See discussions, stats, and author profiles for this publication at: [Link]

net/publication/338735883

A Damping Modification Factor Prediction Model for Horizontal Displacement

Spectrum from Subduction Slab Earthquakes in Japan Accounting for Site
Conditions

Article in Bulletin of the Seismological Society of America · January 2020

DOI: 10.1785/0120190156

CITATIONS READS

8 145

2 authors:

J. Zhou John Zhao

Southwest Jiaotong University 82 PUBLICATIONS 2,229 CITATIONS
8 PUBLICATIONS 149 CITATIONS
SEE PROFILE
SEE PROFILE

All content following this page was uploaded by J. Zhou on 17 December 2020.

The user has requested enhancement of the downloaded file.

 A Damping Modification Factor Prediction Model
for Horizontal Displacement Spectrum from
Subduction Slab Earthquakes in Japan
Accounting for Site Conditions
Jun Zhou1 and John X. Zhao*1,2

ABSTRACT
A damping modification factor (DMF) model without source and path parameters is pre-
sented in this article for subduction slab earthquakes in Japan, using a similar dataset in
the Zhao, Jiang, et al. (2016) study. Site classes based on site periods were used as the site-
effect proxy. DMF models were derived from spectra of 13 damping ratios and 34 spectral
periods, and the DMF can be calculated for any damping ratio between 1% and 30% and at
any spectral period between 0.03 and 5.0 s. A simple fourth-order polynomial for the log-
arithm of the spectral periods and a simple quadratic function of the logarithm of damping
ratios were used to model the effects of spectral periods and damping ratios, respectively.
The model satisfies boundary conditions that require the DMF values equal to 1.0 at very
short spectral periods; at long spectral periods, the DMFs for different damping ratios
appear to converge to 1.0 to satisfy the constant displacement spectrum at long periods.
Model standard deviations are smaller than those for the ground-motion prediction equa-
tions. All standard deviations vary linearly with the increasing logarithm of damping
ratios. The DMFs presented in this study combined with the spectrum from the Zhao,
Jiang, et al. (2016) study produce smoothed displacement spectrum that may be used
for engineering designs. In a spectral period range of 0.2–3.0 s, the DMF values from this
study are close to those by Daneshvar et al. (2016), but, at short periods, the difference is
significant. The residual distribution suggests that DMFs also depend on earthquake
source and path parameters. The model presented in this article does not include the effect
of source and path variables so that this model can be used to scale a 5% damped spectrum
without a known magnitude and a source distance.

Chiou and Youngs (2014), and Idriss (2014). These GMPEs

KEY POINTS were derived from the strong-motion records from California
• We develop a damping modification factor model for and supplemented by the data from shallow crustal earth-
displacement spectra from subduction zone events. quakes in Japan, Taiwan, and China. The GMPEs for subduc-
• The damping modification factor can be predicted by a tion zones based on the strong-motion records from Japan
simple model with no source and path parameters. have been developed by Zhao, Zhou, et al. (2016) for shallow
• The model can be used to calculate horizontal displace- crustal and upper-mantle earthquakes, Zhao, Liang, et al.
ment spectrum with damping ratios other than 5%. (2016) for subduction interface earthquakes, and Zhao,

1. Department of Geotechnical Engineering, School of Civil Engineering, Southwest

INTRODUCTION Jiaotong University, Chengdu, Sichuan, China; 2. Key Laboratory of Building
Structural Retrofitting and Underground Space Engineering, Shandong Jianzhu
Design spectrum from many seismic design codes and the University, Ministry of Education, Jinan, Shandong Province
spectrum from modern ground-motion prediction equations *Corresponding author: johnzhao1000@[Link]
(GMPEs) are usually based on a damping ratio of 5%, for Cite this article as Zhou, J., and J. X. Zhao (2020). A Damping Modification Factor
example, the GMPEs in the Next Generation Attenuation- Prediction Model for Horizontal Displacement Spectrum from Subduction Slab
Earthquakes in Japan Accounting for Site Conditions, Bull. Seismol. Soc. Am. 110,
West2 (NGA-West2) dataset project by Abrahamson et al. 647–665, doi: 10.1785/0120190156
(2014), Boore et al. (2014), Campbell and Bozorgnia (2014), © Seismological Society of America

Volume 110 Number 2 April 2020 [Link] Bulletin of the Seismological Society of America • 647

Downloaded from [Link]

by Chengdu University of Technology, JohnZhao1000
 Jiang, et al. (2016) for subduction slab earthquakes, also for the average response spectral ratio and the associated standard
a damping ratio of 5%. For many structures, the inherent deviations. Using 648 strong-motion records from the western
damping ratio differs from 5% such as seismically isolated United States (primarily from California) and 620 records from
structures or structures with additional damping devices. In the central United States, Cameron and Green (2007) developed
the design for this type of structures, a damping modification a set of DMF models, and the strong-motion records were
factor (DMF) is usually used to scale the 5% damped spectrum divided into a number of groups according to magnitude, source
to obtain the design spectrum with an appropriate damping distance, and site conditions for a series of spectral periods; the
ratio. Three types of design spectra, the relative displacement unsmoothed geometric average of DMFs and standard devia-
(SD ), relative velocity (SV ), and acceleration spectra (SA ) are tions are presented as the model parameters. Cameron and
often used. An acceleration spectrum can be the total acceler- Green (2007) show that tectonic settings, magnitude, and site
ation or pseudoacceleration spectrum (SPA ). SPA can be related conditions have a strong effect on the DMFs, but the source dis-
to SD by tance has relatively little effect. Rezaeian et al. (2014) developed a

2 DMF model based on NGA-West2 dataset, and this model
2π
SPA
T; ζ  SD
T; ζ;
1 accounts for earthquake magnitude and source distance. Based
T
EQ-TARGET;temp:intralink-;df1;53;588

on the ground motions from a stochastic modeling procedure,

in which T is the spectral period, that is, the period of the Palermo et al. (2016) developed analytical solutions for a DMF
single-degree-of-freedom oscillator and ζ is the damping ratio. model. Using 1098 records from shallow crustal earthquakes,
When ζ is small and the spectral period is relatively short, 622 subduction interface records, and 582 subduction slab
SA ≈ SPA and the difference is only for a few percents. However, records, Daneshvar et al. (2016, 2017) developed two relatively
these two types of spectra are not equal when the damping simple DMF models that are continuous functions of damping
ratio is large and the spectral period is long. For a seismically ratio and spectral period. All records are from events with a
isolated structure, SPA is the required horizontal acceleration magnitude of 6.0 or larger, and together with some other data
that will produce a relative displacement of SD , whereas SA is selection criteria, the dataset in that study is not large. So far, few
the total acceleration applied to the seismically isolated build- other models for subduction earthquakes have been developed.
ing and the contents stored in the building. The difference in DMF prediction equations can be divided into three groups:
the forces associated with these two types of accelerations is
resisted by damping devices. This means that different predic- 1. spectral-period-independent DMF models, including
tion models are required for displacement (or pseudoaccelera- Newmark and Hall (1982), Kawashima and Aizawa (1986),
tion) and total acceleration spectrum. Ashour (1987), Tolis and Faccioli (1999), Bommer et al.
Displacement DMF, hereafter, referred to as BD , is defined by (2000), Priestley (2003), Zhou et al. (2003), and Stafford
et al. (2008) developed spectral-period-independent DMF
SD
T; ζ models. Most seismic design codes, including code for
BD  :
2
SD
T; ζ  5% seismic design of buildings (GB50011-2010; Ministry of
EQ-TARGET;temp:intralink-;df2;53;353

Housing and Urban-Rural Development of the People’s

Many DMF models have been published in the literature. In Republic of China, 2010) for China and Eurocode 8 (2004)
most early studies, spectral-period-independent DMF models use this type of DMFs;
were developed, including the models by Newmark and Hall 2. spectral-period-dependent DMF models, including those
(1982) for acceleration, velocity, and displacement spectra based by Wu and Hanson (1989), Lin and Chang (2003, 2004),
on a small number of records from shallow crustal earthquakes. Atkinson and Pierre (2004), Cameron and Green (2007),
Kawashima and Aizawa (1986), Ashour (1987), Tolis and Hubbard and Mavroeidis (2011), Castillo and Ruiz (2014),
Faccioli (1999), Zhou et al. (2003), and Bommer and Mendis Mollaioli et al. (2014), Daneshvar et al. (2016, 2017), and
(2005) developed DMF models based on strong-motion records Pu et al. (2016). Lin and Chang (2003), Atkinson and Pierre
from California and Japan. Using the records from the subduc- (2004), and Cameron and Green (2007) presented their
tion earthquakes in Mexico, Castillo and Ruiz (2014) established models using average values and standard deviation for
a DMF model for acceleration spectrum. Nagao and Kanda each site class (SC) and period group. Site effects were
(2015) proposed a DMF model based on strong-motion records included in some of the models in this group, for example,
from Japan. Hubbard and Mavroeidis (2011), Mollaioli et al. Lin and Chang (2004); and
(2014), and Pu et al. (2016) developed DMF models for near- 3. DMF models accounting for the effects of earthquake
source records with strong velocity impulses. A DMF model source, path, and site conditions are presented by the
from Stafford et al. (2008) accounts for the effect of duration Cameron and Green (2007) and Rezaeian et al. (2014)
and the number of cycles. Using simulated strong-motion models, though many researchers investigated the effect
records, Atkinson and Pierre (2004) developed a DMF model of magnitude and source distance, including Bommer
for eastern North America, and their results are presented in and Mendis (2005) and Hao et al. (2011).

