Content uploaded by Ya-Ting Lee
Author content
All content in this area was uploaded by Ya-Ting Lee on Mar 30, 2015
Content may be subject to copyright.
ORIGINAL PAPER
An empirical equation of effective shaking duration
for moderate to large earthquakes
Ya-Ting Lee •Kuo-Fong Ma •Yu-Ju Wang •Kuo-Liang Wen
Received: 11 May 2014 / Accepted: 18 August 2014 / Published online: 4 September 2014
ÓSpringer Science+Business Media Dordrecht 2014
Abstract The duration of strong shaking is particularly important for assessing building
performance, potential landslides and liquefaction hazards. The results of this investigation
can potentially help reduce related fatalities and economic losses. In this study, we ana-
lyzed the acceleration seismograms of the Taiwan strong motion network to characterize
the strong shaking duration associated with earthquake sources, propagation paths and site
effects. This study proposes a new definition for the strong shaking duration called
‘‘effective shaking duration’’ (ESD), which considers the amplitude and radiation energy
decays. We first consider the window of a time series during which the amplitude is
C0.01 g, and we then defined the ESD as the length of the interval of the dissipated energy
within 5–95 % of the total energy during this time frame. We calculated the strong shaking
duration for 495 inter-plate events with magnitudes of M
L
[5.0 and focal depths \50 km
in the Taiwan region from 1994 to 2012. Using a nonlinear regression procedure, we thus
obtained an empirical equation for strong shaking durations. The equation is a function of
earthquake magnitude, distance and site conditions, which are defined by the V
s
30 value
(the S-wave velocity structure of the top 30 m of the site). The results indicate that the
shaking durations significantly increase with magnitude and also decrease with distance
and V
s
30. Compared with empirical equations from global datasets, our empirical equation
is applicable to earthquakes in other regions and will produce smaller but more applicable
duration values for smaller earthquakes. However, for larger events, our ESD values are
comparable with those derived from other definitions (e.g., significant duration). Although
the empirical relationship is mainly based on Taiwanese events, in view of the massive
dataset, this empirical equation could provide important information to the global com-
munity regarding the ground shaking duration estimation in the ground motion prediction
of future earthquakes.
Y.-T. Lee (&)K.-F. Ma Y.-J. Wang K.-L. Wen
Department of Earth Sciences and Graduate Institute of Geophysics, National Central University,
Jhongli 320, Taiwan, ROC
e-mail: shine2530@gmail.com
Y.-J. Wang
Institute of Earth Sciences, Academia Sinica, Taipei, Taiwan, ROC
123
Nat Hazards (2015) 75:1779–1793
DOI 10.1007/s11069-014-1398-7
Keywords Effective shaking duration Earthquake source Propagation path Site
effect Ground shaking duration estimation
1 Introduction
The duration of strong ground motion is critical to estimating seismic hazards, particularly
for building performance, landslide triggering and liquefaction (Trifunac and Novikova
1995; Rauch and Martin 2000; Hancock and Bommer 2005; Bommer et al. 2006; Kempton
and Stewart 2006). Lee et al. (1972) and Trifunac and Novikova (1995) calculated the
correlations between the duration signal and magnitude and between the duration signal
and distance to obtain the empirical equations for earthquakes in central California. Shoji
et al. (2005) analyzed earthquakes in Japan and obtained an empirical formula for the
strong shaking duration of Japanese earthquakes. The most recent study by Kempton and
Stewart (2006) presented equations for predictions of strong shaking durations (of a sig-
nificant duration) derived from the next generation of attenuation (NGA) global database
of accelerograms for earthquakes with a magnitude range of M5.0–7.6. Additionally,
Bommer et al. (2009) used the NGA dataset and presented empirical predictive equations
for additional duration definitions. Their equations can be used for estimating the strong
shaking durations of shallow crustal earthquakes with M
w
values of 4.8–7.9. Owing to the
dense strong motion network and high seismicity in Taiwan, we investigate the empirical
equation for strong shaking duration in the area in terms of earthquake magnitude,
earthquake distance, geology and local site conditions by utilizing the high-quality motion
data recorded by the Taiwan strong motion instrumentation program (TSMIP). Using a
nonlinear regression procedure, similar to the procedure in Seber and Wild (2003), we
obtained an empirical equation for the prediction of strong shaking durations. Although
some factors related to earthquake sources, near-field effects and rupture directivity may
improve the equation, the empirical equation derived here provides a first-order prediction
of strong shaking durations.
