Bulletin of the Seismological Society of America, Vol. 99, No. 1, pp. 255–267, February 2009, doi: 10.1785/0120080258
Large Effects of Moho Reflections (SmS) on Peak Ground Motion
in Northwestern Taiwan
by Kun-Sung Liu and Yi-Ben Tsai
Abstract
A total of 336 three-component strong-motion recordings from the
Mw 6.35 Niu Dou earthquake of 25 June 1995 at a focal depth of 39.9 km in northern
Taiwan are used to study the effects on strong ground motion due to Moho reflection
of S waves. The residuals of both horizontal peak ground acceleration (PGA) and peak
ground velocity (PGV) recorded from the earthquake are analyzed. The results confirm
that many Class E soft soil stations in the Taipei Basin and the Ilan Plain had the
expected large amplification of about 1.7 and 1.5 times, respectively, the predicted
median PGA values. Surprisingly, a large group of Class C or D dense and stiff
soil sites in Taoyuan (TCU007), Lungtan (TCU013), Guanshi (TCU021), Hsinchu
(TCU095), and Miaoli (TCU047) areas in northwestern Taiwan had unusually large
amplification of about 3.4–8.1 and 1.7–3.3 times the predicted median PGA and PGV
values, respectively. They are interpreted in terms of focusing and interference between SmS waves reflected from the horizontal and inclined portions of an eastdipping Moho discontinuity in this area. This interpretation is supported by the close
agreement between the expected amplitudes and arrival times of the largest shear
waves with the observed data. Our results suggest that when a damaging earthquake
occurs near an inclined Moho boundary, the reflected SmS waves can result in significantly amplified ground motions at distances beginning about 50 km. The exact
distance range will depend on the thickness of the crust and the dip angle of the Moho
boundary.
Introduction
Ground-motion characteristics are known to depend on
the properties of seismic source, wave propagation, and site
response. The most commonly used method for characterizing ground motion is through attenuation relations, in which
the effects of earthquake source, wave propagation, and site
response are typically parameterized by the earthquake magnitude, fault type, source-to-site distance, and site condition. Conversely, we can study the contribution of source,
path, and site effects by examining the residuals of observed
ground-motion data at individual sites with reference to the
values predicted by the empirical attenuation relations.
The path effects on strong ground motion due to crustal
structures have been known for some time. Burger et al.
(1987) investigated the attenuation relations of eastern North
America, which show amplitudes in the distance range of
60–150 km to be higher than that at smaller and greater distances. They showed that the observed interval of relatively
high amplitudes can be attributed to postcritically reflected
S waves from the Moho discontinuity. The presence and location of the interval of relatively high amplitudes is highly
dependent on the crustal velocity structures and may therefore be expected to show regional variations.
Somerville et al. (1990) found significant influence of
critical reflections from the lower crust on ground-motion
attenuation from the large set of strong-motion recordings
of the Saguenay, Quebec, earthquake of 25 October 1988.
At distances beyond 64 km, the peak ground motions occurred at times corresponding not to the direct S wave but
to strong critical reflections from the lower crust. The amplitude of recorded ground motions did not significantly decrease between 50 and 120 km, but it abruptly decreased
beyond 120 km.
In central California, Bakun and Joyner (1984) suggested that the large positive residuals in ML at distances between 75 and 125 km could be due to Moho reflections.
Large-amplitude reflections from the Moho (PmP) were also
observed in seismic refraction data near the source region of
the Loma Prieta earthquake (Walter and Mooney, 1982).
Somerville and Yoshimura (1990) presented evidence of enhanced amplitudes of strong ground motion from the Loma
Prieta earthquake recorded in the San Francisco and Oakland
areas, which were found to exceed the levels predicted by
standard empirical attenuation relations. Their analysis of accelerograms with known trigger times strongly suggests that
255
256
K.-S. Liu and Y.-B. Tsai
enhancement of ground-motion amplitudes in the distance
range of approximately 40–100 km was due to critical reflections from the base of the crust. The effect of critical reflections in amplifying peak accelerations of the Loma Prieta
earthquake in the San Francisco and Oakland regions was
as large as the amplification effect of soft soil site conditions.
McGarr et al. (1991) presented observations in the epicentral distance range of 59–95 km, including the San Francisco International Airport, of the Moho-reflected phases
PmP and SmS from the aftershocks of the Loma Prieta
earthquake to support the hypothesis that the phase SmS accounted much of the enhanced peak ground motion experienced from the mainshock throughout most of the San
Francisco Bay area.
Mori and Helmberger (1996) used closely spaced aftershock data from the 28 June 1992 Landers earthquake to create event record sections that showed clear examples of
varying amplitudes of SmS. Some of the data showed strong
SmS phases to be two to five times larger than the direct S
waves. They interpreted the amplitude variations in terms of
the local crustal and Moho structures in southern California.
Yeh et al. (1988) demonstrated, on the basis of an assumed velocity structure, that focusing of seismic-wave energy in the Taipei Basin can occur for some earthquakes
under favorable conditions. The wave focusing phenomena
were also supported by other strong-motion data collected by
the accelerograph network in Taiwan.
Several later studies have shown additional evidence
that reflected seismic energy can significantly intensify the
ground shaking at an appreciable distance from the epicenter
of an earthquake (Atkinson and Boore, 1995; Catchings
and Kohler, 1996; Atkinson and Boore, 1997; Somerville
et al., 1997; Toro et al., 1997; Chen, 2003). Thus, focusing
of seismic-wave energy can play an important role in enhancing seismic intensity even in regions some distance away
from the epicentral zone. It follows that it may be important
to consider such path effects in engineering practice. The objective of this article is to analyze large effects of the Moho
reflection on the amplification of peak ground motion (peak
ground acceleration [PGA] and peak ground velocity [PGV])
observed at several sites located in northwestern Taiwan during the 1995 Niu Dou earthquake.
must respond accurately in the frequency range from d.c. to
50 Hz in order to faithfully record the near-source ground
motion caused by large earthquakes. In order to record a
wide range of earthquakes on scale, it is required that the
complete system be digital and have at least a 16-bit resolution and a 96 dB dynamic range (Liu et al., 1993). The full
scale of the recording system is 2g at a sample rate of
200 samples=sec. Each recording system is operated in trigger mode with a 20 sec preevent memory and is set to record
an additional 10 sec of data after the signal dropped below a
preset threshold (Liu et al., 1999).
