You are on page 1of 16

Engineering Geology 292 (2021) 106241

Contents lists available at ScienceDirect

Engineering Geology
journal homepage: www.elsevier.com/locate/enggeo

Site-dependent ground-motion prediction equations and uniform hazard


response spectra
Jia-Cian Gao a, b, Chung-Han Chan a, c, Chyi-Tyi Lee a, b, *
a
Earthquake-Disaster & Risk Evaluation and Management Center (E-DREaM), National Central University, Taiwan
b
Graduate Institute of Applied Geology, National Central University, Taiwan
c
Department of Earth Sciences, National Central University, Taiwan

A R T I C L E I N F O A B S T R A C T

Keywords: A novel approach to construct site-dependent ground-motion prediction equations (GMPEs) was implemented to
Ground-motion prediction equation build a uniform hazard response spectra (UHRS) using a reservoir site in southern Taiwan as an example. This
Probabilistic seismic hazard analysis innovation minimizes the GMPE uncertainties to avoid over-estimations in a probabilistic seismic hazard analysis
Uniform hazard response spectra
(PSHA). We selected ground-motion records with similar site conditions from the stations within 70 km of the
Site-dependent
studied site to build a site-dependent ground-motion prediction model for crustal earthquakes in the region. This
site-dependent GMPE set obtains a significantly smaller scattering of residuals compared with the existing
regional GMPEs from the Taiwan data set. This indicates that a set of site-dependent GMPEs could provide a more
accurate and reasonable prediction of ground-motion levels for the design or safety evaluation of important
infrastructures.

1. Introduction always lower than that resulting from the total uncertainties in the
integration (Strasser et al., 2009; Abrahamson and Hollenback, 2012). It
Application of a traditional probabilistic seismic hazard analysis is thus tentatively assume that actually minimizing the total sigma could
(PSHA) is usually ergodic (Anderson and Brune, 1999), which is to say effectively reduce the hazard level.
the ground-motion variability from different earthquake sources recor­ Theoretically, since a PSHA focuses on one source for one site, only
ded by various sites would be included in the analysis. However, the the aleatory variability or the single-path variability should be used in
ground-motion variabilities are not always aleatory, but mixed with the integration. However, global GMPEs (e.g., Abrahamson et al., 2014;
significant epistemic components, including regional differences Boore et al., 2014; Campbell and Bozorgnia, 2014; Chiou and Youngs,
(Anderson et al., 2000). Such an approach causes the PSHA to be sen­ 2014; Idriss, 2014) are obtained by ground-motion data recorded at
sitive to the standard deviation (sigma) of median GMPE residuals, different sites in the world, which results in a large total sigma. Con­
especially for long return periods (Restrepo-Velez and Bommer, 2003; tracting to “full ergodic,” allowing regional differences in the GMPEs,
Bommer and Abrahamson, 2006). partially removes this ergodicity by translating part of the variability
In recent decades, researchers have attempted to decompose the into epistemic uncertainty through the modelling uncertainty in region-
components of ground-motion residuals in GMPEs to distinguish the specific adjustments (Kotha et al., 2016). Regional adjustments are
epistemic and aleatory parts. Only the aleatory part is included in the broken down into smaller geographical units, assuming there are
integration for exceedance probability, whereas the epistemic part repeatable source, path, and site effects for different locations, which, in
should be put into the logic tree (Chen and Tsai, 2002; Atkinson, 2006; principle, can be known and estimated (Lanzano et al., 2016). For site-
Morikawa et al., 2008; Al Atik et al., 2010; Anderson and Uchiyama, specific conditions, Ameri et al. (2017) presents the partially nonergodic
2011; Lin et al., 2011a; Rodriguez-Marek et al., 2011, 2013, 2014; Luzi approach to an infrastructure located in hard-rock site conditions with
et al., 2014; Sung and Lee, 2016; Lanzano et al., 2017). However, when Vs30 (average shear-wave velocity within upper 30 m depth) of about
we incorporate only aleatory uncertainties in the integration and put the 2000 m/s. The target site conditions are defined by shear-wave velocity
epistemic uncertainties in a logic tree, the resulting hazard level is not profiles and high-frequency attenuation based on site investigations and

* Corresponding author at: No. 300, Zhongda Rd., Zhongli District, Taoyuan City 32001, Taiwan.
E-mail address: ct@ncu.edu.tw (C.-T. Lee).

https://doi.org/10.1016/j.enggeo.2021.106241
Received 26 November 2020; Received in revised form 7 June 2021; Accepted 15 June 2021
Available online 18 June 2021
0013-7952/© 2021 Elsevier B.V. All rights reserved.
J.-C. Gao et al. Engineering Geology 292 (2021) 106241

Fig. 2. Spatial distributions of the crustal earthquakes used in this study (red
points) and alignments of active faults (black lines with red edges). (For
interpretation of the references to colour in this figure legend, the reader is
Fig. 1. Selected strong-motion stations for the collection of accelerograms in referred to the web version of this article.)
this study. Open squares are the TSMIP stations, while solid red squares indicate
selected stations. The blue star is the Mudan damsite. Inset shows the general
tectonic setting of Taiwan and the distribution of strong motion stations. (For engineering projects. Although simulation data may be included in the
interpretation of the references to colour in this figure legend, the reader is data set, the control factors used in producing the simulation data still
referred to the web version of this article.) need to be investigated.
Taking a similar approach to developing a deterministic site-
dependent response spectrum (Seed et al., 1976; Mohraz, 1976; New­
the exploitation of earthquake records at seismic stations in the target mark and Hall, 1982), we could construct site-dependent GMPEs for
site area. The (partially) nonergodic GMPEs based on different ap­ developing a uniform hazard response spectrum (UHRS). The site-
proaches for specific regions have been proposed (Dawood and dependent GMPEs were obtained by using strong-motion data with
Rodriguez-Marek, 2013; Kuehn and Scherbaum, 2016; Landwehr et al., similar site conditions and within a certain distance of the study site. The
2016; Lanzano et al., 2016; Ameri et al., 2017; Sedaghati and Pezeshk, resulting data set is much larger than that resulting from a single site,
2017; Abrahamson et al., 2019; Kuehn and Abrahamson, 2020; Kuehn causing the site-dependent GMPEs to be applicable.
et al., 2019; Kohrangi et al., 2021). In this study, to remove the regional as well as local differences, we
In summary, regional characteristics can be considered with respect used the concept of Seed et al. (1976)—the “site-dependent” data-
to the median of global GMPEs and the region-specific sigma. A regional selecting method. We selected strong-motion data recorded at sites
GMPE set (e.g., Lin et al., 2011b) could illustrate regional characteristics within 70 km of the Mudan damsite that have similar site conditions to
directly but include many site-to-site variabilities and path-to-path the damsite to build a set of local GMPEs. Based on the local GMPEs, we
variabilities, which would lead to a large total sigma as well. Howev­ developed a UHRS to catch the response spectral shape and the level for
er, when we compare a regional GMPE with the global ones, the regional the engineering design. Since such local GMPEs are applicable for a
GMPE may exhibit a smaller sigma and reflect the attenuation charac­ specific study site, we called them “site-dependent GMPEs.” We updated
teristics in a region (Lin and Lee, 2008; Lin et al., 2011b). It is expected the earthquake catalog and fault parameters in the vicinity of the study
that the sigma of a local or site-dependent GMPE may even be smaller region, and set up and implemented a site-dependent GMPE scheme in
than the sigma of a regional set and precisely reflect the attenuation this study.
characteristics of a small local area around the study site. Theoretically,
single-station GMPEs would be more applicable to a specific site than 2. Strong-motion database and data selection
site-dependent ones are. A single-station GMPE example obtained by
using the records from only one station has demonstrated the advantage Starting in 1991, the Seismology Center of the Central Weather Bu­
of reducing the sigma by 10% to 20% (Sung and Lee, 2009). It can be reau in Taiwan embarked on a seismic strong motion instrumentation
compared to the single-station sigma of Atkinson (2006) that used a program known as the Taiwan Strong-Motion Instrument Program
regional data set to develop a GMPE and then decomposed the variance (TSMIP) (Liu et al., 1999; Shin et al., 2013). Presently, more than 700
to achieve the single-station sigma. The sigma of a single-station GMPE strong-motion stations have recorded more than 200,000 digital accel­
is more advanced than the sigma calculated using Atkinson’s approach, erograms in three components. These records provide an excellent
because the sigma not only is smaller but also possesses less epistemic database for ground-motion attenuation studies (Liu et al., 1999; Chang
uncertainty. However, one disadvantage of this approach is the avail­ et al., 2001; Wu et al., 2001; Lin and Lee, 2008; Lin et al., 2011b).
ability of data, so the approach is still not applicable to actual We analyzed the database via the following procedure. For the

