The
relations in Figure 3.2 are based on the relations given by Parker. Parker [1990] does
not give a mathematical relation between ;
0
and :
sg0
and between !
:0
and :
sg0
, so
two sets of linear relations between the parameters have been estimated by the
authors of this report:
4
4
5
4
4
6
7
8
< 8 + ' $
< 8 + ' $
<
=
0 sg
0 sg 0 sg
0 sg 0 sg
0 sg
0
5 for 453 0
5 5 2 for 748 0 059 0
5 2 1 for 285 1 274 0
1 for 011 1
:
: :
: :
:
;
.
. . .
. . .
.
(-)
4
4
5
4
4
6
7
8
< 8 + '
< 8 + '
<
=
0 sg
0 sg 0 sg
0 sg 0 sg
0 sg
0
5 for 501 1
5 2 for 000 1 100 0
2 1 for 432 0 384 0
1 for 816 0
:
: :
: :
:
!
:
.
. .
. .
.
(-)
!
:0
;0
1.501
0.453
0 1 2 3 4
:sg0
0.816
1.011
0
1
2
!
:0 .;0

Figure 3.2 Relation between :
sg0
and ;
0
, and the relation between:
sg0
and !
:0
, after
Parker [1990]
Sediment transport formulas
30 University of Twente
3.4 Engelund & Hansen
3.4.1 General
The formula proposed by Engelund & Hansen [1967] is semi-empirical. Although the
formula was originally derived for bed load transport, it proved to be very suitable to
the total sediment transport of small grain sizes. Grain sizes, ranging from 0.19 to
0.93 mm, were used in the experiments on which the formula is based. Engelund &
Hansen note that application of the formula should be done with care for grain sizes
smaller than 0.15 mm.
3.4.2 Uniform transport
The dimensionless sediment transport parameter for uniform sediment transport is
given by:
2
5
50 EH
05 0 ) ( . " ' ' = C transport parameter (-) (3.9)
The total sediment transport in volume omitting pores can be determined using:
3
50
t
D g
q
' ' %
= C
in which:
5
2
2
EH
g
C
)
)
*
+
,
,
-
.
= bed form factor (-)
50
b
50
D g ' ' % '
=
#
/
" Shields parameter (-)
w
i h
u
C
'
= Chzy value (m
1/2
/s)
3.4.3 Fractional transport
No hiding/exposure correction has been developed for this formula. Laguzzi [1994]
suggested that Egiazaroff [1965], modified by Ashida & Michiue [1973], could be
used to correct for the hiding/exposure of the fractions present in bed material. This
method has not been tested and will not be used here. A fractional transport formula
only accounting for the probability of size fractions in the bed is used. The
dimensionless transport parameter yields:
2
5
i EH i i
05 0 p ) ( . " ' ' ' = C fractional transport parameter (-) (3.10)
The transported volume is determined by:
3
i
i t
i
D g
q
' ' %
= C
,

in which:
Sediment transport formulas
University of Twente 31
i
b
i
D g ' ' % '
=
#
/
" fractional Shields parameter (-)
3.5 Meyer-Peter & Mller
3.5.1 General
Meyer-Peter & Mller [1948] proposed an empirical sediment transport formula for
bed load based on extensive experimental research. Both uniform and non-uniform
sediment was used with a grain size ranging from 0.4 to 29 mm.
Egiazaroff [1965] developed a general hiding/exposure factor, which has been
used extensively in combination with the Meyer-Peter & Mller transport formula.
Ashida & Michiue [1973] modified Egiazaroff's hiding/exposure factor.
3.5.2 Uniform transport
The original formula of Meyer-Peter & Mller uses the geometric mean grain size
(D
m
) as the representative grain size. The dimensionless transport parameter is given
by:
2
3
m MPM b
047 0 8 ) . ( $ ' ' = C " transport parameter (-) (3.11)
The relation between the dimensionless transport parameter and the volumetric
sediment transport without pore volume is given by:
3
m
b
b
D g
q
' ' %
= C
in which:
m
b
m
D g ' ' ' %
=
#
/
" Shields parameter for D
m
(-)
2
3
MPM
C
C
)
*
+
,
-
.
0
= bed form factor (-)
The bed form factor should always be smaller or equal to 1, since the grain related
Chzy value (C
90
) can never be smaller than the total Chzy value (C). This restriction
is applied in the present study.
)
)
*
+
,
,
-
. '
' = 0
90
D
h 12
18 C log grain related Chzy value (White-Colebrook) (m

/s)
&
' =
i
i i m
D p D geometric mean grain size (m)
It should be noted that the bed form factor of the Meyer-Peter & Mller formula used
here is not the original bed form factor given by Meyer-Peter & Mller [1948]. In
Appendix 1, the different bed form factors are analysed in detail.
Sediment transport formulas
32 University of Twente
3.5.3 Fractional transport
3.5.3.1 Egiazaroffs hiding/exposure factor
The hiding/exposure factor developed by Egiazaroff [1965] was calibrated with data
from several experiments. Unfortunately, we did not succeed in collecting all
information on the experimental data used by Egiazaroff. This means that the
validity ranges given in Table 3.5 are not complete. Egiazaroff does not explain his
definition of the active layer and his method to determine the composition of the
active layer.
Table 3.5: Validity ranges of experimental data used by Egiazaroff
Experiments D
50

mm
!
g

-
"
50
range
-
Remarks
Nizery & Braudeau [1953] 5.50 4.84 0.666 - 0.917 armouring experiments
Pantlopulos A [1957] 2.55 2.08 0.042 - 0.101 feed experiments
Pantlopulos B [1957] 1.80 2.38 N.A. feed experiments
Pantlopulos C [1957] 2.75 2.54 N.A. feed experiments
Oumarov [1961] N.A. N.A. N.A. -
Ramette [1962] N.A. N.A. N.A. -
The hiding/exposure factor of Egiazaroff [1965] is applied to the critical Shields
parameter. This results in the dimensionless transport parameter per fraction:
2
3
i i MPM i i b
047 0 8 p ) . (
,
' $ ' ' ' = C 3 " fractional transport parameter (-) (3.12)
This parameter is related to the fractional sediment transport (without pore volume):
3
i
i b
i b
D g
q
' ' %
= C
,
,

using:
i
b
i
D g ' ' ' %
=
#
/
" fractional Shields parameter (-)
The hiding/exposure factor of Egiazaroff is semi-theoretically determined. It
considers the balance of forces working on a spherical grain on the threshold of
movement. The correction factor is applied to the critical Shields parameter, which
has a value of 0.047 in the formula of Meyer-Peter & Mller.
( )
2
m i
i
D D 19
19
=
>
?
@
A
B
'
=
log
) log(
3 hiding/exposure factor Egiazaroff (-) (3.13)
3.5.3.2 Ashida & Michiues hiding/exposure factor
For the range D
i
/D
m
< 0.4, Ashida & Michiue [1973] did not find consistency between
their experiments (Table 3.6) and Egiazaroff's correction factor. They present a
correction to Egiazaroff for this range, but this correction must be used with care
since it is based on only one measurement in this range [Ribberink, 1981].
Sediment transport formulas
University of Twente 33
Table 3.6: Validity ranges of the experimental data used by Ashida & Michiue
Experiments D
50

(mm)
!
g

(-)
"
50
range
(-)
Remarks
Ashida & Michiue [1971] 1.7 3.73 0.043 - 0.185 armouring experiments
Hirano A [1970] 0.83 1.24 0.039 armouring experiments
Hirano B [1970] 0.98 2.31 0.038 armouring experiments
The corrected hiding/exposure factor yields:
( )
4
4
5
4
4
6
7
< '
2
=
>
?
@
A
B
'
=
4 0
D
D
for
D
D
85 0
4 0
D
D
for
D D 19
19
m
i
i
m
m
i
2
m i
i
. .
.
log
) log(
3 hiding/exposure factor A&M (-) (3.14)
3.6 Van Rijn
3.6.1 General
Van Rijn [1984a,b] divided his transport formula in a bed load and a suspended load
formula. The total sediment transport is the sum of the bed load and the suspended
load.
s b t
q q q + = total transport (m
2
/s) (3.15)
The suspended load formula is not used in the present study, but presented for
completeness.
3.6.2 Uniform bed load formula
Like Bagnold [1954], Van Rijn assumed in his bed load formula that the bed load
particles primarily move because of saltations or jumps. The bed load formula is
suitable for sediments with a grain diameter ranging from 0.2 2 mm. The
dimensionless bed load parameter is given below:
4
5
4
6
7
2 ' '
< ' '
= C
$
$
3 T for T D 1 0
3 T for T D 053 0
5 1 3 0
1 2 3 0
b
. .
*
. .
*
.
.
bed load parameter (-) (3.16)
The relation between the sediment transport volume without pore volume and the
dimensionless transport parameter yields:
3
50
b
b
D g
q
' % '
= C
in which:
3 1
2
50
g
D D )
*
+
,
-
. ' %
' =
1
*
dimensionless grain size (-)
cr b
cr b b
T
,
,
'
/
/ / $
= bed shear parameter (-)
b b
/ / ' =
'
effective bed shear stress (N/m
2
)
Sediment transport formulas
34 University of Twente
2
C
C
)
*
+
,
-
.
=
'
bed form factor (-)
)
)
*
+
,
,
-
.
'
'
' =
90
D 3
h 12
18 C log ' grain related Chzy value (White-Colebrook) (m
1/2
/s)
cr 50 s cr b
D g " # # / ' ' ' $ = ) (
,
critical bed shear stress (N/m
2
)
The bed form factor can physically not be larger than one because the total Chzy (C)
value must always be smaller than or equal to the grain related Chzy value (C).
However, in this formula the Chzy value cancels out of the equation of the effective
bed shear stress when the following expressions are used:
w
i h C u ' ' =
2
w b
C
u
g i h g )
*
+
,
-
.
' ' = ' ' ' = # # /
This may cause the bed form factor in this formula to become larger than one,
although it is physically not justifiable.
The critical Shields parameter "
cr
can be determined from Figure 3.3.
10
0
10
-1
10
-2
10
0
10
1
10
2
10
3
"
cr
dimensionless particle size, D
*
NO MOTION

Figure 3.3: Relation between D
*
and "
cr
given by Van Rijn [1984a]
The Shields curve given in Figure 3.3 can be approximated by [Van Rijn, 1984a]:
4
4
4
5
4
4
4
6
7
>
8 < '
8 < '
8 < '
8 '
=
$
$
$
150 D 055 0
150 D 20 D 013 0
20 D 10 D 04 0
10 D 4 D 14 0
4 D D 24 0
29 0
10 0
64 0
1
cr
*
*
.
*
*
.
*
*
.
*
* *
.
.
.
.
.
" critical Shields parameter (-) (3.17)
Sediment transport formulas
University of Twente 35
3.6.3 Fractional bed load formula
No specific hiding/exposure correction has been developed for Van Rijns formula.
Although Laguzzi [1994] suggests some options, these have not been verified and
will not be used here. Nevertheless, a fractional bed load formula is given accounting
for the probability of the fractions as described in Section 3.1. This results in the
following dimensionless bed load parameter:
4
5
4
6
7
> ' ' '
< ' ' '
= C
$
$
3 T for T D 1 0 p
3 T for T D 053 0 p
i
5 1
i
3 0
i
i
1 2
i
3 0
i
i b
. .
*
. .
*
,
.
.
fractional bed load parameter (3.18)
The relation between sediment transport volume and the dimensionless transport
parameter now yields:
3
i
i b
i b
D g
q
' % '
= C
,
,

in which:
3 1
2
i
g
D D )
*
+
,
-
. ' %
' =
1
*
dimensionless grain size (-)
i cr b
i cr b b
i
T
, ,
, ,
'
/
/ / $
= fractional bed shear parameter (-)
i cr i s i cr b
D g
, , ,
) ( " # # / ' ' ' $ = critical fractional bed shear stress (N/m
2
)
The critical Shields parameter "
cr,i
can be determined from Figure 3.3 or (3.17). The
other variables in the fractional bed load formula are the same as in the uniform bed
load formula.
3.6.4 Suspended load formula
The suspended load formula determines the volume of the sediment transported as
suspended lead, without pores. Van Rijn [1993] does not give a dimensionless
transport formula. Van Rijns formula for suspended sediment transport yields:
a s
c h u F q ' ' ' = suspended sediment transport (m
2
/s) (3.19)
in which F can only be approximated, as the original differential equation (not given
in this report) cannot be integrated analytically. The following equation gives an
estimation of F [Van Rijn, 1984]:
( ) ' .
'
. '
Z 2 1
h
a
1
h
a
h
a
F
Z
2 1 Z
$ ' )
*
+
,
-
.
$
)
*
+
,
-
.
$ )
*
+
,
-
.
= F-factor (-)
4 0
0
a
8 0
s
2
s
s
c
c
u
w
5 2
u
u
w
2 1
w
Z
. .
*
*
*
. '
)
)
*
+
,
,
-
.
'
)
)
*
+
,
,
-
.
' +
' '
)
)
*
+
,
,
-
.
)
*
+
,
-
.
+
=
D
suspension number (-)
Sediment transport formulas
36 University of Twente
3 0
5 1
50
a
D
T
a
D
015 0 c
.
*
.
. ' ' = reference sediment concentration (-)
c
0
= 0.65 maximum concentration (-)
D = 0.4 constant of Von Karman (-)
The fall velocity w
s
(in m/s) is given by Van Rijn [1993]:
4
4
4
4
5
4
4
4
4
6
7
> ' ' % '
8 8
)
)
)
*
+
,
,
,
-
.
$
' ' % '
+
'
<
'
' ' %
=
mm 1 D for D g 1 1
mm 1 D 1 0 for 1
D g 01 0
1
D
10
mm 1 0 D
18
D g
w
s s
s
2
3
s
s
2
s
s
.
.
.
.
1
1
1

in which D
s
denotes the representative particle size (m) of suspended sediment.
5
6
7
2
< ' $ ' $ ' +
=
25 T for D
25 T for D 25 T 1 011 0 1
D
50
50 s
s
)] ( ) ( . [ !

cr b
cr b b
T
,
,
'
/
/ / $
= bed shear parameter (-)
)
)
*
+
,
,
-
.
+ =
16
50
50
84
s
D
D
D
D
5 0. ! geometric standard deviation of bed material (-)
The reference level (a) is equal to 0.5 times the bed form height (H) or equal to the
overall roughness height (k
s
).
3.7 Hunziker/Meyer-Peter & Mller
3.7.1 General
Hunziker [1995] proposed a fractional formula for bed load. This formula can be
seen as a modified version of the sediment transport formula of Meyer-Peter &
Mller [1948]. Hunziker found that the fractional bed load formula of Meyer-Peter &
Mller with the hiding/exposure factor of Ashida & Michiue [1973] does not lead to
satisfactory predictions of the rotational erosion process of the experiments of
Gnter [1971]. Hunziker states that all fractions in a uniform mixture experience
initial motion at the same critical Shields parameter, but the fine particles are more
mobile than the coarse ones. The formula is calibrated with the experiments of
Gnter [1971] and Suzuki [1992] (Table 3.7).
Table 3.7: Validity ranges of the experimental data used by Hunziker
Experiments D
50

mm
!
g

-
"
50
range
-
Remarks
Gnter: mixture 1 [1971] 1.73 3.08 0.077 - 0.085 armouring experiments
Gnter: mixture 2 [1971] 0.90 3.50 0.127 - 0.148 armouring experiments
Gnter: mixture 3 [1971] 2.54 3.87 0.065 - 0.071 armouring experiments
Suzuki & Hano [1992] 2.10 2.93 0.175 - 0.326 sand feed system
Sediment transport formulas
University of Twente 37
Meyer-Peter & Mller Run 1 [1948] 1.43 2.74 0.062 - 0.197 equilibrium conditions
Meyer-Peter & Mller Run 2 [1948] 3.93 2.62 0.078 - 0.127 equilibrium conditions
Meyer-Peter & Mller Run 3 [1948] 2.62 2.45 0.061 - 0.115 equilibrium conditions
Zarn [1997] N.A. N.A. N.A. equilibrium conditions
Gessler i/5 [1965] 1 4.1 0.081 armouring experiment
3.7.2 Fractional bed load formula
The fractional dimensionless transport parameter of Hunziker/Meyer-Peter & Mller
is given by:
2
3
cm s m i i i b
5 p )) ( (
*
, ,
" " 3 $ ' ' = C transport parameter (-) (3.20)
The relation between the fractional dimensionless transport parameter and the
volumetric sediment transport without pore volume is given by:
3
s m
i b
i b
D g
q
,
,
,
' % '
= C
in which:
E
3
$
)
)
*
+
,
,
-
.
=
s m
i
i
D
D
,
hiding/exposure factor (-) (3.21)
3 0 011 0
5 1
s m
. .
.
*
,
$ ' =
$
" E (-)
s m H s m ,
*
,
" " ' = corrected Shields parameter (-)
2
H
C
C
)
*
+
,
-
.
0
= bed form factor (-)
f
A
g
C = 0 grain related Chzy value (Yalin & Scheuerlein [1988]) (m
1/2
/s)
2
90
f
D 2
h
b
1
A
$
)
)
*
+
,
,
-
.
)
)
*
+
,
,
-
.
'
= ln
D
grain resistance of bed (Yalin & Scheuerlein) (-)
) . ( 1 B 4 0
s
e b
$ '
= (-)
( ) ( )
( ) ( )
2
217 0
s
e 3 5 2 5 8 B
*
Re ln .
*
Re ln . .
$
' $ + = roughness function (Schlichting, [1968]) (-)
1
90
D 2 u ' '
=
*
*
Re Reynolds number (-)
6 2
10 15 Te 00068 0 15 Te 031 0 14 1
$
' $ + $ $ = ] ) ( . ) ( . . [ 1 kinematic viscosity (m
2
/s)
w
i h g u ' ' =
*
shear velocity (m/s)
Hunziker found that using the formula of Strickler [1923] leads to too small values of
the bed form factor at the beginning of motion and uses therefore the formula of
Yalin & Scheuerlein [1988]. This results in a bed form factor closer to one and thus a
larger sediment transport. Hunziker compensated this by applying a smaller value of
Sediment transport formulas
38 University of Twente
the calibration factor in his own formula (5 instead of 8 at Meyer-Peter & Mller),
see (3.20). The bed form factor is restricted to values smaller than 1.
s m
b
ms
D g
,
' ' ' %
=
#
/
" Shields parameter for D
ms
(-) (3.22)
33 0
s m
o m
cr cm
D
D
.
,
,
)
)
*
+
,
,
-
.
' = " " critical Shields parameter corrected for layers in bed (-)
&
' =
i
o i o i o m
D p D
, , ,
geometric mean grain size of the under-layer (m)
&
' =
i
s i s i s m
D p D
, , ,
geometric mean grain size of the surface layer (m)
The critical Shields parameter is determined with the expression given by Iwagaki
[1956]. The expression of Iwagaki is rewritten such that the critical Shields
parameter can be calculated using the dimensionless grain size.
3 1
2
s m
g
D D )
*
+
,
-
. ' %
' =
1
, *
dimensionless grain size (-)
4
4
4
4
5
4
4
4
4
6
7
2
< 8
' ' %
' ' %
< 8
< 8
' ' %
' ' %
<
=
$
$
6 76 D 05 0
6 76 D 8 29
D g
D g 01505 0
8 29 D 3 14 034 0
3 14 D 66 1
D g
D g 1235 0
66 1 D 14 0
s m
s m
s m
s m
cr
22
11
11
3
22
25
32
11
16
3
32
25
. .
. .
) . (
. . .
. .
) . (
. .
*
*
,
,
*
*
,
,
*
1
1
" critical Shields stress (-) (3.23)
Hunziker has made some restrictions for his hiding/exposure correction to
counteract instability in certain parameter ranges. These restrictions are:
E 8 2.3
25 0
D
D
s m
i
.
,
2 in equation 3.21
3.8 Gladkow & Shngen
3.8.1 General
Gladkow & Shngen [2000] developed a fractional bed load formula. Gladkow &
Shngen used several formulas and their modifications as basis for their new bed
load formula. The critical Shields parameter used in the formula is an approximation
of the formula of Knoroz [1958]. Knoroz found a decreasing critical Shields
parameter, which is not in concurrence with Shields [1936], who found a decreasing
critical Shields parameter after which the parameter increases. The formula was
calibrated with field and experimental data. The data used for calibration could not
be acquired, so the validity ranges could not be determined.
Sediment transport formulas
University of Twente 39
3.8.2 Fractional bed load formula
The fractional bed load formula of Gladkow & Shngen calculates the bed load
transport rate without pores as:
4 2
i S
i cr i
4 2
i i b
1
Fr q 0014 0 p q
.
,
.
,
.
)
)
)
)
)
)
*
+
,
,
,
,
,
,
-
.
%
'
'
$
' ' ' ' =
"
" 3
fractional volumetric transport (m
2
/s)
(3.24)
in which:
h g
u
Fr
'
= Froude number (-)
2
i
i
i cr
72 0 D
3 1 D
0266 0
)
)
*
+
,
,
-
.
$
+
' =
.
.
.
*,
*,
,
" critical Shields parameter for fraction i (-)
3
1
2
i i
g
D D )
*
+
,
-
. ' %
' =
1
*,
dimensionless grain size of fraction i (-)
i
i
D
i h
' %
'
= " Shields parameter (-)
4
1
s
s
s
)
)
*
+
,
,
-
. 0
=
D
D
bed form factor of Shngen [1996] (-)
s s s
D D D 0 0 + 0 = total bed roughness according to Shngen [1996] (m)
m
D m s
S 6 1 D ' + = 0 . D grain roughness (m)
&
' =
i i m
D p D mean grain size (m)
&
$ ' =
2
m i i D
D D p S
m
) ( standard deviation grain size (m)
h
h 2
d
s
'
= 0 0 D bed form roughness (-)
( )
2
m
m cr
d
Fr 1 1
6
1
h h $ '
)
)
*
+
,
,
-
.
$ ' ' =
"
"
,
dune height (m)
2
m
m
m cr
72 0 D
3 1 D
0266 0
)
)
*
+
,
,
-
.
$
+
' =
.
.
.
*,
*,
,
" critical Shields parameter for D
m
(-)
3
1
2
m m
g
D D )
*
+
,
-
. ' %
' =
1
*,
dimensionless grain size of D
m
(-)
m
m
D
i h
' %
'
= " Shields parameter for D
m
(-)
Sediment transport formulas
40 University of Twente
( )
4
4
4
4
5
4
4
4
4
6
7
2
< 8
=
>
?
@
A
B
'
< '
=
0 1
D
D
for
D
D
0 1
D
D
4 0 for
D D 19
19
4 0
D
D
for
D
D
85 0
m
i
9 0
i
m
m
i
2
m i
m
i
i
m
i
.
. .
log
) log(
. .
.
3 hiding/exposure factor Shngen (-) (3.25)
3.9 Wu et al.
3.9.1 General
Wu et al. [2000] developed a fractional formula for bed load transport and a
fractional formula for suspended load. The total sediment transport is calculated as
in (eq. 3.15).
3.9.2 Fractional bed load formula
The formula of Wu et al. has been calibrated and tested for a wide range of
experimental and field data. The key element of this formula is the development of a
new hiding/exposure factor. In this factor the grain size of a fraction is compared
with the grain sizes of the other fractions. It is assumed that particles are distributed
randomly on the bed. This leads to the assumption that the exposure height of a
particle is normally distributed. The bed load formula shows large similarities with
the formula of Meyer-Peter & Mller.
Table 3.8: Validity ranges of the experimental data used by Wu et al.
Experimental / Field data D
50

