Australian Earthquake Engineering Society 2016 Conference, Nov 25-27, Melbourne, Vic

Seismic Hazard Modelling for Malaysia

Nelson Lam1, Hing-Ho Tsang2, Daniel Looi3 ,Elisa Lumantarna1 and John Wilson2
1
Department of Infrastructure Engineering, The University of Melbourne, Melbourne, Australia.
2
Department of Civil and Construction Engineering, Swinburne University of Technology, Melbourne,
Australia.
3
Department of Civil Engineering, The University of Hong Kong, Pokfulam Road, Hong Kong.

ABSTRACT
Malaysia is to adopt Eurocode 8 (EC8) in the seismic design of building structures. Some members in
the authorship of this paper have been researching into the seismic hazard modelling of the region
surrounding Malaysia for a long time and have been key contributors to the drafting of the Malaysia
National Annex (NA to MS EN1998-1). The draft document has been approved by the Technical
Committee who has been granted the charter by the Malaysian government to undertake the task. At
the time of writing this paper the draft NA to MS EN1998-1 is undergoing the public ballot process.
The purpose of this paper is to explain the principles underlying the derivation of the elastic response
spectrum model for all parts of Malaysia: Peninsular Malaysia, Sarawak and Sabah. Topics covered
include modelling of distant earthquakes generated from offshore sources and local earthquakes in an
intraplate tectonic setting, decisions on zonation, modelling of earthquake recurrences, and ground
motion modelling.
Keywords: seismic hazard modelling, intraplate earthquakes, distant earthquakes, Eurocode 8,
Malaysia

1. Introduction
The purpose of this paper is to outline and explain the seismic hazard model for different parts of
Malaysia which are characterised by very different seismicity conditions. This paper presents
arguments surrounding decisions on zonation, earthquake recurrence modelling, ground motion
attenuation modelling, and the elastic response spectra as derived from probabilistic seismic hazard
analysis (PSHA). The modelling outcomes have been incorporated into the draft National Annex
(NA) to Eurocode 8 (EC8) for Malaysia (MS EN1998-1:2015): Design of structures for earthquake
resistance – Part 1: General rules, seismic actions and rules for buildings.
Peninsular Malaysia is subject to a combination of earthquake hazards generated from various
offshore sources. Most of the seismological studies and hazard modelling undertaken to date have
been based on ground motions generated from distant earthquakes because of their high representation
in the strong motion database (Balendra et al., 2002; Lam et al., 2009; Megawati & Pan, 2010;
Megawati et al., 2005; Pan & Megawati, 2002; Pan et al., 2007; Petersen et al., 2004; Pappin et al.,
2011). Whilst potential hazards generated locally can be significant very limited amount of data has
been recorded from the monitoring of seismic activities within the peninsula. In view of this unique
pattern of combined seismicity a hybrid modelling approach has been adopted to take into account
both the distant and local seismic hazards. Thus the seismic hazard model stipulated for the
Peninsular is a composite model which encapsulates results from PSHA of long distance earthquakes
(which characterises the high period behaviour of the response spectrum) and local earthquakes
(which characterises the low period behaviour of the response spectrum) based on broad source zone
modelling in accordance with global seismicity data. The hybrid approach best capitalises on the
benefits of abundant data of distant events, whilst obtaining robust estimates of locally generated
hazards based on resourcing global information. Details of modelling of the two types of hazards will
be described in separate sections of the paper.
 Australian Earthquake Engineering Society 2016 Conference, Nov 25-27, Melbourne, Vic