2
J.-C. Gao et al. Engineering Geology 292 (2021) 106241

Fig. 3. Parameters of the implemented strong-motion data: (a) magnitude versus distance, (b) magnitude versus depth, and (c) magnitude versus Vs30.

in Taiwan for Seismology (Kao et al., 2001), and the one summarized by
Wu et al. (2008). The closest distance from the rupture surface (Rrup) is
Table 1 the primary distance parameter implemented in this study, while the
List of the coefficients of the response spectral attenuation equations for hypocentral distance (Rhypo) will be used for the earthquakes without
analyzed period.
finite fault models (Beresnev and Atkinson, 1999). In addition, we
Periods c1 c2 c3 c4 c5 Sigma implemented RotD50, a period-dependent, component-based (not geo­
PGA − 2.822 1.076 − 1.777 0.3828 0.583 0.549 metric mean) measure of ground motion (Boore, 2010). Lee et al. (2001)
0.01 − 2.829 1.076 − 1.776 0.3783 0.583 0.549 conducted the site classification of Taiwan, and Lee and Tsai (2008)
0.02 − 2.827 1.034 − 1.719 0.3615 0.579 0.542 mapped the distribution of Vs30. According to the above information,
0.03 − 2.675 1.034 1.749 0.3992 0.568 0.546
the site type of the Mudan damsite (Vs30 = 529 m/s) belongs to site class

0.04 − 2.414 1.005 − 1.76 0.4492 0.548 0.548
0.05 − 2.032 0.989 − 1.809 0.5508 0.524 0.558 C, with a Vs30 of 360 m/s to 760 m/s.
0.06 − 1.565 0.973 − 1.87 0.6994 0.497 0.574 To reduce the uncertainties from source, path, and site, this study
0.09 0.059 0.979 − 2.16 1.4588 0.43 0.596 aims to analyze the site-dependent conditions and establish site-
0.1 0.725 1.003 − 2.307 1.8782 0.412 0.590 dependent GMPEs. We chose the records from the stations close to the
0.12 0.952 1.021 − 2.341 2.0453 0.413 0.599
0.15 0.913 1.057 − 2.331 2.0003 0.43 0.629
Mudan damsite (within 70 km) with similar site conditions (site class C)
0.17 0.294 1.067 − 2.209 1.5424 0.46 0.648 (red rectangles in Fig. 1). Based on the criteria, we selected a total of 338
0.2 − 0.504 1.086 − 2.056 1.0856 0.505 0.673 records from 61 earthquakes (red points in Fig. 2). Magnitudes of the
0.24 − 1.384 1.104 − 1.905 0.6943 0.556 0.648 selected earthquakes are in the 4.2 to 7.6 range; rupture distances are in
0.3 − 2.774 1.202 1.75 0.3093 0.6585 0.605
the 3.6 km to 300 km range, and hypocentral depths are in the 2.6 km to

0.36 − 3.867 1.313 − 1.681 0.1592 0.7457 0.622
0.4 − 4.451 1.361 − 1.642 0.1100 0.7882 0.643 33.9 km range (presented in Fig. 3a–c, respectively).
0.46 − 5.400 1.435 − 1.566 0.0575 0.865 0.672
0.5 − 5.913 1.487 − 1.543 0.0407 0.908 0.674 3. Site-dependent ground-motion prediction equation
0.6 − 7.080 1.565 − 1.439 0.0156 1.023 0.672
0.75 − 8.071 1.628 − 1.366 0.0069 1.119 0.685
0.85 − 8.866 1.701 − 1.328 0.0036 1.205 0.705
Because the strong-motion records are all taken from rock sites, we
1 − 9.641 1.736 − 1.264 0.0017 1.275 0.717 adopted a simple functional form in the regression analysis of shallow
1.5 − 12.174 1.948 − 1.109 0.0004 1.493 0.699 crustal earthquakes for the peak ground acceleration (PGA) and spec­
2 − 13.658 2.058 − 1.035 0.0001 1.655 0.675 trum accelerations (SAs) with periods from 0.01 s to 5.00 s:
3 − 15.487 2.161 − 0.917 0.0001 1.603 0.711
( ) ( )
4 − 16.248 2.195 − 0.909 0.0001 1.555 0.703 ln yij = C1 + C2 Mi + C3 ln Rij + C4 eC5 Mi + lnε, (1)
5 − 16.950 2.242 − 0.905 0.0001 1.511 0.689

where i is the ith earthquake, j is the jth station record of the ith earth­
quake, yij is the ground-motion level (in g), M is moment magnitude, R is
recorded waveforms, we first made base-line corrections and filtering to the closest distance (in km) to the rupture surface or hypocentral dis­
get true earthquake signals. Then we calculated the corresponding tance, and lnε is random error.
response spectrum (5% damping) for various natural periods. To assure Because the maximum rupture distance is 300 km, we do not include
data completeness, we included earthquakes with records from more the long-distance term (e.g., anelastic attenuation term). The target site
than 10 stations in this study. For magnitude determination, we first does not have the potential source to produce a great earthquake (M >
searched the databases of the Broadband Array in Taiwan for Seis­ 8.0), so we ignore the quadratic term in the magnitude scaling.
mology and the Harvard Centroid Moment Tensor to obtain moment As for the method of regression analysis, we adopted the maximum-
magnitude (MW). Alternatively, we obtained ML from the Central likelihood method to avoid regression bias due to uneven distribution of
Weather Bureau catalog and converted it to MW with the conversion earthquake records on distances. The regression approach is conducted
equation from Lin and Lee (2008). For focal mechanism, we accessed the through the “nlme” package implemented on the R platform (Pinheiro
databases of the Harvard Centroid Moment Tensor, the Broadband Array et al., 2020).

3
J.-C. Gao et al. Engineering Geology 292 (2021) 106241

Fig. 4. Results of the site-dependent GMPE for the PGA and SA1.0: (a) distance scaling for the PGA, (b) distance scaling for SA1.0, (c) magnitude scaling for the PGA,
and (d) magnitude scaling for SA1.0.

