EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS

Earthquake Engng Struct. Dyn. 2010; 39:827–845

Published online 27 October 2009 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/eqe.967

Ground-motion attenuation relationship for the Sumatran

megathrust earthquakes

Kusnowidjaja Megawati1,2, ∗, † and Tso-Chien Pan1

1 School of Civil and Environmental Engineering, Nanyang Technological University, 50 Nanyang Avenue,
Singapore 639798, Singapore
2 Earth Observatory of Singapore, Nanyang Technological University, 50 Nanyang Avenue,

Singapore 639798, Singapore

SUMMARY
A representative attenuation relationship is one of the key components required in seismic hazard assess-
ment of a region of interest. Attenuation relationships for peak ground acceleration, peak ground velocity
and response spectral accelerations for Sumatran megathrust earthquakes, covering Mw up to 9.0, are
derived based on synthetic seismograms obtained from a finite-fault kinematic model. The relationships
derived are for very hard rock site condition and for a long-distance range between 200 and 1500 km. They
are then validated with recorded data from giant earthquakes on the Sumatran megathrust occurring since
year 2000. A close examination of the recorded data also shows that spectral shapes predicted by most
of the existing attenuation relationships and that specified in the IBC code are not particularly suitable
for sites where potential seismic hazard is dominated by large-magnitude, distant, earthquakes. Ground
motions at a remote site are typically signified by the dominance of long-period components with periods
longer than 1 s, whereas the predominant periods from most of the existing attenuation relationships and
the IBC code are shorter than 0.6 s. The shifting of response spectrum towards longer period range for
distant earthquakes should be carefully taken into account in the formulation of future seismic codes for
Southeast Asia, where many metropolises are located far from active seismic sources. The attenuation
relationship derived in the present study can properly reproduce the spectral shape from distant subduction
earthquakes, and could hopefully give insights into the formulation of future seismic codes. Copyright q
2009 John Wiley & Sons, Ltd.

Received 22 August 2008; Revised 17 August 2009; Accepted 21 August 2009

KEY WORDS: Singapore; Sumatra; Mentawai; subduction; giant earthquakes; attenuation; seismic hazard

∗ Correspondence to: Kusnowidjaja Megawati, School of Civil and Environmental Engineering, Nanyang Technological
University, 50 Nanyang Avenue, Singapore 639798, Singapore.
†
E-mail: kusno@ntu.edu.sg

Copyright q 2009 John Wiley & Sons, Ltd.

828 K. MEGAWATI AND T.-C. PAN

1. INTRODUCTION

Seismicity and attenuation relationship are the two key components in seismic hazard analysis for
a region of interest. The regional seismicity is represented by a series of recurrence relationships,
where each relationship describes the average rate at which an earthquake of a certain magnitude
will be exceeded in a particular source zone. The attenuation relationship expresses ground-motion
parameters, such as peak ground acceleration (PGA), peak ground velocity (PGV) and response
spectral acceleration (RSA), as functions of earthquake magnitude, distance and other variables.
In high-seismicity regions, representative attenuation relationships are usually derived empiri-
cally from actual ground motions records. The available records should cover earthquake magni-
tudes, distance range and site characteristics of interest. However, in moderate- and low-seismicity
regions, where the available ground-motion records are often not sufficient to cover earthquake
magnitudes and distance range of interest, derivation of attenuation relationships has to be based
on synthetic seismograms [1–4]. These attenuation relationships derived should then be validated
using the limited available records.
Singapore and peninsular Malaysia face a unique problem in seismic hazard analysis. The local
seismicity itself is very low as evidenced by the fact that all ground tremors felt in the last 50 years
originated from distant Sumatran earthquakes [5, 6]. Intuitively, it could be postulated that the
seismic sources that may potentially affect Singapore and peninsular Malaysia are large-magnitude
earthquakes on the Sumatran fault and the adjacent subduction zone. The closest distances from
Singapore to the fault and megathrust are 400 and 600 km, respectively. The highly segmented,
right-lateral, Sumatran fault could produce earthquakes with Mw up to 7.8 [7], and its attenuation
relationships have been derived [8]. On the other hand, the megathrust has known to repeatedly
produce giant earthquakes with Mw >8.0 [9–11].
The motivation of the present work originated from the fact that the existing attenuation relation-
ships derived for subduction zones in Japan, Taiwan, Cascadia, Alaska, Chile, Mexico, Peru and
other parts of the world [12–17] could not reproduce response spectra from actual data recorded
in Singapore from the Sumatran subduction earthquakes. This will be elaborated further in a
later section. Megawati et al. [18] have derived a set of spectral attenuation relationships for the
Sumatran subduction earthquakes. The study was based on synthetic seismograms produced using
a point-source dislocation model. Although the derived attenuation relationships could estimate
actual recorded data for earthquakes with Mw 7.2, they were not tested for giant earthquakes
because actual recorded data were not available at that time [18].
The objective of the present study is to extend the work by Megawati et al. [18] to cover
larger-magnitude earthquakes. In the present work, synthetic seismograms will be generated using
a finite-fault kinematic model, which have been tested for simulating giant subduction earthquakes
[11]. The validation of the attenuation relationships derived is made possible by the availability
of good quality data recorded in Singapore from giant Sumatran earthquakes in December 2004,
March 2005 and September 2007. Some preliminary results have previously been presented in a
workshop held in Singapore in 2007 [19].

2. SUMATRAN MEGATHRUST

The Sunda arc, extending over 5600 km from the Andaman islands in the northwest to the Banda
arc in the east, was formed by the convergence between the subducting Indian–Australian plate

Copyright q 2009 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2010; 39:827–845
DOI: 10.1002/eqe
 SUMATRAN MEGATHRUST EARTHQUAKES 829

Figure 1. Tectonic setting of Sumatra, together with the epicentres of 12 significant earthquakes occurring
on the Sumatran megathrust, from January 2000 to December 2007. The islands off the western coast
of Sumatra are (from North to South): Sm, Simeulue island; Ni, Nias island; Sb, Siberut island; Sp,
Sipora island; NP, North Pagai island; SP, South Pagai island; En, Enggano island. The Mentawai islands
comprise the four islands of Siberut, Sipora, North Pagai and South Pagai.