648 • Bulletin of the Seismological Society of America [Link] Volume 110 Number 2 April 2020

Downloaded from [Link]

by Chengdu University of Technology, JohnZhao1000
 TABLE 1
Site Class (SC) Definitions Used in the Present Study, the Number of Records in Each SC, and the Approximately Corresponding
National Earthquake Hazards Reduction Program (NEHRP) SCs (Building Seismic Safety Council [BSSC], 2003)

SC Number of Records Description Natural Period V 30 Calculated from Site Period NEHRP SCs

SC I 2022 Rock T < 0:2 s V 30 > 600 A+B+C

SC II 1353 Hard soil 0:2 ≤ T < 0:4 s 300 < V 30 ≤ 600 C
SC III 442 Medium soil 0:4 ≤ T < 0:6 s 200 < V 30 ≤ 300 D
SC IV 878 Soft soil T ≥ 0:6 s V 30 ≤ 200 E+F

The third group of DMF models would have the best predic- For an engineering application, the design spectrum should
tions, but there is a problem for practical applications. The code have a smoothed variation with respect to the spectral period. A
design spectrum is obtained by seismic hazard analyses, and the smoothed spectrum means that the spectrum does not have any
associated earthquake magnitude and distance may not be avail- sharp deviation from the average value within a narrow period
able. Even if a scenario earthquake can be established by a deag- band. Because DMFs contain the effect of earthquake magni-
gregation analysis, the magnitude and the source distance of the tude, source depth, and source distance, the highly damped
scenario earthquake may differ at different spectral periods. In spectrum obtained from a smoothed 5% damped spectrum
these cases, a model without the earthquake source and distance scaled by a smoothed DMF model without these parameters
(the first two groups) will be required. For a seismic hazard is unlikely to be smoothed at all spectral periods. We will dem-
analysis, however, the third group of models would be preferred onstrate these effects in the proposed simple DMF model.
when the source and path parameters are available. There have been few models developed for subduction slab
This article will present a DMF model in the second group. earthquakes. Castillo and Ruiz (2014) presented a DMF model
In a subduction zone like Japan, the tectonic and geological for subduction earthquakes based on the design spectrum with
settings are complex. According to the earthquake location, a spectrum plateau at two fixed spectral periods. For subduc-
focal mechanism, and the subduction-interface geometry, tion earthquake models, Daneshvar et al. (2016) used the
earthquakes in Japan can be divided into four earthquake cat- following equation for the displacement DMF model:
egories, the shallow crustal, upper-mantle, subduction inter-
face, and subduction slab earthquakes by Zhao et al. (2015). EQ-TARGET;temp:intralink-;df3;308;406 BD  1 − f1  a1 − ln
ζa2 g
a3  Ta4 exp
a5 T a6 :
3
Zhao, Jiang, et al. (2016), Zhao, Liang, et al. (2016), and
Zhao, Zhou, et al. (2016) presented three GMPEs, one for shal- Symbols a1 –a6 are the regression coefficients. Equation (3) is
low crustal and upper-mantle events, one for subduction inter- very simple and compact, but the spectral periods were divided
face events, and one for subduction slab events, respectively. into two ranges: 0:05 s ≤ T < 1 s and 1 s < T ≤ 3 s, about 12
Zhao et al. (2019) presented a comparison of DMF ratios for model coefficients for the median model for each SC.
acceleration and displacement spectra, and they found that:
STRONG-MOTION DATASET
1. the effect of earthquake categories is significant at nearly all This study used 4695 strong-motion records from 136 subduc-
spectral periods; tion slab earthquakes recorded by the K-NET and KiK-net in
2. at short periods up to 0.3 s, the effects of earthquake cat- Japan. All recording stations have a measured shear-wave
egory and site conditions for acceleration and displacement velocity profile down to engineering bedrock, allowing for
are reasonably similar; accurate site classification. The earthquakes were classified
3. at spectral periods over 0.3, the effects of earthquake types by Zhao et al. (2015) using the geometry of the subducting slab
and site conditions for acceleration spectrum are much interface from Hayes et al. (2012; the Slab 1.0). SCs were used
larger than those for the displacement spectrum; in this study as the site-effect proxy, and the definition for SCs
4. the effect of site conditions in each type of earthquakes is is presented in Table 1.
significant at many spectral periods; and Table 1 presents the number of records in each SC: 2022
5. the effects of magnitude, source depth, and source distance from SC I sites, 1353 from SC II sites, 442 from SC III sites,
should be accounted for in a DMF prediction model if these and 878 from SC IV sites. Figure 1a shows the distribution of
parameters are available for the 5% damped design spectrum. earthquakes with respect to focal depth and moment magni-
tude. The largest depth is 167 km, the smallest magnitude is
These results suggest that a separate DMF model should be 4.99, and the largest magnitude is 8.25. The distribution of
developed for each group of earthquakes, and we present a events with respect to depth and magnitude is not strongly
model for the subduction slab events using a dataset similar correlated. Figure 1b shows the distribution of records with
to that used by Zhao, Jiang, et al. (2016). respect to source distance and moment magnitude, and a

Volume 110 Number 2 April 2020 [Link] Bulletin of the Seismological Society of America • 649

Downloaded from [Link]

by Chengdu University of Technology, JohnZhao1000
 Figure 1. (a) The distribution of earthquakes used in the present study with Figure 2. A comparison of mean damping modification factors (DMFs)
respect to moment magnitude and fault depth and (b) the distribution of (a) with 1% damping ratio and (b) with 30% damping ratio for four
strong-motion records with respect to magnitude and source distance. The site classes (SCs). The color version of this figure is available only in the
color version of this figure is available only in the electronic edition. electronic edition.

magnitude-dependent cutoff distance was used to exclude the in which ζ m is the damping ratio in percentage, and subscript
records that have a distance larger than the cutoff distance, to m denotes the mth damping ratio. We found that the following
minimize the effects from untriggered stations as discussed by function can be used to model the effect of damping ratios:
Zhao, Zhou, et al. (2016). The shortest distance is 27.4 km and
the largest distance is 300 km. EQ-TARGET;temp:intralink-;df5;320;367 lnBk
T n ; ζ m   d k;n θm  ek;n θ2m ;
5

THE PROPOSED DMF MODELS in which B denotes the average DMF, subscript k is the SC
As described earlier, a model without earthquake source and number (I, II, III, and IV), and n is the period number. The
path terms have to be used to scale a 5% damped design spec- ordinary least-squares method was used to derive the coeffi-
trum that cannot be associated with a given earthquake. cients in equation (5) at each spectral period for each SC.
Figure 2a shows the average DMFs for a damping ratio of Next, the residuals can be decomposed into within-event
1%, and Figure 2b shows the average DMFs for a damping ratio and between-event components
of 30% for four SCs. The differences among different SCs are
reasonably large, especially at short periods between SC I sites EQ-TARGET;temp:intralink-;df6a;320;250 ξ Ti;j;k
T n ; ζ m   lnBi;j;k
T n ; ζ m  − lnBk
T n ; ζ m ;
6a
and the other SCs. The variations of DMFs with spectral periods
are reasonably smooth. The average standard deviation is about
0.25 in the natural logarithm scale for a damping ratio of 30%, EQ-TARGET;temp:intralink-;df6b;320;212 ξ Ti;j;k
T n   ξ i;j;k
T n ; ζ m   ηi
T n ; ζ m ;
6b
and the mean values 1 standard deviation equal the mean
value multiplied or divided by a factor of 1.28. Although the in which B denotes the DMF for displacement spectrum,
differences among four SCs are not statistically significant at superscript T denotes total, subscript i is for the ith earthquake,
some spectral periods, to take the same values for the model and j means the jth record from the ith earthquake. The ran-
parameters among different SCs leads to the difficulty to pro- dom variable ξ i;j;k is the within-event residual with a zero mean
duce a predicted spectrum that varies smoothly with spectral and a standard deviation σ that represents the random error
periods for all SCs. from path and site effects; the random error ηi is the between-
In this study, a damping ratio variable is defined by event residual having a zero mean and a standard deviation of τ,
representing the error mainly associated with earthquake source
EQ-TARGET;temp:intralink-;df4;53;81 θm  ln
ζ m  − ln
5;
4 effect. The total number of damping ratios is 13, as shown in

650 • Bulletin of the Seismological Society of America [Link] Volume 110 Number 2 April 2020

Downloaded from [Link]

by Chengdu University of Technology, JohnZhao1000
 TABLE 2
Selected Damping Ratios

ζ Number ζ ζ Number ζ ζ Number ζ

1 1% 6 7% 11 20%
2 2% 7 8% 12 25%
3 3% 8 9% 13 30%
4 4% 9 10%
5 6% 10 15%

TABLE 3
Selected Spectral Periods

T Number T (s) T Number T (s) T Number T (s) T Number T (s)

1 0.01 10 0.1 19 0.35 28 1.25

2 0.02 11 0.12 20 0.4 29 1.5
3 0.03 12 0.14 21 0.45 30 2
4 0.04 13 0.15 22 0.5 31 2.5
5 0.05 14 0.16 23 0.6 32 3
6 0.06 15 0.18 24 0.7 33 3.5
7 0.07 16 0.2 25 0.8 34 4
8 0.08 17 0.25 26 0.9 35 4.5
9 0.09 18 0.3 27 1 36 5

Table 2, and the total number of spectral periods is 34, as shown When T < 0:06 s, equation (5) can be simplified as
in Table 3, because the first two periods have a DMF value of
close to 1.0. The random variable ξ i;j;k can be decomposed into EQ-TARGET;temp:intralink-;df10a;308;406 dk;n  ak;5 ;
10a
within-site and between-site components. The within-site com-
ponent is mainly associated with path effect, and the between-
EQ-TARGET;temp:intralink-;df10b;308;374 ek;n  bk;5 :
10b
site component is mainly associated with site effect. The total
and within-event standard deviations can be calculated by At any spectral period between 0.03 and 0.06 s, an interpola-
tion based on the logarithm of the spectral period can be used
p
EQ-TARGET;temp:intralink-;df7a;41;328 σT  σ2  τ2;
7a to calculate the coefficients in equations (10a) and 10b).
However, it is not possible to derive all the model parameters
from the least-squares formulation in equations (5) and (9a)
q
EQ-TARGET;temp:intralink-;df7b;41;290 σ σ 2S  τ 2S :
7b and 9b) simultaneously because of error propagation among
different damping ratios. The following iteration method
Symbol τ S denotes the between-site standard deviations, and was used to derive the model parameters:
σ S denotes the within-site standard deviations. All standard
deviations are on the natural logarithm scale. 1. for each spectral period and each SC, the coefficients in
Next, we used the logarithm of the spectral period as a equation (5) were derived using a fixed-effects regression
model variable defined by method and were smoothed with respect to spectral periods;
2. the coefficients in equation (5)) ak;1 –ak;5 were also derived
EQ-TARGET;temp:intralink-;df8;41;185 γn  ln
T n :
8 by a fixed-effects model; and
3. check model prediction, residual distributions, and the shape
When T ≥ 0:06 s the coefficients in equation (5) can be calcu- of displacement and pseudoacceleration spectrum; carry out
lated by model parameter adjustment (smoothing) if necessary.
Iterations were carried out if the model fit was not
EQ-TARGET;temp:intralink-;df9a;41;120 d k;n  ak;1 γn  ak;2 γ2n  ak;3 γ3n  ak;4 γ4n  ak;5 ;
9a satisfactory.