Sarawak is also subject to distant seismic hazard from the Kelawit fault and the Bukit Mersing fault
some 500 kms away from the capital city of Kuching but ground motions predicted from these
identified fault sources are very mild. Consequently, the response spectrum model for Sarawak is
essentially based wholly on the considerations of local hazards for the entire period range (covering
both low and high period structures). The seismic hazard modelling of the effects of local earthquakes
affecting the Peninsular and Sarawak is undertaken jointly given their similarities in terms of the
frequency of earthquake recurrence for these parts of Malaysia.
Sabah is in the proximity of areas of high seismicity, unlike Sarawak and Peninsular Malaysia. Many
fault zones and their focal mechanisms have been identified. These local and regional fault sources
include the Belait fault zone, Jerudong fault zone and Mulu fault zone in the south-west near Brunei;
the Crocker fault zone and Mensaban fault zone which lie at the vicinity of Ranau and Kota Kinabalu
in the central-north; the Labuk bay-Sandakan basin zone near to Sandakan; the Pegasus tectonic line
near Lahad Datu; and the Semporna fault in the Dent-Semporna Peninsula Zone (JMM & MOSTI,
2009). Thus, the seismic hazard model for Sabah in the period range of engineering interests is
essentially based on results generated from conventional PSHA of empirical seismicity data based on
events occurring within Sabah. However, the higher period part of the response spectrum can be
affected by distant offshore sources. References are made to a published Uniform Hazard Spectrum
(UHS) derived from an earlier studies (Lam et al., 2015b; 2016) but the UHS had to be modified by
incorporating the most up-to-date earthquake ground motion attenuation model.
The types of seismic activities characterising the hazards for different parts of Malaysia in the low and
high period ranges are summarised in Table 1.
Table 1 Types of seismic activities characterising the hazards for different parts of Malaysia
Part of Malaysia Peninsular Sarawak and SW NE Sabah
Sabah
Low period wholly controlled by wholly controlled dominated by local and
(T ≤ 1.25s) local seismicity by local seismicity regional seismicity
High period wholly controlled by wholly controlled dominated by distant
(T > 1.25s) distant (offshore) by local seismicity (offshore) seismicity
seismicity

The seismic actions model to be presented is for ordinary (Type II) buildings which is defined herein
as “reference seismic actions”, and is based on a notional 475 year return period (RP) being scaled by
a factor of 2/3 of the benchmarked 2475 year RP earthquake.

2. Zonation and recurrence modelling of local seismic hazards

2.1 Peninsular, Sarawak and South-western Sabah

This section is concerned with the modelling of the spatial distribution and frequency of occurrences
of local earthquakes generated within the low seismicity areas of the Peninsular, Sarawak and the
south-western part of Sabah (which does not include the capital city Kota Kinabalu nor Ranau) and is
not to be confused with long distance hazard generated from the island of Sumatra or from the
subduction fault source off-shore of Sumatra which is the subject matter for considerations in a later
section of the paper.
A seismic hazard map is to divide a region into zones in order that the spatial distribution of
frequency of future earthquake occurrences can be communicated to the designer of future facilities.
Such a map is about the predictions of future activities and is not supposed to be a map which merely
shows the scientific record of historical activities. Although there are clear differences between these
two modelling objectives there is no consensus on how the two types of maps should differ from each
other given that what is shown on many seismic hazard maps is mainly reflecting where historical
earthquakes have occurred. Hazard maps produced for regions of low seismicity have been recognised
to be not robust (Stein et al., 2012).
 Australian Earthquake Engineering Society 2016 Conference, Nov 25-27, Melbourne, Vic

In the Malaysian context, it has been reviewed from a survey of the literature conducted by the
authors that no existing Malaysian seismic hazard maps were able to predict hazards surrounding
Perak in the northern part of the peninsular where a magnitude 4.1 earthquake occurred in 2013. Prior
to the occurrence of this earthquake tremor most of the attention has been drawn to areas surrounding
Bukit Tinggi where tremors had been recorded. This is one of many examples that local intraplate
earthquakes could occur in areas that had no precedents of recognisable earthquake activities.
Difficulties with predicting the location of intraplate earthquakes have been recognised by many code
drafting bodies in affected countries such as Australia (being wholly away from any tectonic plate
boundaries from a long distance).
In Australia, challenges associated with such uncertainties are being attended to through the adoption
of a threshold value of Zmin=0.08 (where Z is the seismic hazard coefficient and can be interpreted as
the notional peak ground acceleration (PGA) for design purposes) in order to ensure a minimum level
of protection and resilience against earthquake shaking (Wilson et al., 2015). A minimum hazard
value of 0.08g in Australia is consistent with current hazard specifications in the two largest
Australian cities of Sydney and Melbourne (AS1170.4-2007). Neither cities had ever experienced any
destructive earthquakes in their respective vicinity in the past and yet the 0.08g threshold has been
decided to be necessary for these cities.
In the study for peninsular Malaysia (together with Sarawak and Southwestern Sabah), a broad source
zone modelling approach was adopted in order that the recurrence modelling of potentially destructive
earthquakes (of magnitude exceeding 5) is predicted directly by the number of M>5 event count
(defined herein as N5). The counting first focused on earthquake events occurring on land in stable
continental areas away from the tectonic plate boundaries around the globe. Earthquakes exceeding
magnitude 5 were used in the counting process since the records are more complete and the intraplate
hazard is contributed mainly by events in the range M5 – M6. For the same reason the event count
was based on a period of observation of 50 years. In view of the generally very low rate of
occurrence of intraplate earthquakes the number of events counted was normalised to a standard land
area of 1,000,000 square kilometres (sq km) which is consistent with conventions adopted by Bird et
al. (2010) and by Bergman and Solomon (1980). The presented statistics of event counts is not
sufficient to indicate an exact global average value. However, it is clear that this global average value
is within the range of 5 – 10 M > 5 events occurring in an area of 1,000,000 sq km in the past 50
years, in a 50 year period (Lam et al., 2016). A parameter KD is introduced herein to represent the rate
of recurrence of intraplate events where KD = 1 refers to five events and KD = 2 refers to ten events.
Amid the uncertainties and lack of adequate local information, it is prudent to err on the safe side.
Thus, “10” (i.e., KD = 2) is a reasonable, and conservative, normalised event count to assume provided
that (validated) local data of earthquake occurrence does not infer a higher value.
The rate of seismic activity is conventionally defined using the Gutenberg-Richter magnitude
recurrence relationship of the form:
log10 N(M) = 𝑎 − 𝑏M 1a)

