Sunteți pe pagina 1din 47

This article was downloaded by: [University of Wollongong]

On: 19 November 2013, At: 21:08


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Journal of Earthquake Engineering


Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/ueqe20

THE EFFECTIVE DURATION OF EARTHQUAKE STRONG


MOTION
a a
JULIAN J. BOMMER & ALEJANDRO MARTÍNEZ-PEREIRA
a
Department of Civil Engineering , Imperial College of Science, Technology & Medicine ,
London SW7 2BU, UK
Published online: 30 Apr 2008.

To cite this article: JULIAN J. BOMMER & ALEJANDRO MARTÍNEZ-PEREIRA (1999) THE EFFECTIVE DURATION OF EARTHQUAKE
STRONG MOTION, Journal of Earthquake Engineering, 3:2, 127-172, DOI: 10.1080/13632469909350343

To link to this article: http://dx.doi.org/10.1080/13632469909350343

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the
publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations
or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any
opinions and views expressed in this publication are the opinions and views of the authors, and are not the
views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be
independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses,
actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever
caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
Journal of Earthquake Engineering, Vol. 3, No. 2 (1999) 127-172
@ Imperial College Press

THE EFFECTIVE DURATION OF EARTHQUAKE


STRONG MOTION

JULIAN J. BOMMER and ALEJANDRO MA~YT~NEZPEREIRA


Department of Civil Engineering, Imperial College of Science, Technology & Medicine,
London S W7 2811, UK

Received 26 June 1997


Downloaded by [University of Wollongong] at 21:08 19 November 2013

Revised 20 March 1998


Accepted 23 April 1998

The duration of strong ground shaking during earthquakes can play an important rdle in
the response of foundation materials and structures, particularly when strength or stiff-
ness degradation is encountered. A thorough seismic hazard assessment should therefore
include an estimation of the expected duration of strong motion, which first requires cri-
teria to define the part of an accelerogram considered to represent t h e duration of strong
ground mot ion. Some 30 different definitions of strong motion duration are reviewed and
classified into generic groups. Problems that arise with the use of these definitions for
duration are highlighted. A new definition of duration is presented using a previously
unexplored option which identifies the part of the record where the main energy is con-
tained and constrains this strong shaking phase by absolute criteria This new definition
is shown to give consistently meaningful durations for strong earthquake accelerograms
from an engineering viewpoint. The correlations between the new definition of duration
and magnitude, soil conditions and distance are explored as a first step towards the
development of predictive equations.

Keywords: strong motion duration, Arias intensity, effective duration, attenuation


relations.

1. Introduction
The assessment of seismic hazard at a site aims to provide a reliable description
of the earthquake motions that can be expected and which, therefore, define the
input into the earthquake-resistant design of projects or the evaluation of the seismic
safety of existing installations. The earthquake motions are usually characterised by
parameters related primarily to the amplitude of the shaking, such as peak ground
acceleration and response spectral ordinates. There are a number of situations in
which the seismic response depends very strongly not only on the amplitude of
the ground motion but also the number of cycles or the duration. In particular,
the response of foundation materials is strongly dependent upon the build up of -
pore water pressure and hence the number of cycles of motion is an important
parameter in the assessment of liquefaction potential [e.g. Seed and Idriss, 19821.
On the structural side, it has been shown that the duration of shaking can have
a significant effect on the inelastic deformationd and energy dissipation demands,
128 J . J . Bommer 43 A. Martinex-Pereinr

especially for relatively weak, short-period structures mahin, 19801. Clearly, for
any structure with stiffness or strength degrading characteristics, the duration of
earthquake shaking will be of critical importance. Therefore, risk assessment for
existing structures requires a reliable estimate of the duration of earthquake shaking
that can be expected to affect the site.
In addition to these situations, it is also necessary to estimate the duration
of strong motion whenever acceleration time-histories are required, such a s in the
calculation of soil response to bedrock motions and in non-linear structural analysis.
Whether the accelerograms used in these cases are selected from databanks of real
earthquake recordings or generated synthetically, it is important to ensure that the
duration of shaking is consistent with the design scenario.
The extent of the influence of the duration in the damage due to earthquake
Downloaded by [University of Wollongong] at 21:08 19 November 2013

motions depends on many factors. Araya and Saragoni [I9801suggest that the level
of acceleration and the density of zero-crossings are the most important factors in
determining damage, and that the strong motion duration is less significant. Liu
and Zhang [I9841 also conclude that the r81e of duration is of second-order in terms
of damage potential. Cabaiias et at. [1997]state that the damage potential is best
estimated using response parameters that consider both amplitude and duration.
Clearly, if during an earthquake a structure is deformed beyond its elastic limit,
then the amount of permanent deformation will partly depend on how long the
shaking is then sustained. In this case, however, the duration is of importance only
because the ground motion is sufficiently strong to cause yielding in the structure.
Therefore, simply comparing the durations of two accelerograms does not reveal
which is more destructive. The curves for assessment of liquefaction potential pub-
lished by Ambraseys 119881 clearly indicate that if the peak acceleration is identical
then the ground motion from the larger magnitude event (with greater duration)
is more likely to induce liquefaction in susceptible soils. Similarly, the curves for
predicting permanent ground displacements of Arnbraseys and Srbulov (19941, for
a given ratio of peak to critical acceleration, show similar dependence on duration,
implied b;y the effect of magnitude. Figure 1 shows two accelerograms recorded in
San Salvador (El Salvador,Central America), one from an offshore earthquake of A&
7.3 in 1982 and the other Erom a local event of magnitude Ms5.4 in 1986. Figure 2
shows the Husid plots for the combined horizontal components for each record and
it can be seen that they reach almost exactly the same level of Arias intensity, but
over very different periods of time. The 1986 earthquake caused more severe dam-
age than the 1982 earthquake, reflecting the fact that in this case the same amount
of energy was imparted to structures in one tenth of the time, thus imposing far
greater demands for dissipation. Whereas for equal accelerations, greater duration
is generally more damaging, for equal energy, shorter duration presents a greater
hazard. Therefore, unqualified statements about the duration of earthquake motion
do not convey much information regarding the damage potential. Nonetheiess, it is
useful to have a stable measure of the duration of strong shaking and, if possible, to
The Eflectiue Dumtion of Earthquake Strong Motion 129
Downloaded by [University of Wollongong] at 21:08 19 November 2013

Fig. 1. Accelerograms from earthquakes of (left) 19 June, 1982 (M,7.3) and (right) 10 October,
1986 (MS5.4) recorded in San Salvador (El S a l d o r , Central America) [Bommer et al., 19971.

0 D 2 0 5 0 4 0 5 0 6 0 7 0 8 0 9 0 '
Time (sec)
Fig. 2. Husid plots (combined horizontal components) of the 1982 (solid) and 1986 (dashed)
accelerograms in Fig. 1.

be able to estimate this duration in future seismic events as an element of seismic


hazard assessment.
The starting point of this study is a review of the various definitions that have
been proposed for strong motion duration.
2. Review of Definitions of Strong Motion Duration
A large variety of approaches to identifying and measuring the strong shaking phase
of accelerogams have been proposed by different researchers. These definitions
of strong motion duration are reviewed and evaluated in this section. In review-
ing definitions of strong motion duration based on earthquake accelerograrns, it is
found that nearly all of the definitions can be classified into three generic groups,
and these are described first. An additional group of definitions is based on the
response of structures to earthquake loading. The definitions are then briefly de-
scribed and subsequently classified according to their generic grouping and other
characteristics.
Downloaded by [University of Wollongong] at 21:08 19 November 2013

2.1. Generic definitions of dumtion


The first group is given the name of "bracketed durations" (Db), and these are
defined as the total time elapsed between the first and last excursions of a specified
level of acceleration, a*, as illustrated in Fig. 3.
One disadvantage of this definition is that it considers only the first and last
peaks that cross the specified threshold and ignores completely the characteristics
of the strong shaking phase, which can result in long durations being estimated
for earthquakes with small sub-events occurring after the main shock motion has
passed. Furthermore, the definition can be rather unstable if low thresholds of
acceleration are employed and, for some accelerogr'ms, a change of the threshold,

i I Time

Fig. 3. Generic definition of "bracketed duration" of an accelerogram.


The Effective h t i o n of Earthquake Stnmg Motion 131

from say 0.03 g to 0.02 g, can result in an increase of the bracketed duration by 20
seconds or more [Pagratis, 19951.
The second group of definitions is given the name ('uniform durations" (D,)and
these are also defined by a threshold level of acceleration, ao, but rather than as
the interval between the first and final peaks that exceed this level, the duration
is defined as the sum of the time intenals during which the acceleration is greater
than the threshold. The concept of "uniform duration" is illustrated in Fig. 4. This
definition is less sensitive to the threshold level than the bracketed duration, but
it has the disadvantage that it does not define a continuous time window during
which the shaking can be considered strong.
Downloaded by [University of Wollongong] at 21:08 19 November 2013

Time

Fig. 4. Generic definition of "uniformduration" of an accelerogram.

The third and final group of record-based definitions are called Usigni6cantdura-
tions" (Ds), and these are based on the accumulation of energy in the accelerograrn
represented by the integral of the square of the ground acceleration, velocity or
displacement. If the integral of the ground velocity is used, then the quantity is
directly related to the energy density [Sarma, 19711,whereas if the integral of the
ground acceleration is employed then the quantity is related to the Arias intensity,
A1 [Arias, 19701.The Arias intensity is defined as

where a ( t ) is the acceleration time-history, t , is the total duration of the acceler*


gram and g is the acceleration due to gravity. The "significant duration" is defined
Downloaded by [University of Wollongong] at 21:08 19 November 2013

Time
Fig. 5. Generic definition of "significant duration" of an accelerogram.

as the interval over which some proportion of the total integral is accumulated.
This is illustrated for arbitrary limits on a plot of the build-up of Arias intensity,
known as a Husid plot [Husid, 19693,in Fig. 5.
The concept of "significant duration" has the advantage that it considers the
characteristics of the entire accelerogram and defines a continuous time window in
which the motion may be considered as strong.
It is worth mentioning at this point the concept of the root-mean-square accel-
eration, a,,, which is defined by

where tl and t z mark the beginning and end respectively of the time interval un-
der consideration. Comparison of Eqs. (1) and (2) shows that ,a over a certain
interval is in fact directly proportional to the square root of the gradient of this
specified interval of the Husid plot. Therefore, any definition of duration based on
the variation of a,, is also considered as a ''sigdicant durationn.
There are a few dehtions of strong motion duration that are based not on
the characteristics of the record of ground motion, but on the characteristics of
the response motion of a specified structure subjected to the pound motion. These
definitions are considered as a fourth generic group, although they can usually also
be classified into the same three groups described above, since in effect they are
applications of the "bracketed",W o r n n or 'sigmficant" duration to the response
time-history. These definitions are referred to as Ustructurdresponse durationsn.
The Eflecttve D u d i o n of Earthquake S t m g Motion 133