During the regression analysis, statistical tests were per-

EQ-TARGET;temp:intralink-;df9b;41;81 ek;n  bk;1 γn  bk;2 γ2n  bk;3 γ3n  bk;4 γ4n  bk;5 :
9b formed for each term. A simple and straightforward test is

Volume 110 Number 2 April 2020 [Link] Bulletin of the Seismological Society of America • 651

Downloaded from [Link]

by Chengdu University of Technology, JohnZhao1000
 equations (9a and 9b) are necessary, and, some can be set
as zero, as shown in Table 5. Figure 4 compares the DMF val-
ues from the data with those from equation (5) for SC I and SC
II sites, and Figure 5 presents those for SC III and SC IV sites.
The smoothed curves fit the data reasonably well across all the
spectral periods.
The physical definition of the displacement response spec-
trum means that, at some long periods (or strictly T  ∞), the
displacement spectrum equals to the maximum displacement
of the strong-motion record, and, therefore, the displacement
spectrum should be the same for all damping ratios, leading to
that the DMF values for all damping ratios and SCs should
equal 1.0. However, it is not possible to use our dataset in this
study to derive the periods for a constant displacement because
the maximum usual period was set as 5.0 s at all records. In our
dataset, there are records with a maximum usable period over
5.0 s, but the number of these records may not be large enough.
More records from Japan since the end of 2012 may be used so
that a model for the starting period for a constant displacement
spectrum can be derived and used as a restraint for the DMF
model. For the model presented in this article, the maximum
Figure 3. Variation of the model coefficients in equation (5) for
spectral period is 5.0 s, and therefore to extend the spectral
(a) θ  ln
ζ − ln
5 and (b) θ2 with spectral period for four SCs. The
color version of this figure is available only in the electronic edition. periods at which constant displacement spectrum is applicable
is a very large extrapolation. To accommodate this physical
problem, we set up a condition for extrapolation. At long peri-
to compute the ratios of the maiden value for a parameter ym ods, for damping ratios less than 5%, DMF  1:0 should be
over the corresponding standard deviation σ y for each model selected when DMF < 1:0; for damping ratios more than 5%,
parameter. If ym =σ y ratio for a particular term is larger than DMF  1:0 should be selected when DMF > 1:0. These condi-
1.96, this term is considered to be significant statistically. tions may be used when no other suitable models are available.
The value for 1.96 is from a standard t-test when the size The extrapolation for the model in Figures 4 and 5 suggests that
of the sample is large. In each step, the variable with the small- the DMFs converge to 1.0 at a spectral period over 8 s.
est ym =σ y ratio will be checked and will be set as zero. After a The random errors for a displacement spectrum and the
parameter is set as 0.0, a regression is then performed again corresponding DMF model can be written as
until all variables have ym =σ y > 1:96.
Figure 3 shows the variations of model coefficients d k;n , and EQ-TARGET;temp:intralink-;df11a;320;315 lnSD
T; ζ  lnSD
T; ζ  ξ;
11a
ek;n with spectral periods, and Table 4 presents the model coef-
ficient at spectral period 0.03, 0.04, and 0.05 s. At spectral peri-
lnSD
T; 5%  lnSD
T; 5  η;
11b
ods of 0.06 s or larger, a function with a fourth order of γ was
EQ-TARGET;temp:intralink-;df11b;320;272

derived, and the solid line was calculated from these coeffi-
cients presented in Table 5. Not all model parameters in EQ-TARGET;temp:intralink-;df11c;320;244 lnBD
T  lnSD
T; ζ − lnSD
T; 5  ξ − η;
11c

TABLE 4
Coefficients in Percentage for the First Three Periods

Site Class T (s) 0.03 0.04 0.05

SC I a1;5 −3.023 −12.509 −19.595

b1;5 −0.545 0.735 0.455
SC II a2;5 −1.200 −6.030 −10.850
b2;5 −0.231 0.684 0.761
SC III a3;5 −1.130 −5.080 −9.560
b3;5 −0.198 0.305 0.862
SC IV a4;5 −1.070 −4.590 −8.240
b4;5 −0.139 0.196 0.506

652 • Bulletin of the Seismological Society of America [Link] Volume 110 Number 2 April 2020

Downloaded from [Link]

by Chengdu University of Technology, JohnZhao1000
 Figure 4. Comparison of the average DMFs calculated from strong-motion Figure 5. Comparison of the average DMFs calculated from strong-motion
records with those predicted by the model developed in this study for (a) SC I records with those predicted by the model developed in this study for (a) SC III
and (b) SC II sites for damping ratios of 1%, 2%, 3%, 8%, 15%, and 30%. and (b) SC IV sites for damping ratios of 1%, 2%, 3%, 8%, 15%, and 30%.
The color version of this figure is available only in the electronic edition. The color version of this figure is available only in the electronic edition.

in which SD denotes the average displacement spectrum, ξ ζ and η are in a 0.6–0.8 range in the natural logarithm scale (from
denotes the random error associated with the displacement spec- GMPEs), and the standard deviation for ζ − η is generally less
trum with a damping ratio of ζ, η is a random error associated than 0.3 for the model presented in this article at most spectral
with the 5% damped displacement spectrum, and ζ − η is the periods. These are not the formal description in terms of statistics.
resultant random error for ln
BD . The standard deviation of Figures 6–9 show the total, between-event, between-site,
η is from the GMPE for a 5% damped spectrum. When ζ and within-site standard deviations for four SCs and damping
and η are fully correlated, ζ − η has a zero mean and a zero stan- ratios of 1% and 30%. All standard deviations are in the natural
dard deviation; when ζ and η are independent, ζ − η has a zero logarithm scale. The standard deviations for a damping ratio of
mean and a compounded standard deviations. When ζ and η are 30% are usually larger than those for a damping ratio of 1% for
partially correlated, the standard deviation for ζ − η would be all SCs and at most spectral periods; the between-event stan-
smaller than the larger one of those for ζ and η. Because dard deviations are much smaller than the within-site standard
SD (T, ζ) and SD (T, 5%) is from the same record, the random deviations at most spectral periods over 0.1 s, suggesting that
errors ζ and η are partially correlated, leading to reduced stan- the earthquake source effect is better modeled than the path
dard deviations for the DMF models. The standard deviations for effect. At short periods up to about 0.2 s, the between-site

TABLE 5
Model Coefficients in Percentage

Site Class γ
p  1 γ2
p  2 γ3
p  3 γ4
p  4 Constant
p  5

SC I a1;p 6.42109 −0.05038 0.04830 0.36625 −28.50586

b1;p 0.75618 0.09181 −0.19901 0.0 −3.50469
SC II a2;p 7.73878 0.86129 −0.70365 0.18966 −28.10100
b2;p 0.70997 0.19563 −0.17218 0.0 −3.28738
SC III a3;p 8.30859 2.38370 −0.95683 0.0 −31.51400
b3;p 1.03193 0.74997 −0.37665 −0.11132 −3.82912
SC IV a4;p 7.24354 2.57147 −0.90603 0.0 −33.73300
b4;p 0.80075 0.49227 −0.32737 −0.08180 −3.65150

Volume 110 Number 2 April 2020 [Link] Bulletin of the Seismological Society of America • 653

Downloaded from [Link]

by Chengdu University of Technology, JohnZhao1000
 Figure 6. Variations of model standard deviations with spectral periods for SC Figure 7. Variations of model standard deviations with spectral periods for SC
I sites, (a) for a damping ratio of 1%, and (b) for a damping ratio of 30%. II sites, (a) for a damping ratio of 1% and (b) for a damping ratio of 30%.
Note that these standard deviations are based on the natural logarithm Note that these standard deviations are based on the natural logarithm
scale. The color version of this figure is available only in the electronic scale. The color version of this figure is available only in the electronic
edition. edition.

Figure 8. Variations of model standard deviations with spectral periods for SC Figure 9. Variations of model standard deviations with spectral periods for SC
III sites, (a) for a damping ratio of 1% and (b) for a damping ratio of 30%. IV sites, (a) for a damping ratio of 1% and (b) for a damping ratio of 30%.
The color version of this figure is available only in the electronic edition. The color version of this figure is available only in the electronic edition.

654 • Bulletin of the Seismological Society of America [Link] Volume 110 Number 2 April 2020

Downloaded from [Link]

by Chengdu University of Technology, JohnZhao1000
 Figure 10. Variations of total standard deviations with spectral periods for four Figure 11. Variations of within- and between-site, between-event, and total
SCs, (a) for a damping ratio of 1% and (b) for a damping ratio of 30%. Note standard deviations with damping ratios for SC I sites at a spectral period of
that these standard deviations are based on the natural logarithm scale. The (a) 0.16 s and (b) 2.5 s. Note that these standard deviations are based on
color version of this figure is available only in the electronic edition. the natural logarithm scale. The color version of this figure is available only
in the electronic edition.

standard deviations are either similar to or larger than the

within-site standard deviations. At spectral periods over 0.5 s, ratio is more than 5%, as shown in Figure 11. The standard
the between-site standard deviations are close to the between- deviations for other spectral periods and SCs also vary linearly
event standard deviations for SC I and II sites. The between- with the logarithm of damping ratios. The following models
site standard deviations are less than the within-site but larger are proposed to predict the standard deviations for each spec-
than the between-event standard deviations at many spectral tral period. The within-site standard deviation σ S and between-
periods over about 0.2 s, suggesting that the site effect is better site standard deviation τ S can be calculated from
modeled than the path effect at those periods. The largest 
g 1k ln
ζ  h1k if ζ < 5
total standard deviations are at about 0.1 s spectral periods σ Sζ;k  ;
12a
r 1k ln
ζ  s1k if ζ > 5
EQ-TARGET;temp:intralink-;df12a;308;315

and vary between 0.2 and 0.32, considerably less than those
of the GMPEs by Zhao, Jiang, et al. (2016) for subduction slab
events, suggesting that the DMF defined in equation (2) can- 
g 2k ln
ζ  h2k if ζ < 5
cels the correlated random errors between the spectrum for a τ Sζ;k  ;
12b
r 2k ln
ζ  s2k if ζ > 5
EQ-TARGET;temp:intralink-;df12b;308;273

given damping ratio and that for a damping ratio of 5%.