or log10 N(M) = 𝑎5 − 𝑏(M − 5) 1b)

where N(M) may be defined as the expected number of earthquakes ≥ M occurring within an area of
1,000,000 km2 over a 50-year period, and a, a5 and b are defined as the seismic constants.
For KD = 1, a5 = 0.7 or a = 5.2 (being 0.7 + 0.9 x 5) assuming b = 0.9. Similarly, for KD = 2, a = 5.5.
Given these seismological parameters and a suite of representative ground motion prediction
equations (GMPEs), PSHA can be undertaken to quantify ground motion intensities in probabilistic
terms as described in details in Lam et al. (2016) and summarised in the later sections of the paper.
The shortcoming of the broad source zone modelling approach as described is its failure to capture
“hot spots” of destructive historical activities. There is the option of superimposing the modelled
hazard (for the identified hot spots) on a map showing uniform hazard (that has been derived from the
broad source zone model) in order that no area is stipulated with a level of hazard which is below a
certain hazard threshold. Although this is precisely the concept of “background seismicity” which is
 Australian Earthquake Engineering Society 2016 Conference, Nov 25-27, Melbourne, Vic

already well known the modelling methodology presented herein provides the rationale (and
transparency) for quantifying this background hazard level. For the Peninsula in particular, and for
Sarawak and South-western Sabah, no such a hot spot which poses a threat to a centre of population
has been identified. Consequently, the level of hazard to be stipulated for this part of Malaysia can be
based entirely on results derived from the broad source zone model.

2.2 Central and Eastern Sabah

The seismicity of Sabah is represented by two zones: (1) a low seismicity zone which is bounded
between the border with Sarawak and a dividing line located to the south-west of Kota Kinabalu (2)
the rest of Sabah to the northeast of the dividing line (refer Fig. 1 which shows the dividing line).
Zone (1) is essentially part of the Sarawak zone whereas the level of hazard for zone (2) is to be
analysed in accordance with earthquake activities that have been recorded for this part of Sabah. The
subject matter of this sub-section is focused upon zone (2) which comprises the eastern and central
parts of Sabah.
Historical seismic activities in the eastern part of Sabah have been attracting much of the attention for
the past 30 years. However, similar level of activities (measured in terms of number of M>5 events
per unit area) have actually been recorded in central Sabah (near Ranau which is only some 50 km
from the capital city of Kota Kinabalu). In addition, a major active fault has been identified in the
central part of Sabah along with a dense network of fault lines in the south. Although many of the
fault lines have been classified as inactive the possibility of a moderate magnitude (M>6) earthquake
occurring in the area in the future cannot be ruled out.

Figure 1 Seismic hazard zonation across the whole of Malaysia

The M>5 earthquake event count of 4 in the 50 year period over an aggregated land area of 6924
sq.km in Central Sabah (Table 2a) is actually consistent with an event count of 5 over an aggregated
land area of 8780 sq. km in Eastern Sabah (Table 2b) when the number of events has been normalised
to the same area of landmass.
In summary, the frequency occurrence of earthquakes in Central and Eastern Sabah have been shown
to be comparable. The statistics presented herein is in support of the rationale of considering the
whole of Central and Eastern Sabah including Kota Kinabalu and Ranau and the stretch of land
between these cities right up to the eastern coast to be of one level of seismic hazard amid lack of data
showing the underlying pattern in the spatial distribution of future earthquake activities. The majority
of locations in these two parts of Sabah should be classified as areas of moderate seismicity. PSHA
for this part of Sabah will be described in details in the next section which shows a reference PGA
value of 0.12g.
 Australian Earthquake Engineering Society 2016 Conference, Nov 25-27, Melbourne, Vic