In this study, we use the TSMIP strong-motion data of
the Mw 6.35 Niu Dou earthquake of 25 June 1995 at a focal depth of 39.9 km in northern Taiwan (Central Weather
Bureau, 1995). A total of 336 three-component accelerograms were recorded at epicentral distances ranging from
4.9 to 244.7 km. We have analyzed these data to study the
dependence of PGA on the seismic-wave propagation path,
especially to examine the effects of Moho reflection. Figure 1
shows a contour map of the mean PGA of two horizontal
components, together with the epicenter location (24.606° N,
121.669° E) of the Niu Dou earthquake. The largest PGA of
all of the records is 259 gal (gal cm=sec2 ). The heavy lines
in the map represent the PGA value of 80 gal. Similarly, Figure 2 plots a contour map of the mean PGV of two horizon-
Strong-Motion Data
The Central Weather Bureau has undertaken the Taiwan Strong-Motion Instrumentation Program (TSMIP) since
July 1991 to collect high-quality instrumental recordings of
strong earthquake shaking. Thus far, more than 640 free-field
accelerograph stations have been deployed in populated
areas of Taiwan. This network has provided large numbers
of recordings to form an excellent database for studying
strong-motion characteristics and for developing attenuation
relations. Each operating free-field station includes triaxial
accelerometers, a digital recording subunit, a power supply,
and a timing system. The transducers for the accelerograph
Figure 1.
Horizontal PGA contour map with the epicenter of the
25 June 1995 Niu Dou earthquake. Nine stations with large amplitude located at Class C or D dense and stiff soil sites are also shown.
Large Effects of Moho Reflections (SmS) on Peak Ground Motion in Northwestern Taiwan
257
tal components. The largest PGV of all of the records is
10:8 cm=sec. The heavy lines in the map represent the
PGV value of 6 cm=sec. From these PGA and PGV contour
maps, a surprising feature is noticed: unusually large PGA
and PGV anomalies were observed at several soft-rock and
dense soil sites (Lee et al., 2001) in the Taoyuan, Lungtan,
Guanshi, Hsinchu, and Miaoli areas in northwestern Taiwan
(see Fig. 1 for the location of the cities).
It is well known that ground-motion characteristics can
be affected by local site response. According to the site classification criteria of Table 1 (Lee et al., 2001), each recording station in Taiwan can be classified as a Class B, C, D, or
E site. Examination of the records with different site classifications confirmed that many Class E soft soil stations in the
Taipei Basin and the Ilan Plain had large PGA values. However, many Class C and D stiff soil sites, especially those
located in the Taoyuan, Lungtan, Guanshi, Hsinchu, and
Miaoli areas, also had unusually large PGA and PGV values
during the 1995 Niu Dou earthquake. As shown in the following, this unusual PGA and PGV amplification can be explained by large effects of Moho reflection.
Attenuation Model and Residual Analysis
Figure 2.
Horizontal PGV contour map with the epicenter of
the 25 June 1995 Niu Dou earthquake. Nine stations with large amplitude located at Class C or D dense and stiff soil sites are also
shown.
A strong-motion attenuation relationship expresses an
earthquake ground-motion parameter as a function of simple
parameters characterizing the earthquake source, the propagation path between the earthquake source and the site, and
the geologic conditions beneath the site. The following equation form is used in this study:
ln Y a lnX h bX cMw d σ;
Table 1
Comparison between the 1997 UBC Provisions and the Simplified Site Classification Used in This Study (after Lee et al. [2001])
Site Class
Site Class Description of 1997 UBC Provisions
A
B
Hard rock, eastern United States sites only, VS > 1500 m=sec.
Rock, V S is 760–1500 (m=sec)
C
Very dense soil and soft rock, V S is 360–760 (m=sec), undrained
shear strength us ≧2000 psf (us ≧100 kPa) or N≧50 blows=ft
Stiff soils, V S is 180–360 (m=sec), stiff soil with undrained shear
strength 1000 psf≦us ≦2000 psf (50 kpa≦us ≦100 kPa) or
15≦N≦50 blows=ft.
Soft soils, profile with more than 10 ft (3 m) of soft clay defined as
soil with plasticity index PI > 20, moisture content w > 40%,
and undrained shear strength us < 1000 psf(50 kPa) or
N < 15 blows=ft
Soils requiring site specific evaluations: (1) soil is vulnerable to
potential failure or collapses under seismic loading: for example,
liquefiable soils, quick and highly sensitive clays, collapsible
weakly cemented soils; (2) peats and/or highly organic clays
(10 ft [3 m] or thicker layer); (3) very high plasticity clays: (25 ft
[8 m] or thicker layer with plasticity index > 75); (4) very thick
soft/medium stiff clays: (120 ft [36 m] or thicker layer)
D
E
F
Note: the provisions of 1997 NEHRP and 1997 UBC are similar.
Site Class Description of Lee et al. (2001)
(not used)
Miocene and older strata, along with limestone, igneous rocks,
metamorphic rocks, etc.
Pliocene and Pleistocene strata, along with conglomerates,
pyroclastic rocks, etc. and geomorphologic lateritic terraces
Late Pleistocene and Holocene strata, geomorphologic fluvial
terrace, along with stiff clays and sandy soils with average
SPTN≧15 in the upper 30 m
Holocene deposits and fills, etc. with average SPT N < 15 in the
upper 30 m
(not classified in the present study and will be studied in the future)
(1)
258
K.-S. Liu and Y.-B. Tsai
Table 2
Regression Coefficients for the PGA and PGV Attenuation Relation
(Liu and Tsai, 2005)
PGA
PGV
a
b
c
d
h
σ
0:852
0:857
0:0071
0:0023
1.027
1.486
1.062
4:472
1.24
1.34
0.719
0.711
Note: lnPGA; PGV a lnX h b X c Mw d σ.
where Y is the ground-motion parameter, X is the hypocentral distance, Mw is the moment magnitude, a is the geometrical spreading coefficient, b is the anelastic attenuation
coefficient, c is the magnitude coefficient, d is a constant, h
is the close-in distance saturation coefficient, and σ is the
standard deviation. The coefficients a, b, c, d, and h are
to be determined by regression from the data (Liu and Tsai,
2005). This equation form is similar to the one used by Joyner and Boore (1981), except for the difference in distance
definitions.