1.1 Definition of effective strong shaking duration
Strong shaking duration is commonly defined as ‘‘bracketed duration’’ (Bolt 1973): The
time interval between the first and last amplitudes greater than the threshold level of the
strong shaking duration value (e.g., Pagratis 1995; Stafford 2008; Bommer et al. 2009;
Kawashima and Aizawa 1989). Another definition is the ‘‘significant duration’’: The time it
takes for a designated percentage of the total energy to arrive (e.g., 5–95 % of the total
energy) (Trifunac and Brady 1975; Martin and Haresh 1979); it has also been widely used
in recent studies (Bommer and Martinez-Pereira 1999; Shoji et al. 2005; Bommer et al.
2009). Here, we combine the two definitions to produce an ‘‘effective shaking duration’’
(ESD) by considering both the strong ground motion amplitude and energy. The ESD was
calculated in two steps. First, we limited the time interval between the first and last
amplitudes by considering those values greater than or equal to a specified threshold value
(0.01 g). Then, the accumulated energy of the three components produced the ESD for the
time window, which has 5–95 % of the total energy within the amplitude threshold
(Fig. 1). The threshold value of 0.01 g was determined by considering the possible trig-
gered landslides PGA value in previous studies, e.g., Del Gaudio and Wasowski (2004),
1780 Nat Hazards (2015) 75:1779–1793
123
Sigaran-Loria et al. (2007) and Rathje and Saygili (2009). To clarify the parameters used in
this study with previous studies, we summarized the type and parameters of the referred
papers and this study in Table 1.
1.2 Strong ground motion data and effective shaking duration (ESD)
The TSMIP network, operated by the Central Weather Bureau (CWB), is composed of
approximately 700 accelerographs at free-field sites (Shin 1993) and has recorded high-
quality strong ground motion data since 1993. In 2000, the National Center for Research on
Earthquake Engineering (NCREE) and the CWB committed to a free-field strong motion
station drilling project to construct an Engineering Geological Database for the TSMIP
(EGDT). A total of 439 free-field stations in the TSMIP network were drilled and com-
pleted the logging measurements. The values of V
s
30 (the average S-wave velocity of the
top 30 mof the strata) of the drilled station were measured by the suspension PS logging
system of Kuo et al. (2012). The suspension PS logging system has two sensors at a fixed
distance of 1 m. The P- and S-waves produced were received by the sensors, and the
S-wave velocities of the shallow strata were then estimated.
Figure 2displays the distribution of the strong motion stations with site classes of the
TSMIP. The free-field strong motion stations of the TSMIP were divided into five site
classes: A (hard rock with V
s
30 [1,500 m/s), B (firm to hard rock with 1,500 m/
sCV
s
30 [760 m/s), C (dense soil and soft rock with 760 m/s CV
s
30 [360 m/s), D
(stiff soil with 360 m/s CV
s
30 C180 m/s) and E (soft soil with V
s
30 \180 m/s) classes
(Kuo et al. 2012). The site classification definition was determined according to the V
s
30-
based provisions of the National Earthquake Hazard Reduction Program (NEHRP). Most
of the stations belong to class C and D sites, and the stations of class C are located around
the Central Mountain and the Coastal Range. As shown in Fig. 2, site class D and E
stations are mainly located within plains and basins (Kuo et al. 2012).
In this study, we noted the earthquakes from 1994 to 2012 with magnitudes of M
L
[5.0
and focal depths \50 km. We considered M
L
rather than M
w
as M
L
is the more complete
catalog and is the magnitude firstly determined in real time upon occurrence of an
earthquake. It, thus, can be utilized further for real-time strong shaking duration prediction.