2.2. Definitions of stmng motion dumtion


There is no universally accepted definition for strong motion duration, and reference
is frequently made to the duration of earthquake motion without speclfylng how
the duration is defined. A large number of researchers have proposed definitions of
earthquake strong motion duration over the last three decades. These definitions
are reviewed in chronological order and each one is classified according to the four
generic groups described in the previous sect ion.
The first study to propose a definition of strong motion duration was by
Rosenblueth and Bustamante (19621.Their definition is a structural response du-
ration, being the duration of an equivalent motion with uniform intensity required
to calculate the effect of damping in a structure. This is defined more precisely as
the duration of uniform motion needed to produce a constant ratio between the
Downloaded by [University of Wollongong] at 21:08 19 November 2013

maximum spectral displacement response of an SDOF (single degree of freedom)


oscillator with a given period without damping and with a specified damping level.
This rather complicated definition was not put forward primarily to identify the
duration of the strong motion in an accelerograrn, but was actually introduced
in a study whose main focus was the influence of structural damping on spectral
ordinates.
Housner (19651discussed the "du~utionof strong shaking" of some early Cali-
fornian accelerograms, but his definition seems to have been based purely on visual
inspection of the records since it is stated that "the duration of the strong phase of
shaking can be distinguished only for ground motions recorded relatively close to the
fault". The definition could be interpreted as a bracketed duration. The criteria for
determining the duration are relative in the sense that unless some portion of the
time-history actually stands out as being stronger than the rest of the record, no
duration is defined. However, Mfunac and Brady [I9751 state that the durations
given by Housner [1965]are comparable to bracketed durations for a threshold of
0.05 g.
The first definition that was proposed explicitly to identify the length of the
strong shaking portion of accelerograms was by Ambraseys and Sarma 119671.They
use the concept of bracketed duration with a threshold acceleration of 0.03 g.
Husid (19691discussed the difEculties in selecting the appropriate value of du-
ration for calculating the a,, and A1 values of an accelerogram and states that "it
is necessary to choose arbitmrily the part of the record which will be considered".
Nonetheless, Husid also defined the minimum duration of the accelerogram that
should be used in the calculation of spectral ordinates, proposing two methods for
its determination. The first is based on identifying the time at which the Husid plot
begins to become horizontal and this is considered as the end of the strong motion
phase. The second method identses the point-at which the plot of a, against time
becomes less steep. Husid argued that the former method was preferable because
of its physical significance, but in fact both these ways of defining the significant
duration are essentially the same.
134 J. J . Bommer 43 A. Martr'nez-Pmim

Kobayashi [1971]used the concept of bracketed duration with thresholds of 50


and 100 gals. Page et al. [I9721 used the concept of bracketed duration with an
acceleration threshold of 0.05 g.a
Donovan (19721defined a significant duration as the interval from the beginning
of the record to the time at which 90% of the total Arias intensity was accumulated.
Bolt [I9731 presented the first study dedicated tomthe duration of earthquake
strong motion in which he introduced the concepts and names of bracketed and
uniform durations. Durations &e calculated not for the original accelerograrns but
after passing each record through a narrow-band filter in order to define the duration
of shaking a t a particular frequency. Thresholds of 0.05 g and 0.10 g were used for '

both types of duration.


Housner (1975)proposed a number of parameters for quantifying the severity of
Downloaded by [University of Wollongong] at 21:08 19 November 2013

earthquake ground motions and duration is introduced in a tweparameter measure


of ground-motion severity. The first quantity, P,, is defined as the "power" of the
earthquake motion and is in fact the square of a,. The second parameter, the
duration, D ,is defined as the interval over which P, is constant, identified as the .
extrapolation of the straight portion of the Husid plot. The "power" is thus defined

where to is the time at the beginning of the strong shaking phase. If the strong
phase is considered to start at the beginning of the record, to is zero, and comparing
Eqs. (1) and (3), there is an obvious relation:

and Z ( t ) is as defined in Eq. (3). D is determined by visual inspection from a plot of


against time. This definition falls into the category of significant durations
although there is a subjective element in its determination.
The study of strong motion duration by Trifunac and Brady (19753 used the
significant duration concept defined for the integrals of the squares of accelera-
tion, velocity and displacement. In each case the duration is dehed as the interval
between the times at which 5% and 95% of the total integral is attained. This def-
inition, using the integral of acceleration squared, was also subsequently employed
by Dobry et al. [1978].
The definition of Hisada and Ando [1976]w s reported by Theofanopulos and
Watabe (19893.The definition is essentially a bracketed duration, being the total
time from the beginning of the accelerograrn to where W e wave amplitude becomes
equd to one tenth of the peak accelemtion", presumably meaning the last peak
which reaches this level.

=The definition of Page et al. El9721 varies very slightly from the generic description given previ-
ously as 'the time intend between the fini and k t a d e m t i o n peaks e&d to m grmtcr tho*
0.05 g."
The Eflectiue Dunztion of E a w e S t m g Motion 135

Trifunac and Westermo I19771 presented a definition of strong motion duration


which is a modification of the si@cant duration of Tkifunac and Brady [1975].The
first modification was to calculate the durations after passing the accelerograms
through narrow-band Nters to find the durations at particular frequencies. The
second modification was to alter the limits of the duration: the duration is still the
time interval in which 90% of the total Arias intensity is attained, but instead of
being found from the 5% and 95% limits, the duration is now the minimum s u m
of the time intervals during which 90% of the total integral is accumulated. This
is illustrated in Fig. 6. Therefore, the definition is the sum of the time intervals of
the steepest parts of the Husid plot that collectively contribute 90% of AI, and in
effect it is a hybrid of the significant and uniform definitions.
Downloaded by [University of Wollongong] at 21:08 19 November 2013

Time

Fig. 6. Definition of strong motion duration by Trifunac and Westermo f19771.

Saragoni [1977Jpresented a model for characterking accelerograms with a three


parameter function that describes the evolution of the acceleration amplitudes with
time. The method considers that the mean square acceleration ~ [ a l ( t )tends
] to
the chi-squared funtion:
& -

E [a2( t )1 = ~ e - ~ ' t Y (5)


The duration of the strong motion region can then be defined as the interval between
the times at which the chi-squared function has its inflection points. Saragoni [I9771
presented two definitions for duration depending upon the value of 7:for 7 > 1,

and for 7 5 1,

McGuire and Barnhard (19791undertook a study of different definitions of strong


motion duration to assess the accuracy with which duration could be predicted
and the usefulness of duration for specifying the severity of seismic shaking. The
definitions used included bracketed duration with absolute thresholds of 0.05 g,
0.10 g, 0.15 g and 0.20 g, and relative thresholds of 0.50, 0.67 and 0.75 of the
Downloaded by [University of Wollongong] at 21:08 19 November 2013

peak acceleration; the significant duration of Trihnac and Brady [1975];and the
definition of Vanmarcke and Lai [1977],which is discussed below. Although McGuire
and Barnhard (19793refer to Bolt (19731for their definition of bracketed duration,
the records in their study are not limited in frequency.
McCann and Shah [1979] defined strong motion duration by plotting the cu-
mulative a,,, of the accelerogram, noting that beyond a certain point it begins
to decay. The end of the strong motion phase, t z , is identified by plotting the
derivative of the cumulative a,,, function against time and noting the time beyond
which it remains negative. The start of the strong motion phase, t l , is identified in
exactly the same way using the reverse acceleration timehistory. The duration is
then (t2 - t and its determination is illustrated in Fig. 7. This definition can be
classified as a significant duration.
Vanmarcke and Lai (1977, 19801 presented a new definition for strong motion
duration. Their duration defines an equivalent interval of shaking over which the
total Arias intensity of the accelerogam is uniformly distributed at a constant
average power a,,. This portion of the accelerogam can be idealised as a stationary
random process. The theory of stationary Gaussian random functions is then used
to provide a prediction of the most probable value of the ratio of peak acceleration,
a
,, to a,, during the steady strong motion interval (i.e. the ratio which has a
probability of e-I of not being exceeded during the interval). For records with two
or more cycles of strong shaking,this results in the following equation for duration,
D:

where I ( t r ) is as defined in Eq. (3) and Tois the predominant period of the motion
found by dividing the length of the strong motion interval by the number of cycles
of strong motion. By assuming a constant ratio between a, and a, of 2.74, the
definition is simplified to the approximate relationship:
The Effective Durnttm of Earthquake Strang Motion
Downloaded by [University of Wollongong] at 21:08 19 November 2013

Time History Reversed Time Histoy

Fig. 7. Definition of strong motion duration by McCann and Shah [1979].

In a study carried out to investigate the displacement responses sustained for


different numbers of cycles, Perez [1980]defined a structural response duration
which is the total time for which the velocity response of a lightly damped SDOF
system is above a specified level. The duration is found by applying the concept of
uniform duration to the structural response motion.
Zahrah and Hall [I9841 proposed another structural response duration defmed
as the time interval during which most or all inelastic deformations of an SDOF
system take place. Explicitly, their "effective duration" is the time interval between
which 5% and 75% of the energy absorbed in a structure is dissipated inelastically,
the assumption being that outside this strong shaking phase the energy is dissipated
through structural damping.
Another definition for strong motion duration was proposed by Zhou and Xie
(1984)and reported by Zhou and Katayama [1985]. The definition is based on the
characteristics of the power distribution of the strong motion record. The first step
is to d e h e the time t , which is effectively the centre of gravity of a2(t) along the
time axis:
The duration of strong motion is then defined as twice the standard deviation of
a2(t)about the time t,:

Shahabi and Mostaghel [I9841 defined strong motion duration as the time inter-
val over which the total energy supplied to d l SDOF systems is imparted with a
uniform peak level which they define as the effective cyclic acceleration. Although
the definition is described in terms of structural response, it is actually determined
Downloaded by [University of Wollongong] at 21:08 19 November 2013

directly from features of the strong motion record. If the Husid plot is enclosed in
a rectangle whose height% I(t,), as defined in Eq. (3), and whose length is t,, the
area of the part of this rectangle that is above the curve, A, is given by:

Shahabi and Mostaghel [I9841then define the strong motion duration as

This implies that the duration is the length of time that would be required t o attain
the final value of Arias intensity if the rate of energy input during the strongest
phase of the shaking were maintained throughout the record (Fig. 8).
Another variation of the significant duration was given by Theofanopoulos and
Drakopoulos (19861.Considering two consecutive points on a Husid plot separated
by a small time interval At, whose ordinates are I ( t ) and I(t + At), a quantity P ( t )