Regressions on the TSMIP data set without differentiating site conditions have resulted in the coefficients of the attenuation relationships, as given in Table 2, for the horizontal
(H) components of PGA and PGV for the whole of Taiwan
Figure 3. Comparison of the horizontal PGA data for the
Mw 6.35 Niu Dou earthquake with the corresponding median
and median σ values derived from the attenuation relationships
of Liu and Tsai et al. (2005).
(Liu and Tsai, 2005). In Table 2, σ refers to the standard deviation of lnY. Assuming a log-normal distribution, this
value can be used to obtain the value of the parameter Y corresponding to different probability levels. Specifically, lnY
at 84% probability (median plus one sigma) may be obtained
by multiplying the predicted median value by eσ.
Figures 3 and 4 compare the observed data from the
1995 Niu Dou earthquake (Mw 6:35) with the PGA and
PGV attenuation relations, respectively, for the Taiwan area.
It can be seen in these figures that most observed data are
distributed within the range between 1standard deviation
of the median attenuation curve. We can also find a cluster
of very high-valued data at a hypocentral distance 66–84 km
from TCU007, TCU013, and TCU021 stations and so on,
which are located in Taoyuan, Lungtan, and Guanshi areas
in northwestern Taiwan, located in Figures 3 and 4.
Examination of the PGA residuals (i.e., the difference
between logarithms of the observed and predicted PGA)
for sites with different soil categories is a useful method
for sets of records where site information is not complete
and hence cannot be included explicitly within the equation
(Abrahamson and Litehiser, 1989). Liu and Tsai (2005) analyzed both the PGA and PGV residuals to study their variations with respect to site conditions. The results showed that
Figure 4.
Comparison of the horizontal PGV data for the
Mw 6.35 Niu Dou earthquake with the corresponding median
and median σ values derived from the attenuation relationships
of Liu and Tsai et al. (2005).
Large Effects of Moho Reflections (SmS) on Peak Ground Motion in Northwestern Taiwan
the residual contour maps have high consistency with both
regional geology and topography of Taiwan. For example,
positive residuals are expected in areas like the Taipei Basin and the Ilan Plain that will experience amplification of
ground motion due to soft soil site conditions.
In order to understand the relation between both PGA
and PGV residuals and local sites as well as propagation path
effects, we plot the residual of horizontal PGA versus epicentral distance in Figure 5 and the residual of horizontal PGV
versus station azimuth in Figure 6. The results confirm that
many Class E soft soil stations in the Taipei Basin and the
Ilan Plain have, as expected, large positive PGA residuals of
about 0.53 and 0.40, corresponding to an amplification factor of 1.7 and 1.5, respectively, relative to the predicted median values.
Surprisingly, residuals at a group of Class C or D dense
and stiff soil sites, especially those in Taoyuan (TCU007),
Lungtan (TCU013), Guanshi (TCU021), Hsinchu (TCU095),
and Miaoli (TCU047) areas, have even larger PGA and PGV
residuals corresponding to an amplification factor of 3.36–
8.13 and 1.73–3.30, respectively, relative to the predicted
median values. As shown in Figures 5 and 6, this group
of unusually large PGA and PGV sites is distributed in a narrow range of epicenral distances (i.e., 50–75 km) and azimuths (i.e., 270°–331°). Figures 1 and 2 clearly show that
Figure 5.
259
these sites form a northeast–southwest-trending narrow band
of anomalously large PGA and PGV in northwestern Taiwan.
Why did these stiff soil sites in northwestern Taiwan,
which were previously shown to have normal local site response (Liu and Tsai, 2005), have such large amplifications
during this earthquake? A plausible answer to this question is
presented in the following section.
Crustal Velocity Structures
We first analyze the path effects, especially regarding
the role of crustal structure in affecting strong ground motion. In order to pick seismic phases and calculate the theoretical travel times of direct and reflected waves from the
hypocenter, an appropriate crustal model for the Taiwan area
is needed. Some attempts to determine the crustal structure
of Taiwan were made in the past. Yeh and Tsai (1981) used
an iterative damped least-squares inversion procedure to find
P-wave velocities of a horizontally layered model for the
crust of central Taiwan. Their crustal velocity model has an
upper crust of two sublayers whose thickness and velocity
are 9 km and 5:8 km=sec for the upper layer and 8 km and
6:1 km=sec for the lower layer, respectively. The thickness
and velocity of the lower crust are 19 km and 6:7 km=sec,
respectively. The velocity of the upper mantle is 7:8 km=sec.
Thus, the overall thickness of the crust in the Central Range
PGA residuals from records at different distances and site conditions. The Taipei Basin and the Ilan Plain, as expected for soft
soil sites, have significantly positive residuals of 1.7 and 1.5 times, respectively, which are the median predictions. Surprisingly, the residuals
for some sites with dense and stiff soils, especially in Taoyuan (TCU007), Lungtan (TCU013), and Guansi (TCU021) areas, are about 3.4–8.1
times the median predictions.
260
K.-S. Liu and Y.-B. Tsai
Figure 6.
PGV residuals from records at different azimuths. Many stations in azimuth ranging from 270° to 335°, especially those in
Taoyuan (TCU007), Lungtan (TCU013), and Guanshi (TCU021) areas, had logarithmic residuals of about 0.55–1.19, corresponding to a
factor of 1.73–3.30 times the predicted median values. These high residual stations were not located at azimuths oriented perpendicular to the
fault plane. The focal mechanism of the Niu Dou earthquake was determined by Ma and Mori (1998) with a dip of 76.2°, a rake of 34.6°, and a
strike of 133.5°.
(CR) and Western Foothill (WF) regions is 36 km. Gravity data were also used to determine the crustal thickness of Taiwan. Liu and Yen (1975) calculated the crustal
thickness by using Bouguer anomaly data and found that
it is thickest 40 3 km in north central Taiwan and
thinnest 27 2 km in the southeastern and southern
coastal areas. The crustal thickness for Taiwan as a whole
is about 34 2 km.