The conversion between M
L
and M
w
for Taiwan region had been examined by Lin and Lee
(2008). The magnitude in M
L
is about 0.2 larger than M
w
for events with M
L
of 5–7. Due to
no sufficient data of intra-plate events, we, thus, chose the crustal and inter-plate events
with focal depths of \50 km. In total, 495 earthquakes were selected (Fig. 3). We applied
the definition of ESD to the records. To avoid contamination with noise, we only chose
stations that had a PGA value [0.015 g. Additionally, the ESD determined should be no
\2 s. For the 495 earthquakes, using the criteria established in the data selection above, a
total of 11,639 records were utilized for our study (details of the data are shown in
Table 2). Of these records, 365 were utilized for site class B.
1.3 Development of the empirical equations for strong shaking duration
The duration of strong ground motion is associated with the earthquake source, propaga-
tion path and site effects.
s¼ssþsDþssite ð1Þ
Nat Hazards (2015) 75:1779–1793 1781
123
Here, sis the strong shaking duration in seconds as recorded by accelerographs at the
free-field sites, s
s
represents the earthquake source duration, s
D
represents the propagation
path dependence, and s
site
represents the site condition dependence. We form regression
Eq. (1) by the following steps.
1. Earthquake source duration, s
s
Hanks and McGuire (1981) and Boore (1983) assumed that the theoretical earthquake
source duration is equal to the reciprocal of the corner frequency that is related to the
seismic moment and stress drop index. Using the theoretical seismic source model
(Abrahamson and Silva 1996; Kempton and Stewart 2006), the regression model for the
source duration was formed as follows:
Fig. 1 Example of ESD estimations of the three component (V, NS and ES) acceleration seismograms at
the CHY006 station (site class C, V
s
30 =423 m/s) for the 1994/01/20 M
L
=5.58 earthquake. The locations
of this earthquake and the CHY006 station are shown in Fig. 3(yellow star and green triangle,
respectively); the earthquake has a hypocentral distance of approximately 151 km. The bottom panel
presents the cumulative energy with time. The blue lines mark the time window of the acceleration C0.01 g.
The red lines mark the time window of the accumulated energy of 5–95 % for the total energy of the
acceleration C0.01 g. The green lines mark the time interval of the SD
1782 Nat Hazards (2015) 75:1779–1793
123
ss¼1
fcðM0;DrIÞ¼1
4:9106b
M0
DrI
1=3
¼
DrI
101:5Mþ16:05
1=3
4:9106bð2Þ
where f
c
is the corner frequency, bis the shear-wave velocity of the crust at the source (set
as 3.2 km/s), and Dr
I
is the stress drop index that is related to the stress drop but not the
true stress drop of the event. The stress drop index is calculated from the duration values
using the source model (Eq. 2). M
0
is the seismic moment (in dyne-cm), which can be
converted from the magnitude (M
L
)asM
0
=10
1.5ML
?16.05
(Hanks and Kanamori 1979).
2. Propagation path dependence, s
D
, and the stress drop index, Dr
I
The logarithm of the strong shaking duration is considered to be a linear decrease with
distance (Kempton and Stewart 2006), written as follows:
log sD¼c1rð3Þ
where c
1
is a regression parameter and ris r
hyp
, which is defined as the hypocenter distance
(source to station distance) of the earthquakes in kilometers. To examine the relationship,
we used the accelerogram dataset of rock site (V
s
30 [760 m/s) recordings of large
earthquakes (M
L
=6.0–7.4) for every 0.2 magnitude interval (Fig. 4). The result fits the
distance decay regression of Eq. (3) well, which suggests the appropriate regression model
was chosen for the propagation path dependence, s
D
, of Eq. (3).