mm
!
g

-
"
50
range
-
Remarks
Samaga [1986a] 0.2 - 0.35 1.91 - 3.79 0.317 - 0.823 Recirculation flume
Kuhnle [1993] 0.44 - 5.58 1.28 - 5.27 - Recirculation flume
Wilcock [1993] 4.53 9.91 0.009 - 0.196 Recirculation flume
Liu [1986] - - - Not available
Sustina river 34.8 - 50.2 1.66 - 2.13 0.059 - 0.127 Gravel bed river
Chultina river 10.6 - 23.6 2.23 - 2.40 0.032 - 0.074 Gravel bed river
Black river 0.38 - 0.53 1.68 - 1.99 0.141 - 0.544 Sand bed river
Toutle river 1.82 - 24.8 1.29 - 3.52 0.051 - 1.387 Gravel bed river
Yampa river 0.47 - 0.70 1.67 - 2.17 0.444 - 4.076 Sand bed river
The dimensionless transport parameter for fractional bed load yields:
2 2
i c
b
2
3
i i b
1
n
n
0053 0 p
.
,
,
'
.
=
=
>
?
@
@
A
B
$ )
*
+
,
-
.
' ' =
/
/
: fractional transport parameter (-) (3.27)
The volumetric bed load transport without pore volume can be determined with:
3
i
i b
i b
D g
q
' ' %
=
,
,
:
in which:
( )
i c i s i c
D g 3 " # # / ' ' ' ' $ =
,
critical shear stress (N/m
2
)
Sediment transport formulas
University of Twente 41
03 0
c
. = " critical Shields parameter (-)
6 0
i h
i e
i
p
p
.
,
,
$
)
)
*
+
,
,
-
.
= 3 hiding/exposure factor (-) (3.28)
&
=
+
=
N
1 j j i
i
j i e
d d
d
p p
,
exposure probability (-)
&
=
+
=
N
1 j j i
j
j i h
d d
d
p p
,
hiding probability (-)
i h g
b
' ' ' = # / bed shear stress (N/m
2
)
20
D
n
6
50
= ' Mannings roughness related to grains (s/m
1/3
)
u
i h
n
2
1
3
2
'
= Mannings roughness (s/m
1/3
)
3.9.3 Fractional suspended load formula
Wu et al. [2000] also proposed a fractional suspended load formula. This formula is
given here for completeness, but it will not be taken into account in the remaining
part of this study. The formula was analysed with one set of experimental data and
two sets of field data.
The fractional transport parameter for suspended sediment transport is:
74 1
i i c
i i s
u
1 0000262 0 p
.
,
,
.
=
=
>
?
@
@
A
B
)
)
*
+
,
,
-
.
$ ' ' =
; /
/
: fractional transport parameter (-) (3.29)
The transport parameter relates to the volumetric sediment transport in the
following way:
3
i
i s
i s
D g
q
' ' %
=
,
,
:
in which:
i h g ' ' ' = # / shear stress (N/m
2
)
i
i
2
i
i
D
95 13 D g 09 1
D
95 13
1 1
; . . . $ ' ' % ' +
)
)
*
+
,
,
-
.
' = settling velocity (m/s)
6 2
10 15 Te 00068 0 15 Te 031 0 14 1
$
' $ + $ $ = ] ) ( . ) ( . . [ 1 kinematic viscosity (m
2
/s)
3.10 Wilcock & Crowe
Wilcock & McArdell [1993] and Wilcock et al. [2001] did extensive research in
experimental flumes after the fractional transport of sand/gravel mixtures. Based on
these experiments, Wilcock & Crowe [2001] developed a fractional transport formula
for mixtures of sand and gravel, using the bed surface as their active layer (the bed
Sediment transport formulas
42 University of Twente
layer determining the transport rate and composition). The experiments showed a
relation between the sand content of the surface composition and the sediment
transport rate.
3.10.1 Fractional formula
The formula is calibrated on the results of five experimental series with 48 runs
(Table 3.9). The bed surface compositions of the experiments were determined using
a point count method. The initial bed material was sieved and different fractions
were painted in different colours. After a run photos were taken of the bed surface
and later the photos were analysed to determine the probability of each fraction in
the bed surface.
Table 3.9: Validity ranges of the experimental data used by Wilcock & Crowe [2001]
Experiments D
50

mm
!
g

-
"
50
range
-
Remarks
Wilcock et al: J06 10.22 2.93 0.028 - 0.134 Recirculation flume
Wilcock et al: J14 8.38 3.69 0.048 - 0.136 Recirculation flume
Wilcock et al: J21 7.06 4.66 0.028 - 0.165 Recirculation flume
Wilcock et al: J27 5.76 5.42 0.012 - 0.201 Recirculation flume
Wilcock & McArdell: BOMC 4.53 9.91 0.009 - 0.196 Recirculation flume
The dimensionless transport parameter of Wilcock & Crowe is given by:
4
4
5
4
4
6
7
2
)
)
*
+
,
,
-
.
$
< '
=
35 1 for
894 0
1 14
35 1 for 002 0
W
5 4
5 0
5 7
i
.
.
. .
.
.
.
*
:
:
: :
dimensionless transport parameter (-) (3.30)
The dimensionless transport parameter is related to the volumetric sediment
transport (without pore volume) according to:
3
i
bi
i
u p
q g
W
*
*
'
' ' %
=
in which:
ri
/
/
: =
w
i h g ' ' ' = # / bed shear stress (N/m
2
)
b
50 s
i
50 rs
ri
D
D
)
)
*
+
,
,
-
.
=
/
/
hiding/exposure function (-) (3.31)
)
)
*
+
,
,
-
.
$ +
=
50 s
i
D
D
5 1 1
69 0
b
. exp
.
(-)
50 s
50 rs
50 rs
D g ' ' ' %
=
#
/
/
*
Shields parameter of median surface grain size (-)
( )
s 50 rs
F 14 013 0 021 0 ' $ ' + = exp . .
*
/
w
i h g u ' ' =
*
shear velocity (m/s)
Sediment transport formulas
University of Twente 43
The parameter F
s
denotes the sand fraction in the active layer. Sand is considered to
have a grain size between 0.062 and 2 mm.
3.11 Ribberink
Ribberink [1998] developed a uniform bed load formula for both steady flow and
oscillatory flow conditions. This formula is largely based on the formula by Meyer-
Peter & Mller [1948], but has some small modifications. Ribberink used laboratory
experiments and field data for the calibration of his formula. Most of these
experiments were conducted in the sheet flow regime with a flat bed. The grain sizes
ranged from 0.19 to 3.8 mm. The effective Shields parameter (part of the bed shear
stress responsible for bed load transport) lies within the range of 0.03 - 7.7. The
uniform formula was modified to a fractional formula. The hiding/exposure factor of
Ashida & Michiue was used.
3.11.1 Uniform formula
The formula of Ribberink uses the median grain size (D
50
) as the representative grain
size. This differs from Meyer-Peter & Mllers formula that uses the mean grain size
(D
m
) as its representative grain size. The dimensionless transport parameter can be
determined by:
67 1
cr 50 b
4 10
.
) ( . " " $ 0 ' = C transport parameter (-) (3.32)
The relation between the dimensionless transport parameter and the volumetric
sediment transport without pore volume is given by:
3
50
b
b
D g
q
' ' %
= C
in which:
( )
50 s
50
D g ' ' $
0
= 0
# #
/
" effective Shields parameter (-)
2
2
C
u
g
0
' = 0 # / bed shear stress caused by grains (N/m
2
)
)
)
*
+
,
,
-
. '
' = 0
s
k
h 12
18 C log grain related Chzy value (m

/s)
( ) ( )
5
6
7
$ + '
'
=
1 6 1 D
D 3
k
50 50
90
s
"
max roughness height (m)
4
4
4
5
4
4
4
6
7
>
8 < '
8 < '
8 < '
8 '
=
$
$
$
150 D 055 0
150 D 20 D 013 0
20 D 10 D 04 0
10 D 4 D 14 0
4 D D 24 0
29 0
10 0
64 0
1
cr
*
*
.
*
*
.
*
*
.
*
* *
.
.
.
.
.
" critical Shields parameter: Van Rijn [1984] (-)
(3.33)
Sediment transport formulas
44 University of Twente
3 1
2
50
g
D D )
*
+
,
-
. ' %
' =
1
*
dimensionless grain size (-)
3.11.2 Fractional formula
The hiding/exposure factor of Ashida & Michiue [1973] was used. This is possible
because the formula of Ribberink is based on the formula of Meyer-Peter & Mller
[1948]. The hiding/exposure factor of Ashida & Michiue is widely used in the
fractional formula of Meyer-Peter & Mller. The dimensionless transport parameter
now yields:
67 1
i cr i i i i b
4 10 p
.
, ,
) ( . " 3 " ' $ 0 ' ' = C fractional transport parameter (-) (3.34)
This parameter is related to the fractional sediment transport (without pore volume):
3
i
i b
i b
D g
q
' ' %
= C
,
,

in which:
( )
i s
i
D g ' ' $
0
= 0
# #
/
" effective Shields parameter (-)
( )
4
4
5
4
4
6
7
< '
2
=
>
?
@
A
B
'
=
4 0
D
D
for
D
D
85 0
4 0
D
D
for
D D 19
19
m
i
i
m
m
i
2
m i
i
. .
.
log
) log(
3 hiding/exposure factor A&M (-)
3 1
2
i
g
D D )
*
+
,
-
. ' %
' =
1
*
dimensionless fractional grain size (-) (3.35)
The critical Shields parameter per fraction ("
cr,i
) is determined using (3.33) and
(3.35). The remaining parameters remain the same as for uniform formula.
3.12 Validity ranges of the fractional formulas
The validity ranges of the fractional formulas were presented in the accompanying
tables in the previous sections. Here, we will give a clearer view on how these validity
ranges compare with each other. Figures 3.4-a,b and 3.5-a,b show that there are
large differences in the validity ranges of the different fractional formulas and
hiding/exposure corrections.
The hiding/exposure correction by Day [1980] and the correction by Proffitt &
Sutherland [1983] were derived especially to transform the sediment transport
formula of Ackers & White [1973] in a fractional formula. These new fractional
formulas with the hiding/exposure corrections were calibrated and verified with new
data. Proffitt & Sutherland divide the data they used in two parts. The data from
Proffitt [1980] were used to derive the hiding/exposure correction and the rest of the
data were used to verify it. The validity ranges of Day and Proffitt & Sutherland can
Sediment transport formulas
University of Twente 45
be seen as validity ranges of the fractional formula of Ackers & White with the
hiding/exposure corrections of Day and Proffitt & Sutherland, respectively.
Both Day and Proffitt & Sutherland have large validity ranges for their formulas.
The fact that Proffitt & Sutherland used field data of gravel bed rivers for verification
of their formula makes it more suitable for predictions with large grain sizes and
large geometric standard deviations. Day, on the other hand, used data of fine
material, which makes it more suitable to material with median grain sizes smaller
than 1 mm.
The hiding/exposure corrections of Egiazaroff [1965] and Ashida & Michiue [1973]
were not developed for the sediment transport formula of Meyer-Peter & Mller
[1948], especially. Both corrections were initially derived for other sediment
transport formulas. Ribberink [1987] was the first who applied these correction
factors to the Meyer-Peter & Mller formula in a modified form. Ribberink calibrated
the fractional formula with experimental data using a bimodal sediment mixture.
These experiments were conducted in a narrow flume and thus the sidewall effects
probably influenced the calibration. This means that the validity ranges given in
Figures 3.4 and 3.5 should not be seen as the validity ranges of the fractional
formula of Meyer-Peter & Mller corrected with Egiazaroff or Ashida & Michiue.
Instead, they should be seen as the validity range of the original fractional formula of
Egiazaroff and the original formula given by Ashida & Michiue. It should also be
noted that we did not manage to collect all the data used by Egiazaroff and thus the
validity ranges are not complete.
The Shields parameter range for the Egiazaroff correction factor is large due to
the experimental data of Nizery and Braudeau [1953] that is taken from a small canal
with a slope of 0.1. Other data that Egiazaroff used are from Pantlopulos [1957],
who did experiments with both carbon grains and sand grains.
Parker [1990] and Hunziker [1995] both developed a fractional formula. This means
that the data they used for calibration and verification directly show the validity
range of their formulas. Parker used data only from rivers with coarse bed material
and a large geometric standard deviation. Hunziker used the data sets of Gnter
[1971] and Suzuki [1992] for calibration of his model and the other data of Table 3.7
for verification.
The data used by Parker et al. [1982] have a small Shields parameter range.
Hunziker uses data from armouring tests, which have a different composition for the
surface layer and the substrate. The original material of the substrate was used for
some experimental series since data on the armour layer were not available.
The formula of Wu et al. has the largest validity ranges of all the formulas analysed.
A large amount of flume and field data has been used for calibration of the formula.
Data with relatively fine material, e.g. experiments of Samaga et al. [2002], were
used as well as data of coarse gravel rivers, e.g. Sustina River. The validity ranges
only cover the data used to calibrate the bed load formula. Some of the data used for
the bed load formula has also been used for the suspended load formula, but we
were not able to acquire all calibration data of the suspended load formula. In the
present study only the bed load formula is tested and analysed.
Sediment transport formulas
46 University of Twente
Wilcock & Crowe have used experimental data of Wilcock & McArdell [1993] and
Wilcock et al. [2002] in the development of their formula. In all these experiments a
sand gravel mixture was used. The experiments were done at relatively low Shields
parameters and with coarse material. Although a lot of experiments were used, the
validity ranges are small since the data covers a narrow range.
It must be noted that the formula of Wilcock & Crowe used the bed surface
composition as the active layer composition, yet in figures 3.4-b and 3.5-b the initial
bed material is used. This was done to allow for better comparison with the other
formulas for which only the initial material composition of the data is known.

Sediment transport formulas
University of Twente 47

Sediment transport formulas
48 University of Twente
Figure 3.4-a: Validity ranges for several fractional formulas and hiding/exposure
corrections given by the median grain size (D
50
) of the bed material versus the
Shields parameters ("
50
).
Sediment transport formulas
University of Twente 49

Sediment transport formulas
50 University of Twente
Figure 3.4-b: Validity ranges for several fractional formulas and hiding/exposure
corrections given by the median grain size (D
50
) of the bed material versus the
Shields parameters ("
50
).

Sediment transport formulas
University of Twente 51

Sediment transport formulas
52 University of Twente
Figure 3.5-a: Validity ranges for several fractional formulas and hiding/exposure
corrections given by the median grain size (D
50
) of the bed material versus geometric
standard deviation (!
g
).

Sediment transport formulas
University of Twente 53

Sediment transport formulas
54 University of Twente
Figure 3.5-b: Validity ranges for several fractional formulas and hiding/exposure
corrections given the median grain size (D
50
) of the bed material versus geometric
standard deviation (!
g
).
Behaviour transport formulas
University of Twente 55
4 Behaviour transport formulas
4.1 Introduction
In this chapter, the sediment transport formulas described in Chapter 3 are analysed
regarding their behaviour under different hydraulic conditions and for different
sediment mixtures. The behaviour of the transport formulas is analysed by carrying
out four series of calculations. The predicted sediment transport rates and the
composition of the transported sediment are analysed.
In Section 4.2 the predicted sediment transport rates and sediment transport
compositions are analysed for the 12 fractional formulas to see the differences
between these formulas. The transport rate predictions of some fractional formulas
are compared with their uniform counterparts (Section 4.3) to find out at which
conditions fractional formulas give significant differences compared to their uniform
counterparts. Section 4.4 is focussed on the behaviour at low Shields parameters
with special attention for incipient motion.
Each of the four series has a different sediment composition. Both the median
grain size (D
50
) and the geometric standard deviation of the bed material (!
g
) have
been varied (Table 4.1).
Table 4.1: Bed material properties of the four series
!
g
(-)
D
50
(mm)
2 5
2 Series 1 Series 2
10 Series 3 Series 4
The sediment mixtures are divided into n fractions of which the grain sizes (D
i
) and
the probabilities (p
i
) are calculated. To do this, an assumption must be made for the
grain size distribution. After Ribberink [1987], it is assumed that the grain size
distribution is log-normal. Ribberinks method of dividing a sediment mixture with a
log-normal grain size distribution into size fractions is given in Appendix 2.
We divided the mixtures into ten grain size fractions. For each series, calculations
were made for the Shields parameter range [0 - 0.6]. The following hydraulic
parameters were constant in the calculations:
i
w
= 0.0005 water surface slope
w = 500 m width of hypothetical river
#
s
= 2650 kg/m
3
sediment density
1 = 0.0000012 m
2
/s viscosity
C
b
= 40 m

/s Chzy value
The width of the hypothetical river was assumed 500 m, so that side wall effects
could be neglected, and the hydraulic radius almost equals the water depth. The
Behaviour transport formulas
56 University of Twente
Chzy value is supposed constant at 40 m

/s. As the Shields parameter and the


median grain size are known, it is now possible to calculate the water depth as
function of a certain Shields parameter:
i
D
h
50 50
% ' '
=
"
water depth (m)
Now the discharge equals:
w
i h h C B Q ' ' ' ' = discharge (m
3
/s)
Now, all input parameters for the calculations are known. The results are shown in
two ways. The transport rate is shown as a function of the Shields parameter. The
predicted composition is shown as the ratio between the geometric mean grain size
of the transported material and the geometric mean grain size of the bed material
(D
m
T/D
m
B), as a function of the Shields parameter.
Not every transport formula is valid for the sediment mixtures used in the four
series. The median grain size and the geometric standard deviation of the bed
material are allowed to have a difference of 20% to increase the number of formulas
that are valid for a sediment mixture used in a series. The 80% - 120% validity ranges
of the two sedimentological parameters for each series are given in Table 4.2,
together with the formulas valid for this series.
Table 4.2: Input parameter ranges of the four model behaviour series and the
formulas valid in that series
D
50
range !
g
range Valid formulas
(mm) (-)
Series 1 1.6 - 2.4 1.6 - 2.4 A&W+D, A&W+P&S, MP&M+Eg, Hunziker, Wu et al.
Series 2 1.6 - 2.4 4.0 - 6.0 A&W+D, A&W+P&S, Hunziker, Wu et al.
Series 3 8.0 - 12 1.6 - 2.4 A&W+D, A&W+P&S, Wu et al.
Series 4 8.0 - 12 4.0 - 6.0 A&W+D, A&W+P&S, Wu et al., Wilcock.
The most important results are discussed in the following sections. The rest of the
results can be found in Appendices 3, 4, and 5.
4.2 Differences between formulas
In this section, the behaviour of the different transport formulas is discussed based
on the calculations of the four series. The formulas are compared for both the
transport rate and the composition of the transported material. The validity ranges
are taken into account in this comparison.
It should be noted that the formulas of Van Rijn and Engelund & Hansen have
never been officially derived for fractional sediment transport. This leads to
exceptionally high transport rates of the finest fractions because no hiding/exposure
correction was introduced.
4.2.1 Transport rate
In this section not all results of all series are given. The median grain size (D
50
) of the
bed material appeared to have little influence on the behaviour of the transport
Behaviour transport formulas
University of Twente 57
formulas. A variation of the geometric standard deviation of the bed material does
influence the transport rate. This is why Series 3 and 4 are chosen to be
representative for the behaviour of the transport formulas under different conditions.
The results for Series 1 and 2 are shown in Appendix 3.

0 0.1 0.2 0.3 0.4 0.5 0.6
10
-8
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
Serie 3: D
50
=10mm, ! !! !
g
=2
"
50
(-)
q
s

(
m
3
/
s
/
m
)
Ackers & White + Day
Ackers & White + Proffitt & Sutherland
Parker surface based
Hunziker/Meyer-Peter & Mller
Wilcock & Crowe
Wu et al.
0 0.1 0.2 0.3 0.4 0.5 0.6
10
-8
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
Serie 3: D
50
=10mm, ! !! !
g
=2
"
50
(-)
q
s

(
m
3
/
s
/
m
)
Engelund & Hansen fractional
Gladkow & Shngen
Meyer-Peter & Mller + Egiazaroff
Meyer-Peter & Mller + Ashida & Michiue
Van Rijn fractional
Ribberink + Ashida & Michiue

Figure 4.1: Series 3 (D
50
= 10 mm, !
g
= 2): Sediment transport rate versus the
Shields parameter for the fractional transport formulas.
From Figure 4.1 it can be seen that the transport formulas of Parker, Wilcock &
Crowe and Engelund & Hansen predict the highest transport rates. The differences
between these three formulas are small for low Shields parameters. Parker predicts a
much higher transport rate for Shields parameters higher than 0.1. It must be noted
that Parkers formula is not valid for all conditions in Series 1 through 4, since the
sediment mixtures are too fine. Nevertheless, Parkers results are comparable to
other formulas at Shields parameters smaller than 0.1.
The formulas of Ackers & White with Day, Ackers & White with Proffitt &
Sutherland, Meyer-Peter & Mller with Ashida & Michiue, Ackers & White with
Egiazaroff and Wu et al. show medium transport rates. There are no noticeable
differences between the two fractional formulas of Meyer-Peter & Mller, which
means that the modification of Ashida & Michiue on the hiding exposure correction
of Egiazaroff has little effect on mixtures with a small gradation. For lower Shields
parameters, Ackers & White with Day predicts a higher transport rate than Meyer-
Peter & Mller with Ashida & Michiue. For higher Shields parameters, the predictions
for the transport rate become almost equal. Note that the correction factor of Ashida
& Michiue was developed for sediment mixtures finer than in Series 1 through 4.
The formulas of Hunziker/Meyer-Peter & Mller, Gladkow & Shngen and the
fractional formula of Van Rijn predict the lowest transport rates. Only at low Shields
parameters predicts the formula of Van Rijn a larger transport rate than most
q
s

(
m
3
/
s
/
m
)

q
s

(
m
3
/
s
/
m
)

Behaviour transport formulas
58 University of Twente
formulas. This is mainly transport of fine fractions due to the lack of a
hiding/exposure correction in this formula. Hunzikers formula starts predicting
sediment transport at rather high Shields parameters. This is partly caused by its bed
form factor, which is smaller than one and partly because the formula uses the equal
mobility concept based on the Shields parameter of the geometric mean grain size
(D
m
). The geometric mean grain size (eq. 2.1) is larger than the median grain size
(D
50
). This results in lower Shields parameters.