In order to ensure precise earthquake location in geologically complex areas and to provide information about
tectonic boundaries or the mechanics of deformation in the
Taiwan area, many three-dimensional velocity models have
been constructed (Roecker et al., 1987; Shin and Ho, 1994;
Chen, 1995; Chen et al., 1995; Rau and Wu, 1995; Ma et al.,
1996; Cheng, 1998; Shih et al., 1998; Yeh et al., 1998; Tomfohrde and Nowack, 2000; Wang, 2004; Kim et al., 2004,
2005; Liao, 2005). These studies all confirm the presence
of significant lateral heterogeneities of seismic velocity in
Taiwan. Tomographic images of the crustal and mantle velocity structures under Taiwan were obtained by Rau and Wu
(1995). They found a root under the CR, which is deeper
in the north and becomes shallower toward the south. Similar results were found by Ma and Song (1997), Song (1997),
and Kim et al. (2005). Ma and Song (1997) investigated
the Pn velocity and Moho depth variations beneath Taiwan.
The results showed that Pn velocity under the CR is about
8:52 km=sec, which is about 5%–8% higher than the other
two provinces: 7:79 km=sec under the Coastal Plain (CP)
and 8:04 km=sec under the WF, respectively. The Moho
depths are about 31, 34.5, and 43 km, respectively, for the
CP, WF, and CR.
Kim et al. (2004, 2005) used the receiver function
method (RFM) and tomographic inversion method (TIM) to
investigate the complex crustal structures in the Taiwan region. The depths of Moho discontinuity were determined as
follows: a trend of crustal thinning starting from the east (at
50–52 km and 55 km by RFM and TIM, respectively) toward
the west (at 28–32 km and 35 km by RFM and TIM, respectively). This is in good agreement with the results from two
east–west-trending deep seismic profiles previously obtained
using airgun sources (Shin et al., 1998; Yeh et al., 1998). In
summary, the depth of Moho under Taiwan varies significantly, especially in the east–west direction. In the western
CP and WF, it is about 28–35 km, deepening gradually eastward to reach a maximum depth of 50–55 km beneath the
eastern CR.
Moho Reflection
In this study, we first considered the one-dimensional
(1D) velocity model of Chen (1995), which is used for rou-
Large Effects of Moho Reflections (SmS) on Peak Ground Motion in Northwestern Taiwan
tine earthquake location by the Central Weather Bureau as
the reference model for calculating the travel times.
Because the hypocenter of the Niu Dou earthquake was
located in northern Taiwan, the crustal velocity structures
under northern Taiwan are discussed in more detail. Shen
(2000) used an iterative damped least-squares inversion procedure to find a P-wave velocity model for the crust of northern Taiwan. This crustal velocity model consists of four
sublayers. The total thickness of the crust varies significantly
from 35 to 43 km, especially along the southwest–northeast
direction. In the western CP and WF, it is about 35 km. It
gradually deepens northeastward to reach a maximum depth
of 43 km beneath the Ilan Plain. Lin (2001) and Lin et al.
(2004) used a tomographic inversion method to determine
the three-dimensional V P and V S velocity structures and to
outline the Moho depth under northern Taiwan. They found a
Moho depth of about 42 km in the area around the Niu Dou
earthquake. Similar results were found by Song (1997) and
Liao (2005). Based on these studies, a velocity model is constructed, incorporating the following features (1) the crustal
thickness for Taiwan as a whole is about 33 km and (2) a
trend of crustal thinning from east (under the Central Mountain Range) toward west (under the Western Coastal Plain
and WF). In addition, the Moho discontinuity under the
Niu Dou area is set at a depth of 42 km and deepens toward
south. The Moho depth under the eastern Central Mountain
Range reaches a maximum depth.
Based on the previous discussion, we have constructed a
crustal velocity model, as given in Table 3 and shown in Figure 7. Basically, this crustal velocity model, following Chen
(1995), consists of 10 layers, except the thicknesses of the
eighth and ninth layers are modified. Thus, the total thickness of the crust, to the base of layer 8, is 42 km beneath
the epicenter of the Niu Dou earthquake. In order to verify
that the Niu Dou earthquake was indeed located within the
crust, we have relocated the earthquake by using the alternative velocity model. By comparing the results in Table 4
for the location of the Niu Dou earthquake, as determined by
using two different velocity models, that is, one previously
developed by Chen (1995) and the other constructed in this
study, as given in Table 3, we conclude that the hypocenter of
261
the Niu Dou earthquake is indeed located within the crust
(C.-H. Chang, personal comm., 2006).
In order to exclude the strong-motion records in the
Taipei Basin, which are known to be subjected to large site
amplification due to soft soils, a profile of recorded time
histories, as shown in Figure 8, is compiled by selecting
53 velocity seismograms restricted in the azimuth range of
245°–335° and epicenter distance range of 5–100 km. The
recordings are from a variety of site conditions. The calculated travel-time curves and the range of distances for the
direct P, S, PmP, and SmS waves are also shown in Figure 8.
The calculated travel-time curves and the range of distances
for the direct P, S waves and the reflected PmPh (SmSh ),
PmPi (SmSi ) waves that represent the waves reflected from
the horizontal and inclined portions of Moho discontinuity,
respectively, are also shown in Figure 8. These curves were
computed using the 2D crustal velocity model whose original
model was given in Table 3; the modified model is shown in
Figure 9.
The source-time function of the Niu Dou earthquake
as teleseismically determined has a pulse with a duration
of about 3 sec (Lin and Ma, 1996). Accordingly, the SmS
arrival-time curves in Figure 8 reappear at a delay of 3 sec
to represent the duration of the strong source pulse teleseismically seen. In Figure 8, the onset of the largest waves at
most stations coincided with the arrival time of the Moho
reflection SmS at distances beyond 50 km. The move-out
of this onset with distance clearly follows the SmS arrivaltime curve and not that of direct S. The observed duration of
strong motion following the SmS arrival-time curves is about
3 sec, which is compatible with the 3 sec duration of strong
source pulse, as determined by Lin and Ma (1996).