and the overriding south-eastern Eurasian plate. The Sumatran megathrust of the Sunda arc lies
250 km off the western coast of Sumatra island (Figure 1), with both Sumatra and Java islands
lying on the Eurasian plate. The convergence is nearly orthogonal to the trench axis south of Java,
but it is highly oblique southwest of Sumatra. Based on velocity vectors derived from the regional
Global Positioning System (GPS) data, the pole of rotation for the relative motion between the
two plates is in East Africa, about 50◦ west of Sumatra [20, 21]. Northern Sumatra is closer to
this pole than southern Sumatra; thus, the orientation and magnitude of the relative-motion vector
vary significantly along the Sumatran portion of the plate boundary, as shown in Figure 1.
Slip vectors of moderate earthquakes along this subduction zone were found to be nearly
perpendicular to the strike of the plate boundary [22, 23]. Most of the strike-slip component of
the oblique convergence between the Indian–Australian plate and the Eurasian plate southwest
of Sumatra is accommodated by right-lateral slip along the trench-parallel Sumatran fault, lying
roughly 250 km northeast of the trench [7, 22–24]. Therefore, the slip along the subduction zone
itself has relatively small strike-parallel components.
Six giant earthquakes (Mw 8.0) have occurred along the Sumatran megathrust in the last 250
years, releasing the strain accumulated by the convergence between the two tectonic plates. The
rupture zones of these earthquakes are depicted in Figure 1. The earliest of these historical events

Copyright q 2009 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2010; 39:827–845
DOI: 10.1002/eqe
830 K. MEGAWATI AND T.-C. PAN

was that of February 1797 [9, 25]. The earthquake had an Mw of 8.7 and ruptured the 370-km
segment from 1◦ S to about 4◦ S [25]. This was followed by the giant earthquake of 1833 (Mw 9.0),
which ruptured a 500-km-long segment south of Siberut island, and another one in 1861 (Mw 8.5)
rupturing a 270-km-long segment beneath the Nias island [9, 25, 26].
Since 1861, no giant earthquake with Mw 8.0 had occurred along the Sumatran megathrust
until 26 December 2004, when the Mw 9.15 Aceh-Andaman earthquake happened [10, 27–29].
This was shortly followed by the Mw 8.6 Nias-Simeulue earthquake on 28 March 2005 [30, 31],
which had a rupture zone coincident with that of the 1861 event. The latest giant earthquake of
Mw 8.4 occurred on 12 September 2007, at 11:10:26 GMT, denoted as Event 10 in Figure 1.

3. SUMATRAN-SUBDUCTION EVENTS RECORDED IN SINGAPORE

The Meteorological Services Division of the National Environment Agency established a network
of digital seismic stations in Singapore, in September 1996. The network comprises one broadband
Global Seismographic Network (GSN) station, four teleseismic stations and two borehole arrays.
The GSN station, situated on a very hard rock site, at the centre of Singapore island, is equipped
with a comprehensive set of sensors to record ground tremors continuously, whereas the other
stations operate based on a triggering system.
Between January 2000 and December 2007, 12 Sumatran subduction earthquakes had generated
substantial ground tremors capable of causing perceptible levels of vibration in Singapore. They
include five earthquakes of Mw 7.9. These significant earthquakes are listed in Table I and their
epicentres are plotted in Figure 1. The ground motions from these significant earthquakes are
recorded at the GSN station in Singapore. Although these 12 sets of ground motions are not sufficient
for deriving a representative attenuation relationship empirically, they are useful for the validation
of the spectral attenuation relationship that will be derived hereafter using synthetic seismograms.

Table I. Twelve significant Sumatran subduction earthquakes that caused perceivable tremors in
Singapore, in the last 8 years.
Epicentre
No. Date Time (GMT) Latitude Longitude Depth (km) Mw R (km)
1 04 Jun 2000 16:28:47 4.730◦ S 101.940◦ E 43.9 7.9 704
2 07 Jun 2000 23:45:35 4.630◦ S 101.820◦ E 16.6 6.7 697
3 11 May 2004 08:28:48 0.415◦ N 97.825◦ E 21.0 6.1 670
4 26 Dec 2004 00:58:50 3.090◦ N 94.260◦ E 28.6 9.1 1600
5 28 Mar 2005 16:09:37 1.670◦ N 97.070◦ E 25.8 8.6 747
6 10 Apr 2005 10:29:11 1.680◦ S 99.540◦ E 12.0 6.7 579
7 10 Apr 2005 11:14:20 1.770◦ S 99.640◦ E 15.0 6.5 576
8 17 Apr 2005 21:23:51 1.660◦ S 99.540◦ E 23.0 5.4 578
9 14 May 2005 05:05:19 0.420◦ N 98.240◦ E 39.0 6.7 625
10 12 Sep 2007 11:10:27 4.438◦ S 101.367◦ E 34.0 8.4 619
11 12 Sep 2007 23:49:04 2.625◦ S 100.841◦ E 35.0 7.9 549
12 13 Sep 2007 03:35:29 2.130◦ S 99.627◦ E 22.0 7.0 602
Note: The information for Events 1–9 is compiled from the Bulletin of the International Seismological Centre
(ISC), and that for Events 10–12 is from USGS National Earthquake Information Center (NEIC). R indicates
distance from Singapore to the centre of the respective fault plane.

Copyright q 2009 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2010; 39:827–845
DOI: 10.1002/eqe
 SUMATRAN MEGATHRUST EARTHQUAKES 831

4. SYNTHETIC SEISMOGRAMS

A representative response-spectral attenuation relationship, which is a key ingredient in any seismic

hazard studies, will be derived based on the synthetic seismograms computed for large-magnitude
subduction earthquakes along the Sumatran megathrust. Six cities in the region, namely Singapore,
Kuala Lumpur, Penang, Medan, Pekanbaru and Palembang, are selected as stations where ground
motions are to be synthesized. The locations of the cities are shown in Figure 1. Kuala Lumpur and
Penang are on the western coast of peninsular Malaysia, while Medan, Pekanbaru and Palembang
are on eastern coast of Sumatra island. The selection of these representative stations is made to
have a wide coverage of distance and azimuth from source to station.

4.1. Ground-motion simulation method

The ground-motion simulation method used in the present study follows a kinematic method, in
which the source rupture is represented using a finite-fault model [4, 11, 32, 33]. The fault plane
is subdivided into several subfaults and each subfault is treated as a point source. The rupture
starts at the hypocentre and propagates radially outward with a certain rupture velocity, triggering
each subfault as the rupture front passes its centre. The ground motions at an observation point
produced by the ruptures of individual subfaults are summed with time lags to account for rupture
propagation on the fault plane.
The crustal structure representing the whole region of Sumatra and peninsular Malaysia is
extracted from the global crustal model CRUST 2.0 [34], which is a 2◦ ×2◦ global model for the
Earth’s crust based on seismic refraction data published in the period of 1948–1995. The one-
dimensional structure is summarized in Table II, where the properties of the structure are obtained
by taking the averages of the properties of all 2◦ ×2◦ grids covering the region.
Green’s functions are based on synthetics derived from elastic wave-propagation model, which
provides proper phasing of body and surface waves.