0 0.1 0.2 0.3 0.4 0.5 0.6
10
-8
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
Serie 4: D
50
=10mm, ! !! !
g
=5
"
50
(-)
q
s

(
m
3
/
s
/
m
)
Ackers & White + Day
Ackers & White + Proffitt & Sutherland
Parker surface based
Hunziker/Meyer-Peter & Mller
Wilcock & Crowe
Wu et al.
0 0.1 0.2 0.3 0.4 0.5 0.6
10
-8
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
Serie 4: D
50
=10mm, ! !! !
g
=5
"
50
(-)
q
s

(
m
3
/
s
/
m
)
Engelund & Hansen fractional
Gladkow & Shngen
Meyer-Peter & Mller + Egiazaroff
Meyer-Peter & Mller + Ashida & Michiue
Van Rijn fractional
Ribberink + Ashida & Michiue

Figure 4.2: Series 4 (D
50
= 10 mm, !
g
= 5): Sediment transport rate versus the
Shields parameter for the fractional transport formulas.
Figure 4.2 shows the transport rate predictions for Series 4 (D
50
= 10 mm, !
g
= 5).
The two fractional formulas of Ackers & White show practically similar transport rate
predictions. These predictions are relatively high compared with the other formulas.
Wilcock & Crowe and Engelund & Hansen predict the highest transport rates. Parker
remarkably gives fluctuating transport rate predictions around the Shields parameter
of 0.1. The formula predicts small transport rates at low Shields parameters and
large rates at high Shields parameters, relative to other formulas. The fractional
formula of Van Rijn shows the opposite behaviour: large transport rates at low
Shields parameters and small transport rates at high Shields parameters.
The formula of Meyer-Peter & Mller with Ashida & Michiue predicts no
sediment transport at all for almost its whole validity range. The formula differs from
Egiazaroffs formula especially when sediment mixtures with a strong gradation are
used. When the hiding/exposure factor of Egiazaroff is used the formula does not
have problems with predicting sediment transport at low Shields parameters. For
conditions with higher Shields parameters, the transport rate is somewhat larger for
Ashida & Michiue than for Egiazaroff.
q
s

(
m
3
/
s
/
m
)

q
s

(
m
3
/
s
/
m
)

Behaviour transport formulas
University of Twente 59
The formulas of Hunziker/Meyer-Peter & Mller and Gladkow & Shngen
predict the smallest transport rates. The formula of Hunziker/Meyer-Peter & Mller
predicts no sediment transport until a Shields parameter of 0.19, which is partly
caused by the fact that the geometric mean grain size (eq. 2.1) is much larger than
the median grain size (D
50
). In the formula of Hunziker, the Shields parameter based
is on D
m
, which, in its turn, is much smaller than the Shields parameter based on D
50
.
Together with a relatively small bed form factor, this causes incipient motion at high
Shields parameters. After the critical Shields parameter is reached, the formula of
Hunziker shows a large increase in the predicted transport rate, but remains the
formula that predicts the smallest amount of sediment transport.
Although there are large differences between the formulas at high Shields
parameters, up to a factor 10, the biggest differences occur at the low Shields
parameter range. There are large differences in the transport rate and in predicting
incipient motion. Some formulas only start predicting sediment transport at high
Shields parameters, while other formulas predict sediment transport at all
conditions.
4.2.2 Transport composition
The composition of the transported material is represented by the ratio between the
geometric mean grain size of the transported material (D
m
T) and the geometric mean
grain size of the bed material (D
m
B):
&
&
'
'
=
i
i i
i
i i
m
m
D B p
D T p
B D
T D

p
i
T probability of fraction i in transported material
p
i
B probability of fraction i in bed material
In this section, again the results of Series 3 and 4 are shown to represent the
behaviour of the several transport formulas under different conditions. The results of
Series 1 and 2 are shown in Appendix 4. Only the fractional formulas of Ackers &
White show sensitivity towards the grain size of the bed material. The D
mT
/D
mB
ratio
becomes larger when the grain size of the bed material increases.

Behaviour transport formulas
60 University of Twente
0 0.1 0.2 0.3 0.4 0.5 0.6
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Serie 3: D
50
=10mm, ! !! !
g
=2
"
50
(-)
Ackers & White + Day
Ackers & White + Proffitt & Sutherland
Parker surface based
Hunziker/Meyer-Peter & Mller
Wilcock & Crowe
Wu et al.
0 0.1 0.2 0.3 0.4 0.5 0.6
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Serie 3: D
50
=10mm, ! !! !
g
=2
"
50
(-)
Engelund & Hansen fractional
Gladkow & Shngen
Meyer-Peter & Mller + Egiazaroff
Meyer-Peter & Mller + Ashida & Michiue
Van Rijn fractional
Ribberink + Ashida & Michiue

Figure 4.3: Series 3 (D
50
= 10 mm, !
g
= 2): Ratio of geometric mean grain size of
transported material and geometric mean grain size of bed material versus the
Shields parameters for the fractional transport formulas.
Most formulas show an asymptotic behaviour of the predicted composition of the
transported material for Shields parameters of about 0.6 (Figure 4.3). Only Van Rijn
still shows a coarsening transport composition at these Shields parameters.
The formulas of Hunziker/Meyer-Peter & Mller, Wu et al. and the two versions
of Ackers & White predict a transport coarser than the bed material at high Shields
parameters.
The formulas of Parker, Wilcock & Crowe, Gladkow & Shngen and both
formulas of Meyer-Peter & Mller predict a transport composition equal to the bed
material composition. This means that the formulas predict equal mobility at
conditions with high enough Shields parameters. There is no difference between the
different formulas of Meyer-Peter & Mller.
The formulas of Engelund & Hansen and Van Rijn show the finest transport
composition. The formula of Engelund & Hansen does not show differences in the
transport composition at different hydraulic conditions. This is caused by the fact
that there is no form of critical shear stress or hiding/exposure factor incorporated
within the fractional formula. Van Rijns formula also predicts very fine sediment
transport due to a lack of a hiding/exposure factor. Both formulas of Engelund &
Hansen and Van Rijn were not officially derived for fractional sediment transport.

D
m
T
/
D
m
B

(
-
)

D
m
T
/
D
m
B

(
-
)

Behaviour transport formulas
University of Twente 61
0 0.1 0.2 0.3 0.4 0.5 0.6
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Serie 4: D
50
=10mm, ! !! !
g
=5
"
50
(-)
Ackers & White + Day
Ackers & White + Proffitt & Sutherland
Parker surface based
Hunziker/Meyer-Peter & Mller
Wilcock & Crowe
Wu et al.
0 0.1 0.2 0.3 0.4 0.5 0.6
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Serie 4: D
50
=10mm, ! !! !
g
=5
"
50
(-)
Engelund & Hansen fractional
Gladkow & Shngen
Meyer-Peter & Mller + Egiazaroff
Meyer-Peter & Mller + Ashida & Michiue
Van Rijn fractional
Ribberink + Ashida & Michiue

Figure 4.4: Series 4 (D
50
= 10 mm, !
g
= 5): Ratio of geometric mean grain size of
transported material and geometric mean grain size of bed material versus the
Shields parameters for the fractional transport formulas.
The series with the larger geometric standard deviation (Series 2 and 4, !
g
=5) show a
larger variation in the predicted transport composition than Series 1 and 3 (see
Appendix 4 for Series 1 and 2). The asymptotic behaviour for !
g
=2 (Series 1 and 3) is
not shown for !
g
=5 (Series 2 and 4). The asymptotic behaviour occurs at higher
Shields parameters since only then the coarsest fractions will start being transported.
Hunziker/Meyer-Peter & Mller predicts a composition of the transported
material as coarse as or coarser than the bed material (Figure 4.4). This is caused by
Hunzikers assumption of equal mobility, combined with his hiding/exposure factor.
The equal mobility concept enables the grain sizes much larger than D
m
to be
transported, while other formulas predict no transport of these large grains.
The formula of Ackers & White with Day predicts the second coarsest transport
composition. The formula of Gladkow & Shngen shows a fast coarsening for low
Shields parameters and asymptotic behaviour for Shields parameters larger than 0.3.
The formula of Parker shows strange behaviour again at Shields parameters around
0.1. The formulas of Engelund & Hansen and Van Rijn show the finest sediment
transport. Engelund & Hansen shows again the same composition for all hydrological
conditions.
The two hiding/exposure factors used for the Meyer-Peter & Mller (Egiazaroff
and Ashida & Michiue) give different results for sediment mixtures with a large
gradation. In these mixtures, the small fractions are sufficiently smaller than the
mean grain size, so that the modification by Ashida & Michiue has effect. The
formula with Egiazaroff shows a very fine transport at the lowest Shields parameters.
This is caused by the fact that if the ratio D
i
/D
m
is smaller than 1/(19
2
) (see eq.
D
m
T
/
D
m
B

(
-
)

D
m
T
/
D
m
B

(
-
)

Behaviour transport formulas
62 University of Twente
3.13), fraction i suffers exposure instead of hiding. This is caused by the
asymptotic behaviour of the correction factor at grain sizes 19 times smaller than the
mean grain size.
4.3 Uniform versus fractional
In this section, the transport rates predicted by the uniform formulas are compared
with those by their fractional counterparts. The goal of this comparison is to
determine for which hydraulic and sedimentological conditions a fractional formula
shows different predictions compared to the uniform formula, and hence it would be
preferred to use the fractional formula. Only the uniform formulas of Ackers & White
and Meyer-Peter & Mller are compared with their fractional versions. The uniform
formula of Parker is not taken into account in this report, as mentioned in Section
3.3. Therefore, a comparison cannot be made. Engelund & Hansen and Van Rijn did
not derive a fractional version of their formulas. The fractional formulas of Engelund
& Hansen and Van Rijn given in this report have not officially been tested and, as
such, they are not taken into account in this section.
An analysis shows that the differences between uniform and fractional
formulas are not sensitive to a change in the median grain size (D
50
) of the bed
material, besides a higher transport rate for predictions with a larger median grain
size. This is because the transport rate is shown as a function of the Shields
parameter in which the grain size is incorporated. A variation in the geometric
standard deviation in the bed material does show significant differences in the
transport predictions. This is the reason why again only Series 3 and 4 are shown
whereas, Series 1 and 2 are shown in Appendix 5. The results are shown as the ratio
between the fractional formula and its uniform counterpart versus the Shields
parameter.
Figure 4.5 shows the ratio of fractional transport rates and uniform transport rates
versus the Shields parameter in Series 3 (!
g
= 2) for four selected transport formulas.
The formula of Ackers & White with Day predicts a much smaller transport rate than
the uniform formula at low Shields parameters, but the transport rate predictions
approach each other at high Shields parameters. The ratio of Proffitt & Sutherland
with the uniform formula increases and then somewhat decreases for Shields
parameters up to 0.14. At higher Shields parameters, the formula of Proffitt &
Sutherland predicts a transport rate, which is a factor 1.5 higher than the uniform
formula of Ackers & White.
The uniform and fractional formulas of Meyer-Peter & Mller predict the same
sediment transport rate at Shields parameters higher than 0.15. The uniform formula
predicts a lower transport rate than the fractional formula at Shields parameters
smaller than 0.15.
There are only large differences between the uniform formulas and the
fractional formulas at the lower Shields parameter range. Most fractional formulas
converge with their uniform counterparts at Shields parameters higher than 0.2.

Behaviour transport formulas
University of Twente 63

Figure 4.5: Series 3 (D
50
= 10 mm, !
g
= 2): Ratio of fractional transport rates with
uniform transport rates versus the Shields parameters for four transport formulas.
Figure 4.6 shows the predicted transport rate ratios versus the Shields parameter in
Series 4 for four combinations of uniform and fractional formulas. At low Shields
parameters, the uniform formula of Ackers & White predicts a higher transport rate
than the fractional formulas. At Shields parameters higher than 0.2 the different
uniform and fractional formulas converge (ratio ! 1).
For Shields parameters smaller than 0.25, the uniform formula of Meyer-Peter
& Mller predicts a much smaller transport rate than the fractional formulas. At
higher Shields parameters, the uniform formula shows more agreement with the
fractional formula of Ashida & Michiue than the one of Egiazaroff.
It can be concluded that the use of fractional formulas is especially important
at low Shields parameters. The higher the gradation of a sediment mixture, the more
a prediction of a fractional formula differs from its uniform counterpart.

Series 3
0.1
1.0
10.0
0.00 0.10 0.20 0.30 0.40 0.50 0.60
Shields D50 (-)
q
s
_
f
r
/
q
s
_
u

(
-
)
Ackers & White + Day
Ackers & White + Proffitt & Sutherland
Meyer-Peter & Muller + Egiazaroff
Meyer-Peter & Muller + Ashida & Michiue
Behaviour transport formulas
64 University of Twente

Figure 4.6: Series 4 (D
50
= 10 mm, !
g
= 5): Ratio of fractional transport rates with
uniform transport rates versus the Shields parameters for four transport formulas.
4.4 Focus on low Shields parameters
In lower Shields parameter ranges (0 - 0.2), the transport formulas show very
different behaviour. The problem is that incipient motion is difficult to define and to
predict. This is why different ways of modelling incipient motion have been
developed, which causes the formulas to differ a lot from each other. In this section,
a short analysis will be given concerning the differences between the uniform
formulas of Ackers & White and Meyer-Peter & Mller. For the formula of Meyer-
Peter & Mller, incipient motion occurs at high Shields parameters. Calculations with
increasing Shields parameters (steps of 0.01) have been done to determine the
actual critical Shields parameter in different sediment mixtures. The actual critical
Shields parameter is the lowest Shields parameter based on the median grain size
(D
50
) at which sediment is transported. Table 4.3 shows the three formulas and their
actual critical Shields parameters (based on D
50
) for the four Series.
Table 4.3: Actual critical Shields parameter (D
50
) of different formulas in the four
series.
Name formula Actual critical Shields parameter (-)
Series 1 Series 2 Series 3 Series 4
Ackers & White 0.05 0.04 0.05 0.04
Meyer-Peter & Mller 0.10 0.19 0.10 0.19
Meyer-Peter & Mller (
MPM
= 1) 0.06 0.12 0.06 0.12
The uniform formula of Meyer-Peter & Mller has large critical Shields parameters,
especially for Series 2 and 4 (!
g
= 5). An important reason for this is the fact that the
bed form factor is already smaller than one at incipient motion. Theoretically, this is
Series 4
0.1
1.0
10.0
0.00 0.10 0.20 0.30 0.40 0.50 0.60
Shields D50 (-)
q
s
_
f
r
/
q
s
_
u

(
-
)
Ackers & White + Day
Ackers & White + Proffitt & Sutherland
Meyer-Peter & Muller + Egiazaroff
Meyer-Peter & Muller + Ashida & Michiue
Behaviour transport formulas
University of Twente 65
not possible since at conditions of incipient motion no bed forms are formed, so the
bed form factor should equal one. It can be that bed forms were formed during
previous flow conditions and are still present. These situations cannot be predicted
by a transport formula since these cannot deal with historical effects. A
morphological model is needed if sediment transport predictions are needed in these
conditions. The effect of a bed form factor below one can be seen by comparing the
predictions of the original Meyer-Peter & Mller formula with the Meyer-Peter &
Mller with a constant bed form factor of one (Figure 4.7 and 4.8).
The second reason why the actual critical Shields parameter differs from 0.047 is
that these transport formulas use a Shields parameter based on the geometric mean
grain size (D
m
) instead of the median grain size (D
50
), which is used in Table 4.3.
Since in the behaviour calculations grain size of the bed material is assumed to be
log-normal distributed, the D
m
can be expressed in terms of D
50
and !
g
:
) . exp(
2
y y m
5 0 D ! + = geometric mean grain size (m) (4.1)
In which:
) ln(
50 y
D =
995 0
g
y
.
) ln(!
! =
Equation 4.1 can be written as:
) . exp(
2
y 50 m
5 0 D D ! ' =
The geometric standard deviation (!
g
) cannot be smaller than one. This means that
the geometric mean grain size (D
m
) cannot be smaller than the median grain size
(D
50
). It must be noted that due to splitting the log-normal distribution into fractions
the geometric mean grain size according to (4.1) does not equal the geometric mean
grain size according to (4.2). For more information see Appendix 2.
&
' =
i
i i m
D p D geometric mean grain size (m) (4.2)
In Table 4.4 the ratio between D
m
and D
50
are given, where D
m
is calculated with
different formulas.
Table 4.4: Ratio between geometric mean grain size (D
m
) and median grain size (D
50
)
for different geometric standard deviations (!
g
) according to (4.1) and (4.2).
!
g
D
m
/D
50
based
on Eq. (4.1)
D
m
/D
50
based
on Eq. (4.2)
2 1.275 1.202
5 3.699 2.527
In this report, (4.2) is used to determine the geometric mean grain size (D
m
). The
uniform formula of Meyer-Peter & Mller uses Shields parameters based on the
geometric mean grain size instead of the median grain size as used in the plots in
this report. Since the Shields parameter is inversely proportional to the grain size,
Behaviour transport formulas
66 University of Twente
the transport formulas show high Shields parameters for conditions of incipient
motion.
The transport rates of three formulas: Ackers & White, Meyer-Peter & Mller
and Meyer-Peter & Mller with bed form factor equal to one for Series 3 and 4 are
shown in Figure 4.7 and 4.8 respectively. The effect of the bed form factor appears
from the differences between the two different Meyer-Peter & Mller formulas. The
differences caused by the difference in geometric standard deviation can be seen by
comparing Figure 4.7 with Figure 4.8.
The transport rates of the Meyer-Peter & Mller formula with the bed form
factor equal to one is only shown at its conditions of incipient motion, because after
sediment is transported, the bed form factor will theoretically be smaller than one
because bed forms will be formed.

Series 3
1.00E-07
1.00E-06
1.00E-05
1.00E-04
1.00E-03
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20
Shields_D50 (-)
q
s

(
m
2
/
s
)
Ackers & White uniform
Meyer-Peter & Muller uniform
Meyer-Peter & Muller uniform (bed form factor = 1)

Figure 4.7: Series 3 (D
50
= 10 mm, !
g
= 2): Sediment transport rate versus the
Shields parameter for the three selected transport formulas at low Shields
parameters.

Behaviour transport formulas
University of Twente 67
Series 4
1.00E-07
1.00E-06
1.00E-05
1.00E-04
1.00E-03
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20
Shields_D50 (-)
q
s

(
m
2
/
s
)
Ackers & White uniform
Meyer-Peter & Muller uniform
Meyer-Peter & Muller uniform (bed form factor = 1)

Figure 4.8: Series 4 (D
50
= 10 mm, !
g
= 5): Sediment transport rate versus the
Shields parameter for the three selected transport formulas at low Shields
parameters.
The problems with modelling incipient motion also occur in the formula of
Hunziker/Meyer-Peter & Mller. The bed form factor of this formula is also not equal
to one at incipient motion. According to Hunziker [1995], the bed form factor of
Yalin and Scheuerlein [1988] is closer to one at conditions of incipient motion. The
use of Yalin and Scheuerlein leads to better predictions than when the bed form
factor of Stickler is used, but even if the bed form factor is equal to one at conditions
of incipient motion, the formula predicts incipient motion at high Shields
parameters. This is an important problem in the formulas of Meyer-Peter & Mller
and Hunziker/ Meyer-Peter & Mller at these conditions. The formula of Hunziker
also uses the geometric mean grain size (D
m
) for its Shields parameter, which leads
to problems in predicting sediment transport with strongly graded sediment
mixtures.
4.5 Conclusions
4.5.1 Transport rate
The transport formulas of Parker [1990] and Wilcock & Crowe [2001] and the
fractional formula of Engelund & Hansen [1967] predict the highest transport rates.
The formulas of Hunziker/Meyer-Peter & Mller [1995], Gladkow & Shngen [2000]
predict low transport rates. The formula of Hunziker predicts lower transport rates
than the different formulas of Meyer-Peter & Mller (Egiazaroff [1965] and Ashida &
Michiue [1973]). The formulas of Ackers & White [1973] with the hiding/exposure
correction of Proffitt & Sutherland [1983] and of Van Rijn [1984a] show relatively
high transport rates for low Shields parameters. Both formulas predict a slowly
Behaviour transport formulas
68 University of Twente
increasing transport rate for higher Shields parameters. Unlike the formulas of
Meyer-Peter & Mller, these show small sediment transport rates at low Shields
parameters and a fast increasing transport rates for higher Shields parameters.
Defining and predicting incipient motion is difficult, which leaded to
differences in modelling incipient motion between the formulas. That is why the
differences between the formulas are larger for small Shields parameters ("
50
< 0.3).
Another reason is the use of different hiding/exposure factors. The differences in the
transport rate predictions become even larger when the geometric standard
deviation of the bed material is larger. This is due to larger differences in incipient
motion when dealing with a widely graded bed mixture instead of dealing with a less
graded mixture.
The influence of the grain size on the behaviour of the different sediment
transport formulas is small. The transport rate increases for calculations with a larger
median grain size of the bed material. This is because the transport rate is shown as
a function of the Shields parameter. The Shields parameter in itself is a function of
the grain size.
Variations in the geometric standard deviation of the grain size result in
different behaviour for the different transport formulas. Engelund & Hansen as well
as Van Rijn show an increase in the transport rate when the standard deviation is
increased, whereas the formulas of Parker and Hunziker/Meyer-Peter & Mller show
a decrease in the transport rate. The transport rate predictions of the other formulas
show little sensitivity concerning the geometric standard deviation of the bed
material.
The largest differences between the uniform and the fractional versions of the
formulas occur in the Shields parameter range up to 0.25. Most uniform and
fractional formulas have converged at this point. The differences between the
uniform and fractional formulas become larger when the geometric standard
deviation of the bed material increases.
The bed form factors in the formulas of Meyer-Peter & Mller and
Hunziker/Meyer-Peter & Mller are an important reason why these formulas show
incipient motion at high Shields parameters. The problem is that the bed form factor
is not equal to one at conditions of incipient motion. Another reason for incipient
motion at high Shields parameters is the fact that the formulas use Shields
parameters based on the geometric mean grain size (D
m
, eq. 2.1) instead of median
grain size (D
50
). The geometric mean grain size is larger than the median grain size
when the geometric standard deviation of a mixture is larger (eq. 4.1).
4.5.2 Transport composition
All formulas show a transport composition finer than the bed composition (D
mT
<
D
mB
) for small Shields parameters. This is caused by the fact that the critical Shields
parameter of the larger grain sizes is not reached so that only the fine material is
transported. The composition of the transported sediment coarsens gradually when
the Shields parameter is increased. For some formulas the composition of the
transported material approximates the composition of the bed for high Shields
parameters. The formulas of Hunziker and Wu et al. show different behaviour: they
predict a sediment transport coarser than the bed material at high Shields
Behaviour transport formulas
University of Twente 69
parameters. This effect is strongest for Hunzikers formula for sediment mixtures
with a strong gradation. At which Shields parameter the formulas predict a transport
composition coarser than the bed material depends on the geometric standard
deviation of the material.
The transport composition predicted by the formulas differs most at low
Shields parameters ("
50
< 0.3). This is caused by the differences in modelling
incipient motion by the different formulas.
Both fractional Ackers & White formulas show sensitivity towards the grain size
of the bed material. The D
mT
/D
mB
ratio becomes larger when the grain size of the bed
material increases. Other formulas do not show changes in the ratio when this grain
size is varied.
The predictions of the transport composition by all formulas are sensitive to
changes in the geometric standard deviation of the bed material. All formulas predict
a finer composition when the geometric standard deviation of the bed material
increases. The asymptotic behaviour of the D
mT
/D
mB
ratio is also smaller in most
cases. The formula of Hunziker/Meyer-Peter & Mller predicts a much coarser
sediment transport when the Shields parameter exceeds 0.3.
Experimental data used for verification
University of Twente 71
5 Experimental data used for verification
5.1 Introduction
In this chapter, ten sets of experiments will be described. In Chapter 6 and 7,
predictions of transport rate and transport composition will be verified with,
respectively, the measured transport rate and transport composition of these
experiments. The experimental data that are analysed in this chapter are:
Blom and Kleinhans, [1999], Non-uniform sediment in morphological
equilibrium situations - Data report Sand Flume experiments 97/98, Research
report CiT 99R-002/MICS-001, Civil Engineering and Management, University
of Twente.
Experimental Series 1 was conducted in the Sand Flume Facility of WL | delft
hydraulics in 1997/1998. Five tests were performed with a sediment mixture
from the river Rhine.
Blom, [2000], Flume experiments with a trimodal sediment mixture Data
report Sand Flume experiments 99/00, Research report CiT: 2000R-004/MICS-
013, Civil Engineering & Management, University of Twente.
Experimental Series 2 was conducted in the Sand Flume Facility of WL | delft
hydraulics in 1999/2000. Four tests were performed with a tri-modal mixture.
Klaassen, [1991], Experiments on the effect of gradation and vertical sorting on
sediment transport phenomena in the dune phase, Proc. Grain Sorting Seminar,
Ascona, Switzerland, pp 127-145.
Six tests with graded sediment were conducted in the Sand Flume Facility of WL
| delft hydraulics in 1990.
Day, [1980], A study of the transport of graded sediments, HRS Wallingford,
Report No. IT 190.
HRS A, experiments conducted by Day. 11 tests with graded sediment were
conducted in the flume of HR Wallingford.
HRS B, experiments conducted by Day. Nine tests with graded sediment were
conducted in the flume of HR Wallingford.
Wilcock and McArdell, [1993], Surface-Based Fractional Transport Rates:
Mobilization Thresholds and Partial Transport of a Sand-Gravel Sediment, Water
Resources Research, Vol. 29, No. 4, pp. 1297-1312.
Bed Of Many Colours (BOMC): Ten tests with a sand gravel mixture. The tests
were carried out at Johns Hopkins University.
Wilcock et al., [2001], Experimental Study of the Transport of Mixed Sand and
Gravel, Water Resources Res., Vol. 37, No. 12, pp. 3349-3358.
Four experimental series with different sand gravel mixtures were conducted at
Johns Hopkins University. 37 runs are used for verification in this report.
Experimental data used for verification
72 University of Twente
5.2 Blom & Kleinhans
A set of flume experiments was carried out in the Sand Flume Facility of WL | delft
hydraulics from November 1997 until January 1998. Blom & Kleinhans [1999] give a
description of the experimental set-up, the measurements and the resulting data.
Sediment was recirculated and uniform flow conditions were maintained in five tests
until a morphological equilibrium was reached. Morphological equilibrium was
defined as: water surface slope equal to the bed slope, constant bed roughness,
constant sediment transport rate and composition, and constant and uniform bed
form dimensions [Blom & Kleinhans, 1999]. Only the equilibrium conditions are
analysed in the present study.
The Sand Flume of WL | delft hydraulics has a width of 1.5 m and a sediment
recirculation system. The D
50
of the initial bed mixture was 1.28 mm and !
g
was
4.28. The geometric mean grain size of the initial bed mixture is 1.68 mm. The
sediment density is 2620 kg/m
3
. All tests were done in the dune phase, except in T1
where longitudinal stripes moved over an armour layer. The conditions measured in
the equilibrium phases of the tests are shown in Table 5.1. The data are corrected
for side wall roughness by the method of Vanoni and Brooks [1956]. The method of
Vanoni and Brooks is given in Appendix 6.
Table 5.1: Equilibrium conditions in experimental series of Blom & Kleinhans
Test Q h R
b
C
b
visc. "
50
q
s
D
mT