By combining Eqs. (2) and (3), the form of the regression model becomes the
following:
log s¼log
DrI
101:5MLþ16:05
1=3
4:9106b
2
6
43
7
5þc1rhyp ð4Þ
To determine the magnitude dependence of the stress drop index, Dr
I
, which was
proposed by Kempton and Stewart (2006), we investigated the magnitude dependence of
Table 1 Type and parameters used in referred papers and this study for prediction equation for strong
shaking duration
Author Duration
parameter
Magnitude
type
Distance
type
Site parameter
Bolt (1973) BD a/n r
hyp
a/n
Trifunac and Brady (1975) SD a/n r
epi
, h Soft alluvium, intermediate
rock and hard rock
Hernandez and Cotton (2000)SD M
w
r
rup
, h Rock, soil
Kempton and Stewart (2006)SD M
w
for M[6
M
L
for M\6
r
hyp
Vs30
Bommer et al. (2009)SDM
w
r
rup
,hVs30
This study ESD M
L
r
hyp
Vs30
Duration parameters: BD bracketed duration; SD significant duration; ESD effective shaking duration
Magnitude parameters: M
w
moment magnitude; M
L
local magnitude
Distance parameters: r
rup
the closest distance from the fault rupture; r
hyp
hypocentral distance, source to
station distance; r
epi
epicentral distance; hhypocentral depth
Site parameter: Vs30 average S-wave velocity of the top 30 m of the site
Nat Hazards (2015) 75:1779–1793 1783
123
Dr
I
of our dataset in 0.25 magnitude bins (e.g., M
L
=5.0–5.25 and 5.25–5.5) and utilized
a nonlinear regression procedure to examine the magnitude dependence of Dr
I
. To opti-
mize the estimation of source parameters, we began by using the accelerogram dataset of
rock site recordings for the regression of Eq. (4). Figure 5shows the relationship of
magnitude and Dr
I
(i.e., the increase of Dr
I
with magnitude). To capture the trend of the
magnitude-dependent stress drop, we therefore referred to the study of Kempton and
Stewart (2006) and adopted the exponential model for Dr
I
. The regression model was thus
rewritten as follows:
log s¼log
exp½b1þb2ðMLMÞ
101:5MLþ16:05
1=3
4:9106b
2
6
43
7
5þc1rhyp;ð5Þ
where M* is the magnitude as the Dr
I
exhibits a jump (Fig. 5). The reference magnitude
M* is set to 5.75; b
1
and b
2
are regression coefficients. By applying the nonlinear
regression procedure of Eq. (5) to the dataset of the rock site, we obtained the regression
coefficients of b
1
=1.1538, b
2
=1.3273 and c
1
=-0.0015. The regression coefficients
of b
1
and b
2
, which are stress drop index-related coefficients, were adopted as constants in
Fig. 2 Distribution of strong motion stations with the site classifications of the TSMIP. The colors denote
site classifications as determined by Kuo et al. (2012). The number of stations for site classes A,B,C,Dand
Eis 1, 29, 200, 193 and 16, respectively
1784 Nat Hazards (2015) 75:1779–1793
123
the final regression. The residuals of the decimal logarithm duration between the observed
(s
obs
) and predictive (s
e
) durations (log sobs log se) exhibit a normal distribution with a
standard deviation of r
eq. 5
=0.229 (Fig. 6). Additionally, in Fig. 7a, b, the model
Fig. 3 Earthquake distribution of selected events from 1994 to 2012 (blue dots) for M
L
=5.0–7.3 and depth
\50 km. The red star indicates the location of the 1999 M
L
=7.3 Chi–Chi earthquake. The yellow star and
green triangle indicate the locations of the example earthquake and the CHY006 station, respectively, as
presented in Fig. 1
Table 2 Number of events and
recordings with different magni-
tude intervals
Magnitude (M
L
) Number of events Number of recordings
5.0–5.2 178 1,950
5.2–5.4 113 1,327
5.4–5.6 55 735
5.6–5.2 50 1,071
5.8–6.0 29 615
6.0–6.2 24 1,211
6.2–6.4 9 620
6.4–6.6 11 1,419
6.6–6.8 16 1,565
6.8–7.0 6 769
7.0–7.2 3 60
7.2–7.4 1 297
Total 495 11,639
Nat Hazards (2015) 75:1779–1793 1785
123
residuals of ESD were plotted as functions of magnitude and distance; they display no clear
bias with magnitude or distance.