Table 2 Listing of local M>5 events in Central Sabah and Eastern Sabah in the 50 year period 1966 - 2016
Region Districts Area Magnitude Approximate Year of
(sq. km) of historical epicentral distance occurrence
earthquakes from Kota Kinabalu
(km)
Central Sabah Kota Kinabalu 816 M5.3 70 1996
Ranau 3556 M5.1 70 1991
Tuaran 1166 M5.9 60 2015
Kota Belud 1386 M5.3 60 2015
Total area = 6924
Eastern Sabah Semporna 1145 M5.0 264 1976
Lahad Datu 6501 M6.2 270 1976
Kunak 1134 M5.3 250 1976
Total area = 8780 M5.6 270 1984
M5.7 300 1994

3. Ground motion modelling of local earthquakes

Once the recurrence behaviour of local earthquakes has been modelled suitable ground motion models
have to be selected forming an important part of the PSHA procedure. The great majority of strong
motion data which many empirical ground motion prediction equations (GMPEs) are based upon were
collected from regions of high seismicity. It is cautioned herein that adapting those GMPEs for use in
low-to-moderate seismicity countries like Malaysia must take into account factors controlling the (a)
waves generation behaviour at the source of the earthquake in a given tectonic setting and (b) waves
modification behaviour of the earth (basement rock) crusts which are not to be confused with the
modification behaviour of near-surface sediments. The Next Generation Attenuation of eastern North
American (NGA-East) database comprises 29000 records from 81 earthquake events that were
recorded from 1379 stations (PEER 2015/04). This database of earthquakes can be taken to be
representative of waves generation behaviour in an intraplate tectonic setting. Ground Motion Models
(GMMs) derived from this database is at present the most elaborate database of intraplate. A literature
review of seismological studies of ground motion models of Eastern North America (ENA) identified
some 40 models that were developed in the period 1983-2014. A subset of 22 models were selected
based on quality and age of data. Further screening managed to reduce the 22 models into 6
representative models (PEER 2015/04). The acronyms for the six selected published ground motion
models (Table 3) are namely: (i) AB95 (ii) SGD02 (iii) A04* (iv) BCA10d (v) BS11 (vi) AB14*.
Results of PSHA showing response spectral acceleration (RSA) values at 0.3 s and 1.0 s based on a
selection of GMMs of NGA-East are superposed on the range of predictions based on the GMMs of
NGA-West2 (Figs. 2a – 2b). Clearly, GMMs namely AB95 and DASG15 are more robust than the
SP15* and PZCT15* models in view of inter-model consistencies. An earlier independent review of
GMMs developed for use in ENA by Ogweno and Cramer (2014) also ranked AB95 favourably in
view of consistencies between the model predictions and field recordings. Predictions from the AB95
and the DASG15 GMMs of NGA-East are overall comparable with predictions from the NGA-West2
and only marginally higher at 0.3s.
The authors had experience of combining the source model of AB95 with the (non-cratonic) crustal
model of generic rock (Boore and Joyner, 1997 which is abbreviated herein as BJ97) for predicting
ground motions generated by intraplate earthquakes. Simulated RSA values for the non-cratonic
version of AB95 based on the classical generic rock class of Boore and Joyner (1997) is representative
of non-cratonic regions. Simulated RSA values for the non-cratonic version of AB95 based on the
classical generic rock class of Boore and Joyner (1997) is representative of non-cratonic regions.
Predictions by the (non-cratonic) model are shown in Fig. 3a and 3b to be significantly higher than
the upper limit of predictions by the NGA-West2 models. The RSA value of 0.25g at 0.3s is translated
to an effective PGA of 0.1g for a return period of 2475 years, or 0.07g (2/3 of 0.1g) for a notional
return period of 475 years.
 Australian Earthquake Engineering Society 2016 Conference, Nov 25-27, Melbourne, Vic

The non-cratonic version of AB95 has also been used to determine the design response spectrum for
Sabah based on modifying PSHA results generated by ARUP (Pappin et al., 2011). A higher
reference PGA value of 0.12g is predicted for a notional return period of 475 years.