4.2. Source parameters

Earthquakes with magnitudes ranging from Mw 5 to 9, with an interval of 0.5, are simulated.
The strike of the Sumatran megathrust is taken to be N320◦ E, and the dip is 12◦ [10, 11]. The
length and width of the fault plane corresponding to each magnitude level are calculated based on
the empirical relationship proposed by Wells and Coppersmith [35], and they are summarized in
Table III. For each magnitude 12 rupture planes are considered, and they are randomly positioned
between latitudes 6◦ N and 6◦ S. The upper side of the fault planes is assumed to be exposed on the
surface of the uppermost crustal layer. The 12 fault planes for each magnitude level have random

Table II. Crustal structure of the region covering Sumatra and peninsular Malaysia.

Layer H (km) v P (km/s) v S (km/s)  (t/m3 ) QP QS

Upper crust 9.6 6.0 3.4 2.7 350 175
Middle crust 9.5 6.6 3.7 2.9 500 250
Lower crust 9.1 7.2 4.0 3.1 650 325
Mantle ∞ 8.2 4.7 3.4 800 400
H = layer thickness; v P = P-wave velocity; v S = S-wave velocity;  = mass density; Q P = quality factor
of P-wave, Q S = quality factor of S-wave.

Copyright q 2009 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2010; 39:827–845
DOI: 10.1002/eqe
 832

Copyright q
Table III. Rupture parameters of simulated Sumatran subduction earthquakes.
Mw 5.0 Mw 5.5 Mw 6.0 Mw 6.5 Mw 7.0 Mw 7.5 Mw 8.0 Mw 8.5 Mw 9.0
Fault plane
Strike N320◦ E N320◦ E N320◦ E N320◦ E N320◦ E N320◦ E N320◦ E N320◦ E N320◦ E
Dip 12◦ 12◦ 12◦ 12◦ 12◦ 12◦ 12◦ 12◦ 12◦
Fault dimensions along the 3×3 7.0×3.5 12×6 22×11 42×18 80×30 150×45 320×100 600×180
strike and downdip (km×km)
Subfault size (km×km) 3×3 3.5×3.5 6×6 5.5×5.5 6×6 10×10 15×15 20×20 20×20

2009 John Wiley & Sons, Ltd.

Slip distribution
Amplitude Normally distributed with a coefficient of variance (COV) of 0.2
Rake angle Random value between 60◦ and 120◦
Slip velocity vd Random value between 0.2 and 0.4 m/s
No. of asperities 0 0 0 2 2 2 2 2 2
Area ratio of asperities∗ — — — 0.25 0.24 0.21 0.23 0.23 0.22
Slip contrast of asperities† — — — 2.0 2.0 2.0 2.0 2.0 2.0
D̄ ‡ (m) 0.11 0.23 0.44 0.73 1.31 2.32 4.64 6.04 7.61
§
D̄asp (m) — — — 1.46 2.62 4.64 9.28 12.08 15.22
¶
D̄0 (m) — — — 0.49 0.90 1.70 3.25 4.24 5.46
K. MEGAWATI AND T.-C. PAN

Rupture propagation
Rupture velocity Random value between 2.4 and 3.0 km/s
Hypocentre Randomly located along the fault plane, but not within the asperities
∗ Area ratio of asperities is the area of the overall asperities as a fraction of the overall rupture area.
† Slip contrast of asperities is defined as the ratio of the average slip in the asperities to the average slip in the overall rupture area.
‡ D̄ is the average slip of the entire rupture plane.
§ D̄
asp is the average slip of the asperities, which is equal to the product of the slip contrast and D̄.
¶ D̄ is the average slip of the remaining rupture area excluding the asperities.
0

DOI: 10.1002/eqe
Earthquake Engng Struct. Dyn. 2010; 39:827–845
 SUMATRAN MEGATHRUST EARTHQUAKES 833

rupture patterns, which are constrained by the parameters described below. Each fault plane is
subdivided into numerous subfaults, where each subfault is modelled using a point dislocation
source. The size of the subfaults for each magnitude is summarized in Table III.
Some rupture parameters, such as rupture directivity, slip distribution, the presence of asperities,
rupture velocity and slip velocity, cannot be determined accurately for a future earthquake. These
parameters have to be treated as random variables and constrained within a range of seismologically
possible values. To be realistic, each fault rupture is assumed to have two asperities, except for
earthquakes with Mw 6.0, for which no asperity is considered. The total area of the asperities
is confined to be about 22% of the total rupture area. The slip contrast, which is defined as the
ratio of the average slip in the asperity area to the average slip in total rupture area, is fixed at
2.0. These parameters have been based on the analyses of fault asperities of 15 shallow crustal
earthquakes with Mw ranging from 5.7 to 7.2 [36]. The individual asperities are randomly located
along the fault plane without overlap.
The slips of the subfaults are lognormally distributed with a coefficient of variance (COV) of
0.2. The average slip of the whole fault plane and that in the asperity are summarized in Table III.
The rupture initiation may take place at any arbitrary location along the fault plane, but not within
the asperities [36]. The rupture propagates radially from the hypocentre with a velocity vr , which
may range from 2.4 to 3.0 km/s.
The source time function of the slip on each subfault is approximated by a ramp function with a
source duration of tr = L s /vr +td , where L s is the length of the subfault, vr is the rupture velocity
and td is the rise time of the local dislocation. The rise time td is equal to Ds /vd , in which Ds
is the slip amplitude and vd is the slip velocity. The source duration tr is to reflect the effects of
rupture propagation within the subfault and the dislocation rise time.