Based on the database and the approach mentioned above, we ob­ Fig. 5 presents the magnitude and distance scaling for various periods.
tained the coefficients of the response spectrum attenuation equations Periods around 0.2 s give the highest predicted shaking levels. The figure
with the site-dependent sigma (the standard deviation of lnε in Eq. (1)) also shows the correlation between adjacent periods, and the depend­
for each period and PGA (shown in Table 1). The distance scaling of the ability in the 28 periods from the PGA to SA5.0 in this study is apparent.
GMPEs is shown in Fig. 4a and b, and the magnitude scaling is shown in Residual analysis represents the difference between ground shaking
Fig. 4c and d. The larger the magnitude and/or the shorter the distance observations and predictions by GMPEs. We examined the residuals
is/are, the higher the level of the PGA or SA1.0 (SA for the period of 1 s) against distance, magnitude, and Vs30 in the site-dependent attenuation
is. Note that the ground shaking level saturates in the near field. That is, model. The attenuation model does not exhibit any systematic bias in
the larger the magnitude is, the smaller the difference in the shaking distance, magnitude, or Vs30, shown in Figs. 6–8, respectively. Histo­
level is. For larger magnitudes, attenuation rate decreases with distance, gram plots indicate that the residuals closely follow lognormal distri­
but actually the attenuation curve appears steeper at longer distances. butions in the PGA, SA0.3, SA1.0, and SA3.0 (shown in Fig. 9a–d,
However, the mean rate is lower in comparison to the almost linear respectively). Compared with regional (Chao et al., 2020; Lin et al.,
curves for smaller magnitudes, and this is consistent with the following 2011b; Phung et al., 2020) and global GMPEs (NGA-West2, M = 6.5, R
consideration that attenuation rate is higher for short period waves. = 30 km), the total sigma in our site-dependent GMPE is significantly

4
J.-C. Gao et al. Engineering Geology 292 (2021) 106241

100 100

101 101
SA (g)

102 102

103 103

0.2 0.5 1 2 5 10 20 50 100 200 400 0.2 0.5 1 2 5 10 20 50 100 200 400

Distance (km) Distance (km)


(a) (b)

100 100

101 101

102 102
SA (g)

PGA
103 SA 0.01 103
SA 0.02 SA 0.50
SA 0.03 SA 0.60
SA 0.04 SA 0.75
SA 0.17 SA 0.85
SA 0.05
104 SA 0.06 104 SA 0.20 SA 1.00
SA 0.24 SA 1.50
SA 0.09 SA 0.30 SA 2.00
SA 0.10 SA 0.36 SA 3.00
SA 0.12 SA 0.40 SA 4.00
105
SA 0.15 105 SA 0.46 SA 5.00

4 5 6 7 8 4 5 6 7 8

Moment Magnitude Moment Magnitude


(c) (d)

Fig. 5. Distance scaling at a magnitude of 6.0 for the site-dependent GMPE: (a) PGA and SA with periods from 0.01 s to 0.15 s, (b) SA with periods from 0.17 s to
5.00 s. Magnitude scaling at a distance of 50 km for the site-dependent GMPE: (c) PGA and SA with periods from 0.01 s to 0.15 s, (d) SA with periods from 0.17 s to
5.00 s.

lower for the response spectra at various periods (Fig. 10), suggesting an is subducting northward beneath the Eurasian plate. In central Taiwan,
improvement of the GMPE regression. there is an active collision zone between the Asiatic continent and the
The regional GMPEs by Lin et al. (2011b) have been divided into four Luzon Volcanic Arc located at the leading edge of the Philippine Sea
categories: soil site, rock site, hanging wall, and footwall. The four Plate (Fig. 1). In southern Taiwan, the South China Sea Subplate of the
equations are obtained independently, with the resulting sigma ranges Eurasian Plate is subducting eastward beneath the southern Taiwan
from 0.628 to 0.651 for the PGA, whereas the sigma of the GMPE setup collision zone, the Luzon Volcanic Arc, and the Philippine Sea Plate. The
with the site-dependent method obtained in this study is 0.549. The arc-continent collisions started in the Late Miocene and are still vigor­
comparison suggests the significant improvement of this site-dependent ously taking place. The whole region of Taiwan is characterized by
method is attributed not only to site classification. active crustal deformation (Bonilla, 1977; Yu et al., 1997) and frequent
earthquakes (Tsai, 1977).
4. Seismotectonics of study region and the Mudan Reservoir The Mudan Reservoir, completed in 1995, is located at the south­
Project ernmost part of Taiwan (Fig. 1). It includes a 65 m-high embankment
dam, a reservoir of 27 million m3, and a water catchment area of 69.2
Taiwan is located at the boundary between the Philippine Sea Plate km2. The reservoir and its vicinity are situated on the Miocene Mudan
and the Eurasian Plate. In northeastern Taiwan, the Philippine Sea Plate Formation, which consists of mudstone and thin alternations of

5
J.-C. Gao et al. Engineering Geology 292 (2021) 106241

Fig. 6. Distribution of residuals versus distance for the site-dependent GMPE: (a) PGA, (b) SA0.3, (c) SA1.0, and (d) SA3.0.

sandstone and mudstone. The strata are dipping at a high angle to the to conduct the seismic hazard analysis.
left bank. The nearest active fault is the Hengchun Fault located about 9
km to the west of the damsite. 5. PSHA model
In the design stage and in the first safety assessment of the Mudan
Reservoir, the southern extension of the Longitudinal Valley Fault This study used a mainshock catalog of Taiwan covering 115 years
(consisting of four segments: No. 43, 44, 45, and 48 indicated in Fig. 2) (1900 to 2014). We adopted the model by Cheng et al. (2015) and
was regarded as the control source for the maximum credible earth­ updated it to implement the PSHA. Four groups of earthquake sources
quake (MCE). The reports calculated a horizontal PGA of 0.41 g at the were identified for evaluating earthquake hazards in Taiwan: (1) crustal
damsite for the MCE. In the second safety assessment, the report faults, (2) crustal areal sources, (3) subduction intraslab earthquake
changed to use the Hengchun Fault as a control source for the MCE and sources, and (4) subduction interface earthquake sources. The fault-
concluded a horizontal PGA of 0.77 g. The analysis of this study was source model was derived from 49 active faults that were identified
completed during the third safety assessment of the Mudan Dam and from cutting through Holocene-Late Pleistocene deposits or expressing
Reservoir in 2019. We updated our data and followed the Taiwan gov­ significant geomorphic features of recent movements and other sup­
ernment’s “Technical Directions for Hydraulic Structures Inspection and porting data (Lee, 1999). Slip rates were summarized from some
Safety Evaluation (Reservoir and Water-Conveying Structures)” of 2018 trenching results and estimated from dated geomorphic surfaces (Cheng

6
J.-C. Gao et al. Engineering Geology 292 (2021) 106241

Fig. 7. Distribution of residuals versus magnitude for the site-dependent GMPE: (a) PGA, (b) SA0.3, (c) SA1.0, and (d) SA3.0.