- Time
Fig. 8. Definition of strong motion duration by Shahabi and Mmtaghel [1984].
The Eflectiue Duration of Earthquake Strong Motion 139

is then defined such that


I ( t )At
P(t) = I ( t a + At) - I(t)- -
t
This value represents the difference between the actual increase in Arias intensity
over the time interval At and the increase that would result from extrapolation of
the average rate of increase of Arias intensity with time from the beginning of the
record. P ( t ) is plotted against time and the strong motion phase is defined as "the
time interval where the function P ( t ) takes high positive values". Although this
definition appears to be rather complicated, the strong motion phase is actually
identified by visual inspection of the plot of P ( t ) .
Elghadamsi et 01. [I9881introduced a definition of duration which is a variation
of that put forward by McCann and Shah [1979].The modification is to select the
Downloaded by [University of Wollongong] at 21:08 19 November 2013

limits of the strong motion phase as the times at which the derivative of the cumu-
lative a,,, functions of the normal and reversed accelerograms reach 1.0 crn/s2/s
and remains lower than this thereafter. Elghadamsi et al. [1988]report that this
results in durations which are on the average 23% shorter than those defined by
McCann and Shah (19791,but the ,a values are only reduced by 7%. Even though
an absolute d u e is defined for the cut-offthreshold, it is a value related to the rate
of change of gradient of the Husid plot rather than its amplitude and in this sense
the criterion is relative rather than absolute.
Xie and Zhang [1988] defined an "engineering duration" which is a bracketed
duration in which the threshold is an absolute value of acceleration, a y , which
causes yielding in an SDOF system. This yielding acceleration is defined by Xie
and Zhang (19881as

where Qv is the yield strength of the structure, M is the total mass of the structure
and P(To)is the ordinate of spectral acceleration with the same damping ratio and
natural period (To) as the structure.
Mohraz and Peng [1989]proposed a definition for strong motion duration which
accounts for structural frequency and damping, although the duration is determined
from a filtered version of the accelerogam rather than a calculated structural re-
sponse. The accelerogram is processed with a low-pass filter with a cut-off frequency
given by

where f, is the natural frequency of the structure and X the damping ratio. The
duration is calculated by applying the definition of McCann and Shah [I9791to the
filtered acceleration timehistory.
Theofanopulos and Watabe (19891presented another variation of significant du-
ration which is very similar to the definition of Zhou and Xie,[1984]. In this new
definition the centre of gravity and standard deviation of the increments of AI along
140 J. J. Bommer 0 A. Murttirez-Perearn

the time axis are also found. As in the definition of Zhou and Xie [1984], the end
of the strong motion interval is the time corresponding to the centre of gravity
plus one standard deviation. The beginning of the strong motion phase in this new
definition corresponds to the first time a specified threshold is exceeded in the ac-
celerogram, the threshold being the increase in A1 over a specified time increment
At that the record has a t the end of the strong motion interval.
Kawashima and Aizawa [I9891 employed bracketed durations with thresholds of
50, 100, 150 and 200 cm/sec2, and fractional durations - which they refer to as
c'normaIised" - with nine equally-spaced thresholds from 0.1 to 0.9 of the peak
acceleration.
Sarma and Casey [1990]presented a definition based on the concept of uniform
duration. For a given accelerogram, the uniform duration is determined for a num-
Downloaded by [University of Wollongong] at 21:08 19 November 2013

ber of fractional levels of the peak acceleration and the logarithms of the uniform
durations are plotted against the acceleration thresholds normalised to the peak.
The results are found to approximate to a straight line between the limits Alo and
Ags,defined as the acceleration levels below which 10% and 95%, respectively, of the
total Arias intensity are contained [Sarma and Yang, 1987).The duration of strong
motion is then defined as the anti-logarithm of the intercept of the extrapolated
straight line segment, as illustrated in Fig. 9. A unique feature of this definition is
that it does not specify the time at which the strong shaking interval begins and
ends and hence its correlation with the accelerogram is unclear.
In a seismic hazard assessment for Greece in terms of strong motion duration,
Papazachos et al. [1992]use bracketed duration with a low threshold of just 0.02 g.
Somerville et al. [I9971 use the significant duration between the limits of 5%
and 75% of the total Arias intensity, which they attribute to Husid [I9691although
Husid did not actually specify limits for the integration in his paper.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1


Acceleration Level (AIPGA)

Fig. 9. Determination of strong motion duration defined by Sarma and Casey (1990].
The Eflective D u d i o n of Earthquake Stmng Motion 141

2.3. Classification of dumtion definitions


In addition to grouping definitions of strong motion duration into the generic cat-
egories described in Sec. 2.1, it is also possible to classify a definition as being
based on an absolute or a relative criterion. For example, the "bracketed durationn
may be defined for a specified absolute level of acceleration (such as 0.05 g) for
the threshold, or alternatively for a threshold which is defined as some proportion
of the peak acceleration, which has been referred to as the "fractional duration"
[McGuire and Barnhard, 19791.
A further classification is necessary because a number of the definitions that have
been proposed are based not on the accelerogram as recorded, but are frequency
limited by first processing the record with a narrow pass-band filter. Therefore, the
Downloaded by [University of Wollongong] at 21:08 19 November 2013

definitions can be considered as frequency-independent, in which case the entire


accelerograrn is used as recorded or digitised, or as frequency-dependent, where the
duration defines the length of strong shaking at a particular frequency.
Table 1 summarises the classification of all of these definitions of strong motion
duration according to the generic groupings described earlier. hrthermore, each
definition is also classified as either frequency-independent or frequency-dependent
according to whether the entire accelerogram is employed in the calculation of
duration or whether it is first passed through a narrow-band fdter. Each definition
is also classified according to whether the criteria used to distinguish the strong
motion portion of the accelerogram are absolute or relative values applied to the
accelerogram, the Husid plot or the response of SDOF systems.

2.4. Assessment of dumtion definitions


Clearly the user is presented with a wide variety of definitions of duration from
which to choose. It is important to recognise that it may not be useful to establish
a single, universal definition for strong motion duration, since different definitions
may be more or less appropriate in different situations. For assessment of the seismic
safety of a particular building, the most suitable definition would be a structural
response based duration. However, for the purposes of developing a general predic-
tive model for strong motion duration that can be employed in hazard assessment,
these definitions are not particularly useful. Furthermore, because they reflect the
chdracteristics of structural response, they are unlikely to correlat e well with easth-
quake parameters. Sucuo~luand Nurtug [1995]point out that for lightly damped
structures, the duration of significant structural response can actually be longer
than the total duration of the recorded g o u n d motion.
The &e of frequency-dependent durations may provide better constrained pre-
diction of variation with distance, since high-frequency motions attenuate more
rapidly than low-fkequency motions. There may also be a use for frequency-
dependent durations if the objective is to generate synthetic time-histories by p r e
dicting the amplitude, duration and phase characteristics of different component
Erequencies of the g o u n d motion. Nonet heless, attenuation relations essentially
142 3. J. Bommer & A. Marthex-Pemim

Table 1. Classification of definitions of strong motion duration.

Bracketed Uniform Significant Structural


Definition Rel. Abs. Rel. Abs. Rel. Abs. Rel. Abs.

Rosenblueth and Bustamente [I9621 0

Wousner (19651
Ambraseys and Sarma [I9671
Husid (19691
Kobayashi (19711
Page et d. (19721
Donovan (19721
Downloaded by [University of Wollongong] at 21:08 19 November 2013

BoIt [1973]
Housner [I9751
Trifunac and Brady (19751
Hisada and Ando [I9761
Trifunac and Westerrno I19771
Saragoni [I9773
k1cGuire and Barnhard [I9791
IbfcCann and Shah [I9791
Perez (I9801 0

Vanmarcke and Lai [1980j


Zahrah and Hall [I9841 0

Zhou and Xie I19841


Shahabi and Mostaghef [1984]
Theofanopoulos and Drakopoulos (19861
Elghadamsi et al. I19881
Xie and Zhang [I9881 0

Mohraz and Peng [I9891


Theo fanopulos and Wat abe (19891
Kawashima and Aizawa f 19891
-7

Sarma and Casey (19901


Papazachus et al. [1992]
Somenrille et (3. [I9971

0 frequency-independent duration o hequency-dependent duration

describe the radiation of seismic energy in earthquakes, and it is therefore more


robust to work with the duration of the complete record rather than isolated fre-
quency components which only carry a limited portion of the wave energy.
On the basis of these arguments, the choice can be reduced to those definitions
corresponding to solid circles in Table 1. This is still a wide airay of possibilities,
so in order to investigate the physical consequences of employing some of these
The Effective Durntion of EarOlquake Strong Motion 143

Table 2. Characteristics of strong motion records in Fig. 10.

am ax AI T
No. Earthquake Station Camp (d (m/s> be4
1 Imperial Valley Cerro 147' 0.166 1.145 64.0
15 Oct, 1979 Prieto

2 Tabas
16 Sept, 1978
3 San Fernando
9 Feb, 1971

4 Managua ESSO EW 0.368 1.613 45.9


Downloaded by [University of Wollongong] at 21:08 19 November 2013

23 Dec, 1972 Refinery

5 hpinia St urno EW 0.319 1.497 60.48


23 Nov, 1980

definitions, durations are calculated for the five horizontal accelerograms listed in
Table 2 and which are shown together with their Husid plots in Fig. 10. These
accelerograms have been chosen to include very strong motion, moderate motion
and very weak motion, and also to include potentially complicated cases with mul-
tiple events. The durations calcuiated using a few representative examples of the
different definitions are surnmarised in Table 3.
A number of important observations can be made from these results. Firstly, the
sensitivity of both bracketed and uniform durations to the thresholds is immediately
apparent. Comparison of the two significant durations at the bottom of the table
show,however, that in general these durations vary less with the thresholds used.
The Vanmarcke and Lai [1980]definition, however, seems to yield consistently low
values for duration. The uniform durations also yield low values especially for higher
thresholds, and very small d u e s for the highest threshold.
The results also reveal features that can arise with strong motion records from
multiple ruptures. On Record No. 2 there is a clear second arrival at about 46 sec-
onds and the bracketed durations include this as part of the strong shaking phase,
which on examination of the Husid plots in Fig. 10 seems perfectly reasonable. How-
ever, although less distinct there is also a second arrival on Record No. 5 at about
43 seconds. This record is from the Irpinia earthquake of 23 November 1980 which
was clearly identified as a multiple event with a number of sub-events [Westaway
and Jackson, l987].The bracketed durations with low thresholds include this sec-
ond event in the strong shaking phase, resulting in durations of 40-50 seconds even
though the motion is very weak for several seconds prior to the arrival of waves from
the subevent. When the threshold is raised from 0.05 g to 0.10 g for this record,
the duration is reduced by almost 35 seconds. It could be argued that on Records
Downloaded by [University of Wollongong] at 21:08 19 November 2013

Fig. 10. Five accclerogtams and their conesponding Husid plots. The characteristics of these
accelerograrns are presented in Table 2. The numbers refer to identification of the records in
Table 2.
The EBective Dumtion of Earthquake Strong Motion 145

Table 3. Durations calculated for accelerograms in Table 2 using different


definitions.