Focusing Effects of an Inclined Moho
In the previous record profile, the Moho reflection from
the base of layer 8 is a strong arrival at the epicentral distance
between 50 and 75 km. It is visible on most of the stations in
this distance range. In the following analysis, we will focus
on the effects of S-wave reflection from the base of an in-
Table 3
The Crustal Velocity Model Used in This Study
Thickness (km)
Depth (km)
V P km=sec
V S km=sec
QP
QS
Density (g=cm3 )
2
2
5
4
4
8
5
12
8
half-space
0–2
2–4
4–9
9–13
13–17
17–25
25–30
30–42
42–50
> 50
3.48
4.48
5.25
5.83
6.21
6.41
6.83
7.29
7.77
8.24
1.96
2.62
3.03
3.35
3.61
3.71
3.95
4.21
4.49
4.76
150
300
300
300
500
500
500
600
1500
2200
75
150
150
150
250
250
250
300
750
1100
2.1
2.5
2.6
2.6
2.8
2.8
2.8
3.0
3.2
3.6
262
K.-S. Liu and Y.-B. Tsai
Figure 7. A crustal velocity model for this study. This crustal
velocity model, basically following Chen (1995), consists of 10 layers, except the thicknesses of the eighth and ninth layers are modified. Thus, the total thickness of the crust, to the base of layer 8, is
42 km beneath the epicenter of the Niu Dou earthquake instead of
35 km as proposed by Chen (1995).
clined Moho discontinuity to explain the enhanced PGA and
PGV recorded in northwestern Taiwan.
We use a ray-tracing method to calculate the travel time
of reflected waves from the Moho discontinuity. As the incident angle increases to 69.7° corresponding to a distance of
76 km, reflection of shear waves from the interface at the
lower crust will reach the critical angle and undergo total
internal reflection. In order to explain and fit the large amplitudes observed at a distance less than 76 km, a refined
two-dimensional (2D) velocity model with inclined Moho
discontinuity is developed.
Let us consider that the Moho discontinuity surface is
not parallel to the free surface but has a dip angle ψ. For
up-dip propagation, the surface is inclined upward from east
to west. The thickness of the crust is not uniform: it is 42 km
beneath the northern Central Mountain Range and 33 km
under the western CP and WF. Thus, the final 2D model for
the velocity structure along the east–west direction consists
of 10 layers, as shown in the upper left part of Figure 9. The
Moho discontinuity is at a depth of 42 km beneath the epicenter of the Niu Dou earthquake. It extends horizontally toward the west for 6.8 km. It then turns up-dip at 24.2° with a
ratio of 20 km horizontally to 9 km vertically. Beyond that
point westward, the Moho discontinuity is kept flat at a constant depth of 33 km.
The curves for the travel time and the incident angle
(theta) as a function of epicentral distance for the dipping
Moho model are shown in Figure 9. Curves A and B show
the relation between incident angle versus the distance for
waves reflected from the horizontal and inclined portions of
the Moho discontinuity, respectively. Curves C and D show
the relation between travel time versus distance for waves
reflected from the horizontal and inclined portions of the
Moho discontinuity, respectively. It is noted that there exists
an overlap distance range from 48 to 86 km over which reflected waves from both discontinuities will simultaneously
arrive, when the theta angle increases from 55.6° to 92.1°. In
the intervals of theta from 55.6° to 72.9° and from 72.9° to
92.1°, the S waves are reflected from the horizontal and dipping portions of the Moho discontinuity, respectively.
In the overlap distance range from 48 to 86 km, simultaneous arrival of the SmS waves reflected from the horizontal and dipping portions of the Moho discontinuities will
cause focusing and interference resulting in enhancement
of ground-motion amplitudes. Figure 10 shows a crustal velocity model for northern Taiwan with ray traces for direct
S waves (in red) and SmS (in blue) reflected from the Moho
discontinuity. We can see that focusing and interference between SmS waves reflected from both the horizontal and inclined portions of the Moho discontinuity in the epicentral
distance range of 48 to 76 km. This is taken as the primary
cause for significantly enhanced PGA and PGV at a group of
stations distributed in this narrow distance range in northwestern Taiwan.
It should be noted that at distances beyond 48 km, the
PGA and PGV occurred at times corresponding to strong
reflections from the Moho discontinuity, instead of direct
S waves. Because of the presence of an inclined Moho discontinuity, the amplitudes of recorded ground motion did
not significantly decrease at hypocentral distances between
66 and 84 km, but decreased rapidly beyond 84 2 km, as
shown in Figure 3.
Discussion
In the preceding analysis, we have demonstrated the focusing and interference effects of reflected S waves from an
inclined Moho discontinuity to be the primary cause for significantly enhanced PGA and PGV recorded in northwestern
Taiwan during the Niu Dou earthquake. However, groundmotion amplitudes can also be affected by a variety of other
factors, such as source or site effects, some of which may
contribute together. It is therefore necessary to distinguish
Table 4
Comparison of the Niu Dou Earthquake Location by Using Two Crustal Velocity Models
Velocity Model
Latitude (°N)
Longitude (°E)
Focal Depth (km)
Moho Depth (km)
ML
Mw
Chen (1995)
This study
24.606
24.603
121.669
121.666
39.88
39.77
35
42
6.50
6.50
6.35
6.35
Large Effects of Moho Reflections (SmS) on Peak Ground Motion in Northwestern Taiwan
263
Figure 9. The travel time and incident angle (theta) versus the
epicentral distance curves of a dipping Moho model. Curves A and
B stand for the incident angle versus distance curves of waves reflected from the horizontal and inclined portions of the Moho discontinuity, respectively. Curves C and D stand for the traveling time
curves of wave reflected from the horizontal and inclined discontinuities, respectively. The crustal velocity model is also shown in the
upper left part of the figure.
Figure 8. A reduced travel-time plot of the north–southcomponent velocity seismograms for stations located in the azimuth
range of 245°–335° and in the epicenter distance range of 5–100 km.
The calculated travel-time curves and the range of distances for
the direct P, S, PmP, and SmS waves are also shown. The calculated travel-time curves and the range of distances for the direct P,
S waves and the reflected PmPh (SmSh ), PmPi (SmSi ) waves that
represent the waves reflected from the horizontal and inclined portions of Moho discontinuity, respectively, are also shown.
the predominant factor that caused the significant enhancement of recorded ground-motion amplitudes.
Seismic radiation patterns can result in geographic
asymmetry of ground motion due to the faulting process that
is closely related to the focal mechanism of the earthquake.