et al., 2015). earthquakes and fault rupture length (Wells and Coppersmith, 1994)
The closest active fault to the Mudan damsite is the Hengchun Fault. was adopted.
A summary of the most related and updated active fault parameters in
the present study is listed in Table 2. Note that the slip rates of the 6. Site-dependent uniform hazard response spectrum
Hengchun Fault were summarized from three references. Cheng et al.
(2015) proposed a minimum slip-rate of 4.2 mm/yr and a median slip- Annual exceedance probability of a specific shaking level at each
rate of 7.5 mm/yr. Wang et al. (2016) reported a minimum slip-rate of structural period is calculated for the damsite by using the computer
5.86 mm/yr and a median slip-rate of 6.15 mm/yr. Finally, an unpub­ code, HAZ (Abrahamson, N., personal communication), and seismic
lished report in 2015 from the Central Geological Survey of Taiwan hazard curves are plotted. Since there are still many epistemic compo­
provided a minimum slip-rate of 1.5 mm/yr, a median slip-rate of 4 mm/ nents mixed in the total variance, we considered only two sigmas for the
yr, and a maximum slip-rate of 6.5 mm/yr. Using the above-mentioned ground-motion prediction equation. Fig. 11 presents the median hazard
parameters, we considered the slip rates of 4.0, 6.0, and 8.0 mm/yr with curves with uncertainties in the 5th and 95th percentiles. The hazards
corresponding weightings of 0.2, 0.6, and 0.2, respectively. This study contributed from various sources around the damsite were also
adopted a characteristic earthquake model (Youngs and Coppersmith, presented.
1985) for its fault sources. As for the determination of magnitude of each Based on the hazard curves, the Hengchun Fault is the control fault
fault source, an empirical relationship between magnitude of for hazards in short periods (Fig. 11a and b), whereas the long-period

7
J.-C. Gao et al. Engineering Geology 292 (2021) 106241

Fig. 8. Distribution of residuals versus Vs30 for the site-dependent GMPE: (a) PGA, (b) SA0.3, (c) SA1.0, and (d) SA3.0.

hazard levels (> 1 s) are dominated by both the Hengchun Fault and damsite and only for crustal sources of earthquakes. Considering the
subduction zone interface sources (Fig. 11c and d). A secondary magnitude and distance ranges of the available data set (Fig. 3), this set
contribution for hazards is from the crustal areal sources (purple curves of site-dependent GMPEs is suitable for a moment magnitude between
in Fig. 11). According to the Taiwanese assessment standard of 2018, the 4.0 and 7.6 and rupture distance less than 300 km.
Mudan Reservoir is ranked high in terms of potential risk. The design
basis earthquake (DBE) and the operational basis earthquake (OBE) 7. Discussion and conclusions
were also analyzed in the 2019 Mudan Reservoir safety assessment
project. The ground-motion levels in the PGA for the return periods of For the selected range of site-dependent data, we initially selected
950 years (DBE) and 475 years (OBE) are 0.40 g and 0.34 g, respectively. the sites within 50 km of the investigated damsite. Due to insufficient
The seismic hazard level with a return period of 10,000 years, which can data for analysis, we increased the search radius to 70 km. To check the
be regarded as the maximum design earthquake, is 0.63 g in PGA at the range of spatial coherence, we selected the larger earthquake, 5 May
damsite. The UHRS for the four different return periods are shown in 2004 M6.4, that occurred near the target site. We calculated the ln(PGA)
Fig. 12. and ln(SA1.0) that were recorded by the site-dependent stations for
This set of site-dependent GMPEs is suitable for use on the Mudan calculating a semivariogram. We used the Gaussian model to fit the

8
J.-C. Gao et al. Engineering Geology 292 (2021) 106241

80 Period = 0.0 80 Period = 0.3


Mean = 0 Mean = 0
Total Sigma = 0.549 Total Sigma = 0.605

60 60
Frequency

Frequency
40 40

20 20

0 0

−3 −2 −1 0 1 2 3 −3 −2 −1 0 1 2 3
Residual Residual
(a) (b)

80 Period = 1.0 80 Period = 3.0


Mean = 0 Mean = 0
Total Sigma = 0.717 Total Sigma = 0.711

60 60
Frequency

Frequency

40 40

20 20

0 0

−3 −2 −1 0 1 2 3 −3 −2 −1 0 1 2 3
Residual Residual
(c) (d)

Fig. 9. Histogram of residuals for the site-dependent GMPE: (a) PGA, (b) SA0.3, (c) SA1.0, and (d) SA3.0.

semivariance. The results show the 70 km search radius is still within the regional records (in Taiwan) for our GMPE regression with magnitudes
range of spatial coherence (Fig. 13). larger than 6.5 and rupture distances less than 40 km (Fig. 14). Based on
The robustness of GMPEs plays an important role in the application the residual analyses for the PGA and SA1.0, there is no systematic bias,
of a PSHA. In this study, we implemented a simple functional form for so we can conclude that the site-dependent GMPE for the Mudan dam­
regression of the GMPE. This is because our strong-motion data sets were site is robust (represented in Figs. 15 and 16, respectively). Also, these
gathered from the same site class (soft rock) in a small region near the equations dramatically decreased the variances of ground-motion pre­
target site, so the deviation from site effect could be minimized. During diction and presented ground-motion level more precisely. The results
the study, we tested different forms for the source term and the path show that the site-dependent GMPE obtains significantly lower vari­
term and compared the effectiveness of different terms in the functional ability as compared with the existing regional attenuation equations
form. At last, we found the simple form can best fit the data with the from the Taiwan data set.
smallest residuals. We also examined the robustness of the site- We further validated our site-dependent GMPEs and other GMPEs by
dependent GMPE by including additional data. We implemented some comparing with observations during the 1999 MW7.6 Chi-Chi, Taiwan,

9
J.-C. Gao et al. Engineering Geology 292 (2021) 106241

Fig. 10. Comparison of the total sigma of the site-dependent GMPE in this study with those of regional and global GMPEs. The regional GMPEs are from Lin et al.
(2011b), Chao et al. (2020) and Phung et al. (2020). HR: hanging wall and rock site. FR: footwall and rock site. The global GMPEs are from NGA-West2 (M = 6.5, R =
30 km).

Table 2
List of the fault sources’ parameters in southern Taiwan.
NO. Fault name Dip Fault type Fault length (km) Rupture depth Slip rate (mm/yr) (weighting) Characteristic magnitude, Mu
(km) (weighting)