Duration of Record
Definition of Duration 1 2 3 4 5
Bracketed (ao = 0.03 g)
Bracketed (ao = 0.05 g)
Bracketed (ao = 0.10 g)
Uniform (a0 = 0.03 g) 15.3 24.8 0.1 10.0 12.2
Uniform (ao = 0.05 g)
Uniform (ao = 0.10 g)
Bracketed (ao = 0.1 a,,,)
Downloaded by [University of Wollongong] at 21:08 19 November 2013

Bracketed (ao= 0.3 urn=)


Bracketed (a0 = 0.5 a
,
)
Uniform (a0 = 0.1 amax)
Uniform (ao = 0.3 amax)
Uniform (ao = 0.5 am-) 2.8 0.5 0.4 0.6 0.6
Significant &90% iA 29.0 20.7 51.8 10.3 20.7
Significant 5%-95% IA 31.2 18.0 55.3 10.9 38.2
Vanmarcke and Lai [1980],Eq. ( 6 ) 20.2 4.9 9.6 5.8 7.2

Nos. 2 and 5 the second event should be removed before calculating the duration,
although it is important to note that when these shocks arrive the ground is not at
rest. firtherrnore, the surface-wave magnitude, Ms,and the seismic moment, Mo,
will actually reflect the combined effect of both shocks and any structure will have
been subjected to both motions. The effect of removing these sub-events - which
are difficult to classify as "aftershocks" - is discussed in Sec. 3.2.
Another very important feature is revealed by the values calculated for Record
No. 3, which was included to represent an example of very weak earthquake motion.
The bracketed durations based on absolute thresholds of 0.05 g and 0.10 g result
in values of zero for this record. If strong motion duration is understood as the
length of the strong shalung phase of the accelerogram, this result suggests that for
this record there is no strong shaking phase. The definitions based on relative cri-
teria, however, give appreciable durations for this accelerogram; the two significant
durations suggest that the strong shaking lasts for almost oneminute. This raises
an important question: whether to use absolute or relative criteria to identify the
limits of the strong shaking part of the accelerogram. Zhou and Katayama [1985]
argue for the use of relative criteria precisely because records with peak accelera-
tions lower than the thrkhold will not have any duration. On the other hand, this
does not seem inconsistent if the objective is to identify and measure the length of .
the "strong" shaking phase. Novikova and Trifunac [I9941point out that relative
definitions are useful since, combined with information about the Fourier spectral
146 J. J. Bommer E4 A. Martinez-Pewziru

amplitudes, they provide a fairly complete description of the ground motion. This
then requires durations to be determined for a large number of individual frequen-
cies, and for a full description of the motion, it would also be necessary to estimate
the phase angle of each component.

3. Effective Strong Motion Duration


In light of the review of the existing definitions for strong motion duration, it was
decided that it was worthwhile exploring a new definition for record-based dura-
tion. Although there was some hesitancy at adding to the confusion by putting
forward yet another approach, shortcomings have been identified in all of the def-
Downloaded by [University of Wollongong] at 21:08 19 November 2013

initions reviewed. For general seismic hazard assessment, a definition is required


that identifies the window of significant shaking of ground motion accelerograms.
This immediately eliminates the uniform and structural response categories, and
also means that the definition should be frequency-independent . Since the brack-
eted duration has been shown to be highly sensitive to the threshold levels used,
the generic category chosen is that of significant durations which have the merit of
being based on the energy in the record. However, all of the existing definitions and
variations of significant duration are based on the use of relative criteria. The new
approach here is to employ the previously unexplored option of significant duration
constrained by absolute thresholds of Arias intensity; a s can be seen from the third
t o last colum in Table 1, none of the existing definitions fall into this category.

3.1. Definition of effective dumtion


The definition presented here is a modification of that presented in an earlier study
[Bornmer and Martinez-Pereira, 19961, which aimed to define a significant dura-
tion based on absolute criteria. However, the definition proposed by Bommer and
Mart inez-Pereira (19961 only used an absolute criterion to ident i@ the beginning
of the strong motion phase. The end of the strong shaking was determined as the
time a t which the additional input of Arias intensity in 1 sec was equal to 1%of
the value accumulated up to that time. This definition did effectively exclude very
weak motions from being assigned strong motion durations, but the relative nature
of the second criterion meant that, for very strong records horn large magnitude
events, the durations were truncated prematurely, thus not including a significant
amount of energy in the strong motion phase. Nonetheless, the dehition appeared
to identlfy the main Swave train for records from earthquakes with relatively simple
ruptures.
The new definition for duration proposed here is also based on the s i d c a n t
duration concept, but both the start and end of the strong shaking phase are
identified by absolute criteria. This duration is given the name "effective duration";
this name has previously been used by Zahrah and Hall [1984], but it is the most
suitable description for this new definition.
The Elgectiue Duration of Earthpuke Strong Motion 147
Downloaded by [University of Wollongong] at 21:08 19 November 2013

Fig. 11. Definition of effective duration.

The definition is based on the Husid plot, as shown in Fig. 11. The objective is
to define criteria by which to identify the times to and tf which mark the beginning
and end of the strong shaking phase. The effective duration, DE,is then defined
simply as

The start of the strong motion phase is considered as the time when the Husid plot
reaches a particular threshold level of Arias intensity, a.
From the inspection of
a large number of strong and weak accelerograms and their corresponding Husid
plots, a value of 0.01 m/s was selected for N o .
The end of the strong shaking phase, tf, is defined as the time at which the
remaining energy in the record is equal to an absolute amount AMI. In the original
definition, the reason for employing the relative criterion described previously was
precisely to avoid making the definition dependent on the tail end of the record and
the Husid plot. The length of this coda (for analogue records) will depend upon
the diligence of the digitiser. The coda may also contain appreciable levels of noise
and its characteristics can vary with the baseline and fltering procedures applied.
However, it has been found that, for acceleration records, a suitable value of AAI!
can be found which yields reasonable results for most records. Values of AMl
between 0.05 m/s and 0.15 m/s were tried and it was found that the duration was
only particularly sensitive for records from events with multiple ruptures. The value
selected for the definition of the effective duration, LIE,is 0.125 m/s. Clearly any
record whose total Arias intensity is less than 0.135 m/s will not have an effective
duration. The implicit assumption, therefore, is that an accelerograrn with an Arias
intensity not greater than 0.135 m/s is not considered to be strong motion from
an engineering perspective. It is important to stress that no correlation is being
attempted here between the level of AI and the damage potential of the motion,
since apart from other considerations it is not clear that A1 is the best indicator of
damage potential, nor what level of AI can be considered as damaging. It can be
stated, however, that a record not attaining this minimum threshold is unlikely to
be of engineering importance.

3.2. Comparison of egective and other durations


For the five accelerograms in Fig. 10, the values of t o and t f , effective durations and
the percent ages of the total Arias intensity contained in the strong shaking phase
are summarised in Table 4. In Fig. 12 these effective durations are compared with
the two most widely used durations, the bracketed (ao = 0.05 g) and significant
[Trifunac and Brady, 19751 definitions.
Downloaded by [University of Wollongong] at 21:08 19 November 2013

In Fig. 12, from the bars that represent the durations, it appears that on vi-
sual inspection the effective duration (the top bar in each case) correlates well in
most cases with the phase of higher amplitude shaking. The significant definition
clearly overestimates the duration for Record No. 3, whose bracketed and effective
durations are both zero. For the very strong Record No. 2, the significant duration
clearly stops short of some very strong shaking and underestimates the length of
strong shaking phase in this case.
An interesting case is Record No. 5 for which the bracketed duration includes
the second wave group arrival that begins at around 43 seconds, It is interesting
to note that the significant duration extends the length of the record up to the
beginning of this second arrival but does not actually include it in the strong shaking
phase. On the other hand, the effective duration stops some 20 seconds earlier,
after the strong shaking phase from the main event. The question is whether the
second event is significant and if it justifies doubling the strong motion duration?
There is probably not a single correct answer to this question, because for some
structures the second event could be the "straw that breaks the camel's back", but
some general considerations can be made. Ftom Table 4 it can be seen that the

Table 4. Characteristics of effective durations for


accelerograms described in TabIe 2.

4 0-65 10.95 10-3 91..6


5 3.33 21.84 ' 18.5 91.0
5' 3.33 15.48 12-2 90.5

'Record terminated prior to arrival of second event.


TAe Eflective Duration of Earthquake Strong Motion 149
Downloaded by [University of Wollongong] at 21:08 19 November 2013

Time (s)
Fig. 12. Effective, significant and bracketed duration of acceIerograrns in Fig. 10. The numbers
refer to the record identification in Table 2. In each case the bars represent the strong motion
phase as defined by the effective duration (top), the Trifunac and Brady [1975] significant duration
(middle) and the bracketed duration with ao = 0.05 g (bottom). For Record No. 3, the effective
and bracketed durations are both zero.

effective duration actually includes 91% of the total energy in the record, whereas
the significant duration, which is twice as long,has only 90% of the total energy. The
definition of significant duration by Trifunac and Westermo [I9771would eliminate
the "silences" between sub-events in multiple event records of this type, but then
the strong shaking phase would be defined as two portions of motion separated by
several seconds. For any situation where there is an accumulative effect, particularly
the generation of excess pore pressure in foundation materials, the removal of these
quiet parts of the strong motion record to represent the shaking by the sum of the
strong intends would be conservative. Nonetheless, to explore this issue further,
Records Nos. 2 and 5 were cut prior to the arrival of the second events at 45 and
42 seconds, respectively. The effective durations for the truncated records are also
given in Table 4, from which it can be seen that the proportion of the total energy
retained by the efective duration is almost unchanged in both cases. The bracketed
durations for the truncated records are 33.7 and 24.5 seconds.respectively, and the
significant durations are 16.4and 16.0 seconds respectively.
150 J . 3. Bomrner 0 A. Martinez-Pereina