Figure 10 compares the total standard deviations from four in which subscript k denotes SC number, 1 for SC I, 2 for SC II,
SCs. For both damping ratios and at very short periods 3 for SC III, and 4 for SC IV sites. The between-event standard
up to 0.1 s, the standard deviations increase with increasing deviation τ and the total standard deviation σ T at each spectral
spectral periods but decrease with increasing SC number or period can be calculated by
with increasing site periods. SC I sites appear to have the largest 
g 3 ln
ζ  h3 if ζ < 5
total standard deviations, and SC IV sites appear to have the τζ  ;
13a
r3 ln
ζ  s3 if ζ > 5
EQ-TARGET;temp:intralink-;df13a;308;171

smallest ones in this short-period range. At spectral periods

over 0.1 s, the total standard deviations among the four SCs
are quite similar. 
g 4k ln
ζ  h4k if ζ < 5
We found that the standard deviations decrease linearly σ T;ζ;k  :
13b
r4k ln
ζ  s4k if ζ > 5
EQ-TARGET;temp:intralink-;df13b;308;129

with the increasing logarithm of the damping ratios when

the damping ratio is within 5%, and increase linearly with In equations (12a,b) and (13a,b), the condition that the stan-
the increasing logarithm of damping ratios when the damping dard deviations should be zero when ζ  5% was not enforced

Volume 110 Number 2 April 2020 [Link] Bulletin of the Seismological Society of America • 655

Downloaded from [Link]

by Chengdu University of Technology, JohnZhao1000
 TABLE 6
Model Parameters for Within-Site, Between-Site, and Total Standard Deviations for SC I Sites

T (s) g11 h11 r 11 s11 g21 h21 r 21 s21 g41 h41 r 41 s41

0.03 −0.004 0.008 0.018 −0.032 −0.002 0.004 0.012 −0.020 −0.012 0.022 0.025 −0.042
0.04 −0.062 0.104 0.051 −0.076 −0.071 0.114 0.065 −0.104 −0.105 0.170 0.090 −0.139
0.05 −0.074 0.125 0.065 −0.096 −0.084 0.137 0.090 −0.146 −0.124 0.206 0.123 −0.192
0.06 −0.075 0.128 0.077 −0.118 −0.082 0.137 0.107 −0.174 −0.127 0.213 0.152 −0.241
0.07 −0.079 0.135 0.085 −0.131 −0.081 0.138 0.122 −0.201 −0.124 0.212 0.171 −0.277
0.08 −0.079 0.135 0.089 −0.137 −0.075 0.129 0.128 −0.214 −0.119 0.205 0.177 −0.288
0.09 −0.078 0.135 0.087 −0.133 −0.062 0.108 0.119 −0.202 −0.108 0.187 0.166 −0.274
0.1 −0.080 0.139 0.086 −0.131 −0.060 0.105 0.119 −0.202 −0.106 0.184 0.162 −0.266
0.12 −0.080 0.138 0.085 −0.128 −0.046 0.082 0.107 −0.183 −0.097 0.170 0.151 −0.247
0.14 −0.082 0.140 0.082 −0.123 −0.039 0.070 0.095 −0.166 −0.096 0.165 0.139 −0.228
0.15 −0.078 0.135 0.082 −0.123 −0.040 0.071 0.095 −0.163 −0.093 0.160 0.138 −0.224
0.16 −0.081 0.139 0.083 −0.125 −0.041 0.073 0.092 −0.157 −0.096 0.165 0.135 −0.219
0.18 −0.080 0.137 0.084 −0.126 −0.033 0.058 0.079 −0.137 −0.091 0.157 0.123 −0.200
0.2 −0.080 0.138 0.084 −0.128 −0.029 0.051 0.075 −0.130 −0.091 0.155 0.121 −0.195
0.25 −0.081 0.140 0.084 −0.127 −0.029 0.049 0.065 −0.112 −0.092 0.158 0.114 −0.181
0.3 −0.079 0.137 0.086 −0.130 −0.032 0.055 0.051 −0.087 −0.092 0.158 0.105 −0.164
0.35 −0.079 0.135 0.089 −0.136 −0.031 0.052 0.056 −0.096 −0.092 0.157 0.108 −0.170
0.4 −0.081 0.139 0.091 −0.140 −0.031 0.052 0.049 −0.083 −0.092 0.157 0.108 −0.168
0.45 −0.079 0.136 0.087 −0.134 −0.031 0.052 0.043 −0.072 −0.094 0.160 0.105 −0.163
0.5 −0.082 0.142 0.087 −0.131 −0.032 0.053 0.040 −0.066 −0.095 0.163 0.106 −0.164
0.6 −0.082 0.139 0.088 −0.136 −0.037 0.062 0.042 −0.066 −0.098 0.167 0.111 −0.172
0.7 −0.079 0.135 0.087 −0.134 −0.039 0.065 0.048 −0.078 −0.098 0.168 0.117 −0.184
0.8 −0.079 0.136 0.088 −0.135 −0.042 0.069 0.050 −0.082 −0.101 0.172 0.121 −0.191
0.9 −0.081 0.139 0.086 −0.130 −0.037 0.064 0.052 −0.085 −0.100 0.171 0.121 −0.190
1 −0.080 0.137 0.086 −0.130 −0.040 0.067 0.051 −0.084 −0.101 0.172 0.120 −0.190
1.25 −0.082 0.142 0.087 −0.132 −0.045 0.073 0.050 −0.080 −0.105 0.180 0.125 −0.196
1.5 −0.077 0.134 0.084 −0.129 −0.045 0.075 0.055 −0.091 −0.102 0.175 0.129 −0.207
2 −0.077 0.131 0.083 −0.129 −0.041 0.069 0.050 −0.082 −0.106 0.180 0.133 −0.214
2.5 −0.071 0.123 0.086 −0.134 −0.042 0.070 0.052 −0.083 −0.102 0.176 0.141 −0.228
3 −0.069 0.120 0.087 −0.137 −0.043 0.072 0.052 −0.083 −0.102 0.176 0.147 −0.238
3.5 −0.064 0.113 0.085 −0.133 −0.038 0.064 0.049 −0.079 −0.096 0.168 0.145 −0.236
4 −0.060 0.106 0.085 −0.134 −0.038 0.064 0.049 −0.080 −0.095 0.166 0.146 −0.239
4.5 −0.057 0.101 0.081 −0.130 −0.031 0.052 0.048 −0.081 −0.091 0.159 0.146 −0.241
5 −0.057 0.100 0.081 −0.131 −0.029 0.051 0.049 −0.081 −0.088 0.154 0.145 −0.239

in deriving those coefficients. Tables 6–10 present the coeffi- 1. the total standard deviation of a GMPE for a damping ratio
cients in equations (12a,b) and (13a,b) for SC I, II, III, and IV less than 5.0% is likely to be similar to or larger than that of
sites, respectively. GMPE with a damping ratio of 5%; and
These standard deviations reflect the goodness of fit for 2. the total standard deviation of a GMPE for a damping ratio
the model proposed in this study and can be used as a com- larger than 5.0% is likely to be similar to or less than that of
parison parameter in model selections. Unfortunately, the GMPE with a damping ratio of 5%.
model standard deviations for DMF are not often presented,
for example, Lin and Chang (2003, 2004), Castillo and Ruiz In the first case, standard deviations from DMF models may
(2014), Daneshvar et al. (2016, 2017), whereas Cameron and be accounted for using the following equation:
Green (2007) presented the standard deviations that are very
similar to those in this article. q
The various standard deviations are useful for model com- EQ-TARGET;temp:intralink-;df14;320;159 σ GT  σ 2T  σ 2GMPET ;
14
parisons as the goodness-of-fit parameter of the DMF model.
However, how these standard deviations can be used in a prob- in which σ GT is the total standard deviation used in probabi-
abilistic seismic analysis for deriving a design spectrum with a listic seismic hazard analyses, σ T is the total standard deviation
damping ratio other than 5% is a different matter. Our inves- for the DMF model, and σ GMPET is the total standard deviation
tigation suggests two possible cases: for the GMPE for a damping ratio of 5%. Equation (14)

656 • Bulletin of the Seismological Society of America [Link] Volume 110 Number 2 April 2020

Downloaded from [Link]

by Chengdu University of Technology, JohnZhao1000
 TABLE 7
Model Parameters for Within-Site, Between-Site, and Total Standard Deviations for SC II Sites