3. Site effect, s
site
We further considered the site condition dependence of the shaking duration equation.
We used the V
s
30 values of the stations for the empirical duration equation. The form of
the regression model for the site condition dependence, s
site
, is based on the study by
Kempton and Stewart (2006) in which the residual of the logarithm duration linearly
decreases with V
s
30. Accordingly, the form of the duration regression equation that con-
siders the source duration, path and site can be written as follows:
log s¼log
exp½b1þb2ðML5:57Þ
101:5MLþ16:05
1=3
4:9106b
2
6
43
7
5þc1rhyp þc2Vs30 þc3;ð6Þ
where c
1
,c
2
and c
3
are regression coefficients. By applying the nonlinear regression
procedure of Eq. (6) to the entire dataset, we obtained our final regression coefficients:
Fig. 4 Logarithm of the ESD time (blue circles) decays with distance at a 0.2 magnitude interval for
M
L
=6.0–7.3. The red dashed lines indicate the best regression of the data
1786 Nat Hazards (2015) 75:1779–1793
123
Fig. 5 Estimated stress drop index and model for the stress drop index as a function of magnitude for the
ESD data
Fig. 6 Probability density of the residuals of the regression model (Eq. 5, where b
1
=1.1538, b
2
=1.3273
and c
1
=-0.0015) for M
L
[5.0 for the rock site data. The standard deviation is r=0.229
Nat Hazards (2015) 75:1779–1793 1787
123
c
1
=-0.0011, c
2
=-0.0004 and c
3
=0.3038; the previously determined constants were
b
1
=1.1538 and b
2
=1.3273. Figure 8displays the probability density function of the
regression model residuals of the duration in decimal logarithm units; it is shown as a
normal distribution with a standard deviation of r
eq. 6
=0.230. The majority of the
residual values were approximately zero. Additionally, the model residuals of ESD are
plotted as functions of magnitude, distance and V
s
30, as shown in Fig. 9a–c, respectively.
The residuals show no significant trends. The results indicated that the derived coefficients
of the empirical duration Eq. (6) provide an adequate basis for the approximate description
of the strong shaking durations of earthquakes (M
L
[5.0 and depth \50 km) in Taiwan.
Fig. 7 Residuals of the regression model of the ESD in decimal logarithm units plotted as a function of
amagnitude and bdistance. The residual data are recorded by the rock site stations of the TSMIP network
for M
L
[5.0. The black dashed lines indicate the residual value at zero
Fig. 8 Probability density of the
residuals of the regression model
(Eq. 6, where b
1
=1.1538,
b
2
=1.3273, c
1
=-0.0011,
c
2
=-0.0004 and c
3
=0.3038)
for M
L
[5.0. The standard
deviation is r=0.230
1788 Nat Hazards (2015) 75:1779–1793
123
2 Discussion
To compare our derived prediction equation, the study by Kempton and Stewart (2006)is
enlisted. We produced two duration calculations for our dataset: One calculation from our
derived equation that provides the ESD, and a second calculation from the derived
equation of Kempton and Stewart that provides the significant duration (SD). The M
L
had
been converted to M
w
accordingly using the conversion derived by Lin and Lee (2008) for
the equation in Kempton and Stewart (2006). The results are shown in Fig. 10 along with
the magnitude scaling for r
hyp
=100 (km) and V
s
30 =450 (m/s) for the two derived
empirical equations. The SD values of global earthquakes from the NGA dataset are also
shown. Generally, the ESD is approximately 20 s less than the SD. Larger events corre-
spond to smaller differences in the values of ESD and SD. However, the ESD and SD of
the Taiwanese dataset nicely fit the derived empirical equations of the individual definition
of durations. In Kempton and Stewart (2006), the definition of SD considers the energy
contained (5–95 %), but it does not consider the amplitude of the ground motion. The SD
can include time series with small amplitudes (amplitude \0.01 g) for small events.