3.3. Engineering evaluation of effective dumtion


The effective duration is illustrated in Fig. 13 for 12 strong motion accelerograrns
listed in Table 5, which gives the characteristics of each record including the propor-
tion of the total energy contained within the strong motion interval corresponding
to the effective duration. The first observation that can be made is that it can be
misleading to identify the strong shaking phase by purely visual inspection since it
is difficult t o assess the amount of energy contained in different parts of the record.
For example, for record B in Fig. 13(b), the effective duration contains 98.5% of
the total energy even though the sub-event at around 30 seconds is excluded. For
this record, it is clear that the sigdicant duration based on relative criteria would
shorten the length of the strong shaking phase, whereas the bracketed duration
Downloaded by [University of Wollongong] at 21:08 19 November 2013

with a threshold such as 0.05 g would imply a much longer strong shaking phase
than indicated by the effective duration.
For the stronger records, the effective duration generally contains more than
80% of the total Arias intensity and frequently more than 90%, such as record
K in Fig. 13(a) for which 97% of the total energy is present. For weaker mo-
tions, t he effective duration inevitably includes smaller proportions of the total
energy.
For example, for record A in Fig. 13(a) and records C and L in Fig. f3(b),
the effective duration contains between 80% and 83% of the total Arias intensity
and excludes the iast of the large acceleration peaks. For these particular records
it could be argued that DE should be extended slightly, which could be achieved
through a small reduction in the value of h A I f from 0.125 cm/sec. However, a
similar situation would then arise with accelerograms that were slightly weaker than
these. Without creating a complex definition in which the criteria for const raining
the duration become functions of the strength of the record, it is inevitable that a
definition based on absolute criteria (to obtain zero durations for weak records) will
yield very small durations for relatively weak records. An example of this is record
E in Fig. 13(a) which has an effective duration of less than 1 sec that contains
62.5% of its Arias intensity. Therefore, a significant proportion of the total energy
in the record is excluded from the strong shaking phase identified by the effective
duration, but the point is that the energy excluded is probably not of particular
engineering importance.
The engineering sigdicance of the new effective duration definition can be ex-
plored through examining its relation to acceleration and displacement response
spectra. The elasto-plastic displacement spectra for periods up to 3 seconds are
calculated for four accelerograms, whose characteristics are given in Table 5. The
spectra are determined using both the entire accelerograrn and using only the por-
tion of the record identified as the strong shaking phase by the effective duration.
The elasteplastic displacement spectra are calculated using a value of the seismic
coefficient, C,,of 0.10. The seismic coefficient is defined as the ratio of the yield
strength to the total weight of the SDOF system considered. The model used is
The E m t i v e Duration of Earthquake Stmng Motion 151
Downloaded by [University of Wollongong] at 21:08 19 November 2013

0 10 20 30 40 50 60 70 80 90 100
Time (s)

0 to 20 30 40 50 60 7C
Time (s)

Fig. 13. Twelve accelerograrns and their strong shaking phases as identified by the effective dur*
tion. The characteristics of these accelerograms are given in Table 5. Letters refer to identification
of records in Table 5.
Downloaded by [University of Wollongong] at 21:08 19 November 2013

Table 5 . Characteristics of accelerograms shown in Figs. 13 and 14.

RI amax A1 DE
Year M Day Time Ctry Mw (km) Station Name (g) (m/s) (c) %AI~
ID^
USA Eur. kderal Bldg.
USA Pacoima Dam Abutment
ITA Tolmezzo 1
IRA Tabas
USA San Martin
USA Oil City
Anticline Rdge, FF
CAN Site 1
USA Lexington Dam,Abut.
UC Santa Cruz Lick Lib.
USA Lucerne
USA Pacoima Dam
JAP Kobe University

I, Source-tcxitu distance, generally from the closest point on the surface projection of the fault rupture.
2. Parcentage of the total Arias intensity contained in the strong-motion corresponding to the effective duration.
3. Record identification in Figs. 13 and 14.
The Eflective Dumtion of h d a q u a k e S t m g Motion 253

0.0 0.3 1.0 13 2.0 ZS 3.0 0.0 03 1.0 13 2.0 2.5 3.0
Period (see) P d d (see)
Downloaded by [University of Wollongong] at 21:08 19 November 2013

0.0 0.5 1.0 1.5 2.0 25 3.0 0.0 0.5 1.0 1.5 LO 25 3.0

Period (see) Period (su)

Fig. 14. Elasteplast ic displacement response spectra for accelerograrns as digitised (thick solid)
and using only the strong motion portion related to the effective duration (thin dashed). T h e
spectra are calculated for an osciIlator with Cy = 0.10, a = 0.1 (see text for explanation) and 5%
of critical damping. The characteristics of these acceIerograms are given in Table 5. Letters refer
to identification of the records in Table 5.

not perfectly plastic after yield, but the stiffness reduces from kl t o k2, and the
hardening is defined by a = k&l.
The elast~plasticdisplacement spectra are shown in Fig. 14 and it can be seen
that the spectra are generally in good agreement, which suggests that the effective
duration does indeed isolate the most iniportant part of the strong motion. For
the strongest record (D), the spectra horn the complete and truncated reccirds are
almost identical; the other records reflect the general trend that as the motion be-
comes weaker, the effect of truncating to the effective duration is more pronounced.
Therefore, for some very weak records, the spectra are appreciably Merent, but
neither the full nor the truncated accelerogram produce spectra that present a
s i m c a n t hazard to engineered structures.

4. Predictive Equations for Strong Motion Duration


In order to be able to estimate expected values of effective duration in seismic
hazard studies, it is necessary to develop attenuation relationships to predict mean
M ~ of Seffective duration as a function of source, path and site parameters. In
this section, the correlations between the effective duration and magnitude, site
154 J. J . Bommer & A. Martinex-Peeint

conditions and distance are investigated to explore both the form of the attenuation
relationship and the feasibility of obtaining robust predictive equations. Before
these exploratory analyses, the existing attenuation relationships for strong motion
duration are briefly reviewed.

4.1. Existing attenuation relationships for stmng motion dumtion


It is not possible to make direct comparisons amongst attenuation relationships
for duration. As with any attenuation relationships, comparisons are complicated
by the use of different magnitude scales and different definitions of distance, and
even for simple parameters such as peak ground acceleration the comparison needs
to take into account the use of the two horizontal components of each accelere
Downloaded by [University of Wollongong] at 21:08 19 November 2013

gram [Ambraseys and Bommer, 1995). For the case of duration, comparisons
are further complicated by the widely differing definitions of duration that have
been employed.
Four studies have produced predictive equations for bracketed duration (ao =
0.05 g) and their predictioiis for a magnitude 6.5 earthquake as a function of distance
are shown in Fig. 15. It is interesting to note the wide range of predicted values of
duration in the near-field.
Figure 16 shows the predicted values of significant duration (5-95% of AI) for
a magnitude 6.5 earthquake according to four different attenuation relationships.
The near-field predictions again differ by a factor of more than three, but the most
interesting feature here is that all four equations predict increasing duration with
distance from the source. This results from the fact that the significant duration is
related t o the geometry of the accelerogram, regardless of its absolute amplitude.
The length of the earthquake signal does indeed grow with distance due to different
wave propagation velocities and multiply reflected and refracted amvals, and this
is reflected in the curves in Fig. 16.This tendency for durations defined by relative
criteria is also observed with the predictive equations of Zhou and Katayama (19851
and Theofanopulos and Watabe [1989]. The one exception is the model of Saragoni
[19771,which predicts increasing durations up to distances of 30 km and decay
beyond that point. For the scenario presented in Fig. 16 the equations of Saragoni
(19773 predict durations of 2.3, 10.8 and 3.5 sec at distances of 0, 30 and 100 krn
respectively.
It is interesting to note that in their study, although they used a definition of
duration based on relative criteria, Theofanopulos and Wat abe (19881actually point
out that "it must also be mentioned that the stmng motion dumtion of o +ewrdfrom
a magnitude 6 w less earthquake at distances greater than approzimately 100 km
i s meaningless". This raises a very important issue and questions the validity of
the predictive relatiomhips in Fig. 16. It is also worth noting that the predictions
of bracketed duration in Fig. 15 also show non-zero durations at 100 km,and
only the equation of Kawashima and Aizawa (19891predicts values tending towards
zero.
The Eflective Duration of Earthquuke Strong Motion 155
Downloaded by [University of Wollongong] at 21:08 19 November 2013

0 X ) 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 O X ) c )
Epicentrd Distmce (km)

Fig. 15. Predicted values of bracketed duration (ao = 0.05 g) as a function of distance for a
magnitude 6.5 earthquake according to different attenuation equations: 1 - McGuire and Barn-
hard [1979];~2 - Kawashima and Aizawa [1983]; 3 - Margaris et al. [1990]; 4 - Papazachos
et al. [1992].For 1, 3 and 4, rock site assumed; 2 is independent of site conditions.

Although it is the opinion of the authors that there are problems associated
with the use of any of the existing attenuation relationships for strong motion
duration, it is worth noting here the independent parameters that they include
as predictor variables. There is universal agreement that the duration depends on
magnitude and many equations relate ground motion duration only to magnitude,
such as Gutenberg and Richter (1942, 19561, Housner [1965],Kobayashi [1971],
Bolt 119731, Hisado and Ando [1976] and Dobry et al. [1978],amongst others. The
first equation to predict duration as a funtion of both magnitude and distance was
presented by Esteva and Rosenblueth (19641, in which the duration of Rosenblueth
and Bustamante [I9621 increased exponentially with magnitude and linearly with
distance. Tkifunac and Novikova [1995b]include the effect of varying geology dong
the travel path and the resulting wave dispersion.
Many attenuation relationships also include the effect of the site geology, usually
simply classifying sites as rock or soil and incorporating the influence of the .site
conditions through a dichotomous variable in the attenuation relationship [McGuire

b ~ ist assumed that the equation is in terms of D rather than ln(D) as presented in the original
paper, since this would otherwise result in durations of the order of weeks!
156 J. J. Bommer 43 A. Martl'nez-Pereim
Downloaded by [University of Wollongong] at 21:08 19 November 2013

Fig. 16. Predicted values of significant duration (5-95% of GI) as a function of distance for a
magnitude 6.5 earthquake according to different attenuation equations: 1 - Trifunac and Brady
119751; 2 - McGuire and Barnhard [1979]; 3 - Kamiyarna [l984]; 4 - Margaris et al. (19901. For
1, 2 and 4, rock site assumed; 3 is independent of site conditions.

and Barnhard, 1979;Papazachos et aL, 19921. Some studies, such as Trifunac and
, Brady [1975],used three categories and a dummy variable taking values of 0,1 and
2 in the equation, although this fixes to two the ratio of the effect of stiff and soft
sites with respect to rock. As discussed in Sec. 4.3, a few studies have recognised
that the classification of the surface geology by itself is not a useful predictor for .

strong motion duration.