Table 3 A selection of ground motion models for use in tectonically stable regions
Acronyms
Literature citations Remarks
in legends
Atkinson and Boore (1995) AB95 BSSA article
Pezeshk, Zandieh, Campbell and Tavakoli (2015) PZCT15 PEER report 2015/04
Darragh et al. (PEER, 2015) DASG15 PEER report 2015/04
Shahjouei and Pezeskh (PEER, 2015) SP15 PEER report 2015/04
Al Noman and Cramer (PEER, 2015) ANC15 PEER report 2015/04
Silva, Gregor and Darragh (2002) SGD02 PEA report 2002
Atkinson (2004) A04 BSSA article
Boore, Campbell and Atkinson (2010) BCA10d BSSA article
Boatwright and Seekins (2011) BS11 BSSA article
Atkinson and Boore (2014) AB14 BSSA article

(a) 0.3s period

(b) 1.0s period

Figure 2a – 2b NGA-East models (cratonic) overlaid on NGA-West2 models for
log10 N = 5.5 – 0.9M
 Australian Earthquake Engineering Society 2016 Conference, Nov 25-27, Melbourne, Vic

(a) 0.3s period

(b) 1.0s period

Figure 3a – 3b Selected NGA-East models (non-cratonic) overlaid on NGA-West2 models for
log10 N = 5.5 – 0.9 M

4. Modelling of distant earthquakes

Earthquake hazards from Sumatra have been generated from two major sources: (1) Sunda Arc
subduction fault source off-shore of Sumatra; and (2) Sumatran strike-slip fault source (Refer Figure
4). The subduction fault source is formed by convergence between the Indian-Australian plate and the
Eurasian plate. Megathrust earthquakes including that of Aceh 2004 (M9.3) and Nias 2005 (M8.7)
events were generated from this fault source. The distance from this fault source to Peninsular
Malaysia is approximately 530 km – 730 km. The Sumatran strike-slip fault source which is located
within the Sumatran island is 1500 km long and is some 300 – 400 km from Kuala Lumpur which is
much closer than from the subduction fault source. The magnitude of recorded historical earthquakes
generated from this fault source within the Sumatran island is limited to around 7.8.
Numerous research groups have contributed to the assessment of the aforementioned far-field seismic
hazards affecting Peninsular Malaysia. Numerous representative GMPEs for predicting ground
motion levels as functions of magnitude and distance have been developed in these studies. The
analysis output depends on the historical earthquake catalogue, completeness criteria, de-clustering
method, source zoning and the use of the logic tree. A literature review undertaken by the authors
provides coverage of some twenty research articles spanning the period 2002 – 2011 (refer a listing of
the literature in the review of Looi et al., 2011). This database features a combination of probabilistic
and deterministic (scenario-based) hazard analysis (PSHA and DSHA) studies. In view of
inconsistencies of the predicted ground motion values from different GMPEs, verification analyses
 Australian Earthquake Engineering Society 2016 Conference, Nov 25-27, Melbourne, Vic