The focal mechanism of the Niu Dou earthquake was determined by Lin and Ma (1996) using teleseismic waveforms. It
was an oblique faulting mechanism with the dip at 76.2°, the
rake at 34.6°, and the strike at 133.5° (Ma and Mori,1998). In
order to understand the relation between the station residuals
of PGV and the seismic radiation patterns, we now refer to
Figure 6 where the residuals of the horizontal PGV are plotted
as a function of site azimuth, that is, the direction from the
seismic source to a recording station. The data confirm that
many Class E soil stations in Chianan Plain (CHY area) have
large positive PGV residuals, above 0.8, corresponding to an
amplification factor of 2.2 relative to the predicted median
values. The large residuals of these stations at azimuth rang-
ing from 220° to 240°, which were close to the azimuths perpendicular to the fault plain, were most likely due to seismic
radiation effects. In the meantime, some of the stations belong to Class E soft soil sites that might have contributed to
part of the large amplification.
Figure 6 also shows that another group of stations at
azimuth ranging from 270° to 331°, especially those in Tao-
Figure 10.
A crustal velocity model for northern Taiwan with
ray traces of direct S waves (in red) and SmS waves (in blue) reflected from the Moho discontinuity. We can see that focusing and
interference between SmS waves reflected from both the horizontal
and inclined portions of the Moho discontinuity in the epicentral
distances ranged from 48 to 76 km. The P and S velocities of each
layer are also shown in the left part of the figure.
264
K.-S. Liu and Y.-B. Tsai
yuan (TCU007), Lungtan (TCU013), Guanshi (TCU021),
Hsinchu (TCU095), and Miaoli (TCU047) areas, had logarithmic residuals of about 0.55–1.19, corresponding to an
amplification factor of 1.73–3.30 over the predicted median
values. These large PGV sites were not located in azimuths
perpendicular to the fault plane. This suggests that the observed large-amplitude late-arrival S waves were unlikely to
be caused by seismic radiation effects.
Moreover, site effects can also play an important role in
amplifying seismic ground motions. A simple way to estimate site effects is made in terms of soil-type classification
(Kawase, 2004). Figures 11 and 12 show a reduced time
plot of the north–south-component acceleration and velocity waveforms, respectively, for the nine stations with PGA
greater than 100 gal, as extracted from Figure 8. According
to Table 5, these nine stations with large PGA and PGV amplitudes are located on Class C or D dense and stiff soils.
Obviously, the unusually large amplification phenomenon
Figure 12. A reduced travel-time plot of the north–southcomponent velocity seismograms for the nine stations with PGA
greater than 100 gal, as extracted from Figure 8. The calculated arrival times for direct P, S, PmP, and SmS phases are also shown.
Station codes and relevant PGV are marked at the heads and tails of
traces, respectively. The relevant PGV data of these records are also
shown in Figures 2, 4, and 6.
Figure 11. A reduced travel-time plot of the north–southcomponent accelerograms for the nine stations with PGA greater
than 100 gal. The calculated arrival times for direct P, S, PmP,
and SmS phases are also shown. Station codes and relevant PGA
are marked at the heads and tails of the traces, respectively. The
relevant PGA data of these records are also shown in Figures 1,
3, and 5.
cannot be explained by site effects, either. This is supported
by normal site response at these stations found in our previous studies using a large data set (Liu and Tsai, 2005).
In summary, we can discount either seismic radiation or local
site response effects as a major cause for unusual amplification observed at these stations.
Furthermore, McGarr et al. (1991) pointed out that if the
highest-amplitude portion of the S-wave train is SmS in the
horizontal seismogram, then the highest-amplitude portion
of the P-wave train should also be found in the PmP phase
on the vertical seismogram. Figure 13 plots a profile of recorded time histories of the vertical component, as compiled
by selecting 53 velocity seismograms. The ranges of azimuth
and epicenter distance of these records are the same as Figure 8. The calculated travel-time curves and the range of distances for the direct P, S, PmP, and SmS waves are also
shown in Figure 13. In the figure, the portion of PmP wave
trains shows similarly enhanced amplitudes just as the enhanced SmS wave trains shown in Figure 8.
Large Effects of Moho Reflections (SmS) on Peak Ground Motion in Northwestern Taiwan
265
Table 5
Station Code, Site Classification, Amplification Factor, and Related Data Used in Figures 1–6, 11, and 12
PGA (cm=sec =sec)
Station
Number
Station
Code
Latitude
(°N)
Longitude
(°E)
Epicentral
Distance (km)
Hypocentral
Distance (km)
Azimuth
(°)
Site
Class
NS
H
1
2
3
4
5
6
7
8
9
TAP038
TCU013
TCU021
TCU007
TCU014
TCU024
TCU095
TCU043
TCU047
25.02
24.87
24.79
25.00
25.05
24.74
24.69
24.69
24.62
121.41
121.20
121.17
121.31
121.31
121.08
121.01
120.95
120.94
53.07
55.04
55.05
56.87
60.90
61.46
66.98
73.11
73.91
66.38
67.97
67.98
69.46
72.80
73.26
77.95
100.8
140.7
330.7
301.5
292.2
320.3
323.2
283.6
278.1
277.3
271.2
D
C
D
D
D
C
C
C
C
179.9
163.1
123.1
286.8
96.7
115.2
137.9
99.2
121.2
155.9
200.9
160.0
259.1
100.7
132.6
176.7
100.0
130.6
Summary
Based on preceding results and discussion, we can summarize as follows:
1. In the present study, we have identified focusing and
interference between SmS waves reflected from a horizontal and inclined portions of the Moho discontinuity
PGV (cm=sec)
Residual Factor
1.53
1.81
1.58
2.10
1.21
1.50
1.87
1.39
1.67
4.60
6.12
4.87
8.13
3.36
4.47
6.48
4.03
5.32
H
10.80
8.95
9.77
10.46
5.23
8.67
9.10
4.87
5.54
Residual Factor
1.19
1.02
1.10
1.19
0.55
1.06
1.17
0.61
0.75
3.30
2.76
3.02
3.30
1.73
2.89
3.23
1.85
2.12
as the primary cause for unusually large peak ground
motion recorded at distances between 50 and 75 km in
northwestern Taiwan. This interpretation is supported by
constructing an inclined Moho model in which the amplitudes and arrival times of the largest shear waves were
in close agreement with the data.