T (s) g12 h12 r 12 s12 g22 h22 r 22 s22 g42 h42 r 42 s42

0.03 −0.002 0.005 0.010 −0.018 0.000 0.001 0.003 −0.006 −0.012 0.020 0.017 −0.027
0.04 −0.055 0.089 0.034 −0.049 −0.051 0.080 0.048 −0.081 −0.088 0.139 0.068 −0.106
0.05 −0.062 0.102 0.047 −0.068 −0.062 0.099 0.055 −0.087 −0.103 0.168 0.090 −0.138
0.06 −0.073 0.121 0.058 −0.085 −0.076 0.124 0.074 −0.116 −0.122 0.201 0.120 −0.186
0.07 −0.076 0.129 0.070 −0.104 −0.084 0.140 0.093 −0.148 −0.125 0.210 0.143 −0.227
0.08 −0.079 0.133 0.077 −0.116 −0.082 0.137 0.110 −0.180 −0.124 0.209 0.157 −0.254
0.09 −0.082 0.140 0.082 −0.123 −0.075 0.130 0.122 −0.202 −0.119 0.204 0.166 −0.268
0.1 −0.083 0.141 0.085 −0.129 −0.074 0.129 0.118 −0.194 −0.117 0.200 0.161 −0.260
0.12 −0.080 0.137 0.087 −0.135 −0.058 0.103 0.108 −0.183 −0.104 0.180 0.153 −0.252
0.14 −0.083 0.144 0.090 −0.137 −0.058 0.104 0.115 −0.194 −0.106 0.184 0.158 −0.257
0.15 −0.081 0.141 0.091 −0.138 −0.048 0.087 0.109 −0.185 −0.099 0.173 0.153 −0.250
0.16 −0.085 0.146 0.092 −0.139 −0.046 0.082 0.104 −0.178 −0.101 0.174 0.149 −0.243
0.18 −0.083 0.143 0.089 −0.135 −0.044 0.078 0.097 −0.164 −0.099 0.170 0.139 −0.226
0.2 −0.081 0.140 0.087 −0.132 −0.033 0.060 0.086 −0.148 −0.093 0.160 0.129 −0.210
0.25 −0.084 0.143 0.084 −0.127 −0.034 0.063 0.088 −0.151 −0.096 0.165 0.128 −0.207
0.3 −0.080 0.139 0.087 −0.133 −0.038 0.066 0.075 −0.127 −0.095 0.164 0.120 −0.190
0.35 −0.079 0.136 0.088 −0.134 −0.043 0.073 0.072 −0.122 −0.098 0.166 0.117 −0.185
0.4 −0.079 0.136 0.090 −0.138 −0.046 0.078 0.071 −0.120 −0.097 0.165 0.118 −0.187
0.45 −0.080 0.138 0.090 −0.139 −0.047 0.080 0.070 −0.115 −0.100 0.172 0.120 −0.190
0.5 −0.079 0.135 0.088 −0.136 −0.050 0.084 0.067 −0.111 −0.100 0.171 0.120 −0.190
0.6 −0.075 0.129 0.084 −0.129 −0.054 0.089 0.067 −0.110 −0.100 0.170 0.119 −0.188
0.7 −0.079 0.136 0.085 −0.130 −0.050 0.085 0.071 −0.118 −0.104 0.178 0.127 −0.201
0.8 −0.075 0.130 0.086 −0.133 −0.050 0.083 0.069 −0.117 −0.102 0.174 0.129 −0.207
0.9 −0.080 0.136 0.087 −0.134 −0.055 0.091 0.073 −0.122 −0.106 0.180 0.132 −0.211
1 −0.078 0.134 0.087 −0.133 −0.055 0.091 0.074 −0.122 −0.106 0.181 0.132 −0.211
1.25 −0.079 0.134 0.089 −0.138 −0.051 0.085 0.075 −0.125 −0.105 0.179 0.138 −0.222
1.5 −0.077 0.133 0.085 −0.132 −0.057 0.096 0.074 −0.122 −0.108 0.184 0.139 −0.224
2 −0.075 0.129 0.085 −0.132 −0.047 0.079 0.070 −0.116 −0.107 0.183 0.143 −0.230
2.5 −0.070 0.122 0.088 −0.138 −0.054 0.091 0.064 −0.104 −0.107 0.184 0.148 −0.239
3 −0.069 0.119 0.086 −0.136 −0.046 0.079 0.065 −0.107 −0.103 0.178 0.152 −0.247
3.5 −0.069 0.119 0.084 −0.130 −0.047 0.081 0.062 −0.098 −0.104 0.180 0.149 −0.241
4 −0.064 0.111 0.082 −0.130 −0.044 0.077 0.064 −0.104 −0.100 0.175 0.151 −0.246
4.5 −0.058 0.103 0.078 −0.122 −0.043 0.074 0.067 −0.112 −0.096 0.168 0.151 −0.249
5 −0.053 0.094 0.079 −0.124 −0.046 0.078 0.062 −0.099 −0.092 0.161 0.148 −0.243

2
assumes that the residuals in the GMPE and the DFM model T
SD  SA ;
15
2π
EQ-TARGET;temp:intralink-;df15;308;289

are not correlated; this assumption is not entirely correct, but

conservative. In the second case, the standard deviation from
the DMF may not be used in the probabilistic seismic hazard in which SA is the 5% damped acceleration spectrum from the
analysis, assuming that variability associated with DMF models GMPE by Zhao, Jiang, et al. (2016), and SD is the approximate
is included in the GMPE already. This method is not entirely displacement spectrum for checking if the DMF model in equa-
correct, but conservative. tion (5) has any abrupt deviation from the mean values in
the variation pattern with spectral periods. Equation (15) is accu-
COMPARISON OF MODEL PREDICTIONS rate at spectral periods up to about 3.0 s, and the accuracy is
For a model that can be used for engineering designs, a design acceptable in a spectral period range of 3.0–5.0 s for our purpose
spectrum should have a smoothed variation with spectral peri- of checking DMF model performance.
ods. Because DMFs are functions of magnitude, source depth, Figure 12 presents the displacement spectra for five damping
and source distance, the smoothed DMFs may not lead to a ratios at SC I sites from an M w 5 event at a distance of 25 km in
smoothed spectrum at all spectral periods in the full range of Figure 12a and 250 km in Figure 12b. About 25 km would be the
these ground-motion parameters, even though the smoothed smallest possible distance for slab earthquake records in Japan.
GMPE and DMF models are used. For the comparison of a dis- The symbols in Figure 12 denote the displacement spectrum for
placement spectrum for a given damping ratio, we calculated the a damping ratio of 5% multiplied by the average DMF values of
displacement spectrum using the following equation: the dataset, and the solid lines denote the spectrum multiplied

Volume 110 Number 2 April 2020 [Link] Bulletin of the Seismological Society of America • 657

Downloaded from [Link]

by Chengdu University of Technology, JohnZhao1000
 TABLE 8
Model Parameters for Within-Site, Between-Site, and Total Standard Deviations for SC III Sites

T (s) g13 h13 r 13 s13 g23 h23 r 23 s23 g43 h43 r 43 s43

0.03 −0.002 0.004 0.008 −0.014 0.000 0.001 0.002 −0.004 −0.011 0.020 0.015 −0.025
0.04 −0.048 0.075 0.030 −0.045 −0.048 0.071 0.024 −0.036 −0.081 0.125 0.052 −0.078
0.05 −0.056 0.093 0.042 −0.062 −0.081 0.122 0.051 −0.081 −0.111 0.177 0.085 −0.131
0.06 −0.067 0.110 0.056 −0.085 −0.088 0.145 0.076 −0.117 −0.126 0.208 0.121 −0.187
0.07 −0.078 0.130 0.073 −0.112 −0.101 0.167 0.098 −0.152 −0.137 0.229 0.148 −0.234
0.08 −0.076 0.131 0.077 −0.117 −0.095 0.160 0.099 −0.153 −0.131 0.223 0.150 −0.236
0.09 −0.078 0.137 0.078 −0.118 −0.078 0.134 0.096 −0.152 −0.118 0.204 0.146 −0.231
0.1 −0.085 0.147 0.079 −0.119 −0.070 0.124 0.110 −0.179 −0.116 0.201 0.152 −0.244
0.12 −0.082 0.141 0.091 −0.142 −0.070 0.119 0.122 −0.204 −0.113 0.192 0.165 −0.272
0.14 −0.074 0.127 0.089 −0.140 −0.053 0.094 0.110 −0.187 −0.096 0.166 0.154 −0.255
0.15 −0.080 0.134 0.090 −0.141 −0.040 0.075 0.107 −0.182 −0.095 0.162 0.151 −0.251
0.16 −0.083 0.142 0.085 −0.131 −0.040 0.075 0.102 −0.171 −0.097 0.168 0.144 −0.235
0.18 −0.077 0.133 0.086 −0.131 −0.045 0.082 0.105 −0.177 −0.094 0.164 0.143 −0.233
0.2 −0.080 0.137 0.086 −0.130 −0.044 0.078 0.091 −0.154 −0.096 0.165 0.133 −0.213
0.25 −0.077 0.136 0.086 −0.131 −0.044 0.078 0.091 −0.154 −0.095 0.166 0.132 −0.212
0.3 −0.081 0.143 0.082 −0.118 −0.048 0.082 0.085 −0.147 −0.100 0.174 0.121 −0.191
0.35 −0.077 0.137 0.085 −0.126 −0.037 0.067 0.094 −0.164 −0.093 0.164 0.129 −0.206
0.4 −0.079 0.140 0.085 −0.124 −0.042 0.073 0.082 −0.137 −0.094 0.166 0.121 −0.187
0.45 −0.079 0.135 0.083 −0.126 −0.038 0.064 0.069 −0.121 −0.095 0.164 0.115 −0.182
0.5 −0.079 0.136 0.089 −0.136 −0.043 0.074 0.060 −0.100 −0.097 0.167 0.117 −0.184
0.6 −0.078 0.135 0.089 −0.136 −0.047 0.078 0.070 −0.114 −0.099 0.170 0.124 −0.195
0.7 −0.080 0.137 0.085 −0.129 −0.053 0.092 0.071 −0.114 −0.105 0.182 0.127 −0.199
0.8 −0.083 0.142 0.083 −0.124 −0.047 0.078 0.069 −0.112 −0.107 0.181 0.126 −0.198
0.9 −0.073 0.124 0.080 −0.125 −0.049 0.084 0.069 −0.112 −0.098 0.167 0.125 −0.201
1 −0.079 0.136 0.080 −0.120 −0.047 0.081 0.078 −0.129 −0.103 0.178 0.130 −0.206
1.25 −0.079 0.137 0.085 −0.131 −0.047 0.081 0.065 −0.107 −0.104 0.179 0.131 −0.208
1.5 −0.077 0.138 0.092 −0.139 −0.053 0.091 0.079 −0.131 −0.105 0.185 0.145 −0.232
2 −0.075 0.128 0.090 −0.142 −0.052 0.088 0.085 −0.141 −0.110 0.187 0.154 −0.250
2.5 −0.074 0.130 0.092 −0.145 −0.049 0.083 0.069 −0.113 −0.107 0.186 0.152 −0.247
3 −0.074 0.126 0.087 −0.137 −0.046 0.080 0.061 −0.098 −0.107 0.183 0.150 −0.244
3.5 −0.065 0.111 0.087 −0.141 −0.041 0.072 0.065 −0.107 −0.098 0.170 0.152 −0.252
4 −0.062 0.106 0.084 −0.134 −0.051 0.088 0.072 −0.119 −0.102 0.177 0.155 −0.255
4.5 −0.054 0.095 0.078 −0.123 −0.041 0.071 0.073 −0.123 −0.093 0.162 0.154 −0.255
5 −0.055 0.094 0.073 −0.116 −0.047 0.079 0.068 −0.115 −0.094 0.162 0.148 −0.246