have been undertaken to identify models which give results that match well with instrumented data
collected from the field (Chandler & Lam, 2004). Two GMPEs reported in the literature have been
validated based on benchmarking against ground motion data instrumentally recorded from a long
distance. These two models are: (1) Component Attenuation Model (CAM) and (2) Megawati
attenuation relationship.
CAM was first developed and coded into program GENQKE for generating synthetic earthquake
accelerograms based on stochastic simulations of the seismological model of Atkinson & Boore
(1995) along with the generic crustal model of Boore & Joyner (1997). Refer also Megawati & Pan
(2010) and Lam et al. (2000) for more details. Whilst CAM was initially developed for predictions of
ground motions generated by local earthquakes, the modelling framework was found to be capable of
predicting ground motions generated by large magnitude earthquakes from the far-field. CAM has
successfully demonstrated as capable of modelling distant mega-magnitude earthquake events that
were generated from the Sunda-Arc subduction source offshore of Sumatra and affecting Singapore
and Peninsular Malaysia (Lam et al., 2009; Balendra et al., 2002 and Balendra & Li, 2008). The
simulation model of CAM could have been used to quantify the reduction of hazard across the
peninsular (from west to east) but it is considered prudent not to do so given that the model has been
validated for a site-source distance of up to 600 km. The modelled attenuation of the distant hazard
with increasing distance is so gradual that the change in the level of distant hazard across the
peninsular is very minor. In perspectives, buildings that are vulnerable to collapse and severe damage
in an earthquake is low-rise and medium-rise buildings as opposed to high rise buildings which
respond to long distance earthquakes. In other words, the part of the response spectrum in the low
period range (and the level of PGA) is governed by local hazards anyway. Thus, the long distance
earthquake hazard affecting Peninsular Malaysia is to be based on one response spectral model.
The Megawati’s attenuation relationship for modelling ground motions generated from the Sumatran
fault source (Megawati et al., 2003) and those from the Subduction fault source (Megawati et al.,
2003) was reported in Pan et al. (2007) and was revised in Megawati & Pan (2010). Synthetic
seismograms which were derived from the analysis of a finite-fault kinematic model have been
verified in a manner similar to what has been carried out with CAM. This attenuation relationship is
based on hard rock conditions and site-source distance ranging between 200 and 1500 km. The use of
the developed relationship for making predictions outside this distance range should be treated with
caution.
In addition to the deterministic studies as described above, Pappin et al. (2011) conducted PSHA for
Malaysia based on historical earthquake data which has been recorded over the past 40 years since
1972, along with the use of the attenuation relationship in Pan et al. (2007). Based on the earthquake
catalogue compiled from the USGS database, the seismic source zone was divided into four categories
of seismogenic depth up to 500 km, and an earthquake database in which small events (M <5) and
aftershocks have been removed.
The response spectrum produced from PSHA is known as the Uniform Hazard Spectrum (UHS) in
which contributions from multiple fault sources have been taken into account (Pappin et al.2011). The
attenuation behaviour of the simulated ground motions in the development of the UHS was based on
GMPEs developed by Pan et al. (2007). Different parts of the UHS can be identified with very
different contributory earthquake scenarios. According to the latest PSHA (Pappin et al., 2011),
seismic hazard level varies across the entire Peninsular Malaysia (due to the different distances
measured from the potential earthquake sources), with Penang indicating the highest hazard. Seismic
zoning map could be prepared for the region, but it is considered unnecessary for two reasons: (1) the
attenuation behaviour of very long period waves that are characteristics of earthquakes generated from
a very long distance is very gradual and (2) the low level of contribution of distant earthquakes to the
total hazard in the low to intermediate period range. Thus, the UHS for Penang has been selected as
the basis of the recommended design spectrum model for the entire Peninsula Malaysia.
The UHS model that has been developed initially requires modifications because of subsequent
improvements made to the accuracies of the regional specific attenuation relationships. The original
attenuation relationship of Pan et al. (2007) has been updated to Megawati & Pan (2010). In parallel
 Australian Earthquake Engineering Society 2016 Conference, Nov 25-27, Melbourne, Vic

with improvements made by the Megawati’s model, CAM has also been shown to be able to simulate
ground motions which match the instrumental field recordings from major events including the Aceh
earthquake of 2004 and the Nias earthquake of 2005.

Figure 4 Offshore earthquake generating sources affecting the Peninsular

To achieve a more robust UHS, the attenuation model has been revised in this study to incorporate
both the updated model of Megawati & Pan (2010) and the latest development of CAM (Lam et al.,
2009). A logic tree weighting factor of 0.5 has been allocated to both attenuation relationships in the
aggregation analysis. The modified UHS as presented in Fig. 5 was obtained by an adjustment
procedure comprising the following steps:
a) The original UHS (for Penang) was firstly scaled down by a notional factor of 2.0 (Musson,
1999) in order to obtain the median UHS.
b) Seven earthquake scenarios were selected by calibrating the response spectra based on the use
of the (original) attenuation model of Pan et al. (2007) with the median UHS at three
reference natural periods of 1 s, 2 s and 5 s.
c) The response spectra of the calibrated earthquake scenarios were then re-calculated using the
updated attenuation model of Megawati & Pan (2010) along with CAM based on equal
weightings. The differences in the spectral parameters were represented by the geometric
mean of spectral ratios (SR) at the three reference periods amongst the seven calibrated
scenarios. The SR for other periods of the UHS were determined accordingly based on
interpolation between the three reference periods.
d) The modified UHS was then obtained by scaling the original UHS by the geometric mean SR.
The revised response spectral values in the long period range of 2 s – 3 s based on probabilistic
analysis is approximately double the response spectral values based on deterministic (median)
predictions as published in Lam et al. (2009).
 Australian Earthquake Engineering Society 2016 Conference, Nov 25-27, Melbourne, Vic

Figure 5 The original and adjusted response spectra representing distant hazard affecting the Peninsula