4.3. Simulation results

Figure 2 presents the velocity time histories simulated at Singapore for earthquakes with three
different Mw values of 5, 7 and 9. The horizontal ground velocities are aligned in the North–South
and East–West directions. The epicentres of the earthquakes are located at about latitude 2◦ S,
and the epicentral distances to Singapore are approximately 650 km. Note that the ground motion
shown for each magnitude level represents only one of the 12 random models simulated for that
particular magnitude level. The upper cut-off frequency of the simulations is 2 Hz. It can be seen in
the figure that the duration of ground motion increases with magnitude, with notable increase from
Mw 7 to 9. This is understandable because the length of the fault plane of the Mw 9 earthquake is
14 times that of the Mw 7 (Table III). With a length of 600 km and a rupture velocity of 2.7 km/s,
it takes 222 s for the rupture of the Mw 9 earthquake to complete if it is unilateral. For the Mw 7
earthquake, it is 16 s.
Figures 3 and 4 summarize the horizontal PGA and RSA with 5% damping ratio at a natural
period of 2 s, respectively, simulated for earthquakes with Mw of 6, 7, 8 and 9. The horizontal
component is represented by the geometric mean value of the NS and EW components. For each
magnitude level, there are 72 data points, resulting from the 12 random rupture models simulated
at the six representative cities. RSA values at different natural periods ranging from 0.5 to 50 s
have also been computed.
Note that the ground motions are simulated for a shear-wave velocity of 3.4 km/s at the ground
surface. This means that the results are for very hard rock sites. For typical rock and soil sites,

Copyright q 2009 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2010; 39:827–845
DOI: 10.1002/eqe
834 K. MEGAWATI AND T.-C. PAN

Figure 2. Velocity time histories from three Sumatran subduction earthquakes with Mw values
of 5, 7 and 9, simulated for Singapore.

such as NEHRP site classes A–E, corresponding site response factors, such as those proposed by
Boore and Joyner [37], should be applied.

5. DERIVATION OF ATTENUATION RELATIONSHIP

The idea of developing an attenuation relationship is to have a relatively simple equation in terms
of earthquake magnitude and source-station distance, which is able to represent the complicated
and time-consuming ground-motion simulations. The functional form adopted for the estimation
of the horizontal ground-motion parameters follows the basic principles of wave propagation in
elastic media as described below:
ln(Y ) = a0 +a1 (Mw −6)+a2 (Mw −6)2 +a3 ln(R)+(a4 +a5 Mw )R +εln(Y ) (1)
where Y is the geometric mean of the horizontal PGA, PGV or RSA values (5% damping ratio)
at various natural periods. The unit for the acceleration values is cm/s2 and that for velocity is
cm/s. Mw is the moment magnitude and R is the distance from the station to the centre of the
corresponding fault plane, in km. The quadratic term of a2 (Mw −6)2 is adopted to account for the
fact that the corner period of earthquake source spectrum increases with the earthquake magnitude
and the source area, and the rate of increase in ground-motion amplitude Y becomes slower for
larger value of Mw [38]. Therefore, the regression coefficient a2 is expected to have negative
values.

Copyright q 2009 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2010; 39:827–845
DOI: 10.1002/eqe
 SUMATRAN MEGATHRUST EARTHQUAKES 835

Figure 3. Attenuation relationship of peak ground acceleration for different Mw values of 6, 7, 8 and 9.

Coefficient a3 in Equation (1) represents the geometrical attenuation rate, whereas a4 and a5
account for the anelastic attenuation. The term εln(Y ) accounts for the variations in the PGA, PGV
and RSA due to the randomness in source parameters considered in the simulations. The term
εln(Y ) has a mean value of 0.0 and a standard deviation value ln(Y ) , representing the standard
deviation of the model due to the randomness in the source process.
The regression coefficients of a0 to a5 are determined to best fit the simulated data using a
least-squares procedure. The procedure minimizes the sum of squared residuals, where the residual
is defined as the difference between the ln(Y ) simulated and that predicted using Equation (1).
The regression coefficients for PGV, PGA and RSA values at various natural periods of engi-
neering interest are summarized in Table IV, together with the standard deviation values. It should
be noted that the values of ln(Y ) , mostly ranging between 0.2 and 0.5, are relatively smaller than
the standard deviation values of other attenuation models for subduction earthquakes. For example,
Youngs et al. [12] propose standard deviation values of 0.6–1.0, Gregor et al. [14] give values of
0.7–0.9, Atkinson and Boore [15] and Lin and Lee [17] suggest 0.5–0.8. It should be understood
that the ln(Y ) in the present study only accounts for the randomness in the source parameters only,
and does not include the randomness in the propagation path. For future uses in regional seismic
hazard analyses, the ln(Y ) values in Table IV need to be increased by about 0.2 to account for the
path effects.

Copyright q 2009 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2010; 39:827–845
DOI: 10.1002/eqe
836 K. MEGAWATI AND T.-C. PAN

Figure 4. Attenuation relationship of response spectral acceleration (5% damping ratio) at a natural period
of 2 s for different Mw values of 6, 7, 8 and 9.

The resulting attenuation relationships for PGA and RSA at a natural period of 2 s are plotted
in Figures 3 and 4, respectively, together with the simulated data. These figures show that the
derived attenuation relationships match closely with the simulated data, indicating the adequacy
of the functional form given in Equation (1).

6. DISCUSSION

6.1. Validation of the attenuation relationships derived

The ground-motion attenuation relationships above were derived based solely on synthetic seismo-
grams. It is therefore necessary to validate these attenuation models. The ground motions recorded
in Singapore from the 12 significant earthquakes listed in Table I are used to validate the attenu-
ation relationships. Geometric-mean of pseudo-acceleration response spectra (5% damping ratio)
of these events is plotted in Figure 5 with thick solid lines. The acceleration response spectra esti-
mated using the derived attenuation relationship are shown by three lines denoted as ‘estimated’ in
each panel of Figure 5. The solid line in the middle indicates the mean spectrum, whereas the two
enclosing dashed lines reflect the mean ± one standard deviation spectra. Note that the standard
deviation values used are the ones that include both the uncertainty in source and path parameters,
namely the values in Table IV plus 0.2. The recorded spectra generally fall within the mean ± one

Copyright q 2009 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2010; 39:827–845
DOI: 10.1002/eqe
 SUMATRAN MEGATHRUST EARTHQUAKES 837