by the DMF values in equations (5), (9a,b) and (10a,b). The pre- from peak ground acceleration smoothly to the peak spectrum
dicted displacement spectra vary smoothly with spectral periods with increasing spectral periods for all damping ratios.
at both distances. Figure 13 presents the predicted displacement Figures 16b and 17b show that for the same events as shown
spectra from SC IV sites for the same magnitude, depth, and in Figures 13a and 14a but at a distance of 250 km, the pseu-
source distances as those in Figure 12. Again, there is no local doacceleration spectrum for a damping ratio of 30% has a small
sharp variation. Figures 14 and 15 present the displacement trough centered at about 0.06 s, and the trough value increases
spectra with the same earthquake depth, distance, and SCs as with increasing distance. For the displacement spectrum, this
for Figures 12 and 13, respectively, but for an M w 7 event. trough is not significant for practical engineering designs.
Again, the spectra vary smoothly with spectral periods without
any sharp local variations. COMPARISON WITH OTHER MODELS
The displacement spectrum has very small values at short Many DMF models have been published, but the only recent
periods, and we calculate the pseudoacceleration spectra using model for subduction earthquakes appears to be from Daneshvar
equation (1) for easy visual inspection on the variation of the et al. (2016, 2017). The function form from that model is pre-
spectrum at short periods (the pseudoacceleration spectrum dif- sented in equation (3), and there are 12 regression coefficients for
fers from the total acceleration spectrum significantly, especially each SC. The spectral periods were divided into two segments,
at long spectral periods and large damping ratios). Figures 16a and all coefficients are continuous with respect to spectral periods.
and 17a show that the pseudoacceleration spectra from an M w 5 Figure 18 compares the DMF values from this study with
event and an M w 7 event at a source distance of 25 km increase those from the Daneshvar et al. (2017) study for two damping

658 • Bulletin of the Seismological Society of America [Link] Volume 110 Number 2 April 2020

Downloaded from [Link]

by Chengdu University of Technology, JohnZhao1000
 TABLE 9
Model Parameters for Within-Site, Between-Site, and Total Standard Deviations for SC IV Sites

T (s) g14 h14 r 14 s14 g24 h24 r 24 s24 g44 h44 r 44 s44

0.03 −0.001 0.002 0.004 −0.008 −0.002 0.004 0.008 −0.015 −0.012 0.020 0.016 −0.026
0.04 −0.034 0.053 0.014 −0.019 −0.025 0.038 0.014 −0.021 −0.061 0.096 0.040 −0.059
0.05 −0.047 0.075 0.028 −0.041 −0.045 0.068 0.026 −0.040 −0.084 0.134 0.066 −0.101
0.06 −0.056 0.091 0.048 −0.073 −0.062 0.099 0.050 −0.077 −0.103 0.169 0.102 −0.160
0.07 −0.065 0.108 0.060 −0.090 −0.080 0.133 0.077 −0.121 −0.115 0.193 0.129 −0.204
0.08 −0.075 0.126 0.073 −0.112 −0.076 0.127 0.089 −0.143 −0.117 0.197 0.142 −0.229
0.09 −0.079 0.134 0.083 −0.126 −0.075 0.128 0.106 −0.174 −0.116 0.198 0.155 −0.250
0.1 −0.078 0.135 0.085 −0.131 −0.082 0.141 0.128 −0.210 −0.119 0.204 0.168 −0.274
0.12 −0.080 0.139 0.090 −0.139 −0.075 0.128 0.129 −0.215 −0.114 0.196 0.170 −0.279
0.14 −0.083 0.141 0.091 −0.142 −0.048 0.084 0.109 −0.185 −0.101 0.172 0.155 −0.255
0.15 −0.079 0.138 0.090 −0.138 −0.046 0.082 0.100 −0.169 −0.097 0.168 0.146 −0.239
0.16 −0.080 0.139 0.086 −0.132 −0.046 0.081 0.095 −0.160 −0.097 0.168 0.140 −0.226
0.18 −0.082 0.141 0.090 −0.136 −0.036 0.063 0.089 −0.157 −0.095 0.162 0.134 −0.218
0.2 −0.079 0.136 0.090 −0.139 −0.040 0.072 0.094 −0.163 −0.093 0.161 0.138 −0.225
0.25 −0.079 0.136 0.084 −0.126 −0.034 0.063 0.075 −0.127 −0.093 0.160 0.120 −0.191
0.3 −0.077 0.135 0.085 −0.128 −0.034 0.058 0.064 −0.111 −0.091 0.157 0.111 −0.176
0.35 −0.079 0.137 0.082 −0.123 −0.030 0.048 0.062 −0.110 −0.092 0.157 0.106 −0.166
0.4 −0.081 0.140 0.083 −0.126 −0.035 0.062 0.061 −0.101 −0.094 0.162 0.108 −0.167
0.45 −0.084 0.143 0.083 −0.125 −0.035 0.060 0.044 −0.070 −0.099 0.169 0.102 −0.155
0.5 −0.080 0.140 0.086 −0.131 −0.031 0.051 0.041 −0.067 −0.093 0.161 0.106 −0.164
0.6 −0.082 0.140 0.080 −0.119 −0.035 0.057 0.038 −0.062 −0.097 0.166 0.102 −0.156
0.7 −0.080 0.137 0.077 −0.114 −0.043 0.071 0.047 −0.076 −0.101 0.172 0.110 −0.169
0.8 −0.080 0.138 0.079 −0.117 −0.048 0.082 0.061 −0.098 −0.104 0.179 0.120 −0.186
0.9 −0.081 0.139 0.086 −0.131 −0.046 0.077 0.061 −0.099 −0.102 0.175 0.125 −0.197
1 −0.083 0.141 0.087 −0.133 −0.044 0.075 0.063 −0.103 −0.105 0.179 0.127 −0.201
1.25 −0.081 0.139 0.088 −0.134 −0.054 0.089 0.065 −0.107 −0.109 0.184 0.132 −0.209
1.5 −0.081 0.138 0.083 −0.126 −0.054 0.091 0.068 −0.110 −0.108 0.185 0.134 −0.214
2 −0.078 0.134 0.088 −0.135 −0.056 0.093 0.081 −0.134 −0.113 0.193 0.150 −0.241
2.5 −0.077 0.134 0.087 −0.134 −0.051 0.085 0.076 −0.126 −0.110 0.190 0.152 −0.247
3 −0.073 0.126 0.090 −0.140 −0.051 0.087 0.078 −0.128 −0.108 0.186 0.159 −0.259
3.5 −0.069 0.121 0.089 −0.139 −0.048 0.081 0.069 −0.114 −0.104 0.180 0.155 −0.253
4 −0.066 0.114 0.086 −0.136 −0.052 0.087 0.074 −0.122 −0.105 0.181 0.157 −0.257
4.5 −0.062 0.108 0.079 −0.123 −0.043 0.073 0.070 −0.116 −0.098 0.171 0.153 −0.251
5 −0.062 0.105 0.075 −0.118 −0.042 0.071 0.063 −0.107 −0.096 0.164 0.147 −0.242

ratios, 1%, and 3% and for two SCs, SC II and SC III, respec- (2017) model. One significant difference in Figure 18b between
tively. The SCs used in the Daneshvar et al. (2016, 2017) study the two models is that the Daneshvar et al. (2017) model has a
are based on V S30 , the travel-time-averaged shear-wave veloc- DMF value of 1.0 at a spectral period less than 5.0 s, whereas
ity of top 30 m soil, whereas the SCs used in this study are the model from this study is approaching 1.0 at about 8.0 s.
based on site period. For a damping ratio of 1% and for an Figure 19 compares the DMFs from this study with those
SC II site, the DMF values from the Daneshvar et al. (2017) from the Daneshvar et al. (2016) study for two damping ratios,
study are moderately larger than those from this study in a 10% and 30% and for two SCs. The SC definitions in these two
period range of 0.06–0.3 s, whereas the two models have very studies differ slightly. For SC II sites, the results from these two
similar predicted values at the other spectral periods, as shown studies are similar in a period band of 0.2–3.0 s, and the
in Figure 18a. For a damping ratio of 3%, the two models have differences increase with decreasing spectral periods, especially
very similar values at nearly the full spectral period range for for a damping ratio of 30% that has an unsmoothed variation
the Daneshvar et al. (2017). For SC III sites and a damping at 1.0 s. The Daneshvar et al. (2016) model does not have a
ratio of 1%, the values from the Daneshvar et al. (2017) are smoothed variation toward 1.0 at short spectral periods,
larger than those from this study in the period range of whereas the models from this study reach 1.0 at a spectral
0.1–0.2 s and are significantly smaller than those from this period of 0.02 s. Similarly, for SC III sites, the DMF values from
study at spectral periods over 1.0 s. Similar to those for SC the Daneshvar et al. (2016) model are similar to those from
II sites, the predicted DMF values from the two models are this study at spectral periods over 0.15 s and are much smaller
very similar at all spectral periods used by the Daneshvar et al. than those from this study at short periods up to 0.1 s. The

Volume 110 Number 2 April 2020 [Link] Bulletin of the Seismological Society of America • 659

Downloaded from [Link]

by Chengdu University of Technology, JohnZhao1000
 TABLE 10
Model Parameters for Between-Event Standard Deviations