2. The residuals of horizontal PGA and PGV recorded from
the Niu Dou earthquake confirmed that many Class E
soft soil stations in the Taipei Basin and the Ilan Plain
had the expected large amplification of about 1.7 and
1.5 times, respectively, the predicted median values. Surprisingly, residuals at many Class C or D dense and stiff
soil sites, especially those in Taoyuan (TCU007), Lungtan (TCU013), Guanshi (TCU021), Hsinchu (TCU095),
and Miaoli (TCU047) areas located in northwestern
Taiwan, had residuals about 3.4–8.1 times the predicted
median values. This is shown primarily due to effects of
reflection at the Moho discontinuity.
3. The propagation path effects, especially those due to
crustal structures, can play an important role in producing
large ground motion while site conditions are clearly not
responsible. Our results suggest that when a damaging
earthquake occurs at depth near the Moho boundary,
the reflected SmS waves can cause significantly amplified ground motions in distances beginning at around
50 km. The exact distance range will depend on the thickness of the crust and the dip angle of an inclined Moho
discontinuity.
Data and Resources
Figure 13.
A reduced travel-time plot of the vertical velocity
seismograms for stations located in the azimuth range of 245°–
335° and in the epicenter distance range of 5–100 km. The calculated travel-time curves and the range of distances for the direct P,
S, PmP, and SmS waves are also shown. The calculated travel-time
curves and the range of distances for the direct P, S waves and the
reflected PmPh (SmSh ), PmPi (SmSi ) waves that represent the
waves reflected from the horizontal and inclined portions of Moho
discontinuity, respectively, are also shown.
Seismograms used in this study were collected as part
of the Taiwan Strong-Motion Instrumentation Program
(TSMIP) using 16-bit accelerographs. The digital TSMIP
strong-motion data can be obtained from the Central Weather
Bureau of Taiwan at www.cwb.gov.tw, available at http://e
-service.cwb.gov.tw/i-sales-web2/Services%20Application/
services_application.htm.
Acknowledgments
We thank the Central Weather Bureau of Taiwan for providing excellent strong-motion data. We are grateful to Joe Litehiser and Art McGarr for
266
K.-S. Liu and Y.-B. Tsai
their critical comments and suggestions that have led to a significant improvement of the article. We are also grateful to Y. H. Tseng, G. S. Chang,
H. C. Pu, and Mr. S. K. Wang for their computer programs. This research
was supported by the Taiwan Earthquake Research Center (TEC) funded
through National Science Council (NSC) with Grant Number NSC96-2625
-Z-244-001 and Number NSC97-2625-M-244-001. The TEC contribution
number for this article is 00037.
References
Abrahamson, N. A., and J. J. Litehiser (1989). Attenuation of vertical peak
acceleration, Bull. Seismol. Soc. Am. 79, 549–580.
Atkinson, G. M., and D. M. Boore (1995). Ground-motion relations for eastern North America, Bull. Seismol. Soc. Am. 85, 17–30.
Atkinson, G. M., and D. M. Boore (1997). Some comparisons between recent ground-motion relations, Seism. Res. Lett. 68, no. 1, 24–40.
Bakun, W. H., and W. B. Joyner (1984). The ML scale in central California,
Bull. Seismol. Soc. Am. 74, 1827–1843.
Burger, R. w., P. G. Somerville, J. S. Barker, R. B. Herrmann, and D. V.
Helmberger (1987). The effect of crustal structure on strong ground
motion attenuation relations in eastern North America, Bull. Seismol.
Soc. Am. 77, 420–439.
Catchings, R. D., and W. M. Kohler (1996). Reflected seismic waves and
their effect on strong shaking during the 1989 Loma Prieta, California,
earthquake, Bull. Seismol. Soc. Am. 86, 1401–1416.
Central Weather Bureau (1995). Seism. Bull. 42, no. 2 (in Chinese).
Chen, K. C. (2003). Strong ground motion and damage in the Taipei
basin from the Moho reflected seismic waves during the March 31,
2002, Hualien, Taiwan earthquake, Geophys. Res. Lett. 30, no. 11,
1551.
Chen, Y. L. (1995). The study of 3-D velocity structure and tectonic subduction in Taiwan, M.S. Thesis, Institute of Geophysics, National Central University, 172 pp. (in Chinese).
Chen, C. H., H. Y. Yen, and Y. H. Yeh (1995). Comparison of earthquake
location by using one- and three-dimensional velocity structures in Taiwan area, TAO 6, 527–540.
Cheng, W. B. (1998). Crust P-wave velocity structure of the eastern Taiwan
region and its tectonic implications, Ph.D. Thesis, Institute of Oceanography, National Taiwan University, 123 pp. (in Chinese).
Joyner, W. B., and D. M. Boore (1981). Peak horizontal acceleration and
velocity from strong-motion records including record from the 1979
Imperial Valley, California, earthquake, Bull. Seismol. Soc. Am. 71,
2011–2038.
Kawase, H. (2004). Site effects on strong ground motion, in International
Handbook of Earthquake and Engineering Seismology, William Lee
(Editor), Academic Press, New York, 1013–1030.
Kim, K. H., J. M. Chiu, H. Kao, Q. Liu, and Y. H. Yeh (2004). A preliminary study of crustal structure in Taiwan region using receiver function
analysis, Geophys. J. Int. 159, 146–164.
Kim, K. H., J. M. Chiu, J. Pujol, K. C. Chen, B. S. Huang, Y. H. Yeh, and P.
Shen (2005). Three-dimensional V P and V S structure models associated with the active subduction and collision tectonics in the Taiwan
region, Geophys. J. Int. 162, 204–220.
Lee, C. T., C. T. Cheng, C. W. Liao, and Y. B. Tsai (2001). Site classification
of Taiwan free-field strong-motion stations, Bull. Seismol. Soc. Am. 91,
1283–1297.
Liao, Y. C. (2005). Analysis of deep crust structure using receiver function in
Taiwan, M.S. Thesis, Institute of Geophysics, National Central University, 91 pp. (in Chinese).
Lin, J. Y. (2001). Conduct the velocity structure using topographic method
and the distribution of possible fluid in northern Taiwan, M.S. Thesis, Institute of Geophysics, National Central University, 168 pp.
(in Chinese).
Lin, C. W., and K. F. Ma (1996). The study of near-field and far-field seismograms from three earthquakes in eastern Taiwan, in Symposium on
Taiwan Strong Motion Instrumentation Program (II), Taipei, Taiwan,
54–57 (in Chinese).