However, in our ESD, before taking into account the energy radiation, we first consider the
time interval with amplitudes C0.01 g; thus, no time interval for amplitudes \0.01 g is
involved. We further demonstrate the differences in ESD and SD for moderate
(M
L
=5.19; the 1995 earthquake) and large events (M
L
=7.3; the 1999 Chi–Chi earth-
quake) (Fig. 11). The SD of 16.7 s is much larger than the ESD value (3.0 s) for a
moderate earthquake (M
L
=5.19). A long time series with amplitudes \0.01 g was
included in SD. For a larger earthquake (M
L
=7.3), the SD of 29.3 s is more similar to the
ESD of 25.0 s. These comparisons suggest that our ESD may be more conservative in
estimating the shaking duration of earthquakes. However, it could be considered a lower
bound of the shaking duration, especially for moderate earthquakes.
Our derived empirical equations may provide predictions for strong shaking durations.
However, many studies have suggested that various factors may impact strong shaking
Fig. 9 Residuals of the regression model of the ESD in decimal logarithm units plotted as a function of
amagnitude, bdistance and cV
s
30. The residual data are recorded by the TSMIP network for M
L
[5.0.
The black dashed lines provide the residual value at zero
Nat Hazards (2015) 75:1779–1793 1789
123
duration predictions. For instance, earthquake sources, near-field effects and rupture
directivity may impact the predictive equations of strong shaking durations (Kempton and
Stewart 2006). Wen and Yeh (1991) used the SMART1 array data in northeastern Taiwan
to discuss the strong shaking durations of acceleration, velocity and displacement
behaviors. They suggested that the variability in duration is primarily caused by the
complicated rupture process of the earthquake source. A study by Trifunac and Brady
(1975) presented the variability of duration increases with epicentral distance; they sug-
gested that the variability of duration was caused by inhomogeneous media through which
the seismic waves propagated. Additionally, Spudich et al. (1999) proposed that the stress
state (extensional or compressive) and the style of faulting may influence the amplitude of
strong ground motion. Ground motion amplitudes increase, and the threshold of the
acceleration level is therefore exceeded for longer periods of time (Bommer et al. 2009).
Somerville et al. (1997) also found that the rupture directivity effect can influence the
strong shaking duration; they indicated that waves in the backward directivity region result
in signals of extended duration. Additionally, many studies have suggested that structural
components are expected to exhibit sensitivity to ground shaking duration (Reinoso and
Guerrero 2000; Hancock and Bommer 2004; Bommer et al. 2004; Hancock and Bommer
2006). In the present paper, we did not include the aforementioned factors in our empirical
equation. To reduce the variance, additional data obtained for large earthquakes are needed
to further address the possible impact of the various factors (e.g., the style of faulting and
Fig. 10 Comparison of the magnitude dependence of strong shaking duration values from this study with
those from the study of Kempton and Stewart (2006). The yellow squares show the ESD of Taiwanese
earthquakes. The blue and red dots show the SD of Taiwan earthquakes and NGA data, respectively. The
data are for distances of 90–110 km and V
s
30 values of 300–600 m/s. The blue and yellow shadows indicate
the standard deviations of the SD and ESD, respectively. The black lines show the empirical equation of
Eq. 6(where r
hyp
=100 km and V
s
30 =450 m/s) and the equation from Kempton and Stewart (2006)
1790 Nat Hazards (2015) 75:1779–1793
123
fault rupture) on strong shaking durations. However, the fits of ESD and SD to the indi-
vidual empirical equations suggest that the empirical equation in this study, despite being
derived from Taiwanese earthquakes for the most part, could be considered a conservative
approximation of the empirical strong shaking duration equation. This empirical equation
may thus be able to provide a good constraint for assessing the potential hazards of the
shaking duration of future earthquakes.