m
3
/s m m m
1/2
/s *10
-6
m
2
/s - m
2
/s mm
T1 0.130 0.201 0.177 48.60 1.20 0.038 4.73 * 10
-7
1.21
T5 0.254 0.245 0.224 37.18 1.20 0.166 3.15 * 10
-5
1.23
T7 0.419 0.354 0.316 35.84 1.20 0.233 5.46 * 10
-5
1.69
T9 0.272 0.260 0.240 34.06 1.20 0.202 3.77 * 10
-5
1.52
T10 0.170 0.193 0.176 41.43 1.20 0.097 1.15 * 10
-5
1.21
The sediment mixture was divided into 12 fractions. The size (D
i
) and the probability
(p
i
) of these fractions are given in Table 5.2.
Table 5.2: Composition of the initial mixture in experimental series of Blom &
Kleinhans
fraction 1 2 3 4 5 6 7 8 9 10 11 12
D
i
(mm) 0.27 0.33 0.39 0.46 0.55 0.65 0.84 1.18 1.67 2.83 5.66 11.3
p
i
(%) 3 9 13 15 13 10 7 6 5 6 6 6
5.3 Blom
Blom [2000] conducted a second set of experiments in the Sand Flume facility of WL |
delft hydraulics. The experimental series can be divided into two test series (A and B)
of each two tests. In Test series A, the initial bed material consisted of a trimodal
sediment mixture with a probability ratio of 1:1:1 of the three size fractions in the
bed. The initial bed of test series B consisted of two layers: a coarse surface layer
and a fine substratum. The surface layer of 3 cm had the same composition as the
initial bed material of test series A and the substratum consisted of only the finest of
the three fractions. Only the data that were gathered in the equilibrium phases of the
Experimental data used for verification
University of Twente 73
tests were used in this study. The tests were maintained until an equilibrium was
reached in: average bed level, water level, sediment transport rate, sediment
transport composition, bed form dimensions and a uniform patterns of bed forms
[Blom, 2000] (Table 5.3). The width of the Sand Flume was reduced to 1 m. A
sediment recirculation system was used.
Table 5.3: Equilibrium conditions in experimental series of Blom
Test Q h R
b
C
b
visc. "
50
q
s
D
mT

m
3
/s m m m
1/2
/s *10
-6
m
2
/s - m
2
/s mm
A1 0.098 0.154 0.140 37.8 1.20 0.082 1.13 * 10
-5
2.25
A2 0.267 0.321 0.271 38.3 1.20 0.137 4.33 * 10
-5
2.41
B1 0.098 0.155 0.140 38.7 1.21 0.077 1.17 * 10
-5
2.06
B2 0.267 0.389 0.351 25.0 1.21 0.626 4.45 * 10
-5
1.01
The initial bed mixture of test series A had a D
50
of 2.1 mm and a !
g
of 2.93. The
geometric mean grain size of the initial mixture was 2.83 mm. The properties of the
sediment mixture used as surface layer in test series B were the same as those of the
initial mixture in test series A. The substratum in test series B had a D
50
of 0.68 mm
and a !
g
of 1.26. The geometric mean grain size of the substratum was 0.71 mm
(Table 5.4). The composition of the bed surface layer is also given in this table.
Table 5.4: Composition of the initial mixture and surface layer in experimental series
of Blom
fraction 1 2 3
D
i
(mm) 0.68 2.1 5.7
p
i
(%) initial material A1, A2 and B1 33.3 33.3 33.3
p
i
(%) initial material B2 80 10 10
p
i
(%) surface material A1 49 44 7
p
i
(%) surface material A2 36 38 26
p
i
(%) surface material B1 52 39 9
p
i
(%) surface material B2 89 6 5
5.4 Klaassen
Klaassen [1990] conducted experiments with graded sediment in the Sand Flume
facility of WL | delft hydraulics. The experiments consisted of six tests. In this study,
only the equilibrium periods of the tests are analysed.
The width of the Sand Flume was reduced to 1.125 m and a sediment recirculation
system was used. The D
50
of the initial bed mixture was 0.66 mm and !
g
was 2.34.
The geometric mean grain size of the initial mixture was 0.93 mm. The sediment
density was assumed to be 2650 kg/m
3
. All tests were done in the dune phase and
no armour layers were formed. The conditions in the equilibrium periods of the tests
are given in Table 5.5. The side wall roughness correction of Vanoni and Brooks
[1956] has been applied to the data.
Table 5.5: Equilibrium conditions in experimental series of Klaassen
Test Q h R
b
C
b
visc. "
50
q
s
D
mT

m
3
/s m m m
1/2
/s *10
-6
m
2
/s - m
2
/s mm
T49 0.110 0.178 0.164 33.8 1.07 0.242 7.92 * 10
-6
0.849
T50 0.240 0.337 0.305 28.6 1.06 0.451 1.59 * 10
-5
0.710
T51 0.300 0.402 0.361 27.4 1.06 0.537 1.82 * 10
-5
0.738
Experimental data used for verification
74 University of Twente
T52 0.241 0.349 0.317 27.2 1.06 0.469 1.36 * 10
-5
0.861
T53 0.111 0.189 0.175 31.6 1.06 0.251 6.45 * 10
-6
0.761
T54 0.050 0.091 0.085 41.9 1.05 0.125 2.80 * 10
-6
0.709
The composition of the initial material has been used as input for the transport
formulas. The initial bed material has been divided into 14 fractions (Table 5.6).



Table 5.6: Composition of the initial mixture in experimental series of Klaassen
fraction 1 2 3 4 5 6 7
D
i
(mm) 0.07 0.1 0.15 0.19 0.25 0.36 0.52
p
i
(%) 0.15 0.11 0.54 2.7 8.5 18 16

fraction 8 9 10 11 12 13 14
D
i
(mm) 0.74 0.95 1.22 1.6 2.4 3.4 4.3
p
i
(%) 15 8.0 13 6.0 7.5 3.4 1.1
5.5 Day: HRS A
Day [1980] conducted his experiments in a 2.46 m wide tilting flume with a sediment
recirculation system. These experiments were conducted at the HR Wallingford and
consisted of 11 tests. After each test, the top few centimetres of the bed were
replaced by initial bed material and the bed was levelled. The standard deviations of
the measured parameters were sufficiently small, so that it can be assumed that the
tests were in equilibrium. No data is available concerning the presence of bed forms.
The experiments were aimed at studying incipient motion of different grain sizes, so
the tests were done in low Shields parameter ranges. The D
50
of the initial mixture
was 1.75 mm and !
g
was 4.21. The geometric mean grain size of the mixture was
2.72 mm. Day did not give a sediment density, and we assume it to be 2650 kg/m
3
.
The conditions in the tests are given in Table 5.7. No side wall roughness correction
was needed for these experiments due to the fact that the width/depth ratio in these
experiments was larger than five.
Table 5.7: Equilibrium conditions in experimental series of Day: HRS A
Test Q h R
b
C
b
visc. "
50
q
s
D
mT

m
3
/s m m m
1/2
/s * 10
-6
m
2
/s - m
2
/s mm
1 0.199 0.166 0.146 48.9 1.24 0.034 4.36 * 10
-7
0.575
2 0.199 0.159 0.141 48.9 1.24 0.038 5.91 * 10
-7
0.586
3 0.199 0.145 0.130 55.1 1.24 0.036 5.43 * 10
-7
0.763
4 0.197 0.169 0.149 47.9 1.24 0.034 2.68 * 10
-7
0.740
5 0.196 0.147 0.131 50.1 1.24 0.040 7.42 * 10
-7
0.927
6 0.197 0.133 0.120 52.6 1.24 0.045 9.07 * 10
-7
1.128
7 0.197 0.123 0.112 47.2 1.24 0.066 2.95 * 10
-6
1.622
8 0.197 0.131 0.118 45.4 1.24 0.063 2.52 * 10
-6
1.659
9 0.196 0.121 0.110 39.8 1.24 0.095 4.71 * 10
-6
2.083
10 0.198 0.112 0.103 41.7 1.24 0.103 1.32 * 10
-5
2.374
11 0.193 0.107 0.098 38.6 1.24 0.124 2.47 * 10
-5
2.626
Experimental data used for verification
University of Twente 75
The composition of the active layer was not determined. Instead, the initial bed
composition was used as input for the bed material for the sediment transport
formulas. Day divided the initial mixture into 19 fractions (Table 5.8).
Table 5.8: Composition of the initial mixture in experimental series of Day: HRS A
fraction 1 2 3 4 5 6 7 8 9 10
D
i
(mm) 0.15 0.21 0.3 0.39 0.46 0.55 0.65 0.78 0.93 1.2
p
i
(%) 1 4 11 6 6 4 3 3 2 5

fraction 11 12 13 14 15 16 17 18 19
D
i
(mm) 1.55 2.03 2.86 4.06 5.56 7.18 8.73 11.1 14.2
p
i
(%) 3 7 10 10 11 7 3 2 0
5.6 Day: HRS B
Day [1980] conducted a second set of experiments (HRS B), which were performed in
the same flume facility as series HRS A. Experiments HRS B consisted of nine tests.
After each test, the top few centimetres of the bed were replaced and the bed was
levelled, like in series HRS A. The conditions at the end of the tests are again
assumed to be in equilibrium.
The experiments were aimed at studying incipient motion. In HRS B the initial bed
mixture was finer and less graded than the mixture used in HRS A. The D
50
of the
sediment mixture was 1.55 mm and !
g
was 3.5. The geometric mean grain size was
1.74 mm. The density of the mixture was again assumed to be 2650 kg/m
3
. The
conditions of the nine tests are given in Table 5.9.
Table 5.9: Equilibrium conditions in experimental series of Day: HRS A
Test Q h R
b
C
b
visc. "
50
q
s
D
mT

m
3
/s m m m
1/2
/s *10
-6
m
2
/s - m
2
/s mm
1 0.203 0.189 0.164 51.1 1.24 0.029 5.04 * 10
-8
1.34
2 0.210 0.184 0.160 54.9 1.24 0.028 9.60 * 10
-8
1.08
3 0.202 0.162 0.143 48.2 1.22 0.043 6.08 * 10
-7
1.17
4 0.202 0.154 0.137 35.9 1.24 0.087 9.53 * 10
-7
1.20
5 0.200 0.145 0.130 37.0 1.24 0.090 2.31 * 10
-6
1.41
6 0.202 0.119 0.109 44.0 1.24 0.096 2.32 * 10
-5
1.62
7 0.204 0.124 0.113 42.6 1.24 0.097 2.11 * 10
-5
1.77
8 0.200 0.117 0.107 41.6 1.21 0.109 2.56 * 10
-5
1.81
9 0.201 0.115 0.105 40.1 1.24 0.123 3.36 * 10
-5
1.83
Too little data is available to determine the composition of the active layer for the
tests. Instead, the composition of the initial bed material is used for the sediment
transport predictions. Day divided the mixture into 16 fractions (Table 5.10).
Table 5.10: Composition of the initial mixture in experimental series of Day: HRS B
fraction 1 2 3 4 5 6 7 8
D
i
(mm) 0.15 0.21 0.28 0.33 0.39 0.46 0.55 0.65
p
i
(%) 2 5 6 6 5 4 3 2

fraction 9 10 11 12 13 14 15 16
D
i
(mm) 0.78 0.93 1.2 1.55 2.03 2.86 4.06 5.56
p
i
(%) 3 3 7 5 17 14 15 1
Experimental data used for verification
76 University of Twente
5.7 Wilcock & McArdell and Wilcock et al.
Wilcock & McArdell [1993] and Wilcock et al. [2001] conducted five sets of
experiments with graded sediments. The 48 tests were done in a tilting flume with
different recirculating systems for water and sediment. Gravel coarser than 16 mm
was recirculated manually. The flume had a width of 60 cm and a working length of
7.9 m. In this report the data from Wilcock et al. has been used for verification
purposes. For each of the five sets of experiments a different sediment mixture of
sand and gravel has been used. The mixtures in the five sets of experiments (J06,
J14, J21, J27 and BOMC) contained 6, 15, 21, 27 and 34 percent sand, respectively.
For the five mixtures, the same gravel has been used. Only the sand in the series
with the 34% sand mixture had a different composition than the sand of the other
series. The bed was remixed and screeded after two or three tests. In case the bed
was not remixed the test was followed by a test with a higher transport rate, so that
history effects of the bed were negligible. After each test the bed surface
composition was determined. Appendix 7 lists the hydraulic and sedimentology data.
Verification of transport rate
University of Twente 77
6 Verification of transport rate
6.1 Introduction
The selected sediment transport formulas have been verified by comparing predicted
transport rates to measured transport rates. Chapter 5 describes the ten sets of
experiments that are used for this verification. In this chapter, the composition of
the initial bed material was used as input for the sediment transport predictions,
since for most experiments only the composition of the initial mixture was known.
However, Chapter 8 does consider transport predictions based on the composition of
the bed surface layer. In Section 6.2, we consider the variations in the measured
transport rates. In Section 6.3, a classification method is given together with the
score table of all the transport formulas. In the Sections 6.4 to 6.13, the results of
the formulas are discussed separately. In Section 6.14, the conclusions concerning
the transport rate predictions are given. Chapter 7 considers the verification of the
transport formulas by comparing the predicted transport composition with the
measured one.
6.2 Measured transport rate
In this section the measured transport is discussed in relation to the Shields
parameter, to show the variation in the measurements. Figure 6.1 shows the
measured transport rates as a function of the Shields parameter. The relation qs
meas

= a "
50
3/2
is shown as well. This power relation is used by Meyer-Peter & Mller and
other researchers. The experiments of Blom & Kleinhans, Klaassen, and Blom (except
test B2) show good agreement with this relation. Tests performed at lower Shields
parameters show a relation with a larger power. This power is higher if the initial bed
material is coarser. E.g., the tests of J06 show agreement with a power of ca. 8, the
tests of BOMC with a power of ca. 3.

Verification of transport rate
78 University of Twente
10
2
10
1
10
0
10
11
10
10
10
9
10
8
10
7
10
6
10
5
10
4
10
3
Measured transport rate

50
()
q
s
m
e
a
s

(
m
3
/
s
/
m
)

50
3/2
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 6.1: Measured transport rates versus the Shields parameters for the
experiments.
6.3 Ranking of the transport formulas
A classification method was developed to be able to compare the performance of the
different formulas. The classification of the formulas is based on the ratio between
the predicted sediment transport and the measured sediment transport. For the
classification of the transport rate predictions, the ratio is defined as:
) (
) (
) (
) (
,
measured q
predicted q
measured q
predicted q
ratio
s
i s
s
s
&
= =
In total, 82 tests were used for the verification. Table 6.1 is the score table of the
predictions for the transport rate. The score is calculated with:
n
j factor
Score
n
1 j
&
=
=
) (

4
5
4
6
7
>
8
=
1 j ratio if
j ratio
1
1 j ratio if j ratio
j factor
) (
) (
) ( ) (
) (
in which n is the total number of tests, and j the specific test. The maximum score is
one.
Table 6.1 gives the scores and the rankings of the different transport formulas. In
the following sections, the results of each transport formula will be explained in
detail.
Table 6.1: Score table for the predictions of the transport rate.
Name Score Ranking
Ackers & White uniform 0.34 5
Ackers & White + Day 0.36 3
Verification of transport rate
University of Twente 79
Ackers & White + Proffitt & Sutherland 0.32 7
Parker 0.21 13
Engelund & Hansen uniform 0.37 2
Engelund & Hansen fractional 0.33 6
Meyer-Peter & Mller 0.24 10
Meyer-Peter & Mller + Egiazaroff 0.27 8
Meyer-Peter & Mller + Ashida & Michiue 0.26 9
Van Rijn uniform 0.22 12
Van Rijn fractional 0.17 15
Hunziker/Meyer-Peter & Mller 0.22 11
Gladkow & Shngen 0.11 17
Wu et al. 0.41 1
Wilcock & Crowe 0.36 3
Ribberink uniform 0.14 16
Ribberink + Ashida & Michiue 0.18 14
6.4 Ackers & White
6.4.1 Ackers & White
The uniform formula of Ackers & White is the 5
th
best formula with a score of 0.34.
Figure 6.2 shows the transport rate ratio versus the Shields parameter of the tests
used for verification. The ratios with values larger than 1000 or smaller than 0.001,
are shown with values of, respectively, 1000 or 0.001. In general, the uniform
formula of Ackers & White has the tendency to overestimate the sediment transport
rate. Particularly at low Shields parameter ranges, the overestimation can be large.
Underestimations are only significant in the experiments of Blom and Wilcock &
McArdell. Especially the latter shows large underestimations at the lowest Shields
parameters. The experiments of Wilcock et al. with the coarsest sediment (J06, J14
and J21) show large overestimations at the lower Shields parameters. The difference
between the measured and predicted transport rate becomes smaller with increasing
Shields parameters. At Shields parameters higher than 0.1, the predicted and
measured transport rates are of the same order of magnitude. The formula predicts
the transport rates for the experiments of Blom & Kleinhans nearly perfectly. Bad
results are there for Day HRS A. The results for the other experimental series are
average.

Verification of transport rate
80 University of Twente
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
3
10
2
10
1
10
0
10
1
10
2
10
3
Ackers & White uniform

50
()
q
s
c
a
l
c
/
q
s
m
e
a
s

(

)
Score: 0.34
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 6.2: Transport rate ratio versus the Shields parameter for the uniform formula
of Ackers & White.
6.4.2 Ackers & White with Day
The fractional formula of Ackers & White with the hiding/exposure correction of Day
shows the 3
rd
best results of all formulas (Figure 6.3). It has a score of 0.36.
The fractional formula of Ackers & White with Day has a large scatter at low
Shields parameters, for which it has a strong tendency to overestimate the transport
rate. The formula overestimates for relatively coarse sediment mixtures (J06, J14 and
J21) and underestimates when fine material is used (HRS B, BOMC). The predictions
show more stable and better results at higher Shields parameters. The formula
shows good results for most experimental series when the Shields parameter is
higher than 0.1.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
3
10
2
10
1
10
0
10
1
10
2
10
3
Ackers & White + Day

50
()
q
s
c
a
l
c
/
q
s
m
e
a
s

(

)
Score: 0.36
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 6.3: Transport rate ratio versus the Shields parameter for the fractional
formula of Ackers & White with Day.
Verification of transport rate
University of Twente 81
6.4.3 Ackers & White with Proffitt & Sutherland
With a score of 0.32 the fractional formula of Ackers & White with the
hiding/exposure factor of Proffitt & Sutherland is the 7
th
best formula (Figure 6.4).

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
3
10
2
10
1
10
0
10
1
10
2
10
3
Ackers & White + Proffitt & Sutherland

50
()
q
s
c
a
l
c
/
q
s
m
e
a
s

(

)
Score: 0.32
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 6.4: Transport rate ratio versus the Shields parameter for the fractional
formula of Ackers & White with Proffitt & Sutherland.
This formula has problems predicting the transport rate at Shields parameters lower
than 0.05 where it underestimates very strongly. This especially occurs for the BOMC
experiments. The transport rate is largely overestimated at low Shields parameters
for experiments J06 and J14. At higher Shields parameters, the predictions become
much better. Only for the experiments of Klaassen a systematic overestimation is
shown. The formula is not valid for fine sediment mixtures as used by Klaassen,
which is confirmed by the relatively bad predictions.
6.5 Parker
With a total score of 0.21, Parkers surface based formula has a 13
th
ranking for the
transport rate predictions (Figure 6.5). Parkers formula shows a systematic
underestimation of the transport rate for low Shields parameters and a systematic
overestimation for high Shields parameters. Only the transport rates of the coarse
material (J06 and J14) are overestimated at low Shields parameters. There are a few
good predictions in the Shields parameter range 0.08 - 0.23. This is similar to the
Shields validity range of the formula (0.06 - 0.16). It must be noted that the formula
is not valid for any of the experimental series used for verification because the
sediment mixtures of the verification data were too fine. The formula shows good
results for the experiments of Blom & Kleinhans only. The results for the BOMC
experiments are very poor. It should be noted that the formula of Parker is based on
the surface layer composition but the initial bed composition is used here as input of
the formula due to lack of surface layer composition data.
Verification of transport rate
82 University of Twente

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
3
10
2
10
1
10
0
10
1
10
2
10
3
Parker surface

50
()
q
s
c
a
l
c
/
q
s
m
e
a
s

(

)
Score: 0.21
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 6.5: Transport rate ratio versus the Shields parameter for the fractional
formula of Parker.
6.6 Engelund & Hansen
6.6.1 Engelund & Hansen uniform
The predictions of the uniform Engelund & Hansen formula are the 2
nd
best of all
formulas (Figure 6.6). It has a score of 0.37. Engelund & Hansens uniform formula
overestimates the transport rate in many cases. This is due to the lack of a critical
shear stress, which results in high predicted transport rates at low Shields
parameters, while the measured transport rates are much lower. The formula shows
some underestimations, most of them in the Shields parameter range 0.09 - 0.23.
The overestimations at low Shields parameters are high, especially for the
experiments of Wilcock et al. with the coarse sediment (J06, J14, and J21).