T (s) g3 h3 r3 s3

0.03 −0.011 0.020 0.013 −0.020

0.04 −0.045 0.071 0.034 −0.052
0.05 −0.054 0.089 0.054 −0.083
0.06 −0.061 0.102 0.075 −0.120
0.07 −0.051 0.088 0.084 −0.139
0.08 −0.048 0.083 0.084 −0.140
0.09 −0.042 0.071 0.078 −0.133
0.1 −0.035 0.060 0.070 −0.121
0.12 −0.032 0.053 0.066 −0.116
0.14 −0.030 0.050 0.062 −0.109
0.15 −0.030 0.050 0.059 −0.105
0.16 −0.030 0.048 0.056 −0.099
0.18 −0.031 0.049 0.048 −0.084
0.2 −0.031 0.049 0.045 −0.079
0.25 −0.034 0.055 0.043 −0.073
0.3 −0.035 0.055 0.034 −0.055
0.35 −0.037 0.061 0.029 −0.044
0.4 −0.031 0.053 0.031 −0.047
0.45 −0.039 0.065 0.039 −0.060
0.5 −0.036 0.061 0.046 −0.074
0.6 −0.040 0.067 0.052 −0.083 Figure 12. A comparison of scaled response spectrum using DMF for an
0.7 −0.044 0.077 0.063 −0.100 Mw 5.0 event at (a) a source distance of 25 km and (b) at a distance
0.8 −0.047 0.080 0.067 −0.109 of 250 km at SC I sites for damping ratios of 1%, 3%, 8%, 15%, and 30%.
0.9 −0.043 0.075 0.067 −0.111 Symbols denote the displacement spectrum for a damping ratio of 5%
1 −0.047 0.080 0.067 −0.111 multiplied by the average DMF values of the dataset. The color version of
1.25 −0.048 0.083 0.075 −0.123 this figure is available only in the electronic edition.
1.5 −0.049 0.084 0.081 −0.136
2 −0.061 0.104 0.091 −0.151
2.5 −0.060 0.104 0.100 −0.166
3 −0.062 0.106 0.106 −0.178
III sites. Figure 20 shows that the distributions of between-
3.5 −0.062 0.107 0.107 −0.180 event residuals with respect to both magnitude and fault depth
4 −0.064 0.111 0.109 −0.182 are biased significantly, as illustrated by the trend line.
4.5 −0.064 0.110 0.112 −0.187 Figure 21 shows that the distribution of within-event residuals
5 −0.060 0.105 0.109 −0.184 is slightly biased with respect to magnitude, whereas the same
distribution is strongly biased with respect to source distance.
Because of the uneven distribution of the strong-motion
differences would largely from using very different datasets, records with respect to magnitude, source depth, and source
that is, the average magnitude and source distance for the data- distance, the slope of the trend line does not equal the added
set in this study differ from those in the Daneshvar et al. (2016) distance coefficient to correct the distance scaling, but a large
dataset, and the effects of different averaged magnitude and slope is a strong indicator that the model can be improved by
source distance are likely to propagate into the magnitude- introducing magnitude, fault depth, and source distance to the
and distance-independent models. model at this particular spectral period.

DISTRIBUTION OF RESIDUAL DISCUSSIONS AND CONCLUSIONS

DMFs are known to vary with earthquake source and path A DMF model without source and path parameters is presented
parameters as shown by Rezaeian et al. (2014). This study does in this article for subduction slab earthquakes from Japan. The
not include these parameters because the DMF model is to be dataset used in this study is similar to that used in the Zhao,
used to scale a 5% damped displacement design spectrum that Jiang, et al. (2016) study with a modification for the maximum
does not have an associated magnitude, source depth, and usable spectral period. About 4695 strong-motion records from
source distance. However, these effects may be significant, 136 subduction slab earthquakes obtained by the K-NET and
and the distribution of the residuals with respect to earthquake KiK-net were used. The recording sites were divided into four
magnitude and source distances will be examined. The largest SCs based on site period calculated from the measured shear-
total standard deviations are at 0.12 s for a damping ratio of wave velocity profile. DMF values for 14 damping ratios and
30%, and the corresponding residuals are presented here for SC 34 spectral periods were used to construct the DMF model,

660 • Bulletin of the Seismological Society of America [Link] Volume 110 Number 2 April 2020

Downloaded from [Link]

by Chengdu University of Technology, JohnZhao1000
 Figure 13. A comparison of scaled response spectrum using DMF for an Figure 14. A comparison of scaled response spectrum using DMF for an
Mw 5.0 event at (a) a source distance of 25 km and (b) at a distance of Mw 7.0 event at (a) a source distance of 25 km and (b) at a distance
250 km at SC IV sites for damping ratios of 1%, 3%, 8%, 15%, and 30%. of 250 km at SC I sites for damping ratios of 1%, 3%, 8%, 15%, and 30%.
The color version of this figure is available only in the electronic edition. The color version of this figure is available only in the electronic edition.

and the DMF value can be calculated for any damping ratio many spectral periods. The largest standard deviations are at
between 1% and 30%, and at any spectral period between about 0.1 s and the standard deviations for a damping ratio
0.03 and 5.0 s. The following conclusions can be reached: of 30% are larger than those among four site classes for the
other damping ratios. The total standard deviations at short
1. at short periods up to about 0.1 s, the DMF values for SC I periods decrease with increasing site periods and are similar
sites differ significantly from those for the other SCs. At among four site classes at spectral periods over about 0.1 s;
spectral periods over 0.3 s, the differences among the DMF 6. all standard deviations can be modeled by a simple linear
values from four SCs for a damping ratio less than 5% are function of the logarithm of damping ratios; the standard
larger than those for a damping ratio more than 5%; deviations decrease linearly with the logarithm of damping
2. a simple quadratic function of the logarithm of damping ratios when the damping ratio is within 5% and increase
ratios can be used to model the effect of damping ratios linearly with the logarithm of damping ratios when the
very well, and the simple function satisfies the condition damping ratio is more than 5%;
that the DMF equals 1.0 at a damping ratio of 5%; 7. the DMF models derived in this study combined with the
3. the effect of spectral periods can be well modeled by a fourth- acceleration spectrum for the Zhao, Jiang, et al. (2016)
order polynomial of the logarithm of spectral periods at spec- produce smoothed displacement spectra at nearly all spec-
tral periods of 0.06 s or longer. The maximum period for the tral periods. At very short spectral periods where displace-
strong-motion records used in this study is 5.0 s, and it is ment spectrum is small, a small trough can be observed for
possible to extrapolate to a longer period if necessary; a damping ratio of 30% from events at a long distance;
4. the model satisfies boundary conditions that require the DMF 8. in a spectral period range of 0.2–3.0 s, the DMF values from
values approaching 1.0 with reducing spectral periods. At this study are close to those by Daneshvar et al. (2016,
spectral periods over 1.0 s, the DMFs for different damping 2017), but at short periods the difference is significant;
ratios appear to converge to 1.0 at spectral periods over about 9. distribution of residuals suggests that DMFs depend on
8 s because of the physical property of constant displacement earthquake source and path parameters;
spectrum at long spectral periods (or strictly T  ∞); 10. the standard deviations can be used to judge of goodness
5. the between-event standard deviations are much smaller of fit of various components in the DMF model and how
than the within-site and between-site standard deviations at these standard deviations are used in the probabilistic

Volume 110 Number 2 April 2020 [Link] Bulletin of the Seismological Society of America • 661

Downloaded from [Link]

by Chengdu University of Technology, JohnZhao1000
 Figure 15. A comparison of scaled response spectrum using DMF for Figure 16. A comparison of scaled pseudoacceleration response spectrum
an Mw 7.0 event at (a) a source distance of 25 km and (b) at a distance using DMF for an Mw 5.0 event at (a) a source distance of 25 km and
of 250 km at SC IV sites for damping ratios of 1%, 3%, 8%, 15%, and 30%. (b) at a distance of 250 km at SC IV sites for damping ratios of 1%, 3%,
The color version of this figure is available only in the electronic edition. 8%, 15%, and 30%. The color version of this figure is available only in the
electronic edition.

Figure 17. A comparison of scaled pseudoacceleration response spectrum

using DMF for an Mw 7.0 event at (a) a source distance of 25 km and Figure 18. Comparison of the average damping ratio factors predicted by the
(b) at a distance of 250 km at SC I sites for damping ratios of 1%, 3%, 8%, model developed in this study and the model developed in Daneshvar et al.
15%, and 30%. The color version of this figure is available only in the (2017) for (a) SC II and (b) SC III sites for damping ratios of 1% and 3%. The
electronic edition. color version of this figure is available only in the electronic edition.

662 • Bulletin of the Seismological Society of America [Link] Volume 110 Number 2 April 2020

Downloaded from [Link]

by Chengdu University of Technology, JohnZhao1000
 Figure 19. Comparison of the average damping ratio factors predicted by the Figure 20. Distribution of between-event residuals for SC III sites for a damp-
model developed in this study and the model developed in Daneshvar et al. ing ratio of 30% at a spectral period of 0.12 s, with respect to (a) magnitude
(2016) for (a) SC II and (b) SC III sites for dampingratios of 10% and 30%. and (b) fault depth. The color version of this figure is available only in the
The color version of this figure is available only in the electronic edition. electronic edition.

seismic hazard analyses depends on the damping ratios of

the target spectrum; and
11. the magnitude range for the dataset used in this study is
4.99–8.25, the depth range is 12.0–166.7 km and the dis-
tance range is 27.4–300.0 km. The model presented in this
article does not contain the earthquake source and path
terms; therefore to scale a design spectrum for event with
a given magnitude, fault depth, and a source distance that
differ significantly from the average magnitude (6.2), the
average depth (34.0 km) of the dataset or a short distance
less than 20.0 km should be cautioned.

DATA AND RESOURCES

The dataset used in this study is similar to that used in the Zhao, Jiang,
et al. (2016) study, but with minor modifications. The maximum usable
period is proportional to T m , the inverse of the high-pass filter corner
frequency at which the scaling factor is 1.0. We modified the maximum
usable period from less than Tm to less than or equal to Tm to calculate
the maximum usable period with the filtering frequency band extended
accordingly. This change leads to more available records for spectral
periods over 2.5 s, and we tested that the increase of the records led
to improved maximum log likelihood after the effect of the increased
number of records on the maximum log likelihood was excluded. All
Figure 21. Distribution of within-event residuals for SC III sites for a damping records were processed manually based on displacement time history
ratio of 30% at a spectral period of 0.12 s, with respect to (a) magnitude and Fourier spectrum variation at low frequencies. The relative values
and (b) source distance. The color version of this figure is available only in for peak displacements in the pre-event recording, and P-wave and the
the electronic edition. S-wave parts of a record were also used as the processing parameters.