3 Conclusions
This study provided an empirical equation for the strong shaking duration of earthquakes
as a function of earthquake magnitude, earthquake distance and site parameter (V
s
30). We
proposed a new definition of strong shaking duration (called effective shaking duration or
ESD) by considering amplitude and energy factors (i.e., the presence of major energy and
Fig. 11 Two examples of the duration of strong ground motion are presented. One example is the
earthquake that occurred in 1995 with a magnitude of M
L
=5.19, and the other example is the Chi–Chi
earthquake (M
L
=7.3) in 1999. The green lines provide the time interval of SD (green shadow). The red
lines provide the time interval of the ESD defined in this study (red shadow)
Nat Hazards (2015) 75:1779–1793 1791
123
amplitudes larger than 0.01 g). We analyzed the strong ground motion of acceleration from
the TSMIP network to obtain the empirical equation. The ESD from our definition is
generally smaller than the values of SD of Kempton and Stewart (2006), but fewer dif-
ferences are exhibited in larger events. The good fits of our dataset to the individual
derived empirical equations for ESD and SD suggest that our derived equation from ESD is
a good approximation. Our strong shaking equation could be considered a conservative yet
more effective parameter for shaking durations in the assessment of possible seismic
hazards. However, we have not yet considered other factors from earthquake sources, near-
field effects and rupture directivity, which may also impact the predictive equation for
strong shaking durations. The duration of strong ground motion is critical to estimating
seismic hazards, particularly for building performance, landslide triggers and liquefaction.
On a preliminary basis, our proposed empirical equations could provide the characteristics
of strong shaking durations. Using the massive dataset from the TSMIP, the empirical
equations derived here could also provide a reference for the global community in esti-
mating ground shaking durations in the ground motion prediction of scenario earthquakes.
Acknowledgements The authors are grateful for research support from both the Ministry of Science and
Technology (MOST) and the Institute of Geophysics, National Central University, Taiwan, ROC. Thanks to
Central Weather Bureau Seismological Network for the strong motion data of Taiwan Strong Motion Imple-
mentation Program (TSMIP). This research was supported by the Taiwan Earthquake Research Center (TEC)
funded through MOST, formerly National Science Council (NSC), with Taiwan Earthquake Model (TEM)
project grant number NSC 102-2119-M-006 –010. The TEC contribution number for this article is 00108.
References
Abrahamson NA, Silva WJ (1996) Empirical ground motion models. Report to Brookhaven National
Laboratory, New York
Bolt BA (1973) Duration of strong ground motion. World conference of earthquake engineering, 5th Rome
6-D paper no 292
Bommer JJ, Martinez-Pereira A (1999) The effective duration of earthquake strong motion. J Earthq Eng
3(2):127–172
Bommer JJ, Magenes G, Hancock J, Penazzo P (2004) The influence of strong shaking duration on the
seismic response of masonry structures. Bull Earthq Eng 2(1):1–26
Bommer JJ, Hancock J, Alarco
´n JE (2006) Correlations between duration and number of effective cycles of
earthquake ground motion. Soil Dyn Earthq Eng 26(1):1–13
Bommer JJ, Stafford PJ, Alarco
´n JE (2009) Empirical equations for the prediction of the significant,
bracketed, and uniform duration of earthquake ground motion. Bull Seismol Soc Am 99(6):3217–3233
Boore DM (1983) Stochastic simulation of high-frequency ground motions based on seismological models
of the radiated spectra. Bull Seismol Soc Am 73:1865–1894
Del Gaudio V, Wasowski J (2004) Time probabilistic evaluation of seismically induced landslide hazard in
Irpinia (Southern Italy). Soil Dyn Earthq Eng 24(12):915–928
Hancock J, Bommer JJ (2004) The influence of phase and duration on earthquake damage in degrading
structures. In: Proceedings, 13th world conference on earthquake engineering, Vancouver, BC, Can-
ada, Paper 1990
Hancock J, Bommer JJ (2005) The effective number of cycles of earthquake ground motion. Earthq Eng
Struct Dyn 34:637–664
Hancock J, Bommer JJ (2006) A state-of-knowledge review of the influence of strong-motion duration on
structural damage. Earthq Spectra 22(3):827–845
Hanks TC, Kanamori H (1979) A moment magnitude scale. J Geophys Res 84:2348–2350
Hanks TC, McGuire RK (1981) The character of high frequency strong ground motion. Bull Seismol Soc
Am 71:2071–2095
Hernandez B, Cotton F (2000) Empirical determination of the ground shaking duration due to an earthquake
using strong motion accelerograms for engineering applications. In: Proceedings, 12th world confer-
ence on earthquake engineering, p 2254/4.