Verification of transport rate
University of Twente 83
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
3
10
2
10
1
10
0
10
1
10
2
10
3
Engelund & Hansen uniform

50
()
q
s
c
a
l
c
/
q
s
m
e
a
s

(

)
Score: 0.37
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 6.6: Transport rate ratio versus the Shields parameter for the uniform formula
of Engelund & Hansen.
6.6.2 Engelund & Hansen fractional
It is noted again that the fractional formula has not been officially developed and
tested. Figure 6.7 shows the transport rate ratios of the fractional formula of
Engelund & Hansen. The score of the formula is 0.33, which gives a 6
th
ranking.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
3
10
2
10
1
10
0
10
1
10
2
10
3
Engelund & Hansen fractional

50
()
q
s
c
a
l
c
/
q
s
m
e
a
s

(

)
Score: 0.33
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 6.7: Transport rate ratio versus the Shields parameter for the fractional
formula of Engelund & Hansen.
The transport rates predicted by the fractional formula are larger than by the uniform
formula due to the lack of a critical shear stress and a hiding/exposure factor. This
leads to a relatively higher transport rates of smaller fractions, thus the total
transport rate of the fractional formula is higher than the one of the uniform
Verification of transport rate
84 University of Twente
formula. The fractional formula overestimates the transport in almost 75% of the
cases. Only the predictions for the experiments by Blom & Kleinhans are good. The
transport rates of the experiments of Wilcock et al. are again strongly overestimated,
especially at the lower Shields parameter range. If the sediment becomes coarser, the
overestimation becomes larger.
6.7 Meyer-Peter & Mller
6.7.1 Meyer-Peter & Mller uniform
The uniform formula of Meyer-Peter & Mller gives the 10
th
best predictions for the
transport rate with a total score of 0.24 (Figure 6.8). This formula has a lot of
problems with predicting the sediment transport rate at Shields parameters smaller
than 0.1. In these conditions, the formula predicts no sediment transport at all. In
Chapter 4, we found that the bed form factors were too small at conditions of
incipient motion, which results here in problems with predicting sediment transport
at the same conditions. The formula shows rather good results when the Shields
parameter exceeds 0.1. In this Shields parameter range, most predictions deviate
little from the measured transport rates.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
3
10
2
10
1
10
0
10
1
10
2
10
3
MeyerPeter & Mller uniform

50
()
q
s
c
a
l
c
/
q
s
m
e
a
s

(

)
Score: 0.24
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 6.8: Transport rate ratio versus the Shields parameter for the uniform formula
of Meyer-Peter & Mller.
6.7.2 Meyer-Peter & Mller with Egiazaroff
With a score of 0.27, the formula is the 8
th
best formula of all formulas; this formula
gives the best predictions of the Meyer-Peter & Mller formulas (Figure 6.9). Like the
uniform formula of Meyer-Peter & Mller, this formula has the tendency to predict no
sediment transport at Shields parameters smaller than 0.1, although the validity
range for the Shields parameter for this formula is 0.04 - 0.9. This is due to the
same problems with the bed form factor as the uniform formula. The predictions for
Verification of transport rate
University of Twente 85
the experiments of Wilcock et al. differ much between each other. For the
experiments with coarse material there is a rather high overestimation, while there is
a strong underestimation for the experiments with fine sediment. The formula is not
valid for the experimental series of Blom & Kleinhans, test B2 by Blom, and Klaassen,
because the sediment mixtures were too fine. Yet, the transport rate predictions for
these experiments are rather good.
Analysing the transport rate predictions show that that the asymptotic
behaviour of the hiding/exposure factor (Section 4.2.2) becomes important when
dealing with strongly graded mixtures. Fractions more than 19 times smaller than
the geometric mean grain size have a smaller hiding coefficient, while fractions of ca.
19 times smaller than the geometric grain size experience a very strong hiding
coefficient. This is also the reason why transport rates are overestimated in some
tests of J06 and J14 with a low Shields parameter. In these tests only the smallest
fraction is transported.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
3
10
2
10
1
10
0
10
1
10
2
10
3
MeyerPeter & Mller + Egiazaroff

50
()
q
s
c
a
l
c
/
q
s
m
e
a
s

(

)
Score: 0.27
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 6.9: Transport rate ratio versus the Shields parameter for the fractional
formula of Meyer-Peter & Mller with Egiazaroff.
6.7.3 Meyer-Peter & Mller with Ashida & Michiue
The fractional transport formula of Meyer-Peter & Mller with the hiding/exposure
correction of Ashida & Michiue gives the 9
th
transport rate predictions of all formulas
with a score of 0.26 (Figure 6.10). It is slightly worse than Meyer-Peter & Mller with
Egiazaroff. The fractional formula with Ashida & Michiue has the same tendency as
the other formulas of Meyer-Peter & Mller to predict no sediment transport at
Shields parameters below 0.1. At higher Shields parameters, the formula shows good
results. The formula has the tendency to underestimate the transport rate except for
the experiments of Wilcock et al. with relatively coarse material: J06, J14 and J21.
The formula is not valid for the experiments of Klaassen, because the sediment
mixture was too fine, but the formulas results can be considered very good for
these experiments.
Verification of transport rate
86 University of Twente
The correction made by Ashida & Michiue on the hiding/exposure of Egiazaroff
has lead to slightly worse verification results. This is due to the fact that the
correction eliminates the asymptotic effect for fractions 19 times smaller than the
geometric grain size. This also means that the smallest fraction is not transported in
all the tests of J06 and J14. It can be concluded that although the correction by
Ashida & Michiue has a better theoretic basis, the original hiding/exposure factor of
Egiazaroff gives better transport rate results due to its asymptotic behaviour.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
3
10
2
10
1
10
0
10
1
10
2
10
3
MeyerPeter & Mller + Ashida & Michiue

50
()
q
s
c
a
l
c
/
q
s
m
e
a
s

(

)
Score: 0.26
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 6.10: Transport rate ratio versus the Shields parameter for the fractional
formula of Meyer-Peter & Mller with Ashida & Michiue.
6.8 Van Rijn
6.8.1 Van Rijn uniform
The uniform formula has a score of 0.22, which means a 12
th
ranking (Figure 6.11).
The uniform formula of Van Rijn overestimates the transport rate for all tests except
for the experiments of Blom and some tests of Wilcock with very low Shields
parameters (<0.03). The transport rates are strongly overestimated at low Shields
parameters (0.03 0.15). The formula shows reasonable results at Shields
parameters higher than 0.15. The formula seems to overestimate the transport rate
stronger if the median grain size increases.

Verification of transport rate
University of Twente 87
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
3
10
2
10
1
10
0
10
1
10
2
10
3
Van Rijn uniform

50
()
q
s
c
a
l
c
/
q
s
m
e
a
s

(

)
Score: 0.22
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 6.11: Transport rate ratio versus the Shields parameter for the uniform
formula of Van Rijn.
6.8.2 Van Rijn fractional
The fractional formula of Van Rijn has not been officially published and tested. It is
noted that Kleinhans & Van Rijn [2001] have developed a fractional bed load formula,
however, it could not be taken into account within the present study. The results of
the fractional formula of Van Rijn are shown in Figure 6.12 (score of 0.17, 15
th

ranking). The predicted transport rates by the fractional formula of Van Rijn are even
larger than the ones of the uniform formula. This is caused by the lack of a
hiding/exposure correction, which could have decreased the large transport of the
fine fractions. The transport rate predictions are especially high at low Shields
parameters. Many experiments have an overestimation of more than 1000 times the
measured transport rate. The formula shows good results for the experiments of
Blom and Klaassen. The results for the other series are very poor.

Verification of transport rate
88 University of Twente
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
3
10
2
10
1
10
0
10
1
10
2
10
3
Van Rijn fractional

50
()
q
s
c
a
l
c
/
q
s
m
e
a
s

(

)
Score: 0.17
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 6.12: Transport rate ratio versus the Shields parameter for the fractional
formula of Van Rijn
6.9 Hunziker/Meyer-Peter & Mller
The formula of Hunziker/Meyer-Peter & Mller has a score of 0.22, which means the
11
th
best formula (Figure 6.14). This formula has the same problems with predicting
sediment transport at low Shields parameters as the formulas of Meyer-Peter &
Mller and in most cases shows the same underestimation of the transport rate. The
formula predicts no sediment transport at Shields parameters below 0.1, where the
formula is sensitive. The two runs of Blom (A1 and B1) show this. These tests had
basically the same conditions, but the predicted transport rates vary strongly. At
conditions with Shields parameters higher than 0.1, the predictions are reasonably
well. The formula gives good results for the experiments of Klaassen, although it is
not valid for the median grain size used in these experiments. Rather good results
are there for Blom & Kleinhans as well, but the formula shows poor results for the
other experimental series.

Verification of transport rate
University of Twente 89
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
3
10
2
10
1
10
0
10
1
10
2
10
3
Hunziker/MeyerPeter & Mller

50
()
q
s
c
a
l
c
/
q
s
m
e
a
s

(

)
Score: 0.22
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 6.14: Transport rate ratio versus the Shields parameter for the fractional
formula of Hunziker/Meyer-Peter & Mller.
6.10 Gladkow & Shngen
The formula of Gladkow & Shngen gives the worst predictions of all formulas. The
formula has a score of 0.11 and Figure 6.15 shows the transport ratios versus the
Shields parameters.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
3
10
2
10
1
10
0
10
1
10
2
10
3
Gladkow & Shngen

50
()
q
s
c
a
l
c
/
q
s
m
e
a
s

(

)
Score: 0.11
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 6.15: Transport rate ratio versus the Shields parameter for the formula of
Gladkow & Shngen.
The formula has a strong tendency to underestimate the sediment transport. This
formula often predicts no sediment transport at conditions with Shields parameters
smaller than 0.1. At higher Shields parameters, the formula underestimates the
transport rate as well. Only for a couple of runs of Wilcock et al., the transport rate is
Verification of transport rate
90 University of Twente
overestimated. The formula does not give good predictions for any of the
experimental series. The transport rate predictions agree better with the
measurements if the sediment mixture is coarser (J06 and J14).
6.11 Wu et al.
The formula of Wu et al. is the best formula with a score of 0.41. Figure 6.16 shows
the transport ratios versus the Shields parameters. The predictions of the formula of
Wu et al. are reasonably good. It shows a large overestimation at low Shields
parameters, but still gives the best results in this Shields parameter range. This
overestimation is the largest for experiments with coarse material, i.e. Wilcock et al:
J06 and J14. The conditions of all tests are within the validity ranges of the formula,
which may be the reason for the good results of the formula. At higher Shields
parameters, the formula slightly underestimates the transport rates.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
3
10
2
10
1
10
0
10
1
10
2
10
3
Wu et al.

50
()
q
s
c
a
l
c
/
q
s
m
e
a
s

(

)
Score: 0.41
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 6.16: Transport rate ratio versus the Shields parameter for the formula of Wu
et al.
6.12 Wilcock & Crowe
The formula of Wilcock & Crowe gives the third best predictions (Figure 6.17). It has
a score of 0.36. The formula of Wilcock & Crowe shows good results. However, this is
not surprising since 57% of the data used for the verification was also used by
Wilcock & Crowe to calibrate the formula, although they used the surface layer as
active layer and in this section the initial bed composition is used. Nevertheless, this
formula shows a large variation at low Shields parameters. The transport rates of the
runs of BOMC (fine material) are strongly underestimated, while the rates of the
experiments J06 and J14 (coarse material) are strongly overestimated. At higher
Shields parameters, the formula overestimates the sediment transport rates.

Verification of transport rate
University of Twente 91
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
3
10
2
10
1
10
0
10
1
10
2
10
3
Wilcock & Crowe

50
()
q
s
c
a
l
c
/
q
s
m
e
a
s

(

)
Score: 0.36
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 6.17: Transport rate ratio versus the Shields parameter for the formula of
Wilcock & Crowe.
6.13 Ribberink
6.13.1 Ribberink uniform
The uniform formula of Ribberink gives the second worst predictions of all formulas.
The formula has a score of 0.14. In Figure 6.18 the results of this formula are shown.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
3
10
2
10
1
10
0
10
1
10
2
10
3
Ribberink uniform

50
()
q
s
c
a
l
c
/
q
s
m
e
a
s

(

)
Score: 0.14
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 6.18: Transport rate ratio versus the Shields parameter for the uniform
formula of Ribberink.
The formula strongly overestimates the sediment transport rates. For some runs the
formula predicts no transport, but this occurs less frequently than with the other
Meyer-Peter & Mller type formulas. At higher Shields parameters, the predictions
Verification of transport rate
92 University of Twente
are good. At low Shields parameters, the transport rates for the experiments of
Wilcock are strongly overestimated. The overestimation is greater when the sediment
is coarser.
6.13.2 Ribberink with Ashida & Michiue
The formula of Ribberink with Ashida & Michiue is the 14
th
best formula with a score
of 0.18. Figure 6.19 shows the transport ratios versus the Shields parameters.
This formula also strongly overestimates the transport rate, although less than
its uniform counterpart. The formula shows good predictions at higher Shields
parameters and strong overestimations at the low Shields parameters (< 0.2).

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
3
10
2
10
1
10
0
10
1
10
2
10
3
Ribberink + Ashida & Michiue

50
()
q
s
c
a
l
c
/
q
s
m
e
a
s

(

)
Score: 0.18
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 6.19: Transport rate ratio versus the Shields parameter for the fractional
formula of Ribberink with Ashida & Michiue.
6.14 Conclusions
The best formula for predicting the sediment transport rate is the fractional formula
of Wu et al. The validity ranges indicate that the formula can be used in a wide range
of conditions. However, the formula should be used cautiously at low Shields
parameters, since it has the tendency to strongly overestimate the transport rate.
The formulas of Wilcock & Crowe and Ackers & White with the hiding/exposure
correction of Day are good alternatives, but also with these formulas caution is
required for low Shields parameters.
The sediment transport rate is extremely difficult to predict at Shields parameters up
to 0.1 with the selected formulas. Some formulas show a strong overestimation of
the transport rate, others predict no sediment transport at all, while again other
formulas show both strong overestimations as underestimations of the transport
rate. In this study, many tests with low Shields parameters have been used for
verification. We deliberately put strong emphasis on these conditions because of the
known problems with predicting the sediment transport in this range.
Verification of transport rate
University of Twente 93
In this range, the formulas of Meyer-Peter & Mller, Hunziker/Meyer-Peter &
Mller and Gladkow & Shngen often predict no sediment transport. In the Meyer-
Peter & Mller type of formulas, this is because the applied bed form factors are too
small at conditions with low Shields parameters. But the formulas of Ribberink, which
can be seen as a type of Meyer-Peter & Mller formula, show a strong
overestimation. This shows that Meyer-Peter & Mller type of formulas are very
sensitive at conditions with low Shields parameters and are thus recommended not
to be used at these conditions.
Other formulas, like Ackers & White with Day, Engelund & Hansen, Van Rijn and
Wu et al. strongly overestimates the transport rate at low Shields parameters. This
overestimation can be more than a factor 1000.
Although only the formulas of Ackers & White with Day and Wu et al. are valid for the
experiments of Klaassen also the formulas of Meyer-Peter & Mller and of
Hunziker/Meyer-Peter & Mller give good results for these experiments. This means
that the formulas of Meyer-Peter & Mller and of Hunziker/Meyer-Peter & Mller are
suitable for transport rate predictions with finer sediment mixtures and higher
Shields parameters.
The fractional formulas of Ackers & White with Day, Meyer-Peter & Mller with
Egiazaroff, Meyer-Peter & Mller with Ashida & Michiue and Ribberink with Ashida &
Michiue give better predictions than their uniform counterparts.
The fractional formulas of Engelund & Hansen and Van Rijn show worse results
than their uniform formulas. A hiding/exposure factor could solve some of the
problems of the fractional formulas, since these show a relatively large transport of
small fractions.
Verification of transport composition
University of Twente 95
7 Verification of transport composition
7.1 Introduction
In this chapter, the fractional formulas are verified by comparing predicted transport
compositions with the measured ones in the experiments discussed in Chapter 5. In
this comparison, like in Chapter 6, the predictions are done using the composition of
the initial bed mixture as input for the formulas. In Chapter 8, the predicted
composition using the bed surface layer as input for the formulas will be discussed.
The composition is represented by its geometric mean grain size. Section 7.2 shows
the measured transport composition related to the initial material composition. A
classification method with the score table of the fractional formulas for the
composition predictions is given in Section 7.3. In Sections 7.4 to 7.13, the results of
the formulas will be discussed separately. The conclusions are given in Section 7.14.
7.2 Measured transport composition
Figure 7.1 shows the ratio of the geometric mean grain sizes of the measured
transport composition and the composition of the initial bed mixture as a function of
the Shields parameter. It can be seen that at low Shields parameters (up to 0.1) the
transport composition is much finer than the initial bed material, which we call
selective transport. At higher Shields parameters, the transport composition
approaches the initial material composition. The transported sediment does not
become systematically coarser than the initial bed material, not even at high Shields
parameters.

10
2
10
1
10
0
10
1
10
0
Measured D
mT
/D
mB

50
()
D
m
T
/
D
m
B

(

)
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 7.1: Ratio of measured geometric mean grain size of transported material and
geometric mean grain size of the initial bed mixture versus the Shields parameters.
Verification of transport composition
96 University of Twente
It must be noted that the initial mixture composition is not the best definition for the
active layer. But due to lack of active layer definitions and measurements in some
experimental series, the composition of the initial bed mixture is used. Chapter 8
will address this problem.
7.3 Ranking of the transport formulas
The ratio between the predicted geometric mean grain size and the measured
geometric mean grain size will be used for this verification. This ratio is calculated as
follows:
&
&
'
)
)
*
+
,
,
-
.
'
= =
i
i i
i
i
total s
i s
meas mT
pred mT
measured D p
predicted D
q
q
D
D
ratio
)) ( (
) (
,
,
,
,

In which:
D
mT
geometric mean grain size (m)
q
s,i
sediment transport rate for fraction i (m
2
/s)
&
=
i
i s total s
q q
, ,
total sediment transport rate (m
2
/s)
p
i
measured probability of fraction i in sediment transport (-)
D
i
grain size of fraction i (m)
The scores of the formulas for the composition predictions are calculated in a similar
way as described in Section 6.3. Table 7.1 lists the scores for the verification of the
composition of the transported material and the ranking of all the fractional
formulas.
Table 7.1: Score table for the transport composition of all fractional formulas
Name Score Ranking
Ackers & White + Day 0.59 4
Ackers & White + Proffitt & Sutherland 0.54 6
Parker 0.58 5
Engelund & Hansen fractional 0.54 6
Meyer-Peter & Mller + Egiazaroff 0.34 11
Meyer-Peter & Mller + Ashida & Michiue 0.27 12
Van Rijn fractional 0.45 9
Hunziker/Meyer-Peter & Mller fractional 0.35 10
Gladkow & Shngen 0.53 8
Wu et al. 0.79 1
Wilcock & Crowe 0.67 2
Ribberink + Ashida & Michiue 0.62 3
Verification of transport composition
University of Twente 97
7.4 Ackers & White
7.4.1 Ackers & White with Day
The fractional formula of Ackers & White with the hiding/exposure factor of Day
shows the 4
th
best results concerning the transport composition with a score of 0.59
(Figure 7.2). The ratios shown in these figures are within the range 0.1 10. If the
ratio is smaller than 0.1 (0 in case of no sediment transport), or if the ratio exceeds
10, it is set on 0.1 or 10, respectively.
At low Shields parameters, the predicted sediment transport by this formula is
often coarser than measured. In this range, there is a large variation in the predicted
composition. This is especially the case in the BOMC experiments. At high Shields
parameters, the variation is smaller and the predicted compositions approximate the
measured composition better, although the predicted geometric mean grain size is
finer than the one measured. The formula shows reasonable results for most
experimental series, and has the largest problems with tests BOMC and J27.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
1
10
0
10
1
Ackers & White + Day

50
()
D
m
,
c
a
l
c
/
D
m
,
m
e
a
s

(

)
Score: 0.59
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 7.2: Transport composition ratio versus the Shields parameter for the
fractional formula of Ackers & White with Day.
7.4.2 Ackers & White with Proffitt & Sutherland
The fractional formula of Ackers & White with Proffitt & Sutherland (score = 0.54)
shows the 6
th
result of all fractional formulas (Figure 7.3). The results are worse than
the formula of Ackers & White with the hiding/exposure of Day.

Verification of transport composition
98 University of Twente
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
1
10
0
10
1
Ackers & White + Proffitt & Sutherland

50
()
D
m
,
c
a
l
c
/
D
m
,
m
e
a
s

(

)
Score: 0.54
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 7.3: Transport composition ratio versus the Shields parameter for the
fractional formula of Ackers & White with Proffitt & Sutherland.
For about half of the tests the fractional formula of Proffitt & Sutherland
overestimates the geometric mean grain size of the transported material, provided
that the sediment transport is not equal to zero. The formula especially
overestimates the grain size for Shields parameters smaller than 0.1. In this range,
the formula shows large variations in the predicted transport composition. In many
tests, the formula does not predict any sediment transport, so the predicted
geometric mean grain size is automatically zero (in figure set on 0.1). At higher
Shields parameters, the formula underestimates the geometric mean grain size, and
the variations become smaller. The formula shows reasonable results for the
experiments of Klaassen, although the formula is not valid for such a fine sediment
mixture. The predictions for BOMC and Wilcock et al. are very poor.
7.5 Parker
The geometric mean grain size of the transported material, predicted by the surface
based formula of Parker, finds good agreement with the measurements. With a score
of 0.58 the formula is the 5
th
best formula (Figure 7.4).

Verification of transport composition
University of Twente 99
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
1
10
0
10
1
Parker surface

50
()
D
m
,
c
a
l
c
/
D
m
,
m
e
a
s

(

)
Score: 0.58
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 7.4: Transport composition ratio versus the Shields parameter for the
fractional formula of Parker.
The formula of Parker underestimates the geometric mean transported grain size for
Shields parameters smaller than 0.2. At Shields parameters larger than 0.2, the
predicted transported mean grain size is similar to the measured one. This behaviour
is in concurrence with the behaviour of the formula for strongly graded sediments
(Figure 4.4). Although, due to their fine sediment mixtures, the formula of Parker is
not valid for the experiments by Blom & Kleinhans, Blom, and Klaassen, the
predictions of the geometric mean grain size are good for these experiments. The
results for the experimental series Day HRS A and HRS B are average. The formula
has problems predicting the composition for the experiments of Wilcock et al. The
BOMC experiments are predicted well at very low Shields parameters, but the results
are poor when the Shields parameter increases.
7.6 Engelund & Hansen
The fractional formula of Engelund & Hansen shows the 6
th
best results for the
transport composition predictions with a score of 0.54 (Figure 7.5).