Finally, it should be noted that a few studies have related strong motion duration
to macroseismic intensity [Margaris et d., 19901 or to intensity and other parameters
such as distance and soil conditions [Trifunac and Brady, 1975; Westerno, 1980;
W a c and Westerno, 1982; Novikom and 'IWunx, 1993a, 1993b], If seismic
hazard is evaluated directly from the history of observed seismic intensities in a
region, then the first type of equation m a y be useful. Otherwise, nothing is gained *

by expressing duration, which is a response variable, as a function of intensity (which


is also a response variable) rather than purely in terms of independent predictor
variables such as magnitude, distance and soil conditions.
The Eflectiue Dumtion of Earthquuke Strong Motion 157

4.2. Infience of magnitude on strong motion dumtion


At a rock site in the near-field of an earthquake, the duration of strong shaking
should be closely related to the duration of the fault rupture. If a constant velocity
of rupture of 2.5 km/s is assumed, then horn the relationship of Wells and Copper-
smith [I9941between fault rupture length and moment magnitude, the following
equation for the rupture duration, DR,can be derived:

For very large magnitude earthquakes it would be expected that the duration of
ground shaking would cease to increase with increasing magnitude, simply because
the rupture dimensions become so long that energy released from distant portions
Downloaded by [University of Wollongong] at 21:08 19 November 2013

of the fault will attenuate before arriving at sites tens or hundreds of kilomet res
away along the rupture.
To explore the relationship between effective duration and moment magnitude, a
dataset of 32 records was compiled for rock sites at distances no greater than 10 krn
from the earthquake source. These records were taken from the datasets presented
by Ambraseys et al. (19961and Somerville et al. [1997] and the effective duration is
taken as the mean of the values calculated for the two horizontal components. In
Fig. 17 the values of the logarithm of effective duration are plotted against moment
magnitude, Evl, together with the predicted values of D R from Eq. (18). The plot

45 5.0 55 6.0 6.5 7.0 7.5


Magnitude Mw
Fig. 17. Effective durations (mean of horizontal components) as a function of moment magnitude
for records from rock sites at distances of less than 10 km. Solid line is the best-fit found by
regression analysis. Dashed line is the rupture duration predicted by Eq. (18).
158 J. J. Bommer EI A. Martinez-Pereirn

also shows the best fit from regression of the d u e s of log(DE) on Mw and it can
be seen that these are consistently smaller than the values of DR.In part this
arises from the fact that the derivation of Eq. (18) assumes that the fault rupture
propagates in one direction only. Clearly, for bilateral fault ruptures such as the 1989
Lorna Prieta [Wald et al., 19911 and the 1995 Hyogo-ken Nanbu [Kanamori, 19951
earthquakes, the duration of rupture would be half of that predicted by Eq. (18).
To account for this effect, a factor was defined based on the distances L1 and L2
which are simply the longer and shorter distances (measured parallel to the surface
break) horn the epicentre to the ends of the fault rupture respectively. For pure
unilateral rupture, L1 would be equal to the total rupture length, L, and L2 would
be zero; for pure bilateral rupture, L1 and L2 would both be equal to L / 2 . The
Downloaded by [University of Wollongong] at 21:08 19 November 2013

correction factor, F, for the effective duration calculated from an accelerogram is


then defined simply as

b1
Figure 18 shows the data of Fig. 17 corrected by the factors F which are presented
in Table 6. For many aftershocks it has not been possible to determine the value of
F and the same is true for most earthquakes smaller than M, = 5. The best fit to
the data is now given by the equation

Fig. 18. As Fig. 17, but durations adjusted by ia-r F for non-unilaterd rupture. Solid line is
Eq-(20)-
The Efle-ctiue Dumtiun of Earthquake Stmng Motion 159

Table 6. Correction factors for non-unilateral rupture.

Year b1 D Time Earthquake MW h F


Name (km)

Imperial Valley
Eureka
Parkfield
Koyna
San Fernando
Acona
Ancona
Acona
Downloaded by [University of Wollongong] at 21:08 19 November 2013

Ionian
Azores
Oroville
Orovilie
Friuli
Friuli (A.S)
Uzbekistan
Denizii
Fkiuli (A.S)
Friuli (A.S)
Friuti (A.S)
F'riuli
Sicily
Tabas-e-Golsham
Montenegro
Montenegro (A.S)
~ursunbe~
Coyote Lake
Valnerina
Imperial Valley
Marnmoth Lakes
Campano L. (A.S)
Alkion
Preveza
Coalinga
Balikesir
h4org& Hill
Lazio Abruzzo
Kiparisia
160 J. J. Bommer 4Y A. Martinez-Pmirn

Table 6. (Continued)

Year M D Time Earthquake MW h F


Name (km)
Nahanni
No. Pdm Springs
Kalamata
Kalamata (A.S)
San Salvador
Killini
Lorna Prieta
Downloaded by [University of Wollongong] at 21:08 19 November 2013

Chenoua
Dzhava-Racha
Dzhava-Racha (A.S)
Erzincan
Landers
Pyrgos
Patras
Nort hridge
Hyogo ken-Nanbu

with a standad deviation of 0.28. It can be seen Erom Fig. 18 that the predicted val-
ues of effective duration are always 20% smaller than the predicted values of rupture
duration. This model agrees with the assumption of Trifunac and Novikova [1995a]
that the duration of the rupture process is an exponential function of magnitude.
Somenrille et a2. [I9971have developed factors to .correct peak ground and re-
sponse spectra accelerations and strong motion durations for the effects of rupture
directivity. Although the definition of duration used by Somerville et d.119971 dif-
fers from effective duration, the factors presented in their paper, and additional
factors calculated in the same way for records from Ambraseys et al. [1996],are
applied to the data in Fig. 18. The adjusted data is presented in Fig. 19 and it can
be seen that now the mean values of effectiveduration are almost identical to the
predicted values of rupture duration, although the scatter is actually increased by
the application of these factors.
The c w e s in Fig. 18 suggest that a physically meaningfid prediction of DE as a
function of magnitude can be made, although the user will need to be aware that the
predicted value could be overestimated by a factor of two if the postulated future
event involves bilateral rupture. Clearly, this is something that is very difficult to
assess a in seismic hazard assessment. It is also important to note that there
is a sigmficant scatter in Eq. (18),which may be the result of other features of the
fault rupture that are very d i f f ~ d tto predict for future events, including rupture
The Effective Durntion of Earthquclke S t m g Motion 161
Downloaded by [University of Wollongong] at 21:08 19 November 2013

4.5 5.0 5.5 6.0 6.5 7.0 7.5


Magnitude Mw
Fig. 19. As Fig. 18, but durations also adjusted by the directivity factor of Somerville et al. [l997].

velocities different from 2.5 km/s and the non-uniform rupture history, including the
rise-time and stopping and starting phases. It has been suggested that as a result
of non-uniform rupture propagation, a simple relationship between magnitude or
moment and the duration of ground motion may not exist [Das and Richards, 19793.
For effective duration, it appears that there is a simple correlation between the two
parameters, but it is associated with large scatter.
As a final obsemtion, Figs. 20 and 21 show the same data as in Fig. 18, again
corrected by the F factors fiom Table 6, but for bracketed (ao = 0.05 g) and
significant (5-95% of AI) durations respectively. Although the significant durations
show less scatter than the effective durations in Fig. 18, the large differences between
the best-fit regressions and the predicted d u e s of rupture duration in Figs. 20 and
21 confirm that effective duration has a superior physical basis.

4.3. Infience of site conditions on strong motion duration

All existing attenuation relationships for strong motion duration that include the
influence of site geology predict longer shaking on soil sites than rock sites. In order
to explore the dependence of effective duration on site conditions, a dataset of 90
accelerograms obtained a t distances of less than 10 km at sites not classified as
rock was compiled from the datasets of Ambraseys et al. [I9961 and Somenrille
et al. 119971. The effective duration of these records are plotted against moment
162 J. J. Bommer EI A. Martinez-Penzirn
Downloaded by [University of Wollongong] at 21:08 19 November 2013

45 5.0 5.5 6.0 6.5 7.0 ' 75


Magnitude Mw
Fig. 20. Bracketed durations (ao = 0.05 g) as a function of moment magnitude for records from
rock sites at distances of less than 10 km. Solid line is the best-fit found by regression analysis.
Dashed line is the rupture duration predicted by Eq. (18).

5 30.0-
.-
Y

#
no-
C
&
.- 6.0-
*
'6
.- 3.0-
UY

10-
0.6 -

45 50 55 6a 6s 7.0 75
Magnitude Mw

Fig. 21. Significant durations (595% of AI) as a function of moment magnitude for records born
rock sites at d i c e s of less than 10 km. Solid line is the best-fit found by regression analysis.
Dashed line is the rupture duration prdined by Eq. (18).
The Eflective Dunation of Earthquake S t m g Motion 163
Downloaded by [University of Wollongong] at 21:08 19 November 2013

O . I ~ l i r i ~ i i r i i 1 1 1 1 1 r ~ ~ ~ 1 i ~ ~ ~ ~ r l i 1 1

45 5.0 5.5 6.0 6.5 7.0 7.5


Magnitude Mw
Fig. 22. Effective durations (mean of horizontal components) as a function of moment magnitude
for records from soil sites at distances of less than 10 km. Solid line is the best-fit found by
regression analysis. Dashed line is the rupture duration predicted by Eq. (18).

magnitude in Fig. 22 and again in Fig. 23 after the application of the F factors in
Table 6. The best-fit line in Fig. 23 has the equation

with a standard deviation of 0.30. As for the rock records considered previously,
the agreement with the predicted mean values of D R is good, but the scatter in
the D E values is even larger in this case. This is not surprising, since in this case,
in addition to the factors related to rupture propagation discussed in the previous
section there are also other factors not accounted for in the very simple grouping
of the records into a single 'soil' category.
Equations (20) and (21) can be combined to show that the ratio of the mean
values of DE on soil and rock sites is given by

Therefore, durations for soil sites are fractionally greater (20% for a magnitude 7.5 -
earthquake), but the differencebetween the two predictions is considerably smaller
than the uncertainty in either. Physically, it would not necessarily be expected that
the presence of soil layers at a site would significantly increase the effective du-
ration even if the amplitudes of the motion are magnified. This is illustrated by
164 J. J. Bummer @ A. Martinex- Pereirn
Downloaded by [University of Wollongong] at 21:08 19 November 2013

Fig. 23. As Fig. 22, but durations adjusted by factor F in Table 6. Solid line is Eq. (21).

Fig. 24 which shows accelerograms from rock and soil sites recorded during the
1986 San Salvador and 1994 Northridge earthquakes. The amplitudes of the accel-
eration peaks are increased by the response of the soil layers, but the durations do
not increase by more than a small amount. For the length of shaking to increase
significantly at a site, it is necessary for seismic wave energy to become trapped
by multiple reflections at the ground surface and at the interface between the soil
layer and the underlying bedrock. Therefore, significant prolongation of the shaking
cannot be predicted simply on the basis of the type of ground at the surface, but
rather it depends on the thickness and softness of the soil layer and the impedance
contrast between the uppermost layer and its base [Kamiyama, 19841. Some pre-
dictive models for duration have incorporated explicitly the depth of sediments at
the site [Westerno, 19801.This parameter is not known for more than a few of the
sites in the dataset used in this study, so this is not explored further at this stage.