5. Elastic response spectra for different parts of Malaysia

The response spectrum models of Malaysia do not follow the generic EC8 code which stipulates Type
1 and Type 2 spectrum, as per Cl. 3.2.2.2 (2)P of EC8. The model proposed for Peninsular Malaysia
(Fig. 6a and 6b) is a composite (hybrid) model which encapsulates results from the PSHA of recorded
regional earthquakes as well as from the predictions of the local earthquakes based on broad source
zone modelling as described in the earlier part of the paper. This approach best capitalises on the
benefits of abundant data of distant events, whilst obtaining robust estimates of locally generated
hazards.
The response spectrum model for Sarawak and South-western Sabah (Fig. 7a and 7b) is essentially
based on the considerations of local hazards only. The values of PGA for the notional 475 year RP
and the benchmarked 2475 year RP are 0.07g and 0.1g respectively, as for Peninsular Malaysia but
differs in the higher period range (>1.25 s) due to different frequency of occurrence of regional
seismicity. The response spectrum model for the central and eastern parts of Sabah northeast of the
dividing line is essentially based on results generated from conventional PSHA based on recorded
seismicity data (Fig. 8a and 8b). The values of PGA for the notional 475 year RP and the
benchmarked 2475 year RP are 0.12g and 0.18g respectively. Refer to the earlier parts of the paper for
description of the modelling methodology.

(a) (b)
Figure 6 Elastic Response Spectrum on rock (a) displacement and (b) acceleration for Peninsular
Malaysia for Type II ordinary buildings (Design PGA = 0.07g, RP = 475 years)
 Australian Earthquake Engineering Society 2016 Conference, Nov 25-27, Melbourne, Vic

(a) (b)
Figure 7 Elastic Response Spectrum on rock (a) displacement and (b) acceleration for Sarawak and
South-western Sabah for Type II ordinary buildings (Design PGA = 0.07g, RP = 475 years)

(a) (b)
Figure 8 Elastic Response Spectrum on rock (a) displacement and (b) acceleration for Central and Eastern
Sabah for Type II ordinary buildings (Design PGA = 0.12g, RP = 475 years)

6. Summary
The key challenge of quantifying seismic hazard for Malaysia is in how to (i) quantify the recurrence
behaviour of local earthquakes based on the results of a survey of M>5 earthquakes across many
countries in five different continents and (ii) select the most suitable ground motion model (GMM)
which best represents the seismic environment of Malaysia for conducting probabilistic seismic
hazard analysis (PSHA). The additional challenge for modelling the design response spectrum for
Peninsular Malaysia is in encapsulating both long distance (subduction) earthquakes and local
intraplate earthquakes in one response spectrum. Arguments in support of the adopted broad zonation
approach in the modelling of local intraplate hazards have also been presented.

7. References
Atkinson, G.M. (2004), “Empirical attenuation of ground motion spectral amplitudes in southeastern Canada
and the northeastern United States”, Bulletin of the Seismological Society of America 94(3): 1079-1095.
Atkinson, G.M. and Boore, D.M. (2014), “The attenuation of Fourier amplitudes for rock sites in eastern North
America”, Bulletin of the Seismological Society of America 104(1): 513-528.
Atkinson, G.M. & Boore, D.M. (1995), “Ground Motion Relations for Eastern North America”, Bulletin of the
Seismological Society of America 85: 17-30.
Australian Standard: AS 1170.4 (2007), Structural Design Actions – Part 4 Earthquake Actions.
Balendra, T., Lam, N.T.K., Wilson, J.L. & Kong, K.H. (2002), “Analysis of long-distance tremors and base
shear demand for buildings for Singapore”, Engineering Structures 24: 99-108.
Balendra, T. & Li, Z. (2008), “Seismic Hazard of Singapore and Malaysia”, EJSE Special Issue: Earthquake
Engineering in the low and moderate seismic regions of SEA and Australia.
Bergman, E.A. & Solomon, S.C. (1980), “Oceanic intraplate earthquakes: implications for local and regional
interpolate stress”, Journal of Geophysical Research 85(BI0): 5389-5410.
 Australian Earthquake Engineering Society 2016 Conference, Nov 25-27, Melbourne, Vic