17 Chelungpu 40E T1 90 0–20 7.5 (0.2) 15.0 (0.6) 22.5 (0.2) 7.5 (0.2) 7.7 (0.6) 7.9 (0.2)
18 Changhua 30E T1 85 2–15 7.5 (0.2) 15.0 (0.6) 22.5 (0.2) 7.4 (0.2) 7.6 (0.6) 7.8 (0.2)
19 Tamopu- 45E T1 70 0–15 0.4 (0.2) 0.8 (0.6) 1.2 (0.2) 7.2 (0.2) 7.4 (0.6) 7.6 (0.2)
Hsuangtung
20 Shuilikeng 50E T1 32 0–15 0.3 (0.2) 0.6 (0.6) 0.9 (0.2) 6.8 (0.2) 7.0 (0.6) 7.2 (0.2)
21 Chenyulanchi 50E T1 36 0–15 0.4 (0.2) 0.8 (0.6) 1.2 (0.2) 6.8 (0.2) 7.0 (0.6) 7.2 (0.2)
22 Chiuchiungkeng 35E T1 24 0–15 9.5 (0.2) 10.0 (0.6) 10.5 (0.2) 6.8 (0.2) 7.0 (0.6) 7.2 (0.2)
23 Kukeng 90 S4 10 0–15 1.5 (0.2) 3.0 (0.6) 4.5 (0.2) 6.1 (0.2) 6.3 (0.6) 6.5 (0.2)
24 Meishan 90 D2 14 0–15 3.0 (0.2) 6.0 (0.6) 9.0 (0.2) 6.3 (0.2) 6.5 (0.6) 6.7 (0.2)
24BT5 Chaiyi blind thrust 30E T1 36 2–15 6.0 (0.2) 12.0 (0.6) 18.0 (0.2) 7.0 (0.2) 7.2 (0.6) 7.4 (0.2)
25 Chukou 40E T1 70 0–15 5.0 (0.2) 10.0 (0.6) 15.0 (0.2) 7.3 (0.2) 7.5 (0.6) 7.7 (0.2)
26 + 27 Muchiliao + Liuchia 35SE T1 30 0–15 4.0 (0.2) 8.0 (0.6) 12.0 (0.2) 6.9 (0.2) 7.1 (0.6) 7.3 (0.2)
28 Tsochen 90 S4 12 0–15 1.25 (0.2) 2.5 (0.6) 3.75 (0.2) 6.2 (0.2) 6.4 (0.6) 6.6 (0.2)
29 Hsinhua 90 D2 12 0–15 2.5 (0.2) 5.0 (0.6) 7.5 (0.2) 6.2 (0.2) 6.4 (0.6) 6.6 (0.2)
30 Houchiali 60 W T1 12 0–15 2.5 (0.2) 5.0 (0.6) 7.5 (0.2) 6.2 (0.2) 6.4 (0.6) 6.6 (0.2)
30BT5 Tainan blind thrust 30E T1 36 2–15 5.0 (0.2) 10.0 (0.6) 10.5 (0.2) 7.0 (0.2) 7.2 (0.6) 7.4 (0.2)
31 Hsiaokangshan 50SE R3 12 0–15 0.8 (0.2) 1.5 (0.6) 2.3 (0.2) 6.3 (0.2) 6.5 (0.6) 6.7 (0.2)
32 Chishan 45E R3 60 0–15 1.5 (0.2) 3.0 (0.6) 4.5 (0.2) 7.1 (0.2) 7.3 (0.6) 7.5 (0.2)
33 Yueikangshan 80S R3&S4 12 0–10 0.5 (0.2) 1.0 (0.6) 1.5 (0.2) 6.0 (0.2) 6.2 (0.6) 6.5 (0.2)
34 Yuchang-Lunchuan 50E R3 50 0–15 3.0 (0.2) 5.0 (0.6) 7.0 (0.2) 7.0 (0.2) 7.2 (0.6) 7.4 (0.2)
35 Yenwu-Fenshan 50E R3 21 0–15 0.8 (0.2) 1.5 (0.6) 2.3 (0.2) 6.6 (0.2) 6.8 (0.6) 7.0 (0.2)
36 Liukuei 50E R3 20 0–15 0.2 (0.2) 0.5 (0.6) 0.8 (0.2) 6.5 (0.2) 6.7 (0.6) 7.0 (0.2)
37a Chaochou (north) 50E R3 55 0–20 1.0 (0.2) 3.0 (0.6) 5.0 (0.2) 7.2 (0.2) 7.4 (0.6) 7.6 (0.2)
37b Chaochou (south) 50E R3 40 0–20 1.0 (0.2) 3.0 (0.6) 5.0 (0.2) 7.0 (0.2) 7.2 (0.6) 7.4 (0.2)
38 Hengchun 50E R3 40 0–20 4.0 (0.2) 6.0 (0.6) 8.0 (0.2) 7.0 (0.2) 7.2 (0.6) 7.4 (0.2)
43 Ueimei 50E R3&S4 45 0–30 10.0 (0.2) 20.0 (0.6) 30.0 (0.2) 7.3 (0.2) 7.5 (0.6) 7.7 (0.2)
44 Yuli 50E R3&S4 48 0–30 8.0 (0.2) 16.0 (0.6) 24.0 (0.2) 7.3 (0.2) 7.5 (0.6) 7.7 (0.2)
45 Chihshang 50E R3&S4 30 0–30 10.0 (0.2) 20.0 (0.6) 30.0 (0.2) 7.1 (0.2) 7.3 (0.6) 7.5 (0.2)
46 Yuli West 50 W R3&S4 40 0–25 5.0 (0.2) 10.0 (0.6) 10.5 (0.2) 7.1 (0.2) 7.3 (0.6) 7.5 (0.2)
47 Luyeh 50E R3 18 0–25 4.0 (0.2) 6.0 (0.6) 8.0 (0.2) 6.7 (0.2) 6.9 (0.6) 7.1 (0.2)
48 Lichi 50E R3&S4 24 0–25 8.0 (0.2) 16.0 (0.6) 24.0 (0.2) 6.9 (0.2) 7.1 (0.6) 7.3 (0.2)
49 Chimei 50E R3&S4 25 0–30 2.0 (0.2) 4.0 (0.6) 6.0 (0.2) 7.0 (0.2) 7.2 (0.6) 7.4 (0.2)

Note: 1: Thrust, 2: Dextral, 3: Reverse, 4: Sinistral, 5: Blind thrust.

10
J.-C. Gao et al. Engineering Geology 292 (2021) 106241

0
10

(a) (b)

-1
Total hazard
10 5% and 95%
Hengchun fault
Annual Probability of Exceedance Chaochou fault
Areal sources
-2 South-Interface
10
South-Intraslab

475yr 475yr

-3
10 950yr 950yr

2475yr 2475yr

-4
10 10000yr
10000yr

10 -50.0 0.2 0.4 0.6 0.8 1.0 1.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2
PGA(g) SA0.3(g)

0
10

(c) (d)

-1
10
Annual Probability of Exceedance

-2
10

475yr 475yr

-3
10 950yr 950yr

2475yr 2475yr

-4
10
10000yr 10000yr

10 -50.0 0.2 0.4 0.6 0.8 1.0 1.2 0.0 0.2 0.4
SA1.0(g) SA3.0(g)

Fig. 11. Hazard curves and de-aggregation for each seismogenic source at the Mudan damsite: (a) PGA, (b) SA0.3, (c) SA1.0, and (d) SA3.0.

earthquake (Fig. 17). The implemented regional GMPEs are from Lin comparison of median spectra of the eight GMPEs. The median spectra
et al. (2011b), Chao et al. (2020) and Phung et al. (2020), and the global of the new site-dependent GMPEs have lower levels of acceleration,
GMPEs are from NGA-West2 with regional adjustment (Abrahamson especially for short periods (< 1 s). A comparison of sigmas is provided
et al., 2014; Boore et al., 2014; Campbell and Bozorgnia, 2014; Chiou in Fig. 10, where the sigmas of the site-dependent GMPEs show signif­
and Youngs, 2014; Idriss, 2014). We used the observations in the Chi-Chi icant lowest values. Fig. 19 shows the comparison of the sigmas of a
earthquake, assumed the site condition in Vs30 of 560 m/s (the mean single-station GMPE from Sung and Lee (2019) in the black points. The
value of C class) and calculated Ztor, Z1.0 and Z2.5 using the transform red point is the site-dependent sigma from this study. The blue points are
function by Kaklamanos et al. (2011), set the region flag to “Taiwan” sigmas for the regional GMPEs from Lin et al. (2011b). The results show
without the hanging wall effect and basin effect. The regional GMPE that the site-dependent sigma is roughly equal to or a bit larger than the
exhibits clearly a lower level of acceleration as compared with the global sigma of the single-station GMPE in the PGA. The sigma of single-station
ones, and the site-dependent GMPE shows the lowest level of accelera­ GMPE is sensitive to the number of ground-motion recordings, so the
tion in the near field. The site-dependent GMPE could also fit with the sigma would be high when the number is smaller than 20. The site-
Chi-Chi earthquake data in the PGA and SA1.0. Fig. 18 shows the dependent method in this study will avoid this issue and be applicable

11
J.-C. Gao et al. Engineering Geology 292 (2021) 106241

Fig. 12. UHRS developed in this study for different return periods.