Table IV. Regression coefficients of attenuation relationships for PGV (cm/s), PGA (cm/s2 ) and RSA
(cm/s2 ) with 5% damping ratio for Sumatran subduction earthquakes.
Period (s) a0 a1 a2 a3 a4 a5 ln(Y )
PGV 2.369 2.0852 −0.23564 −0.87906 −0.001363 0.0001189 0.3478
PGA 3.882 1.8988 −0.11736 −1.00000 −0.001741 0.0000776 0.2379
0.50 4.068 1.9257 −0.12435 −0.99864 −0.001790 0.0000564 0.2410
0.60 4.439 1.9094 −0.13693 −0.99474 −0.002462 0.0001051 0.2496
0.70 4.836 1.8308 −0.13510 −0.99950 −0.003323 0.0001945 0.2565
0.80 4.978 1.8570 −0.12887 −1.00000 −0.003054 0.0001475 0.2626
0.90 5.108 1.9314 −0.13954 −0.98621 −0.002986 0.0001075 0.2424
1.00 4.973 1.9547 −0.13913 −0.97603 −0.002851 0.0001106 0.2343
1.20 2.729 2.0316 −0.13658 −0.60751 −0.002570 0.0000409 0.2436
1.50 2.421 1.8960 −0.07075 −0.59262 −0.002453 0.0000668 0.2614
2.00 2.670 1.8182 −0.07657 −0.62089 −0.002190 0.0000674 0.2780
3.00 1.716 1.7922 −0.01895 −0.61167 −0.001177 0.0000121 0.2944
5.00 −0.060 1.8694 −0.09103 −0.32688 −0.001765 0.0000529 0.3963
7.00 0.518 2.1948 −0.24519 −0.47529 −0.001064 0.0000189 0.4206
10.00 0.044 2.3081 −0.29060 −0.50356 −0.000848 0.0000125 0.5183
15.00 −0.525 2.5297 −0.41930 −0.52777 −0.001454 0.0001435 0.4495
20.00 −1.695 2.5197 −0.42807 −0.42096 −0.001575 0.0001498 0.4543
30.00 −2.805 2.6640 −0.42674 −0.43304 −0.001576 0.0001568 0.3686
50.00 −4.340 2.2968 −0.27844 −0.38291 −0.002564 0.0002540 0.3946

standard deviation spectra for most of the events, indicating that the derived spectral attenuation
relationship can predict actual recorded spectra within an acceptable level of uncertainty.
Table V summarizes six existing attenuation relationships that have been derived from inter-plate
subduction earthquakes in Cascadia, Taiwan, Japan and other parts of the world. These attenuation
relationships are well established and have been validated extensively. The spectra resulting from
these six attenuation relationships are to be plotted in Figure 5 to study whether they can be
extrapolated to predict spectra recorded in Singapore from the Sumatran subduction earthquakes.
However, before comparing the spectra predicted by the six attenuation relationships with the
recorded spectra, some adjustments have to be made with regard to difference in source-site
distance definition and site effect.
The two different definitions of source-site distance used by these attenuation relationships,
namely the closest distance to the rupture plane and the hypocentral distance, are different from
the distance definition used in Equation (1), which is the distance to the centre of the rupture plane.
As the present study deals with long-distance earthquakes, this difference in distance definition
would have marginal effects, except for earthquakes with very large rupture area such as the
26 December 2004 earthquake. Nevertheless, this difference in distance definition is considered
when computing response spectra from different attenuation relationships.
Site effect is the second factor that has to be corrected. The six existing attenuation relationships
selected are derived for typical rock sites (NEHRP site class B with shear-wave velocity values
ranging from 760 to 1500 m/s). On the other hand, the ground motions were recorded at the GSN
station, which is located on a very hard rock site with a shear-wave velocity value exceeding
3000 m/s. Boore and Joyner [37] proposed the site responses, which are the combined effects
of amplification and attenuation, for NEHRP site classes B, C and D. The factor for NEHRP
site class B ranging between unity and 1.6 within a frequency band of 0.01–10 Hz (Figure 11 of

Copyright q 2009 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2010; 39:827–845
DOI: 10.1002/eqe
838 K. MEGAWATI AND T.-C. PAN

Figure 5. Recorded and estimated acceleration response spectra (5% damping ratio) of the 12 significant
earthquakes listed in Table I.

Copyright q 2009 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2010; 39:827–845
DOI: 10.1002/eqe
 SUMATRAN MEGATHRUST EARTHQUAKES 839

Table V. Six existing attenuation relationships for subduction earthquakes in other regions.
Mw range Distance definition and range
Reference Model description considered considered
Youngs et al. [12] Regression of recorded ground 5.0–8.2 Rrup =The closest distance to the
motions from inter-plate earth- rupture plane = 10−500 km
quakes occurring in subduc-
tion regions of Alaska, Chile,
Cascadia, Japan, Mexico, Peru
and the Solomon islands
Petersen et al. [13] Using the attenuation relation- 5.0–8.2 Rrup =The closest distance to the
ship derived by Youngs et al. rupture plane = 200−500 km
[12] and scaling the resulting
spectrum down by a factor of
e−0.0038(R−200)
Gregor et al. [14] Regression of synthetic ground 8.0–9.0 Rrup =The closest distance to the
motions from the Cascadia rupture plane = 10−500 km
subduction zone earthquakes
simulated based on a stochastic
finite-fault model
Atkinson and Boore [15] Regression of recorded ground 5.0–8.3 Rrup =The closest distance to the
motions from inter-plate earth- rupture plane = 10−500 km
quakes occurring in subduc-
tion regions of Alaska, Chile,
Cascadia, Japan, Mexico, Peru
and the Solomon islands
Kanno et al. [16] Regression of strong ground 5.5–8.2 Rrup =The closest distance to the
motions recorded in Japan from rupture plane = 1−400 km
1963 to 2003, supplemented by
data from 10 earthquakes in
U.S.A. and Turkey
Lin and Lee [17] Regression of recorded ground 5.5–8.1 Rhypo =The hypocentral
motions from inter-plate subduc- distance = 10−400 km
tion earthquakes in Taiwan and
other regions

Reference [37]) is used to reduce the response spectra predicted by the six attenuation relationships
in Table V before comparing them with the recorded spectra.
In Figure 5, it can be seen that the spectra calculated based on the attenuation relationships
of Youngs et al. [12], Gregor et al. [14], Kanno et al. [16] and Lin and Lee [17] overestimate
the recorded spectra in all the 12 events by one-order of magnitude. The resulting spectra from
Gregor et al. [14] for earthquakes with Mw 7 have a jagged shape, indicating that the attenuation
relationship was derived only for large-magnitude subduction earthquakes (Table V) and should
not be extrapolated to lower magnitude range. Petersen et al. [13] used a scaled-down attenuation
relationship from Youngs et al. [12] in a probabilistic seismic hazard analysis for Sumatra and
peninsular Malaysia. The distance-dependent scaling factor of exp[0.0038(R −200)] was obtained
by fitting the predicted spectral values at a natural period of 1 s from Youngs et al. [12] to the actual
data recorded in Singapore from Sumatran subduction earthquakes [13]. Therefore, the resulting
spectra have identical shape with those from Youngs et al. [12], and fit well with the recorded
spectral values at the natural period of 1 s.