Verification of transport composition
100 University of Twente
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
1
10
0
10
1
Engelund & Hansen fractional

50
()
D
m
,
c
a
l
c
/
D
m
,
m
e
a
s

(

)
Score: 0.54
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 7.5: Transport composition ratio versus the Shields parameter for the
fractional formula of Engelund & Hansen.
For most tests, the fractional formula of Engelund & Hansen underestimates the
geometric mean grain size of the transported material. As mentioned before, this
formula has never been officially derived, and thus, it has not been calibrated for
fractional transport predictions. It should be noted that, under different hydraulic
conditions, the formula predicts the same transport composition for identical bed
composition (see Section 4.2.2). Therefore, the ratio of the calculated and measured
geometric mean grain sizes of the transported material for experiments with the
same initial bed material only differs because of differences in the measured
transport composition. Since the measured composition becomes coarser if the
Shields parameter increases, the ratio decreases. The predictions are not good for
any of the experimental series, but on average the formula predicts the composition
rather well.
7.7 Meyer-Peter & Mller
7.7.1 Meyer-Peter & Mller with Egiazaroff
The fractional sediment transport formula of Meyer-Peter & Mller with the
hiding/exposure correction of Egiazaroff is the 11
th
best formula to predict the
transport composition with a score of 0.34 (Figure 7.6).
In all cases, the formula of Meyer-Peter & Mller with Egiazaroff predicts a
finer geometric mean grain size than measured. At low Shields parameters, the
underestimation tends to be larger than at high Shields parameters. The formula has
large problems predicting the sediment transport composition at Shields parameters
up to 0.1: it often predicts no transport at all (Section 6.7.2). This is the most
important factor for the poor result of this formula. At higher Shields parameters the
composition predictions are quite good.

Verification of transport composition
University of Twente 101
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
1
10
0
10
1
MeyerPeter & Mller + Egiazaroff

50
()
D
m
,
c
a
l
c
/
D
m
,
m
e
a
s

(

)
Score: 0.34
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 7.6: Transport composition ratio versus the Shields parameter for the
fractional formula of Meyer-Peter & Mller with Egiazaroff.
7.7.2 Meyer-Peter & Mller with Ashida & Michiue
With a score of 0.27, the fractional formula of Meyer-Peter & Mller with the
hiding/exposure correction of Ashida & Michiue is the worst formula for predicting
the transport composition (Figure 7.7).

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
1
10
0
10
1
MeyerPeter & Mller & Ashida & Michiue

50
()
D
m
,
c
a
l
c
/
D
m
,
m
e
a
s

(

)
Score: 0.27
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 7.7: Transport composition ratio versus the Shields parameter for the
fractional formula of Meyer-Peter & Mller with Ashida & Michiue.
The predicted geometric mean grain size of the transported sediment is smaller than
the measured one. The geometric mean grain size predicted by the combination of
Meyer-Peter & Mller and Ashida & Michiue is mostly slightly smaller than by the
combination of Meyer-Peter & Mller and Egiazaroff. This leads to larger
Verification of transport composition
102 University of Twente
underestimations and thus worse predictions. The difference between the two
formulas is largest at low Shields parameters.
The differences between the predictions using the two different
hiding/exposure factors are caused by the asymptotic behaviour of the
hiding/exposure factor of Egiazaroff. This behaviour makes the hiding factor of
some small fractions so high that these fractions are never transported according to
the formula. This means that the composition of the transported material becomes
coarser.
7.8 Van Rijn
The fractional version of Van Rijns formula gives the 9
th
best result concerning the
sediment transport composition. The formula has a score of 0.45 and its results are
shown in Figure 7.8. It should be noted that this formula has not been officially
derived and tested by Van Rijn [1984].

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
1
10
0
10
1
Van Rijn fractional

50
()
D
m
,
c
a
l
c
/
D
m
,
m
e
a
s

(

)
Score: 0.45
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 7.8: Transport composition ratio versus the Shields parameter for the
fractional formula of Van Rijn.
The fractional formula of Van Rijn always predicts a finer sediment transport
composition than measured, except in some runs of BOMC at very low Shields
parameters. In many tests, the formula predicts that the geometric mean grain size
of the transported material is less than half of the measured one. The results for the
BOMC tests are good, whereas the results for the experiments of Klaassen and Day
HRS A are average. The results for the other experiments are poor. The predictions
of the transport composition are somewhat better for higher Shields parameters.
7.9 Hunziker/Meyer-Peter & Mller
The predictions of the transport composition of the formula of Hunziker/Meyer-
Peter & Mller are the 10
th
best with a score of 0.35 (Figure 7.9). For small Shields
Verification of transport composition
University of Twente 103
parameters, the formula of Hunziker predicts a finer sediment transport composition
than measured. For Shields parameters higher than about 0.15, the formula predicts
a coarser transport composition than measured. This concurs with the results of
Section 4.2.2. Like the fractional formulas of Meyer-Peter & Mller, Hunzikers
formula has problems predicting sediment transport for small Shields parameters
(smaller than 0.1), for which the formula often predicts no sediment transport at all.
For tests where the formula does predict sediment transport to occur, the
composition ratio is mostly within the 0.5 - 2 range, which is quite good.
The hiding/exposure factor of Hunziker (eq. 3.21) predicts an inverse
hiding/exposure effect (i.e. small grains are assumed to experience a higher shear
stress, and large grains a smaller shear stress than they would experience in a bed
with uniform material of the same grain size) at grain roughness related Shields
parameters (
*
ms
" ) smaller than 0.11. In combination with the equal mobility concept,
this can be a reason for the predicted transport composition being too fine at low
Shields parameters, although the predictions are still reasonable.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
1
10
0
10
1
Hunziker/MeyerPeter & Mller

50
()
D
m
,
c
a
l
c
/
D
m
,
m
e
a
s

(

)
Score: 0.35
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 7.9: Transport composition ratio versus the Shields parameter for the formula
of Hunziker/Meyer-Peter & Mller.
7.10 Gladkow & Shngen
The score of the formula of Gladkow & Shngen is 0.53, which means an 8
th
ranking.
Figure 7.10 shows the results of this formula.
This formula overestimates as well as underestimates the geometric mean
grain size of the transported material. For many tests, it does not predict any
sediment transport to occur, which results in a less than average ranking. For the
tests for which sediment transport does occur, the composition is predicted rather
well or slightly too coarse (BOMC experiments).

Verification of transport composition
104 University of Twente
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
1
10
0
10
1
Gladkow & Shngen

50
()
D
m
,
c
a
l
c
/
D
m
,
m
e
a
s

(

)
Score: 0.53
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 7.10: Transport composition ratio versus the Shields parameter for the
formula of Gladkow & Shngen.
7.11 Wu et al.
The formula of Wu et al. gives the best predictions of all formulas. With a score of
0.79 it is much better than the other formulas (Figure 7.11).

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
1
10
0
10
1
Wu et al.

50
()
D
m
,
c
a
l
c
/
D
m
,
m
e
a
s

(

)
Score: 0.79
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 7.11: Transport composition ratio versus the Shields parameter for the
formula of Wu et al.
The formula of Wu et al. gives very good predictions for the composition of the
transported material. Only for Shields parameters smaller than 0.08, the formula has
a tendency to predict a finer transport than measured. At higher Shields parameters,
the composition is predicted very well with both small overestimations and small
underestimations of the geometric grain size.
Verification of transport composition
University of Twente 105
7.12 Wilcock & Crowe
The formula of Wilcock & Crowe gives the 2
nd
best predictions of all formulas. It is
has a score of 0.67 and its results are shown in Figure 7.12. The formula has the
tendency to predict the sediment transport being finer than measured. For most
tests, the composition ratio lies within the 0.5 1 range. This stable pattern results
in the good classification of this formula. For many of the BOMC experiments, the
composition is predicted coarser than measured, but for the most other tests the
predicted composition is mostly somewhat finer.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
1
10
0
10
1
Wilcock & Crowe

50
()
D
m
,
c
a
l
c
/
D
m
,
m
e
a
s

(

)
Score: 0.67
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 7.12: Transport composition ratio versus the Shields parameter for the
formula of Wilcock & Crowe.
7.13 Ribberink with Ashida & Michiue
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
1
10
0
10
1
Ribberink + Ashida & Michiue

50
()
D
m
,
c
a
l
c
/
D
m
,
m
e
a
s

(

)
Score: 0.62
Blom & Kleinhans
Blom
Klaassen
Day: HRS A
Day: HRS B
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Verification of transport composition
106 University of Twente
Figure 7.13: Transport composition ratio versus the Shields parameter for the
fractional formula of Ribberink with Ashida & Michiue.
The fractional formula of Ribberink with the hiding/exposure correction of Ashida &
Michiue gives the 3
rd
best predictions for the transport composition, with a score of
0.62 (Figure 7.13). For most tests, the predicted composition is finer than the
measured composition. Only in a couple of tests of HRS A and BOMC the predicted
composition is coarser. The formula gives the best composition predictions of the
Meyer-Peter & Mller type formulas. This is because the formula nearly always
predicts sediment transport to occur.
7.14 Conclusions
The formula of Wu et al. gives the best transport composition predictions. For
conditions with low Shields parameters, the formula shows in general small
underestimations of the geometric mean grain size.
The formula of Wilcock & Crowe gives the second best composition
predictions. This is not really surprising since 57% of the verification data used in
this report was used for calibrating the formula. But the formula shows good results
for the other experimental data as well.
All formulas have problems predicting the sediment transport composition for low
Shields parameters. For this range, some formulas do not predict any sediment
transport at all, and others show a large variation in the predicted compositions.
Only the formula of Wu et al. shows good and stable predictions for the transport
composition in this range.
All fractional formulas predict the transport composition of the experiments of
Klaassen well. Only the formulas of Ackers & White with Day and Wu et al. are valid
for these experiments, but the other formulas seem also suitable for these fine
sediment mixtures and high Shields parameters.
The formulas of Meyer-Peter & Mller, Hunziker/Meyer-Peter & Mller and Gladkow
& Shngen have problems predicting any sediment transport for low Shields
parameters. This is the main reason why the scores of the transport composition of
the formulas are poor.
Also the formula of Hunziker/Meyer-Peter & Mller has problems predicting
sediment transport for low Shields parameters. The formula overestimates the
geometric mean grain size for Shields parameters larger than 0.15. Hunziker found
that, at these high Shields parameters, the transported material was coarser than the
bed material. This assumption was based on armouring tests of Suzuki.

Surface layer as active layer
University of Twente 107
8 Surface layer as active layer
8.1 Introduction
An important input parameter for fractional sediment transport formulas is the
composition of the part of the bed layer that determines the rate and composition of
the transported sediment, the so-called active layer. Wilcock et al. [2001] state that
the bed surface layer is the bed material that determines the transport rate and
composition. However, for some of the experiments described and used in this
study, only the composition of the initial bed mixture was known. That is why we
chose to base our verification in Chapter 6 and 7 on the composition of the initial
bed mixture. In this chapter we will use the composition of the surface layer as input
for the transport predictions, for the tests for which we do know this surface layer
composition.
The next paragraphs are based on Wilcock [2001] and the accompanying
discussion by Kleinhans and Blom [2001]. A definition of the active layer is
relatively straightforward when one is speaking of flat bed situations: the surface
layer. When bed forms are present, different definitions are given in literature. One
definition states that the active layer can be considered as the layer over which the
bed forms migrate. This is usually a relatively coarse layer, which is only exposed to
the flow in the trough areas.
Another definition of the active layer can be considered the actual bed
surface: the top layer of bed forms and trough areas. The second definition gives the
best basis for sediment transport predictions since the shear stress acting by the
flow interacts only with the bed surface.
In the following sections the results will be discussed. The differences between
the transport rate using the initial bed material composition or the surface
composition are discussed in Section 8.3. The differences in the transport
composition will be discussed in Section 8.4. This is followed by conclusions in
Section 8.5.
8.2 Measurements
Wilcock & McArdell [1993] (BOMC) and Wilcock et al. [2001] (J06, J14, J21, and J27)
directly measured their surface layer compositions. In all of these tests the bed was
essentially planar [Wilcock et al., 2001]. Blom [2000] took box core samples and
measured the bed elevation distributions. By comparing the vertical profiles of bed
composition and the bed elevation probabilities, the surface layer can be estimated.
The surface compositions of Blom [2000] are shown in Section 5.3. Appendix 7 gives
the bed surface compositions for the tests of Wilcock & McArdell [1993] and Wilcock
et al. [2001].
The surface layer composition measured in experimental series J06 and J14 is
much coarser than the composition of their initial bed material. In test series J21 and
Surface layer as active layer
108 University of Twente
J27, it became on average slightly coarser than the initial material. The surface
compositions in the BOMC experiments are much finer than its initial composition,
and the experiments of Blom show a surface composition slightly finer than the
initial material. Table 8.1 gives the scores of the formulas using as input the initial
bed or surface composition for both the transport rate and the transport
composition.
Table 8.1: Score table for the predictions of the transport rate and transport
composition for two active layer compositions: initial bed mixture and surface layer
composition.
Name transport rate transport composition
Initial bed surface Initial bed surface
Ackers & White uniform 0.27 0.24 - -
Ackers & White + Day 0.23 0.27 0.52 0.51
Ackers & White + Proffitt & Sutherland 0.22 0.26 0.49 0.39
Parker 0.19 0.25 0.52 0.76
Engelund & Hansen uniform 0.29 0.34 - -
Engelund & Hansen fractional 0.25 0.29 0.53 0.68
Meyer-Peter & Mller 0.11 0.15 - -
Meyer-Peter & Mller + Egiazaroff 0.19 0.19 0.27 0.26
Meyer-Peter & Mller + Ashida & Michiue 0.12 0.19 0.20 0.23
Van Rijn uniform 0.14 0.13 - -
Van Rijn fractional 0.08 0.10 0.42 0.56
Hunziker/Meyer-Peter & Mller 0.10 0.13 0.25 0.26
Gladkow & Shngen 0.16 0.13 0.52 0.51
Wu et al. 0.30 0.36 0.74 0.85
Wilcock & Crowe 0.27 0.53 0.65 0.77
Ribberink uniform 0.06 0.06 - -
Ribberink + Ashida & Michiue 0.08 0.08 0.58 0.69
8.3 Transport rate
Most sediment transport formulas show better transport rate predictions when the
surface layer composition is used as input for these formulas instead of the initial
bed composition. But there are also formulas that give worse predictions. Only the
formula of Wilcock & Crowe shows a big improvement when the surface layer is used
as input for the formula. The score of the formula is 0.53 for the transport
predictions, which is the highest of all formulas. This is not a coincidence since the
formula of Wilcock & Crowe was calibrated on the tests of Wilcock & McArdell and
Wilcock et al. using the surface layer composition as input for the transport formula.
With a score of 0.36, the formula of Wu et al. gives the 2
nd
best result. In this
section only the results of the two best formulas will be discussed, because the
differences are small for most other formulas. The transport rate decreases for the
tests for which the active layer composition becomes coarser (J06, J14, J21, and J27)
and increases for the tests for which the active layer becomes finer (Blom and
BOMC).

Surface layer as active layer
University of Twente 109
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
3
10
2
10
1
10
0
10
1
10
2
10
3
Wilcock & Crowe

50
()
q
s
c
a
l
c
/
q
s
m
e
a
s

(

)
Score: 0.27
Blom
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
3
10
2
10
1
10
0
10
1
10
2
10
3
Wilcock & Crowe

50
()
q
s
c
a
l
c
/
q
s
m
e
a
s

(

)
Score: 0.53
Blom
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 8.1: Transport rate ratio versus the Shields parameter for the formula of
Wilcock & Crowe (left: using initial bed material composition, right: using bed surface
composition)
Figure 8.1 shows the transport rates predicted by the formula of Wilcock & Crowe for
both active layer definitions: the initial bed composition (left) and the surface layer
composition (right). The large variation of the predicted transport rates at small
Shields parameters is much smaller when the surface layer composition is used. This
is caused by the fact that for the tests with an overestimation of the transport rate
using the initial mixture composition, i.e. J06, J14 and J21, the surface layer
composition is coarser than the initial material composition. This reduces the
transport rate and comes closer to the measured transport rate. For the tests with an
underestimation of the transport rate, i.e. BOMC, the surface layer composition is
finer than the initial bed mixture composition, which leads to higher transport rates.
The transport rates predicted by the formula of Wu et al. are shown in Figure 8.2. On
the left are shown the results when using the initial bed material composition and on
the right the results when using the surface layer composition. The formula of Wu et
al. is also sensitive to changes in the active layer composition for the transport rate
predictions. The overestimations at low Shields parameters become smaller, but are
still large, yet they do show a significantly smaller scatter.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
3
10
2
10
1
10
0
10
1
10
2
10
3
Wu et al.

50
()
q
s
c
a
l
c
/
q
s
m
e
a
s

(

)
Score: 0.3
Blom
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
3
10
2
10
1
10
0
10
1
10
2
10
3
Wu et al.

50
()
q
s
c
a
l
c
/
q
s
m
e
a
s

(

)
Score: 0.36
Blom
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 8.2: Transport rate ratio versus the Shields parameter for the formula of Wu et
al. (left: using initial bed material composition, right: using bed surface composition)
Surface layer as active layer
110 University of Twente
8.4 Transport composition
The predictions for the transport composition are more influenced by changes in the
active layer than those of the transport rate. Almost all the formulas result in better
predictions when the surface layer composition is used instead of the initial mixture
composition. All formulas predict a coarser or finer sediment transport when the
active layer composition becomes coarser or finer, respectively. In this section the
results of the three best formulas will be discussed. The best formula is the one of
Wu et al. with a score of 0.85. The second best formula is the one of Wilcock &
Crowe with a score of 0.77. The surface formula of Parker is the third best formula
with a score of 0.76.
The formula of Wu et al. already gave good predictions of the transport composition,
but they improve when the surface layer composition is used as input for the
predictions (Figure 8.3). If the composition of the active layer becomes finer or
coarser, the transport becomes finer or coarser, respectively. This gives better
results for the experiments BOMC, where the active layer becomes finer and tests
J06, J14 and J21, where the active layer becomes coarser.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
1
10
0
10
1
Wu et al.

50
()
D
m
,
c
a
l
c
/
D
m
,
m
e
a
s

(

)
Score: 0.74
Blom
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
1
10
0
10
1
Wu et al.

50
()
D
m
,
c
a
l
c
/
D
m
,
m
e
a
s

(

)
Score: 0.85
Blom
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 8.3: Transport composition ratio versus the Shields parameter for the formula
of Wu et al. (left: using initial bed material composition, right: using bed surface
composition).
The formula of Wilcock & Crowe shows improvements comparable to the formula of
Wu et al. (Figure 8.4). The left picture shows the predictions using the initial material
composition and the right the predictions using the surface layer composition.

Surface layer as active layer
University of Twente 111
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
1
10
0
10
1
Wilcock & Crowe

50
()
D
m
,
c
a
l
c
/
D
m
,
m
e
a
s

(

)
Score: 0.65
Blom
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
1
10
0
10
1
Wilcock & Crowe

50
()
D
m
,
c
a
l
c
/
D
m
,
m
e
a
s

(

)
Score: 0.77
Blom
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 8.4: Transport composition ratio versus the Shields parameter for the formula
of Wilcock & Crowe (left: using initial bed material composition, right: using bed
surface composition).
The formula of Parker gives much better results when the surface layer composition
is used for the active layer composition instead of the initial bed material
composition (Figure 8.5). The improved results are mostly found in the experimental
series J06, J14 and J21. The other three experimental series (Blom, J27, and BOMC)
do not show a big improvement. Parkers formula was calibrated using the
compositions of surface layers. The results show that the transport composition of
the formula is very sensitive to changes in the active layer composition and better
results are achieved when the surface layer composition is used instead of the initial
bed composition.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
1
10
0
10
1
Parker surface

50
()
D
m
,
c
a
l
c
/
D
m
,
m
e
a
s

(

)
Score: 0.52
Blom
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10
1
10
0
10
1
Parker surface

50
()
D
m
,
c
a
l
c
/
D
m
,
m
e
a
s

(

)
Score: 0.76
Blom
Wilcock et al: J06
Wilcock et al: J14
Wilcock et al: J21
Wilcock et al: J27
Wilcock & McArdell: BOMC

Figure 8.5: Transport composition ratio versus the Shields parameter for the formula
of Parker (left: using initial bed material composition, right: using bed surface
composition).
8.5 Conclusions
Most formulas give better sediment transport predictions using the surface layer
composition as input for these formulas instead of using the composition of the
initial bed material. It is logical that using the surface layer composition leads to
better results since the shear stress caused by the water flow is applied directly to
Surface layer as active layer
112 University of Twente
the bed surface, which gives the direct availability of the different grain size
fractions. The surface layer composition can differ from the composition of the initial
bed mixture in equilibrium situations due to vertical sorting.
The formula of Wilcock & Crowe [2001] shows the best results for the transport
rate predictions. The score of the formula is almost twice as good when the surface
layer composition is used instead of the initial bed material composition. The
formula of Wu et al. [2000] also shows better results, there is less scatter in the
predictions and the predictions are better.
The transport composition is very well predicted by the formula of Wu et al.
using the surface layer composition. The formulas of Wilcock & Crowe and Parker
[1990] give significantly better predictions for the transport composition when the
surface composition is used.