Fig. 13. Semivariogram of the ln(PGA) and ln(SA1.0) that were recorded by the site-dependent stations for the 5 May 2004 M6.4 earthquake.

for real engineering projects. (2008).


The site-dependent GMPE is validated to be applicable. The pro­ Ameri et al. (2017) used partially nonergodic GMPEs to analyze the
cedure is straightforward and implementable. Using the site-dependent site term’s adjustment and limited their work to a specific region in
GMPE in the PSHA, we got a significantly lower UHRS in each structural France. Similar to our outcomes, their PSHA results showed a 50%
period than the one built using regional GMPEs. A design response to60% reduction in the hazard level for short periods for the MCE.
spectrum developed in the design stage by the deterministic method for Since the design spectrum was developed using a traditional statis­
MCE is shown as the black curve in Fig. 20 for comparison. Comparing tical method from strong-motion records of medium and small earth­
the new site-dependent UHRS to the original design spectrum, the new quakes, an update using a new method and a new data set is necessary.
spectrum is slightly higher than the old one at low and medium periods The new site-dependent UHRS was naturally and reasonably adopted in
(< 2 s). The UHRS derived from the regional GMPE, however, is the third safety evaluation project of the Mudan Dam and Reservoir.
significantly higher than that of the site-dependent UHRS and the This UHRS was further used as a target spectrum to generate a time
original design spectrum. For periods longer than 1 s, the results of site- history of artificial synthetic acceleration by using the spectral matching
dependent and regional UHRSs are identical due to the contribution method (Al Atik and Abrahamson, 2010) with suitable seismograms that
from the subduction sources based on the same GMPE by Lin and Lee were recorded at similar condition sites and of suitable magnitudes and

12
J.-C. Gao et al. Engineering Geology 292 (2021) 106241

Fig. 14. Additional data used for checking purposes (red ones): (a) magnitude versus distance, (b) magnitude versus depth, and (c) magnitude versus Vs30. (For
interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 15. Residuals between observations and predictions in the PGA for all records with additional data are in red. (For interpretation of the references to colour in
this figure legend, the reader is referred to the web version of this article.)

Fig. 16. Residuals between observations and predictions in SA1.0 for all records with additional data are in red. (For interpretation of the references to colour in this
figure legend, the reader is referred to the web version of this article.)

13
J.-C. Gao et al. Engineering Geology 292 (2021) 106241

Fig. 17. Comparison of distance scaling among the site-dependent GMPE, regional GMPEs, and global GMPEs: (a) PGA and (b) SA1.0 for the Chi-Chi earthquake of
MW 7.6. Note: Ztor = Depth to top of coseismic rupture. Z1.0 = Depth to Vs equal to 1.0 km/s. Z2.5 = Depth to Vs equal to 2.5 km/s.

R = 10(km), Mw = 7, Ztor = 5.36(km), Z1.0 = 0.13(km) R = 30(km), Mw = 7, Ztor = 5.36(km), Z1.0 = 0.13(km)
Z2.5 = 1(km), Vs30 = 560(m/s), Dep = 9.7(km) Z2.5 = 1(km), Vs30 = 560(m/s), Dep = 9.7(km)
100 100
SA (g)

SA (g)

101 This study 101


Lin2011b_HR
Lin2011b_FR
ASK14
BSSA14
CB14
CY14
I14
Chao2020
Phung2020
102 102

0.01 0.03 0.1 0.3 1 3 5 0.01 0.03 0.1 0.3 1 3 5


Period (sec) Period (sec)
(a) (b)

Fig. 18. Comparison of median response spectrum predicted by the site-dependent GMPE, regional GMPEs, and global GMPEs: (a) MW 7 with distance 10 km, (b) MW
7 with distance 30 km.

distances indicated from de-aggregation analysis after the PSHA. The strong-motion stations not only at the engineering site but also at
completed acceleration time histories were used as input ground mo­ nearby locations of the same site class would be necessary during the
tions required for the dynamic analysis of the dam. feasibility study stage of major infrastructures (e.g., nuclear power
This study demonstrates the effectiveness of implementing site- plants and high dams).
dependent GMPEs for crustal earthquakes. Further developing a set of
site-dependent GMPEs for subduction zone earthquakes is necessary to Declaration of competing interest
estimate more precise hazard levels at periods longer than 1 s (Fig. 20).
Since the development of a site-dependent GMPE requires many strong- No conflict of Interest.
motion seismograms recorded near the target site, installing free-field

14
J.-C. Gao et al. Engineering Geology 292 (2021) 106241

to express our heartfelt thanks to them.