Copyright q 2009 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2010; 39:827–845
DOI: 10.1002/eqe
840 K. MEGAWATI AND T.-C. PAN

Among the six existing attenuation relationships, only the relationships derived by Atkinson
and Boore [15] and Kanno et al. [16] produce spectra with consistent shape to that of the recorded
ones. It predicts spectra with a long predominant natural period of 2 s, which agrees well with that
shown in the recorded spectra and are consistent with the values expected for distant earthquakes
[8, 11, 18, 32, 33]. The other four attenuation relationships produce spectra with short predominant
natural periods, below 0.6 s, which seem too short for distant earthquakes. It shows that these
attenuation relationships were not well constrained at long distance and they are only meant for
near-field earthquakes although the data used covered up to a distance of 500 km.

6.2. Estimation of hazard from future earthquakes on the Sumatran megathrust

Several lines of evidence have indicated that the Sumatran megathrust between 0.5 and 4◦ S
(Figure 1) is very likely to rupture within the next few decades. Large sections of the Sumatran
megathrust have failed progressively over the past 9 years, involving three giant earthquakes
(Mw 8) of Mw 9.15 in December 2004, Mw 8.6 in March 2005 and Mw 8.4 in September 2007
[10, 30, 31, 39, 40]. This sequence of big earthquakes left a portion of the megathrust beneath the
Mentawai islands, between 0.5 and 4◦ S, unruptured in the current episode of strain release. This so-
called Mentawai segment last ruptured in 1797 and 1833, and the interseismic strain accumulated
has approached or exceeded the levels relieved in 1797, indicating that its rupture is very likely
within the next few decades [25, 39].
The expected earthquake magnitude if this segment ruptures depends on the size of the area
that ruptures and the amplitude of the slip involved. The length and width of the segment are 400
and 240 km, respectively. The potential slip on the Mentawai segment is not uniform, but varies
according to the degree of coupling between the subducting and overriding plates [41], the rate of
convergence orthogonal to the trench and the amount of slip released by the series of earthquakes
in September 2007. The potential slip on the segment, as computed by Sieh et al. [39] and shown in
Figures 4 and S44 of Reference [39], varies between 4 and 8 m. This is equivalent to an earthquake
with Mw 8.8.
Figure 6 shows the acceleration response spectrum (5% damping ratio) in Singapore resulting
from the Mw 8.8 earthquake in Singapore. The distance R is taken as 600 km. The acceleration
response spectra estimated using the derived attenuation relationship are shown by the set of three
lines, where the solid line in the middle indicates the mean spectrum and the two enclosing dashed
lines reflect the mean±one standard deviation spectra.
Current building design code for structures in Singapore has been developed largely based on
the BS 8110 [42], which does not have any provision for seismic loading. It does, however, require
that all buildings to be capable of resisting a notional ultimate lateral design load applied at each
floor level simultaneously for structural robustness. These static lateral loads are equal to 1.5% of
the characteristic dead weight of the structure. The minimum capacity of buildings can therefore
be taken as constant at 1.5%g (15 cm/s2 ) across the entire natural period range, as shown by the
horizontal line in Figure 6.
From Figure 6, it is understood that the risk for typical buildings with natural periods ranging
from 0.5 to 5.0 s on rock sites is very low as the demand spectrum is lower than the capacity. For
example, at T = 2 s, corresponding to buildings with 20 storeys widely found in Singapore, the
probability that the demand response spectral value exceeds 1.5%g is only 15.9%. However, the risk
to buildings founded on soft soil sites will be much higher because the demand spectrum would
increase due to soil amplification but the capacity would remain the same. It is therefore important

Copyright q 2009 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2010; 39:827–845
DOI: 10.1002/eqe
 SUMATRAN MEGATHRUST EARTHQUAKES 841

Figure 6. Pseudo-acceleration response spectrum (5% damping ratio) in Singapore resulting from the Mw
8.8 earthquake in the Mentawai segment.

to quantify the site response in Singapore because the ground motion from future earthquake in the
Mentawai segment is likely to affect structures located in soft soil sites and not structures founded
on rock.

6.3. Spectral shape

The shape of a response spectrum depends on two key factors, namely the magnitude of the
earthquake and the distance from the site to the source of the earthquake. It is well accepted that
earthquakes with larger magnitude produce ground motions with richer long-period components
than do smaller magnitude earthquakes. Ground motions also attenuate with distance, where
the high-frequency components attenuate faster compared with the low-frequency counterparts.
Combining these two factors, it is expected that ground motion from a large-magnitude subduction
earthquake at a remote site would have a long predominant period. The recorded response spectra
in Figure 5, all showing predominant periods longer than 1 s, support this argument.
Standard seismic design codes, such as the IBC 2000 [42] and the Indonesian Seismic Code
(SNI-1726-2002), specify the design base shear for buildings as

IC
Vb = W (2)
R
where W is the total dead weight of the structure, R is the strength reduction factor to account
for ductility capacity and inelastic performance of structures and I is the importance factor, which
is taken to be 1.0 for ordinary structures. The period-dependent seismic coefficient C depends on
the location of the structure, which is characterized by the seismic hazard at the site and the local
site condition. The spectral shape of seismic coefficient C is governed by the ordinates of the
pseudo-acceleration at a short natural period of 0.2 s, A(Tn = 0.2 s), and that at 1.0 s, A(Tn = 1.0 s).

Copyright q 2009 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2010; 39:827–845
DOI: 10.1002/eqe
842 K. MEGAWATI AND T.-C. PAN

Figure 7. Normalized acceleration response spectra from the IBC 2000 and three giant
Sumatran subduction earthquakes.

The normalized design spectrum from the IBC 2000 is shown in Figure 7, where the corner period
Tc indicates the intersection between the constant-acceleration and constant-velocity branches of
the spectrum. As the constant-acceleration branch is defined by A(Tn = 0.2 s) and the constant-
velocity branch is determined by A(Tn = 1.0 s), the corner period Tc always lies between 0.2
and 1.0 s. The Tc values for rock sites typically range between 0.4 and 0.6 s. Note that this
spectral shape was derived from statistical analyses of numerous ground motions recorded near the
epicentres, and thus it is suitable for sites where the seismic hazards are controlled by near-field
earthquakes.
The normalized response spectra of three giant earthquakes (Mw 8.0) in Table I are plotted in
Figure 7. Comparing these spectra with the normalized IBC spectrum indicates that the latter is not
particularly suitable for sites where the potential seismic hazard is dominated by large-magnitude,
distant, earthquakes because it may under-predict the hazard at long-period range. Unfortunately,
this spectral shape with Tc = 0.5 s is currently used for the whole territory of Indonesia, regardless
of the local seismicity. For example, eastern coast of Sumatra (Figure 1), where the seismic hazard
is controlled by distant earthquakes on the Sumatran fault running along the western coast and the
subduction zone off the coast, is assigned the same design spectral shape as the high-seismicity
western coast.
Building authorities in Singapore and Malaysia are currently considering incorporation of
earthquake-resistant design to existing building codes. It should be understood that characteristics
of ground motions in Singapore and the Malay Peninsula are unique and do not resemble those in
high-seismicity regions. The shifting of response spectrum towards longer period range should be
carefully taken into account in the formulation of future seismic codes.