Volume 110 Number 2 April 2020 [Link] Bulletin of the Seismological Society of America • 663

Downloaded from [Link]

by Chengdu University of Technology, JohnZhao1000
 The earthquake classification is from the Zhao et al. (2015) study based ground motion and response spectra, Earthq Spectra 30, no. 3,
on the best ground-motion prediction equation (GMPE) goodness-of- 1087–1116.
fit parameter, and all relevant earthquake source parameters were pre- Daneshvar, P., N. Bouaanani, K. Goda, and G. M. Atkinson (2016).
sented by Zhao et al. (2015). All strong-motion records used in this Damping reduction factors for crustal, inslab, and interface earth-
study can be downloaded from [Link] (last quakes characterizing seismic hazard in southwestern British
accessed November 2019). Source distance and site classification for Columbia, Canada, Earthq. Spectra 32, no. 1, 45–74.
all records used in this study can be supplied on request. The regression Daneshvar, P., N. Bouaanani, K. Goda, and G. M. Atkinson (2017).
analyses used in this study can be performed by commercial code like Damping modification factors for deep inslab and interface sub-
MATLAB. duction earthquakes, 16th World Conf. on Earthquake, Santiago,
Chile, 9–13 January 2017, Paper Number 2827.
ACKNOWLEDGMENTS Eurocode 8 (2004). Design of Structures for Earthquake Resistance,
The authors would like to thank Associate Editor Kawase for his effort Part 1: General Rules, Seismic Actions and Rules for Buildings,
in organizing the review and his constructive comments. Sreeram EN 2004-1-1, CEN, Brussels, Belgium.
Reddy Kotha provided an extensive review that is appreciated greatly. Hao, A. M., D. Y. Zhou, Y. M. Li, and H. Zhang (2011). Effects of
The authors would also thank an anonymous reviewer for construc- moment magnitude, site conditions and closest distance on damping
tive comments and suggestions. The work reported here is partially sup- modification factors, Soil Dynam. Earthq. Eng. 31, no. 9, 1232–1247.
ported by research grants from the National Science Foundation of Hayes, G. P., D. J. Wald, and R. L. Johnson (2012). Slab1.0: A three-
China (51578470 and 51878396) and by Shandong Jianzhu University. dimensional model of global subduction zone geometries, J.
Geophys. Res. 117, no. B01302, doi: 10.1029/2011JB008524.
REFERENCES Hubbard, D. T., and G. P. Mavroeidis (2011). Damping coefficients
Abrahamson, N. A., W. J. Silva, and R. Kamai (2014). Summary of the for near-fault ground motion response spectra, Soil Dynam.
ASK14 ground motion relation for active crustal regions, Earthq. Earthq. Eng. 31, no. 3, 401–417.
Spectra 30, no. 3, 1025–1056. Idriss, I. M. (2014). An NGA-West2 empirical model for estimating
Ashour, S. A. (1987). Elastic Seismic Response of Buildings with the horizontal spectral values generated by shallow crustal earth-
Supplemental Damping, Department of Civil Engineering, quakes, Earthq. Spectra 30, no. 3, 1155–1177.
University of Michigan, Ann Arbor, Michigan. Kawashima, K., and K. Aizawa (1986). Modification of earthquake
Atkinson, G. M., and J. R. Pierre (2004). Ground-motion response response spectra with respect to damping ratio, Proc. of the 3rd US
spectra in eastern north America for different critical damping National Conf. on Earthquake Engineering, Charleston, South
values, Seismol. Res. Lett. 75, no. 4, 541–545. Carolina, 24–28 August 1986, 1107–1116.
Bommer, J. J., and R. Mendis (2005). Scaling of spectral displacement Lin, Y. Y., and K. C. Chang (2003). Study on damping reduction factor
ordinates with damping ratios, Earthq. Eng. Struct. Dynam. 34, for buildings under earthquake ground motions, J. Struct. Eng.
no. 2, 145–165. 129, no. 2, 206–214.
Bommer, J. J., A. S. Elnashai, and A. G. Weir (2000). Compatible Lin, Y. Y., and K. C. Chang (2004). Effects of site classes on damping
acceleration and displacement spectra for seismic design codes, reduction factors, J. Struct. Eng. 130, no. 11, 1667–1675.
presented at the Proc. 12th World Conf. on Earthquake Ministry of Housing and Urban-Rural Development of the People’s
Engineering, Auckland, New Zealand, 30 January–04 February Republic of China (2010). Code for Seismic Design of Buildings
2000, Paper Number 207. (GB50011-2010), Ministry of Housing and Urban-Rural
Boore, D. M., J. P. Stewart, E. Seyhan, and G. M. Atkinson (2014). Development of the People’s Republic of China.
NGA-West2 equations for predicting PGA, PGV, and 5% damped Mollaioli, F., L. Liberatore, and A. Lucchini (2014). Displacement
PSA for shallow crustal earthquakes, Earthq. Spectra 30, no. 3, damping modification factors for pulse-like and ordinary records,
1057–1086. Eng. Struct. 78, no. 1, 17–27.
Building Seismic Safety Council (BSSC) (2003). The 2003 NEHRP Nagao, K., and J. Kanda (2015). Estimation of damping correction
Recommended Provisions for New Buildings and Other Structures factors using duration defined by the standard deviation of phase
(FEMA 368/369), Part I (Provisions) and Part II (Commentary), difference, Earthq. Spectra 31, no. 2, 761–783.
Washington, D.C. Newmark, N. M., and W. J. Hall (1982). Earthquake Spectra and
Cameron, W. I., and R. A. Green (2007). Damping correction factors Design, EERI Monograph Series, Earthquake Engineering
for horizontal ground-motion response spectra, Bull. Seismol. Soc. Research Institute, Oakland, California.
Am. 97, no. 3, 934–960. Palermo, M., S. Silvestri, and T. Trombetti (2016). Stochastic-based
Campbell, K. W., and Y. Bozorgnia (2014). NGA-West2 ground damping reduction factors, Soil Dynam. Earthq. Eng. 80, 168–176.
motion model for the average horizontal components of PGA, Priestley, M. J. N. (2003). Myths and Fallacies in Earthquake
PGV, and 5% damped linear acceleration response spectra, Earthq. Engineering, Revisited, IUSS Press, Pavia, Italy.
Spectra 30, no. 3, 1087–1116. Pu, W., K. Kasai, E. K. Kabando, and B. Huang (2016). Evaluation of
Castillo, T., and S. E. Ruiz (2014) Reduction factors for seismic design the damping modification factor for structures subjected to near-
spectra for structures with viscous energy dampers, J. Earthq. Eng. fault ground motions, Bull. Earthq. Eng. 14, no. 6, 1519–1544.
18, no. 3, 323–349. Rezaeian, S., Y. Bozorgnia, I. M. Idriss, N. A. Abrahamson, K. W.
Chiou, B. S.-J., and R. R. Youngs (2014). Update of the Chiou and Campbell, and W. J. Silva (2014). Damping scaling factors for elas-
Youngs NGA model for the average horizontal component of peak tic response spectra for shallow crustal earthquakes in active

664 • Bulletin of the Seismological Society of America [Link] Volume 110 Number 2 April 2020

Downloaded from [Link]

by Chengdu University of Technology, JohnZhao1000
 tectonic regions: “Average” horizontal component, Earthq. Spectra horizontal component from subduction earthquakes, Bull. Seismol.
30, no. 2, 939–963. Soc. Am. doi: 10.1785/0120190105.
Stafford, P. J., R. Mendis, and J. J. Bommer (2008). Dependence of Zhao, J. X., S. L. Zhou, P. J. Gao, T. Long, Y. B. Zhang, H. K. Thio, M.
damping correction factors for response spectra on duration Lu, and D. A. Rhoades (2015). An earthquake classification scheme
and numbers of cycles, Struct. Eng. 134, no. 8, 1364–1373. adapted for Japan determined by the goodness of fit for ground-
Tolis, S. V., and E. Faccioli (1999). Displacement design spectra, J. motion prediction equations, Bull. Seismol. Soc. Am. 105, no. 5,
Earthq. Eng. 3, no. 1, 107–125. 2750–2763.
Wu, J., and R. D. Hanson (1989). Study of inelastic response spectra Zhao, J. X., S. L. Zhou, P. J. Gao, Y. B. Zhang, J. Zhou, M. Lu, and D. A.
with high damping, J. Struct. Div. 115, no. 6, 1412–1431. Rhoades (2016). Ground-motion prediction equations for shallow
Zhao, J. X., F. Jiang, P. Shi, H. Xing, Y. Zhang, P. C. Yu, M. Lu, and D. A. crustal and upper mantle earthquakes in Japan using site class and
Rhoades (2016). Ground-motion prediction equations for subduc- simple geometric attenuation functions, Bull. Seismol. Soc. Am.
tion slab earthquakes in Japan using site class and simple geometric 106, no. 4, 1552–1569.
attenuation functions, Bull. Seismol. Soc. Am. 106, no. 4, 1535–1551. Zhou, F., L. Wenguang, and Z. Xu (2003). State of the art on appli-
Zhao, J. X., X. Liang, F. Jiang, H. Xing, M. Zhu, R. Hou, Y. Zhang, X. cations, R&D and design rules for seismic isolation in China, Proc.
Lan, D. A. Rhoades, Y. Fukushima, et al. (2016). Ground-motion of the 8th World Seminar on Seismic Isolation, Energy Dissipation
prediction equations for subduction interface earthquakes in Japan and Active Vibration Control of Structures, Yerevan, Armenia,
using site class and simple geometric attenuation functions, Bull. 6–10 October 2003, 113–121.
Seismol. Soc. Am. 106, no. 4, 1518–1534.
Zhao, J. X., Q. S. Yang, K. W. Su, J. G. Liang, J. Zhou, H. Zhang, and X.
Yang (2019). Effects of earthquake source, path and site conditions Manuscript received 21 June 2019
on damping modification factor for the response spectrum of the Published online 21 January 2020

Volume 110 Number 2 April 2020 [Link] Bulletin of the Seismological Society of America • 665

Downloaded fromView
[Link]
publication stats
by Chengdu University of Technology, JohnZhao1000