References

Abrahamson, N.A., Hollenback, J.C., 2012. Application of single station sigma ground
motion prediction equations in practice. In: Proceedings of the 15th World
Conference on Earthquake Engineering, Lisbon, Portugal paper 2536.
Abrahamson, N.A., Silva, W.J., Kamai, R., 2014. Summary of the ASK14 ground motion
relation for active crustal regions. Earthquake Spectra 30 (3), 1025–1055. https://
doi.org/10.1193/070913EQS198M.
Abrahamson, N.A., Kuehn, N.M., Walling, M., Landwehr, N., 2019. Probabilistic seismic
hazard analysis in California using nonergodic ground-motion models probabilistic
seismic hazard analysis in california using nonergodic ground-motion models. Bull.
Seismol. Soc. Am. 109 (4), 1235–1249.
Al Atik, L., Abrahamson, N.A., 2010. An improved method for nonstationary spectral
matching. Earthquake Spectra 26 (3), 601–617. https://doi.org/10.1193/
1.3459159.
Al Atik, L., Abrahamson, N.A., Bommer, J., Scherbaum, F., Cotton, F., Kuehn, N., 2010.
The variability of ground-motion prediction models and its components. Seismol.
Res. Lett. 81 (5), 794–801. https://doi.org/10.1785/gssrl.81.5.794.
Ameri, G., Hollender, F., Perron, V., Martin, C., 2017. Site-specific partially nonergodic
PSHA for a hard-rock critical site in southern France: adjustment of ground motion
prediction equations and sensitivity analysis. Bull. Earthq. Eng. 15 (10), 4089–4111.
Anderson, J.G., Brune, J.N., 1999. Probabilistic seismic hazard analysis without the
ergodic assumption. Seismol. Res. Lett. 70 (1), 19–28. https://doi.org/10.1785/
gssrl.70.1.19.
Anderson, J.G., Uchiyama, Y., 2011. A methodology to improve ground-motion
prediction equations by including path corrections. Bull. Seismol. Soc. Am. 101 (4),
1822–1846. https://doi.org/10.1785/0120090359.
Anderson, J.G., Brune, J.N., Anooshehpoor, R., Ni, S.D., 2000. New ground motion data
and concepts in seismic hazard analysis. Curr. Sci. 79 (9), 1278–1290.
Fig. 19. Comparison of the total sigmas. Each black point is the total sigma of a Atkinson, G.M., 2006. Single-station sigma. Bull. Seismol. Soc. Am. 96 (2), 446–455.
https://doi.org/10.1785/0120050137.
single-station GMPE from Sung et al. (2019). The blue points are the regional
Beresnev, I.A., Atkinson, G.M., 1999. Generic finite-fault model for ground-motion
sigmas from Lin et al. (2011b). The red point is the site-dependent sigma in this prediction in eastern North America. Bull. Seismol. Soc. Am. 89 (3), 608–625.
study. (For interpretation of the references to colour in this figure legend, the Bommer, J.J., Abrahamson, N.A., 2006. Why do modern probabilistic seismic-hazard
reader is referred to the web version of this article.) analyses often lead to increased hazard estimates? Bull. Seismol. Soc. Am. 96 (6),
1967–1977. https://doi.org/10.1785/0120060043.
Bonilla, W.B., 1977. Summary of Quaternary faulting and elevation changes in Taiwan.
Mem. Geol. Soc. Chaina 2, 43–55.
Boore, D.M., 2010. Orientation-independent, nongeometric-mean measures of seismic
intensity from two horizontal components of motion. Bull. Seismol. Soc. Am. 100
(4), 1830–1835. https://doi.org/10.1785/0120090400.
Boore, D.M., Stewart, J.P., Seyhan, E., Atkinson, G.M., 2014. NGA-West2 equations for
predicting PGA, PGV, and 5% damped PSA for shallow crustal earthquakes.
Earthquake Spectra 30 (3), 1057–1085. https://doi.org/10.1193/070113EQS184M.
Campbell, K.W., Bozorgnia, Y., 2014. NGA-West2 ground motion model for the average
horizontal components of PGA, PGV, and 5% damped linear acceleration response
spectra. Earthquake Spectra 30 (3), 1087–1115. https://doi.org/10.1193/
062913EQS175M.
Chang, T.Y., Cotton, F., Angelier, J., 2001. Seismic attenuation and peak ground
acceleration in Taiwan. Bull. Seismol. Soc. Am. 91 (5), 1229–1246. https://doi.org/
10.1785/0120000729.
Chao, S.H., Chiou, B., Hsu, C.C., Lin, P.S., 2020. A horizontal ground-motion model for
crustal and subduction earthquakes in Taiwan. Earthquake Spectra 36 (2), 463–506.
Chen, Y.H., Tsai, C.C.P., 2002. A new method for estimation of the attenuation
relationship with variance components. Bull. Seismol. Soc. Am. 92 (5), 1984–1991.
https://doi.org/10.1785/0120010205.
Cheng, C.T., Hsieh, P.S., Lin, P.S., Yen, Y.T., Chan, C.H., 2015. Probability seismic hazard
mapping of Taiwan. Encycl. Earthq. Eng. 10, 1–25. https://doi.org/10.1007/978-3-
642-36197-5_100-1.
Chiou, B.S.J., Youngs, R.R., 2014. Update of the Chiou and Youngs NGA model for the
average horizontal component of peak ground motion and response spectra.
Earthquake Spectra 30 (3), 1117–1153. https://doi.org/10.1193/072813EQS219M.
Dawood, H.M., Rodriguez-Marek, A., 2013. A method for including path effects in
ground-motion prediction equations: An example using the Mw 9.0 Tohoku
earthquake aftershocks. Bull. Seismol. Soc. Am. 103 (2B), 1360–1372.
Idriss, I.M., 2014. An NGA-West2 empirical model for estimating the horizontal spectral
Fig. 20. This study’s 10,000-year UHRS using site-dependent GMPE (red values generated by shallow crustal earthquakes. Earthquake Spectra 30 (3),
1155–1177. https://doi.org/10.1193/070613EQS195M.
curve), that developed using regional GMPEs (blue curve), and the original
Kaklamanos, J., Baise, L.G., Boore, D.M., 2011. Estimating unknown input parameters
design spectrum (black curve). (For interpretation of the references to colour in when implementing the NGA ground-motion prediction equations in engineering
this figure legend, the reader is referred to the web version of this article.) practice. Earthquake Spectra 27 (4), 1219–1235.
Kao, H., Liu, Y.H., Jian, P.R., 2001. Source parameters of regional earthquakes in
Taiwan: January-December, 1997. Terr. Atmos. Ocean. Sci. 12 (2), 431–439.
Acknowledgements Kohrangi, M., Kotha, S.R., Bazzurro, P., 2021. Impact of partially non-ergodic site-
specific probabilistic seismic hazard on risk assessment of single buildings.
We extend our sincere thanks to the Seismic Information Center, Earthquake Spectra 37 (1), 409–427.
Kotha, S.R., Bindi, D., Cotton, F., 2016. Partially non-ergodic region specific GMPE for
Central Weather Bureau, and Ministry of Transportation and Commu­ Europe and Middle-East. Bull. Earthq. Eng. 14 (4), 1245–1263.
nications, Taiwan for providing us with strong-motion data. This work Kuehn, N.M., Abrahamson, N.A., 2020. Spatial correlations of ground motion for non-
was supported by the Water Resources Agency, Ministry of Economic ergodic seismic hazard analysis. Earthq. Eng. Struct. Dyn. 49 (1), 4–23.
Kuehn, N.M., Scherbaum, F., 2016. A partially non-ergodic ground-motion prediction
Affairs, Taiwan, and Liming Engineering Consultants. In the research equation for Europe and the Middle East. Bull. Earthq. Eng. 14 (10), 2629–2642.
process, many people helped us in innumerable ways, and we also want