7. CONCLUSION

One of the key components required in seismic hazard assessment of a region is a set of repre-
sentative attenuation relationships. A new set of attenuation relationships for Sumatran subduction
earthquakes, covering magnitudes up to 9, have been derived based on synthetic seismograms.

Copyright q 2009 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2010; 39:827–845
DOI: 10.1002/eqe
 SUMATRAN MEGATHRUST EARTHQUAKES 843

It is necessary to develop the attenuation relationships because the existing relationships derived
for other regions were not well constrained for large earthquake magnitudes and/or long-distance
range. They systematically over-predicted the recorded data from the distant Sumatran subduction
earthquakes.
The attenuation relationships derived are for very hard rock site condition and for a long-distance
range between 200 and 1500 km. The derived attenuation relationships should not be extrapolated
to distances shorter than 200 km, for which the relationships derived by Atkinson and Boore [15]
and Kanno et al. [16] may be more appropriate.
The newly derived attenuation relationships have been validated with recorded data from giant
earthquakes on the Sumatran megathrust occurring in the last 9 years. The attenuation relationships
are also used to estimate ground-motion intensity in Singapore due to a highly expected earthquake
scenario on the Sumatran megathrust, and the result shows that the risk level to structures founded
on rock is rather insignificant. The earthquake may, however, pose some problems to structures
founded on soft soil deposits because of significant site amplification.
These recorded data also show that the IBC spectral shape may not be suitable for sites where
the potential seismic hazard is dominated by large-magnitude, distant, earthquakes. The IBC design
spectra, which have been adopted by many countries, were derived for near-field earthquakes
and may under-predict spectral values from large-magnitude, distant, earthquakes, in long natural-
period range. The shifting of response spectrum towards longer period range for distant earthquakes
should be carefully taken into account in the formulation of future seismic codes in Singapore and
Malaysia, where seismic hazard is determined by distant Sumatran earthquakes. The attenuation
relationship derived in the present study can properly reproduce the spectral shape from distant
subduction earthquakes, and could hopefully give insights into the formulation of future seismic
codes.
Although the attenuation relationships derived here are intended for the Sumatran subduction
earthquakes, they may be used as a reference for other regions facing similar threats from distant
earthquakes.

ACKNOWLEDGEMENTS
The authors thank the Building and Construction Authority of Singapore and Dr Jack Pappin of Arup for
the continuous discussions and collaboration. The ground motions recorded at the GSN station in Singapore
were provided by the Meteorological Services Division, National Environment Agency, Singapore.

REFERENCES
1. Atkinson GM, Boore DM. Ground-motion relations for Eastern North America. Bulletin of the Seismological
Society of America 1995; 85:17–30.
2. Toro GR, Abrahamson NA, Schneider JF. Model of strong ground motions from earthquakes in Central and
Eastern North America: best estimates and uncertainties. Seismological Research Letters 1997; 68:41–57.
3. Campbell KW. Prediction of strong ground motion using the hybrid empirical method and its use in the
development of ground-motion (attenuation) relations in Eastern North America. Bulletin of the Seismological
Society of America 2003; 93:1012–1033.
4. Megawati K. Hybrid simulations of ground motions from local earthquakes affecting Hong Kong. Bulletin of the
Seismological Society of America 2007; 97:1293–1307.
5. Pan TC, Sun JC. Historical earthquakes felt in Singapore. Bulletin of the Seismological Society of America 1996;
86:1173–1178.
6. Pan TC, Megawati K, Brownjohn JMW, Lee CL. The Bengkulu, southern Sumatra, earthquake of 4 June 2000
(Mw = 7.7): another warning to remote metropolitan areas. Seismological Research Letters 2001; 72:171–185.

Copyright q 2009 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2010; 39:827–845
DOI: 10.1002/eqe
844 K. MEGAWATI AND T.-C. PAN