Discussion
University of Twente 113
9 Discussion
9.1 Modifications to Van Rijn
In this study, the formula of Van Rijn adapted to the case of non-uniform sediment
did not give good predictions. Kleinhans & Van Rijn [2001] propose a number of
modifications to this formula and create a fractional bed load formula based on a
stochastic representation of the grain shear stress. The formula was calibrated on
the tests of Blom & Kleinhans [1999] and according to Kleinhans & Van Rijn, it gave
good predictions for the sediment transport in two rivers. This formula could not be
verified within the time frame of this project.
9.2 Bed form factors
The bed form factors in the formulas of Meyer-Peter & Mller and Hunziker/Meyer-
Peter & Mller result in some difficulties (Sections 4.4, and Appendix 1). Bed form
factors denote the part of the bed roughness that is related to grain roughness and
not to bed form roughness. This means that at conditions of incipient motion, where
no bed forms are present, the bed form factor should be one. However, in the
formulas mentioned above, the bed form factor is much smaller than one at
conditions of incipient motion.
The bed form factors of Strickler, Yalin & Scheuerlein, and White-Colebrook all
have this problem. This means that the formulas of Meyer-Peter & Mller and
Hunziker/Meyer-Peter & Mller should be used carefully at conditions of incipient
motion.
Conclusions
University of Twente 115
10 Conclusions
In the first part of the study, validity ranges of a number of selected fractional
formulas were determined. In the second part, the behaviour of the transport
formulas was analysed for different hydraulic and sedimentological conditions. The
final part deals with the verification of the transport formulas with measurements of
ten sets of experimental data. In this chapter, the conclusions of each separate part
will be given, followed by some overall conclusions.
10.1 Validity ranges of fractional formulas
The validity ranges of the original fractional formulas were determined. Data used for
the original calibration of the formulas were analysed concerning the ranges of the
median grain size (D
50
), the geometric standard deviation of the bed material (!
g
),
and the Shields parameter ("
50
). The formula of Wu et al. [2000] has the largest
validity ranges of all analysed formulas. This formula is based on a large set of
experimental and field measurements. The fractional formulas of Ackers & White
[1973] with Day [1980] and of Ackers & White with Proffitt & Sutherland [1983] have
large validity ranges as well. Wilcock & Crowe [2002] and Hunziker [1995] used large
data sets for calibration and verification of their formulas, but the conditions in their
experiments did not vary much, so that their validity ranges are small.
10.2 Behaviour analysis
The behaviour of the transport formulas was analysed in terms of transport rate and
the transport composition as a function of the Shields parameter. This was done by
computing the transport rate and composition under different hydraulic and
sedimentological conditions. Four series of calculations were done with different bed
compositions: the bed mixtures varied in median grain size (2 or 10 mm) and
geometric standard deviation (2 or 5). Calculations were done for Shields parameters
up to 0.6.
10.2.1 Transport rate
The formulas of Wilcock & Crowe [2001], Parker [1990] and Engelund & Hansen
[1967] predict the highest sediment transport rates in their calculations. The
formulas of Van Rijn [1984a], Gladkow & Shngen [2000] and Hunziker/Meyer-Peter
& Mller [1995] predict the lowest transport rates.
Defining and predicting incipient motion is difficult, resulting in great
differences in modelling transport rates around incipient motion between the
formulas, i.e. for small Shields parameters ("
50
< 0.3).
The behaviour of the transport formulas appears not very sensitive to changes
in the grain size of the bed material. This is due to the fact that the transport is
shown as a function of the Shields parameter, in which the grain size is incorporated.
Conclusions
116 University of Twente
A variation in the geometric standard deviation of the bed material does influence
the predicted transport rates. The formulas of Engelund & Hansen and Van Rijn show
an increase in the transport rate when the geometric standard deviation (!
g
) is
increased, while Parker and Hunziker/Meyer-Peter & Mller show a decrease in the
transport rate. The formulas of Ackers & White and Meyer-Peter & Mller are not very
sensitive to a change in the geometric standard deviation of the bed material.
The formulas of Meyer-Peter & Mller and Hunziker/Meyer-Peter & Mller
predict incipient motion at high Shields parameters. One reason is the fact that the
bed form factors used in these formulas are smaller than one at conditions of
incipient motion. However, theoretically, the bed form factor should be equal to one
at these conditions: all bed shear stress is related to the grains since there are no
bed forms at incipient motion. However, even if the bed form factor is one at
conditions of incipient motion, incipient motion still occurs at relatively high Shields
parameters for the formulas of Meyer-Peter & Mller and Hunziker/Meyer-Peter &
Mller. This is an important problem of these formulas. The formulas use the
geometric mean grain size (D
m
) (eq. 2.1) for its Shields parameter, which leads to
problems in predicting sediment transport for strongly graded sediment mixtures.
There are little differences between the uniform formulas and their fractional
counterparts. The largest differences occur at Shields parameters below 0.25. Most
fractional formulas converge with their uniform formula at higher Shields
parameters.
10.2.2 Transport composition
For low Shields parameters, all formulas predict a transport composition that is much
finer than the bed composition. The composition of the transported material
coarsens gradually for higher Shields parameters. In some cases, the composition of
the transported material approximates the composition of the bed material when the
Shields parameter increases. For these high Shields parameters, the formulas of Wu
et al. and Hunziker/Meyer-Peter & Mller generally predict a transport composition
coarser than the bed composition.
The transport compositions predicted by the different formulas differ most for
Shields parameters smaller than 0.3, as predicting sediment transport at these
conditions of incipient motion is difficult. This has led to many different approaches
of modelling incipient motion.
Only the predictions of the transport composition by the two fractional
formulas of Ackers & White are sensitive to variations in the median grain size of the
bed material. All formulas are sensitive to variations in the geometric standard
deviation in the bed material. They all predict a finer transport composition when the
geometric standard deviation is increased. Only the formula of Hunziker/Meyer-
Peter & Mller shows opposite behaviour.
10.3 Verification with experimental data
Predictions of the sediment transport formulas were verified with measurements
from ten sets of experiments. These experiments were all carried out in flumes with
sediment recirculation systems. Only data of the equilibrium phases of the
Conclusions
University of Twente 117
experiments were used for verification. Both the transport rate and transport
composition were used for verification.
Since, the surface layer of the bed material determines the rate and composition of
the transported material, the composition of the surface layer should be used as
input for the transport formulas. However, for some experiments used for
verification, only data on the composition of the initial bed mixture was available. For
this reason, initially, we used the composition of the initial bed mixture as
composition of the active layer (Section 10.3.1). After that, for the experiments for
which data on the composition of the surface layer was available, we compared
prediction on the basis of i) the initial bed mixture and ii) the surface layer
(Section10.3.2).
10.3.1 Initial bed mixture as active layer
Transport rate
The formula of Wu et al. gives the best transport rate predictions. The formula
should be used with care at conditions with low Shields parameters and relatively
coarse material. At other conditions the formula gives good predictions.
The formula of Wilcock & Crowe is a good second option, yet this formula is
not reliable in the low Shields parameter range as well.
The sediment transport is extremely difficult to predict at Shields parameters smaller
than 0.1. The formulas of Gladkow & Shngen, Meyer-Peter & Mller and
Hunziker/Meyer-Peter & Mller frequently predict no sediment transport at all,
whereas formulas like Engelund & Hansen, Wu et al., Ribberink with Ashida & Michiue
and Van Rijn strongly overestimate the transport rate for many tests.
The bed form factors in the formulas of Meyer-Peter & Mller and of
Hunziker/Meyer-Peter & Mller seem to be too small in conditions with low Shields
parameters. This results in predicting no sediment transport, while, according to
measurements, sediment transport did occur.
Although only the formulas of Wu et al. and Ackers & White with Day are valid for the
conditions of the experiments of Klaassen, also the formulas of Meyer-Peter & Mller
and Hunziker/Meyer-Peter & Mller give good results for these experiments. This
means that these formulas are also suitable for transport rate predictions with
sediment mixtures finer and Shields parameters higher than their validity range.
The fractional formulas of Ackers & White with Day, Ackers & White with Proffitt &
Sutherland, Meyer-Peter & Mller with Egiazaroff and Meyer-Peter & Mller with
Ashida & Michiue give better predictions than their uniform counterparts.

Conclusions
118 University of Twente
Transport composition
Also for the transport composition the formula of Wu et al. shows the best
predictions. For conditions with low Shields parameters, the formula shows small
underestimations of the geometric mean grain size of the transported mixture.
All formulas have problems predicting the transport composition for low Shields
parameters. Some formulas predict no sediment transport at all, while other
formulas show a large variation in the predicted compositions. Only the formula of
Wu et al. shows good and stable predictions for the transport composition.
All fractional formulas predict the transport composition of the experiments of
Klaassen well. Only the formulas of Ackers & White with Day, Wu et al. are valid for
these experiments, but the other formulas are also suitable for predictions with
sediment this fine and Shields parameters this high.
The formulas of Meyer-Peter & Mller, Hunziker/Meyer-Peter & Mller, and Gladkow
& Shngen have problems predicting sediment transport at low Shields parameters.
This is the main reason why the scores for the transport compositions of the
formulas are poor.
10.3.2 Surface layer as active layer
When the bed surface is used as the active layer, which is defined as the part of the
bed that determines the rate and composition of the transported sediment, the
transport predictions of most formulas are better. The surface layer composition was
only determined for the experiments of Blom [2000], Wilcock & McArdell [1993], and
Wilcock et al. [2001]. Wilcock & Crowe [2001] now lead to the best transport rate
predictions of all the formulas. Nonetheless, we need to remark that the
measurements of the surface composition in the experiments by Wilcock & McArdell,
and Wilcock et al. were used for the development of their formula. When the surface
layer composition is used the large variation at the lower Shields parameters has
strongly decreased. The formula of Wu et al. [2000] shows better results with the
surface layer as active layer, as well. It still strongly overpredicts the transport rate
at low Shields parameters, yet it shows much less scatter.
The transport composition predicted by Wu et al. improves when the bed
surface composition is used as input. The predictions agree very well with the
measurements, although at low Shields parameters the predicted sediment
composition is finer than measured. The formula of Wilcock & Crowe shows good
results too, although somewhat worse than Wu et al., because it shows more
variation in the Shields parameter range 0.05 - 0.25.
The use of the surface layer as active layer does not only give better results,
there is also a theoretical basis why the surface layer should be used as input for the
sediment transport predictions. Only the material present directly at the bed surface
is exposed to the water flow and can be transported.
Conclusions
University of Twente 119
10.4 Final conclusions
Among the transport formulas analysed in this study, the formula of Wu et al. is the
best formula for sediment transport predictions. The formula has large validity
ranges due to the large data set that was used for the original calibration of the
formula. The formula is the best formula to predict the transport rate and the best to
predict the transport composition when the composition of the initial bed mixture is
used as input for the transport formulas. If the bed surface composition is used, the
formula of Wu et al. is also the best formula, together with the formula of Wilcock &
Crowe.
The formula of Meyer-Peter & Mller with the hiding/exposure correction of Ashida
& Michiue [1973] shows good results for the transport rate predictions. However, the
formula does not perform well when predicting the composition of the transported
sediment. The validity range of this formula is much smaller than the formula of Wu
et al., but it shows good predictions outside its validity range as well. The formula is
not suitable to conditions with Shields parameters below 0.1. The formula often
predicts no sediment transport in this Shields parameter range, whereas experiments
of Day [1980], Wilcock & McArdell [1993] and Wilcock et al. [2001] show that
sediment transport did occur in these conditions.
Using the bed surface composition instead of the composition of the initial bed
mixture as the active layer composition leads to better predictions for most
formulas. The use of the surface layer has also a better theoretical basis because the
shear stress acts directly on the bed surface.
10.5 Recommendations
In this research, data from only flume experiments was used for the verification of
the sediment transport formulas. Although many tests were used with large ranges
in hydraulic and sedimentological conditions, it is recommended that the transport
formulas be also verified with field measurements. It is important to know how the
formulas behave in a natural environment where uncertainties in the measurements
are much larger. However, when selecting field data, it is important to know how
reliable the measurements are. The hydraulic conditions can be measured reasonably
well, but the composition of the bed surface layer introduces a lot of problems.
Three measurements are needed for verification: hydraulic conditions, sediment
transport, and the active layer composition. It is difficult to find combined
measurements that were taken at the same time and at the same place.
More research is needed for a definition of the active layer, i.e. the material that
determines the transport rate and composition of the transported material. The
experiments of Wilcock & McArdell [1993], Wilcock et al. [2001], Blom & Kleinhans
[1999], and Blom [2000] give a good basis, but more coupled measurements of the
composition of the surface layer, sediment transport (rate and composition), and
hydraulic conditions are required.
Conclusions
120 University of Twente
The influence of the geometric mean grain size of the mixture in the active layer on
the predicted sediment transport should be investigated. The commonly used
definition of the geometric mean grain size leads to problems predicting the
sediment transport of widely graded sediment. Parker (1990) gives a different
definition, which seems to solve a part of our problems.
The modified formula of Van Rijn [Kleinhans & Van Rijn, 2001] shows promising
results according to the developers of this formula. It is recommended to verify this
formula with a larger data set and compare its results with results from other
formulas. Introducing the stochastic approach of Kleinhans & Van Rijn [2001] into
other transport formulas might lead to better results as well. Especially in conditions
of incipient motion, the stochastic approach seems very useful.
It is useful to test the formula of Wu et al. [2000] in a graded morphological model.
This will show if the formula is also able to predict morphological changes and if the
formula introduces stability problems into morphological models.

References
University of Twente 121
References
Ackers, P. and White, W.R., (1973), Sediment transport: a new approach and analysis,
Proc. ASCE, J. of the Hydr. Div., Vol. 99, HY11, pp. 2041-2060.
Ashida, K. and Michiue, M., (1971), An investigation of river bed degradation
downstream of a dam, Proc. 14
th
Congress IAHR, Vol. 13, Paris.
Ashida, K. and Michiue, M., (1973), Studies on bed load transport rate in alluvial
streams, Trans. JSCE, Vol. 4.
Blom, A. and Kleinhans, M., (1999), Non-uniform sediment in morphological
equilibrium situations - Data report Sand Flume experiments 97/98, Research
report CiT 99R-002/MICS-001, Civil Engineering and Management, University of
Twente.
Blom, A., (2000), Flume experiments with a trimodal sediment mixture - Data report
Sand Flume experiments 99/00, Research report CiT: 2000R-004/MICS-013,
Civil Engineering & Management, University of Twente.
Blom, A. and Ribberink, J.S., (1999), Non-uniform sediment in rivers: vertical
sediment exchange between bed layers, Proc. IAHR Symposium on River, Coastal
and Estuarine Morphodynamics, Genova, Italy 6-10 September 1999, pp. 45-54.
Blom, A., Ribberink, J.S. and Scheer, P. van der, (2000), Sediment transport in flume
experiments with a trimodal sediment mixture, CD-rom contribution, Gravel-bed
Rivers Workshop, New Zealand, 27 Aug-2 Sept.
Bredius, J., (1998), Verificatie van transportformules voor gegradeerd sediment, (in
Dutch), M.Sc. thesis Civil Engineering and Management, University of Twente.
Colebrook C. F., (1939), Turbulent flow in pipes, with particular reference to the
transition region between the smooth and rough pipe laws, Institution of Civ.
Eng. journal v11, paper no. 5204.
Day, T.J., (1980), A study of the transport of graded sediments, HRS Wallingford,
Report No. IT 190.
Egiazaroff, I.V., (1965), Calculation of non-uniform sediment concentrations, J. of
Hydr. Div., ASCE, Vol. 91, No. 4, pp. 225-248.
Engelund, F. and Hansen, E., (1967), A monograph on sediment transport in alluvial
streams, Teknisk Forlag, Copenhagen, Denmark.
Gladkow, G.L., and Shngen, B., (2000), Modellierung des Geschiebetransports mit
unterschiedlicher Korngre in Flssen, Mitteilungsblatt der Bundesanstalt fr
Wasserbau, Vol. 82, pp 123-129.
Gnter, A., (1971), Die kritische mittlere Sohlenschubspannung bei
Geschiebemischungen unter Bercksichtigung der Deckschichtbildung und der
turbulenzbedingten Sohlenschub-spannungsschwankungen, Mitteilungen Nr. 3
der Versuchsanstalt fr Wasserbau, Hydrologie und Glaziologie, ETH Zrich.
References
122 University of Twente
Hirano, M., (1970), On the river bed degradation downstream of a dam and the
armoring phenomena, Proc. 14
th
Conference on Hydraulics, (in Japanese).
Hunziker, R.P., (1995), Fraktionsweiser Geschiebetransport, Ph.D. thesis Mitteilungen
Nr. 138 der Versuchsanstalt fr Wasserbau, Hydrologie und Glaziologie, ETH
Zrich, Switserland.
Iwagaki, Y., (1956), Hydrodynamical Study on Critical Tractive Force, Trans. of JSCE,
Vol. 41, Tokyo.
Klaassen, G.J., (1991), Experiments on the effect of gradation and vertical sorting on
sediment transport phenomena in the dune phase, Proc. Grain Sorting Seminar,
Ascona, Switzerland, pp 127-145.
Kleinhans, M.G. and Blom, A., (2001), Discussion of (The Flow, The Bed, and the
Transport: Interaction in the Flume and the Field, by Wilcock, 2001), proc. Gravel-
Bed Rivers V, edit. Mosley, M.P., The Caxton Press, pp. 212-214, Christchurch.
Kleinhans, M.G. and Rijn, L.C. van, (2002), Stochastic prediction of sediment
transport in sand-gravel bed rivers, J. of Hydr. Eng., Vol. 128, No. 4, pp. 412-
425.
Kuhnle, R.A., (1993), Fluvial transport of sand and gravel mixtures with bimodal size
distributions, Sedimentary Geology, Vol. 85, pp. 17-24.
Laguzzi, M., (1994), Modelling of sediment mixtures, Report No. Q 1660, WL | delft
hydraulics.
Meyer-Peter, E. and Mller, R., (1948), Formulas for bed-load transport, Proc. 2
nd

Congress IAHR, Stockholm, Sweden.
Nizery, A. and Braudeau, G., (1953), Variation de la Granulometrie de Charriage dans
une section de Riviere, Proc. 5
th
Congress IAHR, Minnesota, pp 49-60.
Pantlopulos, J., (1955), Note sur la granulometrie de charriage et la loi du debit
solide par charriage de fond dun melange de materiaux, Proc. 6
th
Congress IAHR,
The Hague, Vol 4, pp D10/1-D10/11.
Pantlopulos, J., (1957), tude Exprimentale du mouvement par harriage de fond
dun mlange de matriaux recherches sur la similitude du charriage, Proc. 7
th

Congress IAHR, Lissabon, pp D30/1-D30/24.
Parker, G., Klingeman, P.C. and McLean, D.G., (1982), Bedload and size distribution
in paved gravel-bed streams, Proc. ASCE, J. of the Hydr. Div., Vol. 108, HY 4, pp.
544-571.
Parker, G. and Klingeman, P.C., (1982), On why gravel bed streams are paved, Water
Resources Res., Vol. 18, No. 5, pp. 1409-1423.
Parker, G., (1990), Surface-based bedload transport relation for gravel rivers, J. of
Hydr. Res., Vol. 28, No. 4, pp. 417-436.
Proffitt, G.T. and Sutherland, A.J., (1983), Transport of non-uniform sediments, J. of
Hydr. Res., Vol. 21, No. 1, pp. 33-43.
References
University of Twente 123
Ribberink, J.S., (1981), Bed-load formulae for non-uniform sediment, Delft University
of Technology, Department of Civil Engineering, Fluid Mechanics Group, Internal
Report no. 4-78, The Netherlands.
Ribberink, J.S., (1987), Mathematical modelling of one-dimensional morphological
changes in rivers with non-uniform sediment, Ph.D. thesis, Report No. 87-2,
Comm. On Geot. And Hydr. Eng., Delft Univ. of Technology, The Netherlands.
Ribberink, J.S., (1998), Bed-load transport for steady flows and unsteady oscillatory
flows, Coastal Eng., Vol. 34, pp. 59-82.
Rijn, L.C. van, (1993), Principles of sediment transport in rivers, estuaries and coastal
seas, Aqua Publications, Amsterdam, The Netherlands.
Rijn, L.C. van, (1984a), Sediment transport, Part I: Bed Load Transport, J. of Hydr.
Eng., ASCE, Vol. 110, No. 10, pp. 1431-1456.
Rijn, L.C. van, (1984b), Sediment transport, Part II: Suspended Load Transport, J. of
Hydr. Eng., ASCE, Vol. 110, No. 11, pp. 1613-1641.
Samaga, B.R., Ranga Raju, K.G. and Garde, R.J., (1986a), Bed Load Transport of
Sediment Mixtures, J. Hydr. Eng., Vol. 112, No. 11, pp. 1003-1018.
Samaga, B.R., Ranga Raju, K.G. and Garde, R.J., (1986b), Suspended Load Transport
of Sediment Mixtures, J. Hydr. Eng., Vol. 112, No. 11, pp. 1019-1035.
Scheer, P. van der, (2000), Transport Formulae for Graded Sediment - Verification
with Flume Data 1999-2000, M.Sc. thesis, University of Twente, Enschede.
Strickler, A., (1923), Beitrge zur Frage der Geschwindigkeitsformel und der
Rauhigkeitszahlen fr Strme, Kanle und geschlossene Leitungen, Mitt. No. 16
des Amtes fr Wasserwirtschaft, Eidgenssisches Departement des Innern, Bern.
Suzuki, K. and Hano A., (1992), Grain Size change of bed surface layer and sediment
discharge of an equilibrium river bed, Proc. Grain Sorting Seminar, Ascona,
Switzerland, pp 151-156.
Wallingford, (1990), Sediment transport: the Ackers & White theory revised, Report
SR 237, HR Wallingford.
Wilcock, P.R. and McArdell, B.W., (1993), Surface-Based Fractional Transport Rates:
Mobilization Thresholds and Partial Transport of a Sand-Gravel Sediment, Water
Resources Res., Vol. 29, No. 4, pp. 1297-1312.
Wilcock, P.R., (2001), The Flow, The Bed, and the Transport: Interaction in the Flume
and the Field, proc. Gravel-Bed Rivers V, edit. Mosley, M.P., The Caxton Press, pp.
183-219, Christchurch.
Wilcock, P.R., Kenworthy, S.T. and Crowe, J.C., (2001), Experimental study of the
transport of mixed sand and gravel, Water Resources Res., Vol. 37, No. 12, pp.
3349-3358.
Wilcock, P.R. and Crowe, J.C. (2001), A surface-based transport model for sand and
gravel, submitted to J. Hydr. Eng.
References
124 University of Twente
Wilcock, P.R. and Kenworthy, S.T., (2001), A two fraction model for the transport of
sand/gravel mixtures, submitted to Water Resources Res.
Williams, G.P. and Rosgen, D.L., (1989), Measured total sediment loads (suspended
loads and bedloads) for 93 United States streams, USGS Open_File Report 89-67,
Denver, CO.
Wu, W., Wang, S.S.Y. and Jia, Y, (2000), Nonuniform sediment transport in alluvial
rivers, J. of Hydr. Res., Vol. 38, No. 6, pp 427-434.
Yalin, M.S., Scheuerlein, H., (1988), Friction Factors in alluvial rivers, Bericht Nr. 59,
Instituts fr Wasserbau und Wassermengenwirtschaft, Oskar van Miller Institut,
Obernach.
Zarn, B., (1997), Einfluss des Flussbettbreite auf die Wechselwirkung zwischen
Abfluss, Morphologie und Geschiebetransportkapazitt, Mitteilung Nr. 154 der
Versuchsanstalt fr Wasserbau, Hydrologie und Glaziologie, ETH Zrich,
Switserland.
Appendix 1
University of Twente A-1
Appendix 1: Bed roughness analysis
In general, three expressions for the bed roughness are used. These are given by:
2
1
2
1
i h
f
g 8
u ' ' = Darcy-Weisbach
2
1
3
2
i h
n
1
u ' ' = Manning
2
1
2
1
i h C u ' ' = Chzy
The relations between the different roughness coefficients is:
f
g 8
n
h
C
6
1
= =
C Chzy value [m
1/2
/s]
n Manning coefficient [s/m
1/3
]
f Darcy-Weisbach coefficient [-]
The formulas of Meyer-Peter & Mller [1948] use the following expression for the
bed form factor:
2
3
k
k
)
*
+
,
-
.
=
'
bed form factor [-]
The formula of Strickler [1923] is used for the calculation of k and k:
6
h
c
k = total roughness according to Strickler [m
1/3
/s]
6
90
D
c
k = ' grain roughness according to Strickler [m
1/3
/s]
It should be noted that Stricklers roughness is the inverse roughness of Mannings
roughness. The constant c is not dimensionless, but has the dimension [m
1/2
/s]. The
value of c varies in expressions of different authors. Meyer-Peter & Mller [1948] use
c = 26, while Van Rijn [1993] uses c = 25.
The total bed shear stress (/) can be divided in a shear stress related to the grains (/)
and a shear stress related to the bed forms (/):
' ' ' / / / + =
This relation can be rewritten for Chzy values which results in:
2 2 2
C
1
C
1
C
1
' ' '
+ =
The bed form factor of Meyer-Peter & Mller can now be rewritten as:
Appendix 1
A-2 University of Twente
2
3
2
3
2
3
C
C
n
n
k
k
)
*
+
,
-
.
= )
*
+
,
-
.
= )
*
+
,
-
.
=
'
'
'

The grain roughness by Strickler can now be rewritten to the other grain roughness
predictors:
6
90
6
D
26
h
C
n
1
k = = =
'
'
'
Which leads to:
6
1
90
D
h
26 C
)
)
*
+
,
,
-
.
= ' (1)
Stricklers equation approximates the White-Colebrook [1939] formula. This
approximation can only be used in the Chzy value range 40 to 70 m
1/2
/s. The
White-Colebrook relation for the grain related Chzy value yields:
)
)
*
+
,
,
-
.
=
90
D
h 12
18 C log ' grain related Chzy value according to White-Colebrook (2)
The equation above can be seen as the White-Colebrook expression of the original
Meyer-Peter & Mller expression for the grain roughness.
Note that the expression for the grain related Chzy value used in the bed load
formula of Van Rijn [1984a] differs from the one used in the Meyer-Peter & Mller
formula.
)
)
*
+
,
,
-
.
'
'
=
90
D 3
h 12
18 C log ' grain related Chzy value used by Van Rijn [1984a]
In the figure below, the differences between equations 1 and 2 can be seen.