15
J.-C. Gao et al. Engineering Geology 292 (2021) 106241

Kuehn, N.M., Abrahamson, N.A., Walling, M.A., 2019. Incorporating nonergodic path Rodriguez-Marek, A., Montalva, G.A., Cotton, F., Bonilla, F., 2011. Analysis of single-
effects into the NGA-West2 ground-motion prediction equations. Bull. Seismol. Soc. station standard deviation using the KiK-net data. Bull. Seismol. Soc. Am. 101 (3),
Am. 109 (2), 575–585. 1242–1258. https://doi.org/10.1785/0120100252.
Landwehr, N., Kuehn, N.M., Scheffer, T., Abrahamson, N.A., 2016. A nonergodic ground- Rodriguez-Marek, A., Cotton, F., Abrahamson, N.A., Akkar, S., Al Atik, L., Edwards, B.,
motion model for California with spatially varying coefficients. Bull. Seismol. Soc. Montalva, G.A., Dawood, H.M., 2013. A model for single-station standard deviation
Am. 106 (6), 2574–2583. using data from various tectonic regions. Bull. Seismol. Soc. Am. 103 (6),
Lanzano, G., D’Amico, M., Felicetta, C., Puglia, R., Luzi, L., Pacor, F., Bindi, D., 2016. 3149–3163. https://doi.org/10.1785/0120130030.
Ground-motion prediction equations for region-specific probabilistic seismic-hazard Rodriguez-Marek, A., Rathje, E.M., Bommer, J.J., Scherbaum, F., Stafford, P.J., 2014.
analysis. Bull. Seismol. Soc. Am. 106 (1), 73–92. Application of single-station sigma and site-response characterization in a
Lanzano, G., D’Amico, M., Felicetta, C., Luzi, L., Puglia, R., 2017. Update of the single- probabilistic seismic-hazard analysis for a new nuclear site. Bull. Seismol. Soc. Am.
station sigma analysis for the Italian strong-motion stations. Bull. Earthq. Eng. 15 104 (4), 1601–1619. https://doi.org/10.1785/0120130196.
(6), 2411–2428. Sedaghati, F., Pezeshk, S., 2017. Partially nonergodic empirical ground-motion models
Lee, C.T., 1999. Neotectonics and active faults in Taiwan. In: Workshop on Disaster for predicting horizontal and vertical PGV, PGA, and 5% damped linear acceleration
Prevention/Management and Green Technology, Foster City, California, U.S.A., response spectra using data from the Iranian plateau. Bull. Seismol. Soc. Am. 107 (2),
pp. 61–74. 934–948.
Lee, C.T., Tsai, B.R., 2008. Mapping Vs30 in Taiwan. Terr. Atmos. Ocean. Sci. 19 (6), Seed, H.B., Ugas, C., Lysmer, J., 1976. Site-dependent spectra for earthquake-resistant
671–682. design. Bull. Seismol. Soc. Am. 66 (1), 221–243.
Lee, C.T., Cheng, C.T., Liao, C.W., Tsai, Y.B., 2001. Site classification of Taiwan free-field Shin, T.C., Chang, C.H., Pu, H.C., Lin, H.W., Leu, P.L., 2013. The geophysical database
strong-motion stations. Bull. Seismol. Soc. Am. 91 (5), 1283–1297. https://doi.org/ management system in Taiwan. Terr. Atmos. Ocean. Sci. 24 (1), 11–18. https://doi.
10.1785/0120000736. org/10.3319/TAO.2012.09.20.01(T).
Lin, P.S., Lee, C.T., 2008. Ground-motion attenuation relationships for subduction-zone Strasser, F.O., Abrahamson, N.A., Bommer, J.J., 2009. Sigma: issues, insights, and
earthquakes in northeastern Taiwan. Bull. Seismol. Soc. Am. 98 (1), 220–240. challenges. Seismol. Res. Lett. 80 (1), 40–56. https://doi.org/10.1785/gssrl.80.1.40.
https://doi.org/10.1785/0120060002. Sung, C.H., Lee, C.T., 2009. Single site strong-motion attenuation relationship. In:
Lin, P.S., Chiou, B.S.J., Abrahamson, N.A., Walling, M., Lee, C.T., Cheng, C.T., 2011a. Proceeding of the Next Generation of Research on Earthquake-Induced Landslides -
Repeatable source, site, and path effects on the standard deviation for empirical An International Conference in Commemoration of 10th Anniversary of the Chi-Chi
ground-motion prediction models. Bull. Seismol. Soc. Am. 101 (5), 2281–2295. Earthquake, Jhongli, Taiwan, pp. 284–292.
https://doi.org/10.1785/0120090312. Sung, C.H., Lee, C.T., 2016. A new methodology for quantification of the systematic path
Lin, P.S., Lee, C.T., Cheng, C.T., 2011b. Response spectral attenuation relations for effects on ground-motion variability. Bull. Seismol. Soc. Am. 106 (6), 2796–2810.
shallow crustal earthquakes in Taiwan. Eng. Geol. 121 (3–4), 150–164. https://doi. https://doi.org/10.1785/0120160038.
org/10.1016/j.enggeo.2011.04.019. Sung, C.H., Lee, C.T., 2019. Improvement of the quantification of epistemic uncertainty
Liu, K.S., Shin, T.C., Tsai, Y.B., 1999. A free-field strong-motion network in Taiwan: using single-station ground-motion prediction equations. Bull. Seismol. Soc. Am. 109
TSMIP. Terr. Atmos. Ocean. Sci. 10 (2), 377–396. (4), 1358–1377.
Luzi, L., Bindi, D., Puglia, R., Pacor, F., Oth, A., 2014. Single-station sigma for Italian Tsai, Y.B., 1977. Tectonic implications of the seismicity in the Taiwan region. Mem. Geol.
strong-motion stations. Bull. Seismol. Soc. Am. 104 (1), 467–483. https://doi.org/ Soc. China 2, 13–41.
10.1785/0120130089. Wang, Y.J., Chan, C.H., Lee, Y.T., Ma, K.F., Shyu, J.B.H., Rau, R.J., Cheng, C.T., 2016.
Mohraz, B., 1976. A study of earthquake response spectra for different geological Probabilistic seismic hazard assessment for Taiwan. Terr. Atmos. Ocean. Sci. 27 (3),
conditions. Bull. Seismol. Soc. Am. 66 (3), 915–935. 325–340. https://doi.org/10.3319/TAO.2016.05.03.01(TEM).
Morikawa, N., Kanno, T., Narita, A., Fujiwara, H., Okumura, T., Fukushima, Y., Wells, D.L., Coppersmith, K.J., 1994. New empirical relationships among magnitude,
Guerpinar, A., 2008. Strong motion uncertainty determined from observed records rupture length, rupture width, rupture area, and surface displacement. Bull. Seismol.
by dense network in Japan. J. Seismol. 12 (4), 529–546. Soc. Am. 84 (4), 974–1002.
Newmark, N.M., Hall, W.J., 1982. Earthquake Spectra and Design. Earthquake Wu, Y.M., Shin, T.C., Chang, C.H., 2001. Near real-time mapping of peak ground
Engineering Research Institute, Berkeley, CA. acceleration and peak ground velocity following a strong earthquake. Bull. Seismol.
Phung, V.B., Loh, C.H., Chao, S.H., Chiou, B.S., Huang, B.S., 2020. Ground motion Soc. Am. 91 (5), 1218–1228. https://doi.org/10.1785/0120000734.
prediction equation for crustal earthquakes in Taiwan. Earthquake Spectra 36 (4), Wu, Y.M., Zhao, L., Chang, C.H., Hsu, Y.J., 2008. Focal-mechanism determination in
2129–2164. Taiwan by genetic algorithm. Bull. Seismol. Soc. Am. 98 (2), 651–661. https://doi.
Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D., Team, R.C., 2020. Nlme: Linear and org/10.1785/0120070115.
Nonlinear Mixed Effects models. R package version 3.1-149. Youngs, R.R., Coppersmith, K.J., 1985. Implications of fault slip rates and earthquake
Restrepo-Velez, L.F., Bommer, J.J., 2003. An exploration of the nature of the scatter in recurrence models to probabilistic seismic hazard estimates. Bull. Seismol. Soc. Am.
ground-motion prediction equations and the implications for seismic hazard 75 (4), 939–964.
assessment. J. Earthq. Eng. 7 (spec01), 171–199. Yu, S.B., Chen, H.Y., Kuo, L.C., 1997. Velocity field of GPS stations in the Taiwan area.
Tectonophysics 274 (1–3), 41–59.

16

You might also like