4.4. I n . e n c e oj distance o n st- motion dumtion


. The dependence of strong motion duration on distance is complex. On one hand, as
the wave train separates due to different propagation velocities and scattering, the
duration will increase with increasing distance. On the other hand, any definition
constrained by absolute levels of acceleration or energy will decrease with distance
due to attenuation of the motion. For the effective duration, which in the near-field
The Efective Dutation of Earthquake S t m g Motion 165
Downloaded by [University of Wollongong] at 21:08 19 November 2013

Time (see) Time (see)

Fig. 24. Records from two earthquakes, showing amplification by soil deposits without increase
in effective duration. Top: 1986 San Salvador (El Salvador) earthquake at HSH (left) and IVU
(right); bottom: 1994 Northridge (California)at Pacoima Kagel Canyon (left) and Sylmar Hospital
(right).

is related mainly to the S-wave train for genuinely strong motions, the attenuation
effect is likely to dominate.
In order to examine the dependency of effective duration on distance, the data
from a number of well recorded earthquakes are plotted in Fig. 25. In each case,
the average effectiveduration at each station, corrected by the F factor, are plotted
against distance. The distance is measured horizontally from the closest point on
the surface projection of the fault rupture. It can be seen that there is a very clear
tendency for the durations to increase with distance, although some soil records at
distances of around 40 km in the Whittier Narrow, Lorna Prieta and Northridge
earthquakes, show surprisingly high durations. The case of the Loma Prieta earth-
quake is particularly interesting: the appreciable durations on both rock and soil
sites at distances of between 60 km and 80 km are the result of exceptionally soft
soil deposits in San Francisco Bay and also to "Moho bouncen [Somerville and
Yoshimura, 19901.In all cases there is very significant scatter and no clear trend of
durations at soil sit& being consistently larger than those at rock sites. It is clear .
that in many cases, soil sites result in significant increase of the effective duration,
but with this simple binary site classification it is not possible to isolate the records
for which this is the case. It is also possible that at short distances, directivity is
responsible for part of the scatter; however, the factors of Somenrille et aL (19971
166 J . J . Bommer & A. Martinez-Pemim

San Fernando, 9 February 1971


-7

Lorna Rieta, 18 October 1989


=J
Downloaded by [University of Wollongong] at 21:08 19 November 2013

Impend Valley. 15 October 1979


3 4
0 ,
0
7 , .

Q 2 O J O 4 0 5 O s O 7 O s O o O * K 1
. - y m , s , 7 , a i

Distmcs (km)

Nonhridge, 17 January 1994


-3
a O ! r r . , - . . - . , a 6 . I , . 3 I . # - I
0 D 2 Q X I W S ~ 7 O Q ~ W
Distonca (krn)

Whittier N a m w r 1 October 1987

AD.
0
,:,

~
, ,Or

2 0
,-, ;;
X ~
0

!
;;
;
% O
. ;mT-

6
,

0
, -,

7 0 ~ W ~
Diitmm (km)

Fig. 25. Effective durations, corrected by factors for non-unilateral rupture, as a function of
distance from the surface projection of the rupture for five earthquakes. Circles represent rock
sites, triangles soil sites.

have not been applied in this case because the definition of duration that they
employed is Merent from the effective duration.
An important feature to be observed in Fig. 25 is that even at relatively short
distances,there are many aceelemgrams for which the effective duration is zero.This
The Effective D u d i o n of Earthquuke Strong Motion 167

is an important feature and any predictive equation for effective duration must be
able to indicate the magnitudedistance combinations beyond which the value of DE
is zero. This is a departure from the usual practice in deriving attenuation relations.
For example, PGA values are obtained from instruments that produced records and
the resulting equations, for distant events, will predict very low values, including
values lower than the triggering level of accelerogaphs. For effective duration the
situation is quite different, because for a dataset including records outside a certain
magnitude-distance space there will be many zero values of DE.If a logarithmic
form of the regression model is used, this creates obvious problems. Clearly, if the
zero duration records were simply ignored it would result in an overestimation of
the durations at longer distances. One approach in some PGA studies has been
to eliminate horn the dataset of each earthquake, all the records from distances
Downloaded by [University of Wollongong] at 21:08 19 November 2013

greater than the distance to the closest non- triggered accelerograph [Joyner and
Boore, 19811. The equivalent practice for effective duration would be to eliminate
all data a t distances greater than that corresponding to the first record with zero
duration. This would at least cause the curves to attenuate more rapidly in the far-
field. Another possibility is the use of maximum likelihood methods, which have
been used to incorporate non-triggered instruments into regression analyses for
PGA [McLaughlin, 1991].

5. Discussion

In this paper, some 30 definitions of strong motion duration have been reviewed and
classified. In light of the inconsistencies that can arise with these existing definitions
and the requirement of a record-based definition that can be used to develop gen-
eral predictive equations for use in seismic hazard assessments, a new duration is
proposed. The new definition is based on the previously unexplored option of using
absolute critieria to identify the limits of the strong shaking phase from the Husid
plot. This effective duration, D E , is shown to consistently constrain the portion of
strong earthquake accelerograms that is of greatest engineering significance. The
effective duration definition could be refined by varying the values of the thresh-
olds used to constrain the strong shaking phase. It is clear, nonetheless, that it
provides a simple and robust way to define a continuous time window on strong
accelerograms that contain the motion of real engineering importance.
In this preliminary study, the effective duration has been defined for single
components of ground motion. However, since it is based on the accumulation
of Arias intensity, it could easily be applied to both horizontal components by
summing their indivudal Arias intensities and adjusting the values of Ab and AAIf
accordingly.
It would be possible to derive attenuation equations for effective duration on
rock sites using a simple function of magnitude and distance. However, at this stage
the authors have prefered to explore the correlations between duration and these
predictor variables. It is clear that a good correlation exists between magnitude and
168 J. J. Bommer & A. M a d n e z - P k r n

effective duration in the near-field and that this correlation has a sound physical
basis. The scatter associated with the relationship could be improved by deriving
factors to account for rupture directivity effects, as Sornerville et at. [I9971have
done, although the rock site dataset is relatively small. However, it is also important
to note from the outset that the durations can vary by up to a factor of two
depending on whether the fault rupture is bilateral or unilateral, which cannot easily
be predicted for future events. It would be possible to apply the F factor to the
data and then perform regressions to determine mean values on the upper bounds
of durations, effectively normalking the dataset to unilateral rupture. However, as
explained at the beginning of this paper, in many situations it is the lower bound
on the duration that will be of greater interest from an engineering point of view.
Regarding the distance-dependence of effective duration, it is first necessary to
Downloaded by [University of Wollongong] at 21:08 19 November 2013

identify those records for which the site characteristics play a clear rolein extending
the duration. This will involve incorporating the depth of soil deposits and possibly
other geotechnical site characteristics into the model. It is unlikely that a generalised
model could be developed which incorporates Moho bounce, so records affected
by this phenomenon could legitimately be removed from the dataset. It will then
be possible to obtain equations to predict mean values of effective duration as a
function of distance.

Acknowledgements
At various stages in the development of this work, the authors have benefitted horn
fruitful discussions with Nicholas Ambraseys, Art uro Arias, Amr Elnashai, Matt hew
Free, Enrique Martinez-Rueda, Sarada K. Sarma, Keith Simpson and Milutin Sr-
bulov. The paper has also benefitted very much from reviews made by David Boore
and three anonymous referees. We are also very grateful to Nancy Smith for provid-
ing important strong motion records. The support of COLCIENCIAS is gratefully
acknowledged.

References
Ambraseys, N.N. [I9881 "Engineering seismology," Earthq. Engtg. Struct. Dyn 17,1-105.
Ambraseys, N. N. and Bommer, J. J. (19951 "Attenuation relations for use in Europe:
An overview," P m . Fifth SECED Conference on European Seismic Design Practice,
Chester, England, pp. 67-74.
Ambraseys, N.N.and Sanna,S. K. 119671 'Response of earth dams to strong earthquakes,"
Ge'otechnique 17(2), 181-283.
Ambraseys, N. N., Simpson, K. A. and Bommer, 3. J. [I9961 'Rediction of horizontal
spectra in Europe," Earthq. E m . Struct. Dyn. 25, 371-400.
Ambraseys, N. N. and Srbulov, M. (19941 uAttenuation of earthquake-induced ground
displacements," Earthq. Engrg. Stnrct. Dyn. 2 3 , 4 6 7 4 7 .
Araya, R. and Saragoni, G. R. (1980) "Earthquake accelerogram destructiveness poten-
tial factor," Prac. Seventh World C m f m on Earthquake En@neefing,Istanbul,
pp. 483-490.
The Eflective Dvnrtion of Earthquake Strong Motion 169

Arias, A. [1970] "A measure of earthquake intensity," in Seinnic Design for Nuclear Power
Plants, ed. Hansen, R. (MITPress,Cambridge, Massachusetts), pp. 438-483.
Bolt, B;A. (19731 "Duration of strong ground motions," Proe. Fiflh World Conference on
Earthquake Engineering, Vol. 1, Rome,pp. 1304-1313.
Bommer, J. J. and Martinez-Pereira, A. 11996) "The prediction of strong-motion duration
for engineering design," Pmc. Eleventh World Conference on Earthquake Engineering,
Acapulco, Paper No. 84.
Bommer, J. J., Udias, A., Cepeda, J. M., Hasbun, J. C., Salazar, W. M., S u t e z , A.,
Ambraseys, N. N., Buforn, E., Cortina, J., Madariaga, R., Minda, P., Macua, J.
and Papastamatiou, D. I19971 UA new digital accelerograph network for El Salvador,"
Seismological R w e a d Letters 68(3), 426-437.
C a b a i k , L., Benito, B. and Herraiz, M. (19971 "An approach to the measurement of the
potential structural damage of earthquake ground motions," EarUlq. Engq. Stmet.
Dyn. 26, 79-92.
Downloaded by [University of Wollongong] at 21:08 19 November 2013