7. Sieh K, Natawidjaja D. Neotectonics of the Sumatran fault, Indonesia. Journal of Geophysical Research 2000;
105:28295–28326.
8. Megawati K, Pan TC, Koketsu K. Response spectral attenuation relationships for Singapore and the Malay
Peninsula due to distant Sumatran-fault earthquakes. Earthquake Engineering and Structural Dynamics 2003;
32:2241–2265.
9. Newcomb KR, McCann WR. Seismic history and seismotectonics of the Sunda Arc. Journal of Geophysical
Research 1987; 92:421–439.
10. Chlieh M, Avouac JP, Hjorleifsdottir V, Song TRA, Ji C, Sieh K, Sladen A, Hebert H, Prawirodirdjo L, Bock Y,
Galetzka J. Coseismic slip and afterslip of the great Mw 9.15 Sumatra–Andaman earthquake of 2004. Bulletin
of the Seismological Society of America 2007; 97:S152–S173.
11. Megawati K, Pan TC. Regional seismic hazard posed by the Mentawai segment of the Sumatran megathrust.
Bulletin of the Seismological Society of America 2009; 99:566–584.
12. Youngs RR, Chiou SJ, Silva WJ, Humphrey JR. Strong ground motion attenuation relationships for subduction
zone earthquakes. Seismological Research Letters 1997; 68:58–73.
13. Petersen MD, Dewey J, Hartzell S, Mueller C, Harmsen S, Frankel AD, Rukstales K. Probabilistic seismic hazard
analysis for Sumatra, Indonesia and across the southern Malaysian peninsula. Tectonophysics 2004; 390:141–158.
14. Gregor NJ, Silva WJ, Wong IG, Youngs RR. Ground-motion attenuation relationships for Cascadia subduction
zone megathrust earthquakes based on a stochastic finite-fault model. Bulletin of the Seismological Society of
America 2002; 92:1923–1932.
15. Atkinson GM, Boore DM. Empirical ground-motion relations for subduction-zone earthquakes and their application
to Cascadia and other regions. Bulletin of the Seismological Society of America 2003; 93:1703–1729.
16. Kanno T, Narita A, Morikawa N, Fujiwara H, Fukushima Y. A new attenuation relation for strong ground motion
in Japan based on recorded data. Bulletin of the Seismological Society of America 2006; 96:879–897.
17. Lin PS, Lee CT. Ground-motion attenuation relationships for subduction-zone earthquakes in Northeastern Taiwan.
Bulletin of the Seismological Society of America 2008; 98:220–240.
18. Megawati K, Pan TC, Koketsu K. Response spectral attenuation relationships for Sumatran-subduction earthquakes
and the seismic hazard implications to Singapore and Kuala Lumpur. Soil Dynamics and Earthquake Engineering
2005; 25:11–25.
19. Pan TC, Megawati K, Lim CL. Seismic shaking in Singapore due to past earthquakes. Proceedings of 2007
Workshop on Earthquakes and Tsunamis: From Source to Hazard Singapore, Singapore, 7–9 March 2007. Paper
No. 5.
20. Larson KM, Freymueller JT, Philipson S. Global plate velocities from the global positioning system. Journal of
Geophysical Research 1997; 102:9961–9981.
21. Prawirodirdjo L, Bock Y, Genrich JF, Puntodewo SSO, Rais J, Subarya C, Sutisna S. One century of tectonic
deformation along the Sumatran fault from triangulation and global positioning system surveys. Journal of
Geophysical Research 2000; 105:28343–28361.
22. McCaffrey R. Slip vectors and stretching of the Sumatran fore arc. Geology 1991; 19:881–884.
23. McCaffrey R. Oblique plate convergence, slip vectors, and forearc deformation. Journal of Geophysical Research
1992; 97:8905–8915.
24. Fitch TJ. Plate convergence, transcurrent faults, and internal deformation adjacent to Southeast Asia and the
western Pacific. Journal of Geophysical Research 1972; 77:4432–4460.
25. Natawidjaja DH, Sieh K, Chlieh M, Galetzka J, Suwargadi BW, Cheng H, Edwards RL, Avouac JP,
Ward SN. Source parameters of the great Sumatran megathrust earthquakes of 1797 and 1833 inferred from
coral microatolls. Journal of Geophysical Research 2006; 111(Art. No. B06403).
26. Zachariasen J, Sieh K, Taylor FW, Edwards RL, Hantoro WS. Submergence and uplift associated with the giant
1833 Sumatran subduction earthquake: evidence from coral microatolls. Journal of Geophysical Research 1999;
104:895–919.
27. Ammon CJ, Ji C, Thio HK, Robinson D, Ni SD, Hjorleifsdottir V, Kanamori H, Lay T, Das S, Helmberger D,
Ichinose G, Polet J, Wald D. Rupture process of the 2004 Sumatra–Andaman earthquake. Science 2005;
308:1133–1139.
28. Lay T, Kanamori H, Ammon CJ, Nettles M, Ward SN, Aster RC, Beck SL, Bilek SL, Brudzinski MR, Butler R,
DeShon HR, Ekstrom G, Satake K, Sipkin S. The great Sumatra–Andaman earthquake of 26 December 2004.
Science 2005; 308:1127–1133.
29. Subarya C, Chlieh M, Prawirodirdjo L, Avouac JP, Bock Y, Sieh K, Meltzner AJ, Natawidjaja DH, McCaffrey R.
Plate-boundary deformation associated with the great Sumatra–Andaman earthquake. Nature 2006; 440:46–51.

Copyright q 2009 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2010; 39:827–845
DOI: 10.1002/eqe
 SUMATRAN MEGATHRUST EARTHQUAKES 845

30. Briggs RW, Sieh K, Meltzner AJ, Natawidjaja D, Galetzka J, Suwargadi B, Hsu YJ, Simons M, Hananto N,
Suprihanto I, Prayudi D, Avouac JP, Prawirodirdjo L, Bock Y. Deformation and slip along the Sunda Megathrust
in the great 2005 Nias-Simeulue earthquake. Science 2006; 311:1897–1901.
31. Konca AO, Hjorleifsdottir V, Song TRA, Avouac JP, Helmberger DV, Ji C, Sieh K, Briggs R, Meltzner A. Rupture
kinematics of the 2005 Mw 8.6 Nias-Simeulue earthquake from the joint inversion of seismic and geodetic data.
Bulletin of the Seismological Society of America 2007; 97:S307–S322.
32. Megawati K, Chandler AM. Distant-earthquake simulations considering source rupture propagation: refining the
seismic hazard of Hong Kong. Earthquake Engineering and Structural Dynamics 2006; 35:613–635.
33. Megawati K, Pan TC. Prediction of the maximum credible ground motion in Singapore due to a great Sumatran
subduction earthquake: the worst-case scenario. Earthquake Engineering and Structural Dynamics 2002; 31:
1501–1523.
34. Laske G, Masters G, Reif C. CRUST 2.0: a new global crustal model at 2 × 2 degrees. Institute of
Geophysics and Planetary Physics, The University of California, San Diego. Available from: http://mahi.ucsd.edu/
Gabi/rem.dir/crust/crust2.html [19 January 2008].
35. Wells DL, Coppersmith KJ. New empirical relationships among magnitude, rupture length, rupture width, rupture
area, and surface displacement. Bulletin of the Seismological Society of America 1994; 84:974–1002.
36. Somerville P, Irikura K, Graves R, Sawada S, Wald D, Abrahamson N, Iwasaki Y, Kagawa T, Smith N, Kowada A.
Characterizing crustal earthquake slip models for the prediction of strong ground motion. Seismological Research
Letters 1999; 70:59–80.
37. Boore DM, Joyner WB. Site amplifications for generic rock sites. Bulletin of the Seismological Society of America
1997; 87:327–341.
38. Fukushima Y. Scaling relations for strong ground motion prediction models with M 2 terms. Bulletin of the
Seismological Society of America 1996; 86:329–336.
39. Sieh K, Natawidjaja DH, Meltzner AJ, Shen CC, Cheng H, Li KS, Suwargadi BW, Galetzka J, Philibosian B,
Edwards RL. Earthquake supercycles inferred from sea-level changes recorded in the corals of west Sumatra.
Science 2008; 322:1674–1678.
40. Konca AO, Avouac JP, Sladen A, Meltzner AJ, Sieh K, Fang P, Li Z, Galetzka J, Genrich J, Chlieh M,
Natawidjaja DH, Bock Y, Fielding EJ, Ji C, Helmberger DV. Partial rupture of a locked patch of the Sumatra
megathrust during the 2007 earthquake sequence. Nature 2008; 456:631–635.
41. Chlieh M, Avouac JP, Sieh K, Natawidjaja DH, Galetzka J. Heterogeneous coupling on the Sumatra megathrust
constrained from geodetic and paleogeodetic measurements. Journal of Geophysical Research 2008; 113:B05305.
42. International Code Council. International Building Code, 2000.

Copyright q 2009 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2010; 39:827–845
DOI: 10.1002/eqe