Appendix 1
University of Twente A-3
Comparison grain roughness predictors
0
10
20
30
40
50
60
70
80
0 50 100 150 200 250 300 350 400
h/D90 (-)
G
r
a
i
n

r
o
u
g
h
n
e
s
s

(
m
1
/
2
/
s
)
Strickler
White-Colebrook


Appendix 2
University of Twente A-5
Appendix 2: Splitting a sediment mixture into
n fractions
In this appendix, a summary is given of the equations used to split a sediment
mixture with a log-normal distribution in n size fractions. Only the equations used in
the MATLAB model are given in this appendix. Ribberink [1987] gives a more
thorough description. The log-normal probability distribution of grain diameter D
50

can be written as:
{ }
2
y
2
y 50
50 y
50
D 5 0
2 D
1
D f !
F !
) (ln . exp ) ( $ $ = log-normal probability distribution
In which:
50 y
D ln = mean value of y
995 0
g
y
.
ln!
! = standard deviation of y
The standard normal distribution is given by z. A lower and upper boundary of the
grain size distribution (D
l
and D
u
) were chosen according to:
% . ) ( ) ( 28 2 D D P D D P
u l
= 2 = 8
These boundaries coincide with the following values of z:
z
l
= -2 lower boundary z
z
u
= 2 upper boundary z
The region between these boundaries is divided into n equal parts %z:
n
z z
z
l u
$
= %
The z-value of each fraction i (z
i
) is:
z
2
1 i 2
z z
l i
% '
$
+ =
) (

This can be transformed into D
i
using:
) exp(
y y i i
z D ! + = grain size of fraction i (m)
In order to determine the probabilities of each size fraction, an upper (z
ui
) and lower
(z
li
) boundary of each fraction i is necessary:
z 5 0 z z
i li
% $ = .
z 5 0 z z
i ui
% + = .
The probability of fraction i (p
i
) can be determined with:
Appendix 2
A-6 University of Twente
{ } ) . ( ) . (
.
.
li ui i
z 2 5 0 erf z 2 5 0 erf
9545 0
5 0
p ' $ ' =
Note: the probability of fraction i is divided by a factor 0.9545 to correct for the
upper and lower boundaries of the grain size distribution (1 - 2 * 0.0228 G 0.9545).
The sediment transport formulas use a number of representative grain size
diameters, which are calculated as follows:
) . exp( 995 0
D
D
y
50
16
$ ' $
=
!

) . exp( 385 0
D
D
y
50
35
$ ' $
=
!

) . exp( 995 0
D
D
y
50
84
' $
=
!

) . exp( 281 1
D
D
y
50
90
' $
=
!


Note: Due to the discretisation of the log-normal distribution some errors are
introduced when calculating the geometric mean grain size. There is an error
between the two ways the geometric mean grain size can be calculated.
) . exp(
2
y 50 m
5 0 D D ! ' = continuous geometric mean grain size (m)
and
&
' =
i
i i m
D p D discrete geometric mean grain size (m)
The error caused by the discretisation is caused by the lower and upper boundary (z
l

and z
u
) and the number of fractions. If these boundaries are set on higher values
(e.g. -10 and 10 respectively) and the number of fractions is set very large (e.g. 100),
the discrete geometric mean grain size approaches the continuous geometric mean
grain size.

Appendix 3
University of Twente A-7
Appendix 3: Transport rate Series 1 and 2
0 0.1 0.2 0.3 0.4 0.5 0.6
10
-8
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
Serie 1: D
50
=2mm, ! !! !
g
=2
"
50
(-)
q
s

(
m
3
/
s
/
m
)
Ackers & White + Day
Ackers & White + Proffitt & Sutherland
Parker surface based
Hunziker/Meyer-Peter & Mller
Wilcock & Crowe
Wu et al.
0 0.1 0.2 0.3 0.4 0.5 0.6
10
-8
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
Serie 1: D
50
=2mm, ! !! !
g
=2
"
50
(-)
q
s

(
m
3
/
s
/
m
)
Engelund & Hansen fractional
Gladkow & Shngen
Meyer-Peter & Mller + Egiazaroff
Meyer-Peter & Mller + Ashida & Michiue
Van Rijn fractional
Ribberink + Ashida & Michiue

0 0.1 0.2 0.3 0.4 0.5 0.6
10
-8
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
Serie 2: D
50
=2mm, ! !! !
g
=5
"
50
(-)
q
s

(
m
3
/
s
/
m
)
Ackers & White + Day
Ackers & White + Proffitt & Sutherland
Parker surface based
Hunziker/Meyer-Peter & Mller
Wilcock & Crowe
Wu et al.
0 0.1 0.2 0.3 0.4 0.5 0.6
10
-8
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
Serie 2: D
50
=2mm, ! !! !
g
=5
"
50
(-)
q
s

(
m
3
/
s
/
m
)
Engelund & Hansen fractional
Gladkow & Shngen
Meyer-Peter & Mller + Egiazaroff
Meyer-Peter & Mller + Ashida & Michiue
Van Rijn fractional
Ribberink + Ashida & Michiue

q
s

(
m
3
/
s
/
m
)

q
s

(
m
3
/
s
/
m
)

q
s

(
m
3
/
s
/
m
)

q
s

(
m
3
/
s
/
m
)

Appendix 4
University of Twente A-9
Appendix 4: Transport composition Series 1
and 2
0 0.1 0.2 0.3 0.4 0.5 0.6
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Serie 1: D
50
=2mm, ! !! !
g
=2
"
50
(-)
Ackers & White + Day
Ackers & White + Proffitt & Sutherland
Parker surface based
Hunziker/Meyer-Peter & Mller
Wilcock & Crowe
Wu et al.
0 0.1 0.2 0.3 0.4 0.5 0.6
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Serie 1: D
50
=2mm, ! !! !
g
=2
"
50
(-)
Engelund & Hansen fractional
Gladkow & Shngen
Meyer-Peter & Mller + Egiazaroff
Meyer-Peter & Mller + Ashida & Michiue
Van Rijn fractional
Ribberink + Ashida & Michiue

0 0.1 0.2 0.3 0.4 0.5 0.6
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Serie 2: D
50
=2mm, ! !! !
g
=5
"
50
(-)
Ackers & White + Day
Ackers & White + Proffitt & Sutherland
Parker surface based
Hunziker/Meyer-Peter & Mller
Wilcock & Crowe
Wu et al.
0 0.1 0.2 0.3 0.4 0.5 0.6
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Serie 2: D
50
=2mm, ! !! !
g
=5
"
50
(-)
Engelund & Hansen fractional
Gladkow & Shngen
Meyer-Peter & Mller + Egiazaroff
Meyer-Peter & Mller + Ashida & Michiue
Van Rijn fractional
Ribberink + Ashida & Michiue

D
m
T
/
D
m
B

(
-
)

D
m
T
/
D
m
B

(
-
)

D
m
T
/
D
m
B

(
-
)

D
m
T
/
D
m
B

(
-
)

Appendix 5
University of Twente A-11
Appendix 5: Uniform versus fractional for
Series 1 and 2


Serie 1
0.1
1.0
10.0
0.00 0.10 0.20 0.30 0.40 0.50 0.60
Shields D50 (-)
q
s
_
f
r
/
q
s
_
u

(
-
)
Ackers & White + Day
Ackers & White + Proffitt & Sutherland
Meyer-Peter & Muller + Egiazaroff
Meyer-Peter & Muller + Ashida & Michiue
Serie 2
0.1
1.0
10.0
0.00 0.10 0.20 0.30 0.40 0.50 0.60
Shields D50 (-)
q
s
_
f
r
/
q
s
_
u

(
-
)
Ackers & White + Day
Ackers & White + Proffitt & Sutherland
Meyer-Peter & Muller + Egiazaroff
Meyer-Peter & Muller + Ashida & Michiue
Appendix 6
University of Twente A-13
Appendix 6: Vanoni and Brooks
Vanoni and Brooks [1957] made a correction for the sidewall roughness for flume
experiments with a width/depth ratio smaller than 5. This correction was made
because the sidewall roughness has a big influence under these circumstances and
since only the bed roughness is relevant for the sediment transport formulas. The
correction factor is applied to the hydraulic radius and the Chzy coefficient.
w b b
i R g u ' ' =
*,
shear velocity related to the bed (m/s)
R
f
f
R
b
b
' = hydraulic radius related to the bed (m)
h 2 b
h b
R
' +
'
= hydraulic radius (m)
2
u
u
8 f )
*
+
,
-
.
' =
*
friction coefficient (-)
w
i R g u ' ' =
*
shear velocity (m/s)
h b
Q
u
'
= mean flow velocity (m/s)
) (
b b
f f
b
h 2
f f $ '
'
+ = friction coefficient related to the bed (-)
1884 0
f
0428 0
f
0026 0 f
2
w
.
Re
log .
Re
log . + )
*
+
,
-
.
' $
)
)
*
+
,
,
-
.
)
*
+
,
-
.
' = friction coefficient
related to smooth sidewalls (10
5
" Re/f < 10
8
) (-)
1
R u 4 ' '
= Re Reynolds number (-)
h water depth (m)
i
w
water surface slope (-)
b width (-)
Q discharge (m
3
/s)
1 kinematic viscosity coefficient (m
2
/s)
Appendix 7
University of Twente A-15
Appendix 7: Data Wilcock
Hydraulic data Wilcock & McArdell [1993] and Wilcock et al. [2001]
Test Q h Rb Cb visc "
50
qs DmT
m3/s m m m1/2/s m2/s - m2/s mm
J06-1 0.047 0.104 0.095 26.42 1.11E-06 0.028 8.51E-11 5.40
J06-2 0.052 0.108 0.102 27.75 1.11E-06 0.032 9.69E-10 6.14
J06-3 0.058 0.104 0.094 26.53 1.11E-06 0.059 3.55E-08 10.29
J06-4 0.062 0.102 0.086 28.61 1.11E-06 0.082 4.37E-07 16.23
J06-5 0.054 0.103 0.094 27.9 1.11E-06 0.042 7.70E-09 9.12
J06-6 0.063 0.103 0.092 30.28 1.11E-06 0.058 1.65E-07 10.59
J06-7 0.073 0.106 0.103 25.28 1.11E-06 0.102 5.56E-06 17.19
J06-8 0.047 0.105 0.099 31.49 1.11E-06 0.036 1.27E-09 6.73
J06-9 0.077 0.109 0.101 27.71 1.11E-06 0.117 1.13E-05 16.48
J06-10 0.08 0.108 0.118 25.11 1.11E-06 0.134 7.82E-05 17.60
J14-1 0.076 0.117 0.102 26.34 1.11E-06 0.143 1.97E-05 14.18
J14-2 0.075 0.109 0.112 25.87 1.11E-06 0.140 2.85E-05 14.50
J14-3 0.05 0.107 0.107 30.87 1.11E-06 0.048 1.15E-08 6.61
J14-4 0.061 0.104 0.094 30.79 1.11E-06 0.082 6.82E-07 8.84
J14-5 0.066 0.106 0.114 25.71 1.11E-06 0.113 4.79E-06 12.15
J14-6 0.047 0.102 0.101 36.45 1.11E-06 0.033 7.36E-09 7.27
J14-7 0.057 0.106 0.106 30.69 1.11E-06 0.065 5.90E-07 7.94
J14-8 0.055 0.106 0.097 31.23 1.11E-06 0.060 2.10E-07 8.10
J14-9 0.08 0.117 0.131 25.19 1.11E-06 0.136 4.41E-05 14.83
J21-1 0.076 0.118 0.105 26.58 1.11E-06 0.161 5.17E-05 12.71
J21-2 0.047 0.108 0.096 35.85 1.11E-06 0.041 1.87E-07 4.84
J21-3 0.053 0.102 0.095 33.45 1.11E-06 0.064 2.34E-06 5.39
J21-4 0.06 0.105 0.097 28.64 1.11E-06 0.105 4.56E-06 7.38
J21-5 0.044 0.109 0.1 36.27 1.11E-06 0.033 5.10E-08 4.06
J21-6 0.054 0.104 0.094 31.73 1.11E-06 0.071 4.90E-06 5.32
J21-7 0.039 0.099 0.09 38.94 1.11E-06 0.028 6.40E-09 4.43
J21-8 0.067 0.102 0.096 27.16 1.11E-06 0.153 5.82E-05 10.81
J27-1 0.039 0.102 0.098 38.03 1.11E-06 0.032 1.85E-07 2.81
J27-2 0.054 0.101 0.101 33.54 1.11E-06 0.076 1.40E-05 4.12
J27-3 0.03 0.11 0.112 42.51 1.11E-06 0.012 1.12E-09 1.76
J27-4 0.034 0.101 0.098 35.71 1.11E-06 0.028 7.85E-08 2.35
J27-5 0.049 0.093 0.09 34.01 1.11E-06 0.074 1.06E-05 5.00
J27-6 0.045 0.098 0.094 38.33 1.11E-06 0.045 2.15E-06 3.45
J27-7 0.062 0.106 0.101 34.17 1.11E-06 0.091 2.56E-05 5.35
J27-8 0.068 0.106 0.098 34.14 1.11E-06 0.112 4.02E-05 6.17
J27-9 0.075 0.106 0.082 34.77 1.11E-06 0.163 1.30E-04 9.03
J27-10 0.078 0.111 0.106 27.79 1.11E-06 0.201 2.98E-04 10.85
BOMC-14c 0.017 0.111 0.098 33.86 1.11E-06 0.009 8.85E-10 0.60
BOMC-7a 0.021 0.11 0.098 33.42 1.11E-06 0.013 1.26E-08 0.55
BOMC-14b 0.022 0.109 0.096 35.57 1.11E-06 0.014 1.48E-08 0.54
BOMC-7b 0.024 0.111 0.098 34.38 1.11E-06 0.017 3.65E-08 0.54
BOMC-7c 0.029 0.105 0.094 35.79 1.11E-06 0.024 1.64E-07 0.57
BOMC-1 0.04 0.12 0.104 41.01 1.11E-06 0.030 2.22E-06 0.56
BOMC-2 0.04 0.112 0.102 33.32 1.11E-06 0.049 2.72E-06 0.83
BOMC-6 0.047 0.096 0.088 33.29 1.11E-06 0.091 4.79E-05 1.90
BOMC-4 0.049 0.094 0.086 33.81 1.11E-06 0.099 6.02E-05 2.07
BOMC-5 0.057 0.088 0.083 29.53 1.11E-06 0.195 2.19E-04 5.97
Initial bed and surface layer composition of the tests of Wilcock & McArdell [1993] and
Wilcock et al. [2001].

Appendix 7
A-16 University of Twente
fraction 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Size (mm) 0.355 0.75 1.2 1.7 2.4 3.4 4.85 6.85 9.65 13.65 19.3 27.3 38.65 54.65
test p_1 p_2 p_3 p_4 p_5 p_6 p_7 p_8 p_9 p_10 p_11 p_12 p_13 p_14
J06 initial 0.001 0.034 0.016 0.012 0.037 0.084 0.120 0.097 0.083 0.091 0.143 0.139 0.084 0.060
J06-1 0.000 0.001 0.000 0.002 0.004 0.038 0.103 0.125 0.118 0.132 0.199 0.178 0.065 0.035
J06-2 0.000 0.001 0.000 0.002 0.004 0.038 0.103 0.125 0.118 0.132 0.199 0.178 0.065 0.035
J06-3 0.000 0.000 0.000 0.001 0.003 0.068 0.069 0.111 0.117 0.133 0.195 0.186 0.074 0.043
J06-4 0.000 0.000 0.000 0.001 0.004 0.077 0.052 0.111 0.104 0.131 0.211 0.192 0.078 0.041
J06-5 0.000 0.000 0.000 0.001 0.003 0.038 0.075 0.104 0.106 0.128 0.193 0.174 0.104 0.073
J06-6 0.000 0.000 0.000 0.001 0.005 0.038 0.088 0.098 0.102 0.120 0.177 0.187 0.109 0.075
J06-7 0.000 0.000 0.000 0.001 0.006 0.040 0.103 0.117 0.095 0.103 0.179 0.183 0.094 0.080
J06-8 0.000 0.000 0.000 0.000 0.005 0.037 0.105 0.133 0.108 0.118 0.175 0.172 0.090 0.057
J06-9 0.000 0.000 0.000 0.002 0.008 0.041 0.101 0.094 0.089 0.103 0.172 0.217 0.102 0.074
J06-10 0.000 0.000 0.000 0.001 0.005 0.055 0.128 0.117 0.101 0.095 0.148 0.183 0.101 0.067
J14 initial 0.001 0.077 0.037 0.034 0.033 0.076 0.109 0.088 0.076 0.083 0.130 0.126 0.076 0.054
J14-1 0.000 0.001 0.001 0.015 0.016 0.085 0.054 0.110 0.091 0.105 0.166 0.203 0.102 0.050
J14-2 0.000 0.001 0.003 0.007 0.012 0.076 0.067 0.105 0.112 0.096 0.163 0.197 0.105 0.056
J14-3 0.000 0.000 0.000 0.006 0.012 0.118 0.068 0.129 0.117 0.118 0.152 0.154 0.077 0.049
J14-4 0.000 0.002 0.002 0.013 0.010 0.092 0.051 0.118 0.099 0.122 0.163 0.169 0.095 0.066
J14-5 0.000 0.001 0.001 0.007 0.013 0.113 0.054 0.108 0.089 0.102 0.178 0.175 0.088 0.072
J14-6 0.000 0.002 0.003 0.012 0.015 0.124 0.070 0.134 0.108 0.103 0.132 0.151 0.089 0.058
J14-7 0.000 0.002 0.004 0.012 0.021 0.121 0.067 0.114 0.107 0.099 0.131 0.165 0.096 0.063
J14-8 0.000 0.001 0.001 0.011 0.015 0.110 0.062 0.130 0.111 0.111 0.152 0.165 0.085 0.046
J14-9 0.000 0.000 0.001 0.010 0.006 0.086 0.054 0.126 0.091 0.105 0.160 0.201 0.099 0.063
J21 initial 0.001 0.106 0.051 0.048 0.031 0.071 0.102 0.082 0.071 0.077 0.121 0.117 0.071 0.051
J21-1 0.000 0.003 0.001 0.030 0.027 0.080 0.188 0.124 0.079 0.082 0.129 0.149 0.068 0.040
J21-2 0.000 0.021 0.004 0.046 0.031 0.116 0.130 0.115 0.097 0.088 0.123 0.127 0.060 0.041
J21-3 0.000 0.021 0.010 0.027 0.025 0.091 0.117 0.127 0.103 0.099 0.119 0.138 0.075 0.047
J21-4 0.000 0.005 0.003 0.026 0.026 0.120 0.113 0.114 0.097 0.092 0.122 0.142 0.085 0.056
J21-5 0.000 0.018 0.012 0.044 0.031 0.155 0.114 0.140 0.094 0.092 0.109 0.097 0.061 0.034
J21-6 0.000 0.016 0.016 0.040 0.028 0.154 0.091 0.132 0.099 0.085 0.111 0.107 0.074 0.046
J21-7 0.000 0.015 0.015 0.045 0.030 0.176 0.069 0.120 0.102 0.090 0.125 0.135 0.051 0.029
J21-8 0.000 0.057 0.041 0.067 0.038 0.148 0.034 0.068 0.063 0.075 0.115 0.168 0.083 0.043
J27 initial 0.001 0.140 0.068 0.065 0.029 0.065 0.094 0.076 0.065 0.071 0.112 0.109 0.066 0.038
J27-1 0.000 0.074 0.038 0.069 0.029 0.155 0.073 0.103 0.074 0.079 0.101 0.113 0.054 0.039
J27-2 0.000 0.036 0.039 0.080 0.038 0.153 0.063 0.107 0.079 0.072 0.106 0.120 0.059 0.047
J27-3 0.000 0.078 0.051 0.082 0.041 0.158 0.087 0.107 0.068 0.079 0.087 0.091 0.044 0.028
J27-4 0.000 0.079 0.035 0.062 0.028 0.160 0.078 0.128 0.091 0.080 0.087 0.098 0.046 0.028
J27-5 0.000 0.041 0.041 0.098 0.039 0.148 0.063 0.111 0.083 0.079 0.097 0.108 0.057 0.036
J27-6 0.000 0.078 0.040 0.076 0.035 0.141 0.058 0.110 0.076 0.080 0.108 0.117 0.055 0.026
J27-7 0.000 0.100 0.059 0.068 0.026 0.113 0.051 0.091 0.067 0.074 0.119 0.139 0.063 0.031
J27-8 0.000 0.085 0.050 0.077 0.031 0.133 0.051 0.084 0.070 0.080 0.119 0.129 0.062 0.030
J27-9 0.000 0.119 0.066 0.092 0.032 0.140 0.040 0.068 0.062 0.059 0.085 0.127 0.073 0.039
J27-10 0.000 0.065 0.045 0.069 0.032 0.157 0.055 0.100 0.074 0.078 0.102 0.117 0.067 0.040











fraction 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Size (mm) 0.355 0.75 1.2 1.7 2.4 3.4 4.85 6.85 9.65 13.65 19.3 27.3 38.65 54.65
test p_1 p_2 p_3 p_4 p_5 p_6 p_7 p_8 p_9 p_10 p_11 p_12 p_13 p_14
BOMC initial 0.177 0.104 0.033 0.029 0.026 0.060 0.086 0.069 0.059 0.065 0.102 0.099 0.053 0.038
BOMC-14c 0.198 0.130 0.059 0.051 0.029 0.085 0.114 0.088 0.069 0.055 0.062 0.048 0.010 0.003
BOMC-7a 0.246 0.174 0.037 0.043 0.028 0.067 0.110 0.077 0.055 0.052 0.055 0.038 0.014 0.005
Appendix 7
University of Twente A-17
BOMC-14b 0.210 0.119 0.048 0.063 0.035 0.076 0.114 0.088 0.069 0.055 0.062 0.048 0.010 0.003
BOMC-7b 0.242 0.114 0.040 0.045 0.041 0.102 0.094 0.080 0.059 0.063 0.054 0.043 0.018 0.007
BOMC-7c 0.150 0.148 0.032 0.049 0.036 0.089 0.141 0.091 0.069 0.061 0.057 0.053 0.019 0.007
BOMC-1 0.325 0.158 0.034 0.020 0.014 0.052 0.065 0.061 0.051 0.047 0.068 0.061 0.027 0.018
BOMC-2 0.216 0.180 0.044 0.026 0.014 0.047 0.066 0.071 0.058 0.057 0.082 0.071 0.046 0.020
BOMC-6 0.205 0.194 0.056 0.033 0.019 0.064 0.056 0.056 0.046 0.056 0.081 0.082 0.027 0.025
BOMC-4 0.217 0.197 0.053 0.039 0.022 0.057 0.070 0.064 0.048 0.047 0.076 0.073 0.027 0.009
BOMC-5 0.332 0.171 0.054 0.039 0.023 0.068 0.072 0.048 0.031 0.028 0.042 0.047 0.032 0.014

S-ar putea să vă placă și