Das, S. and Richards, P. G. 119791 "Effects of non-uniform spontaneous rupture p r o p


agation on the level and duration of earthquake ground motion," Pmc. Second US
National Conference on Earthquake Engineering, Stanford, pp. 203-212.
Dobry, R., Idriss, I. M. and Ng, E. I19781 "Duration characteristics of horizontal compe
nents of strong-motion earthquake records," Bull. SeiP. Soc. Am. 68(5), 1487-1520.
Donovan, N. C. [1972]"Earthquake hazards for buildings," Building Science Series 46,
82-111.
Elghadamsi,F. E., Mohraz, B., Lee, C. T. and Moayyad, P. (19881 'Timedependent power
spectral density of earthquake ground motion," Soil Dyn. Earthq. Engrg. 7 , 15-21.
Esteva, L. and Rosenblueth, E. [I9641 "Espectros de temblores a distancias moderadas y
grades," Boletin Sociedad Meziurno de ingenieria Shmicn 2(1), 1-18.
Gutenberg, B. and Richter, C. F. [I9421 "Earthquake magnitude, intensity, energy and
acceleration," Bull. Seis. Soc. Am. 32(3), 163-191.
Gutenberg, B. and Richter, C. F. [I9561 "Earthquake magnitude, intensity, energy and
acceleration (second paper)," Bull. Seis. Soc. Am. 46(2), 105-145.
Hisada; T. and Ando, H. 119761 "Relation between duration of earthquake ground motion
and magnitude," Kujtma ICT, Japan, June issue.
Housner, G. W. 119651 "Intensity of ground shaking near the causative fault," Pmc.
Third World Confezence on Emthquake Engineering, Vol. 1 , Auckland, New Zealand,
pp. 81-94.
Housner, G. W. (19751 "Measures of severity ofground shaking," Pmc. US National Con-
/emace on Earthquake Engineering, Ann Arbor, Michigan, pp. 25-33.
Husid, L. R. (1969)"Caracteristicas de terremot os. An6lisis general," Revisto del I D E M
8, Santiago de Chile, 21-42.
Joyner, W. B. and Boore, D. M. I19811 "Peak horizontal acceleration and velocity from
strong-motion records including records born the 1979 Imperial Valley, California,
earthquake," Bull. Seis. Soc. Am 71(6), 2011-2038.
Kamiyama, M. (19841 "Effects of subsoils conditions and other factors on the duration of
earthquake ground shakings," P m c Eighth World Conference on Earlhquake Engi-
neering, Vol. 2, San Francisco, pp. 793-800.
Kanamori, R. (19953 "The Kobe (Hyogdsen Nanbu), Japan, earthquake of January 16,
1995," Seismologid Reseanlr Letters 66(2), 6-10.
KawaPhima, K. and h w a , K. I19891 "Bracketed and normalized duration8 of earthquake
ground acceleration," Ea7th.q. Engrg. Stmct. Dyn. 18, 1041-1051.
Kobaydu, Y. [1971] "Effects of earthquakes on ground (11)," J. Phys. Earth 19,
231-241.
170 J. J. Bommer & A. Martt'nez-Pmim

Liu, H. and Zhang, Q. 119841 "Ground-motion .features and earthquake destructive-


ness," Pmc. Eighth World C o n f e ~ n c eo n Earthquake Engineering, San Francisco,
pp. 881-888.
Mahin, S. A. 119801 "Effects of duration and aftershocks on inelastic design earth-
quakes," Pmc. Seventh World Conference on Earthquake Engineering, Voi. 5, Istan-
bul, pp. 677-680.
Margaris, V. N., Theodulidis, N. P., Papaioannou, Ch. A. and Papazachos, B. C. (19901
"Strong mot ion duration of earthquakes in Greece," P m . X X I I General Assembly of
the European Seismological Commission, Barcelona.
McCann, M. W. and Shah, H.C. (1979)"Determining strong-motion duration of earth-
quakes," Bull. Seis. Soc. Am. 69, 1253-1265.
McGuire, R. K. and Barnhard, T.P.(19791 "The usefulness of ground motion duration in
prediction of severity of seismic shaking," Pmc. Sewnd US IVsrtional Conference on
Earthquake Engineering, Stanford, pp. 713-722.
Downloaded by [University of Wollongong] at 21:08 19 November 2013

McLaughlin, K. L. [I9911 ''Maximum likelihood estimation of strong-motion attenuation


relationships," Earthq. Speck 7, 267-279.
Mohraz, B. and Peng, M.-H. [I9891 'The use of low-pass filter in determining the dura-
tion of strong ground-motion," J. Ptessure Vessels and Piping Division ASCE 182,
197-200.
Novikova, E. I. and Tkifunac, M. D. [1993a] "Modified Mercalli intensity scaling of the
frequency dependent duration of strong ground motion," Soil Dyn. Earthq. Engrg.
12,209-225.
Novikova, E. I. and Trifunac, M. D. [1993b]"The Modified M e r c d intensity and the
geometry of the sedimentary basin as scaling parameters of the frequency dependent
duration of strong ground motion," Soil Dyn. Earthq. Engq. 12, 309-322.
Novikova, E.I. and Trifunac, M. D. (19941 tiDuration of strong ground motion in terms of
earthquake magnitude, epicentral distance,site conditions and site geometry," Earthq.
Engrg. Struct- Dyn. 23, 1023-1043.
Page, R. A., Boore, D. M., Joyner, W. B. and Coulter, H. W. [I9721 "Ground motion
values for use in seismic design of the trans-Alaska pipeline system," US GeoIogicuf
Sumey Circular 672.
Pagratis, D. [I9951 "Prediction of earthquake strong ground-motion for engineering use,"
M.Sc. Dissertation, Imperial College, London.
Papazachos, B. C.,Papaiannou, Ch. A., Margaris, V. N. and Theodulidis, N. P. [I9921
"Seismic hazard assessment in Greece based on strong motion duration," Proc. Tenth
World Conference on Earthquake Engineering, Vol. 2, Madrid, pp. 425-430.
Perez, V. [19801 "Spectra of amplitudes sustained for a given number of cycles: An in-
terpretat ion of. response duration for strong-motion earthquake records," Bull. Seis.
Soc. Am. 70(5), 1943-1954.
Rosenblueth, E.and Bustamante, J. I. 11962) 'Pistribution of structural response to earth-
quakes," J. Engy. Mefh. Div. ASCE 88(EM3),75-106.
Saragoai, G. R [19?7]T h e a& method for the characterization of earthquake accelero-
grams," Proc. Sidh World Confemce un Earthquake Engineering,Vol. I , New Dehli,
pp. 357-362.
S a m , S. K. [197ll %nergy flux of strong earthquakes," Tectonophysics 11,15%173.
Sarma, S. K. and Casey, B. J. [I9901 T h a t i o n of strong motion earthquakes,"
Prrrc Ninth European Conftmmx on Eadquake Engineering, Vol. 1&A, Moscow,
pp. 174-183.
S a w S. K. and Yang, K. S. (19873 UAn evaluation of strong motion records and a new
parameter Ass," Earthq. E m . Stnrcf- Dyn, 15, 11+132.
The Eflectiwe h t i o n of Earthqwke Strong Motion 171

Seed,H. B. and Idriss, I. M. [I9821 Gmund Motions and Sod Liquefaction During Earth-
quakes (EERI Monograph, Earthquake Engineering Research Institute, Oakland,
CA) .
Shahabi, M. and Mostaghel, N.(19841 %trong ground motion duration and effective cyclic
acceleration," Proc Eighth World Conference o n Earthquake Engineering, Vol. 2, San
Francisco,pp. 843-850.
Somerville, P. G., Smith, N.F.,Graves, R.W. and Abrahamson, N. A. [1997]"Modification
of empirical strong ground motion attenuation relations to include the amplitude
and duration effects of rupture directivity," Seismological Reseumh Letters 68(l),
199-222.
Somerville, P.G. and Yoshimura, J. [I9901 'The influence of critical Moho reflections on
strong ground motions recorded in San Francisco and Oakland during the 1989 Lorna
Prieta earthquake," Geophysical Research Letters 17, 1203-1206.
Sucuo#lu, H. and Nurt ug, A. [I9951 'Zarthquake ground mot ion characteristics and seismic
Downloaded by [University of Wollongong] at 21:08 19 November 2013

energy dissipation," Eorthq. Engsg. and Struct. Dyn. 24, 1195-1213.


Theofanopoulos, N. and Drakopoulos, J. I19861 "A study of strong ground motion du-
ration of earthquakes in Greece," Proc. Eighth European Conference o n Earthquake
Engineering, Vol. 1, Lisbon, pp. 3.2165-3.2167.
Theofanopulos, N. A. and Watabe, M. [1989] UAnew definition of strong motion duration
and comparison with other durations," Stmct. Engrg. Earthq. Engrg. 6(1),111s-122s
(Proc.of JSCE No. 404/I-11).
Trifunac, M. D. and Brady, A. G. 11975) "A study on the duration of strong earthquake
ground motion," Bull. Seis. Soc. Am. 65, 581-626.
Trifunac, M. D. and Novikova, E. I. (1995aI "Duration of earthquake fault motion in
California," Earthq. Engrg. Stmct. Dyn. 24, 781-799.
Trifunac, M. D. and Novikova, E. I. [1995b] "State of the art review on strong motion du-
ration," Pmc. Tenth European Conference on EaTUIquake EngmeerCng, Vol. 1,V i e ~ a ,
pp. 131-140.
Trifunac, M. D. and Westermo, B. 119771 'A note on the correlation of frequency-
dependent duration of strong earthquake ground motion with the Modified hlercalli
intensity and the geologic conditions at the recording site," Bull. Seip. Soc. Am. 67(3),
917-927.
Trifunac, M. D. and Westermo, B.D. [I9821 "Duration of strong earthquake shaking,"
Soil Dyn. Earthq. Engrg. l(3), 117-121.
Vanmarcke, E. H. and Lai, S.-S. P. 119771 "Strong motion duration of earthquakes,"
Research Report R 77-16, Department of Civil Engineering, MIT, Cambridge,
Massachussets.
Vanmarcke, E. H. and Lai, S . 3 . P. [I9801 "Strong-motion duration and RMS amplitude
of earthquake records," Bull. Seis- Soc. Am. 70(4), 1293-1307.
Wdd, D.J., Helmberger, D. V. and Heaton, T. H. [I9911 'LRutpure model of the 1989 Lorna
Priet a earthquake horn the inversion of strong-motion and broadband t eleseismic
data," Bull. Seis. Soc. Am. 81,1540-1572.
Wells, D.L.and Coppersmith, K. J. (19941'New empirical relationships among magnitude,
rupture length, rupture width, rupture area, and surface displacement," Bull. Seis.
Soc. A m 84(4), 974-1002.
Westaway, R. and,Jackson, J. (19871 'The earthquake of 23 November 1980 in Campania-
Basilicata (southern Italy) ," Notun 312, 4 3 W 8 .
Westenno, B. [I9801 T h e duration of strong motion and its dependence on rmording
site geology," Proc Seventh World Conference on Earthquuke Engineering, Vol. 2,
Istanbul,pp. 263-266.
Xie, L.-L. and Zhang, X. [I9881 "Engineering duration of strong-motion and its effect on
seismic damage," Pmc. Ninth World Conjerence on Earthquake Engineering, Vol. 2,
TokpKyoto, pp. 307-312.
Zahrah, T.and Hall, W. J. (19841 "Earthquake energy absorption in SDOF structures,"
J. Stmct. Engtp. Diu. ASCE 11O(ST8), 1757-1772.
Zhou, Y. and Katayama, T. [1985] ''Effects of magnitude, epicentral distance and site
conditions on the duration of strong ground motion," Seisan Ken+ 37(12), 1&13.
Zhou, Y. and Xie, L. [I9841 "A new definition of duration of strong ground motion," Pmc.
Chinese National Confezme on Earthpuke Engineering, Shanghai, (in Chinese).
Downloaded by [University of Wollongong] at 21:08 19 November 2013

S-ar putea să vă placă și