Lin, J. Y., S. K. Hsu, and J. C. Sibuet (2004). Melting features along the
western Ryukyu slab edge (northeast Taiwan): tomographic evidence,
J. Geophys. Res. 109, B12402.
Liu, C. C., and T. P. Yen (1975). Bouguer anomaly, surface elevation and crustal thickness in Taiwan, Petrol. Geol. Taiwan 12,
97–107.
Liu, K. S., and Y. B. Tsai (2005). Attenuation relationships of peak ground
acceleration and velocity for crustal earthquakes in Taiwan, Bull.
Seismol. Soc. Am. 95, 1045–1058.
Liu, K. S., T. C. Shin, and Y. B. Tsai (1999). A free-field strong motion
network in Taiwan: TSMIP, TAO 10, no. 2, 377–396.
Liu, K. S., T. C. Shin, W. H. K. Lee, and Y. B. Tsai (1993). Taiwan strong
motion instrumentation program: the characteristics comparison of
free-field accelerographs, Meteorol. J. Cent. Weather Bur. 39, no. 3,
132–150 (in Chinese).
Ma, K. F., and J. Mori (1998). Rupture process of the June 25, 1995 Ilan
earthquake, in Symposium on the Taiwan Strong Motion Instrumentation Program (III), Taipei, Taiwan, 141–149 (in Chinese).
Ma, K. F., and D. R. Song (1997). Pn velocity and Moho depth in Taiwan,
J. Geol. Soc. China 40, 167–184.
Ma, K.-F., J.-H. Wang, and D. Zhao (1996). Three-dimensional seismic
velocity structure of the crust and uppermost mantle beneath Taiwan,
J. Phys. Earth 44, 85–105.
McGarr, A., M. Celebi, E. Sembera, T. Noce, and C. Muller (1991).
Ground motion at the San Francisco International Airport from the
Loma Prieta earthquake sequence, 1989, Bull. Seismol. Soc. Am. 81,
1923–1944.
Mori, J., and D. Helmberger (1996). Large-amplitude Moho reflection
(SmS) from Landers aftershocks, southern California, Bull. Seismol.
Soc. Am. 86, 1845–1852.
Rau, J.-R., and F.-T. Wu (1995). Tomographic imaging of lithospheric structures under Taiwan, Earth Planet. Sci. Lett. 133, 517–532.
Roecker, S. W., Y. H. Yen, and Y. B. Tsai (1987). Three-dimensional P and
S wave velocity structure beneath Taiwan: deep structure of an arccontinent collision. J. Geophys. Res. 92, no. 10B, 10547–10570.
Shen, C. H. (2000). Using ray tracing to investigate the seismic velocity
structure in northwestern Taiwan, M.S. Thesis, Institute of Geophysics,
National Central University, 103 pp. (in Chinese).
Shih, R. C., C. H. Lin, H. L. Lai, Y. H. Yeh, B. S. Hwang, and H. Y. Yen
(1998). Preliminary crustal structures across central Taiwan from
modeling of the onshore-offshore wide angle seismic data, TAO 9,
317–328.
Shin, T. C., and M. Y. Ho (1994). Three dimensional velocity structure of
western Taiwan, Meteorol. J. Cent. Weather Bur. 40, no. 3, 216–234
(in Chinese).
Somerville, P. G., and J. Yoshimura (1990). The influence of critical Moho
reflections on strong ground motions recorded in San Francisco and
Oakland and during the 1989 Loma Prieta earthquake, Geophys.
Res. Lett. 17, 1203–1206.
Somerville, P. G., J. P. McLaren, C. K. Saikia, and D. V. Helmberger (1997).
The 25 November 1988 Saguenay, Quebec, earthquake: source parameters and the attenuation of strong ground motion, Bull. Seismol. Soc.
Am. 80, 1118–1143.
Song, D. R. (1997). The study of the crust characteristics of Orogenic zone
in Taiwan, M.S. Thesis, Institute of Geophysics, National Central University, 134 pp. (in Chinese).
Tomfohrde, D. A., and R. L. Nowack (2000). Crustal structure beneath Taiwan using frequency-band inversion of receiver function waveforms,
Pure Appl. Geophys. 157, 737–764.
Toro, G. R., N. A. Abrahamson, and J. F. Schneider (1997). Model of strong
ground motion attenuation relationships for subduction zone earthquakes, Seism. Res. Lett. 68, no. 1, 41–57.
Walter, A. W., and W. D. Mooney (1982). Crustal structure of the Diablo and
Gabilian ranges central California: a reinterpretation of existing data,
Bull. Seismol. Soc. Am. 72, 1567–1590.
Wang, Yu-Ju (2004). Three dimensional S-wave attenuation model of the
crust and uppermost mantle beneath arc-continent collision, Taiwan,
Large Effects of Moho Reflections (SmS) on Peak Ground Motion in Northwestern Taiwan
M.S. Thesis, Institute of Geophysics, National Central University,
112 pp. (in Chinese).
Yeh, Y. H., and Y. B. Tsai (1981). Crustal structure of central Taiwan from
inversion of P-wave arrival times, Bull. Inst. Earth Sci. Acad. Sin. 1,
83–102.
Yeh, Y. H., C. H. Chen, and T. L. Teng (1988). Study on focusing of
seismic wave energy in Taipei basin: I. a dynamic ray-tracing approach, in Proceedings of the Coordination Council for North American Affairs–American Institute in Taiwan Joint Seminar on Research
and Application for Multiple Hazards Mitigation, Taipei, Taiwan,
107–120.
Yeh, Y. H., R. C. Shih, C. H. Lin, C. C. Liu, H. Y. Yen, B. S. Huang, C. S.
Liu, P. Z. Chen, C. S. Huang, C. J. Wu, and F. T. Wu (1998). Onshore/
offshore wide-angle deep seismic profiling in Taiwan, TAO 9,
301–316.
General Education Center and Hazard Mitigation Research Center
Kao Yuan University
Kaohsiung, Taiwan, Republic of China
[email protected]
(K.-S.L.)
Pacific Gas and Electric Company
San Francisco, California
[email protected]
(Y.-B.T.)
Manuscript received 10 October 2007
267