1792 Nat Hazards (2015) 75:1779–1793
123
Kawashima K, Aizawa K (1989) Bracketed and normalized durations of earthquake ground acceleration.
Eng Struct Dyn 18(7):1041–1051
Kempton JJ, Stewart JP (2006) Prediction equations for significant duration of earthquakes ground motions
considering site and near source effects. Earthq Spectra 22(4):985–1013
Kuo CH, Wen KL, Hsieh HH, Lin CM, Chang TM, Kuo KW (2012) Site Classification and Vs30 estimation
of free-field TSMIP stations using the logging data of EGDT. Eng Geol 129–130:68–75
Lee WHK, Bennett RE, Meagher KL (1972) A method of estimating magnitude of local earthquakes from
signal duration. USGS open file report, pp 1–28
Lin PS, Lee CT (2008) Ground-motion attenuation relationships for subduction-zone earthquakes in
northeastern Taiwan. Bull Seismol Soc Am 98(1):220–240
Martin WMJ, Haresh CS (1979) Determining strong-motion duration of earthquakes. Bull Seismol Soc Am
69(4):1253–1265
Pagratis D (1995) Prediction of earthquake strong ground-motion for engineering use. MSc Dissertation,
Imperial College London
Rathje EM, Saygili G (2009) Probabilistic assessment of earthquake-induced sliding displacements of
natural slopes. Bull N Z Soc Earthq Eng 42(1):18
Rauch AF, Martin JR (2000) EPOLLS model for predicting average displacements on lateral spreads.
J Geotech Eng 126:360–371
Reinoso EMO, Guerrero R (2000) Influence of strong ground-motion duration in seismic design of struc-
tures. In: Proceedings, 12th world conference on earthquake engineering, p 1151
Seber GAF, Wild CJ (2003) Nonlinear regression. Wiley-Interscience, Hoboken
Shin TC (1993) Progress summary of the Taiwan strong-motion instrumentation program. In: ‘‘Symposium
on the Taiwan strong-motion program’’. Central Weather Bureau, pp 1–10
Shoji Y, Tanii K, Kamiyama M (2005) A study on the duration and amplitude characteristics of earthquake
ground motions. Soil Dyn Earthq Eng 25(7–10):505–512
Sigaran-Loria C, Kaynia AM, Hack R (2007) Soil stability under earthquakes: a sensitivity analysis. In:
Proceedings of the 4th international conference on earthquake geotechnical engineering, Thessaloniki,
Greece
Somerville PG, Smith NF, Graves RW, Abrahamson NA (1997) Modification of empirical strong ground
motion attenuation relations to include the amplitude and duration effects of rupture directivity.
Seismol Res Lett 68(1):199–222
Spudich P, Joyner WB, Lindh AG, Boore DM, Margaris BM, Fletcher JB (1999) SEA99: a revised ground
motion prediction relation for use in extensional tectonic regimes. Bull Seismol Soc Am
89(5):1156–1170
Stafford PJ (2008) Conditional prediction of absolute durations. Bull Seismol Soc Am 98(3):1588–1594
Trifunac MD, Brady AG (1975) A study of the duration of strong earthquake ground motion. Bull Seismol
Soc Am 65(3):581–626
Trifunac MD, Novikova EI (1995) Duration of earthquake fault motion in California. Earthq Eng Struct Dyn
24(6):781–799
Wen KL, Yeh YT (1991) Characteristics of strong motion durations in the SMART1 array area. Terr Atmos
Ocean 2(3):187–201
Nat Hazards (2015) 75:1779–1793 1793
123