Bird, P., Kreemer, C. & Holt, W.E (2010), “A long-term forecast of shallow seismicity based on the global
strain rate map”, Seismological Research Letter 81(2): 184-194.
Boatwright , J. & Seekins, L. (2011), “Regional spectral analysis of three moderate earthquakes in northeastern
North America”, Bulletin of the Seismological Society of America 101: 1769-1782.
Boore, D.M. & Joyner, W.B. (1997), “Site amplifications for generic rock sites”, Bulletin Seismological Soc of
America 87: 327-341.
Boore, D.M., Campbell, K.W. & Atkinson, G.M. (2010), “Determination of stress parameters for eight well-
recorded earthquakes in eastern North America”, Bulletin Seismological Soc of America 100: 1632-1645.
Chandler, A.M. & Lam, N.T.K. (2004). “An attenuation model for distant earthquakes”, Earthquake
Engineering and Structural Dynamics 33: 183–210.
Eurocode 8 - Design of structures for earthquake resistance. EN 1998-1 Part 1: General rules, seismic actions
and rules for buildings.
JMM & MOSTI (2009), Final report on seismic and tsunami hazards and risks study in Malaysia, Academy of
Sciences Malaysia.
Lam, N.T.K., Balendra, T., Wilson, J.L. & Venkatesan, S. (2009), “Seismic load estimates of distant subduction
earthquakes affecting Singapore”, Engineering Structures 31: 1230-1240.
Lam, N.T.K., Wilson, J.L. & Hutchinson, G. (2000), “Generation Of Synthetic Earthquake Accelerograms
Using Seismological Modelling: A Review”, Journal of Earthquake Engineering 4(3): 321-354.
Lam, N.T.K., Tsang, H.H., Lumantarna, E. & Wilson, J.L. (2016), “Minimum loading requirements for areas of
low seismicity”, Earthquake and Structures 11(4): 539-561.
Lam, N.T.K., Tsang, H.H., Lumantarna, E. & Wilson, J.L. (2015a), “An Alternative Probabilistic Seismic
Hazard Assessment Method in Areas of Low-to-Moderate Seismicity”, Procs ASEM15 conf South Korea.
Lam, N.T.K., Tsang, H.H., Lumantarna, E. & Wilson, J.L. (2015b), “Results of Probabilistic Seismic Hazard
Analysis Assuming Uniform Distribution of Seismicity”, Procs 10th Pacific conf Earthquake Engr, Sydney.
Looi, D.T.W., Hee, M.C., Tsang, H.H. & Lam, N.T.K. (2011), “Probabilistic Approach for Determination of
Design Spectrum Model for Peninsular Malaysia: Part One (Background & Methodology) and Part Two
(Literature Review)”, Proceedings of 1-day symposium and 1-day workshop on earthquake engineering in
Malaysia and Asia Pacific Region, 6 – 7 December 2011, Armada Hotel Petaling Jaya, Malaysia.
MS EN1998-1:2015 Eurocode 8: Design of structures for earthquake resistance – part 1: General Rules, Seismic
Actions and Rules for Buildings
Megawati, K. & Pan, T.C. (2010), “Ground-motion attenuation relationship for the Sumatran megathrust
earthquakes”, Earthquake Engineering and Structural Dynamics 39: 827-845.
Megawati, K., Pan, T.C. & Koketsu, K. (2003), “Response spectral attenuation relationships for Singapore and
the Malay Peninsula due to distant Sumatran-fault earthquakes”, Earthquake Engineering and Structural
Dynamics 32: 2241-2265.
Megawati, K., Pan, T.C. & Koketsu, K. (2005), “Response spectral attenuation relationships for Sumatran-
subduction earthquakes and the seismic hazard implication to Singapore and Kuala Lumpur”, Soil Dynamics
and Earthquakes Engineering 25: 11-25.
Musson, R.M.W. (1999), “Determination Of Design Earthquakes In Seismic Hazard Analysis Through Monte
Carlo Simulation”, Journal of Earthquake Engineering 3(4): 463-474.
Ogweno, L.P. & Cramer, C.H. (2014), “Comparing the CENA GMPEs Using NGA-East Ground Motion
Database”, Seismological Research Letters 85(6): 1377-1393.
Pacific Earthquake Engineering Research Centre (2015), NGA-East: Median Ground Motion Models for the
Central and Eastern North America Region, PEER Report No. 2015/04 University of California, Berkeley
PEER, website: http://peer.berkeley.edu/ngawest/index.html.
Pan, T.C. & Megawati, K. (2002), “Estimation of PGA of the Malay Peninsula due to Distant Sumatra
Earthquakes”, Bulletin of the Seismological Society of America 92:3: 1082-1094.
Pan, T.C., Megawati, K. & Lim, C.L. (2007), “Seismic shaking in Singapore due to past Sumatran earthquakes”,
Journal of Earthquake and Tsunami 1(1): 49-70.
Pappin, J.W., Yim, P.H.I. & Koo, C.H.R. (2011), “An approach for seismic design in Malaysia following the
principles of Eurocode 8”, IEM Jurutera Magazine, 22-28.
Petersen M.D., Dewey J., Hartzell S., Mueller C. et al. (2004). “Probabilistic seismic hazard analysis for
Sumatra, Indonesia and across the Southern Malaysian Peninsula”, Tectonophysics 390: 141-158.
Silva, W.J., Gregor, N. and Darragh, R.B. (2002), “Development of regional hard rock attenuation relations for
central and eastern North America”, Report to Pacific Engineering and Analysis, El Centro, C.A.
Stein, S., Geller R.J. & Liu M. (2012). “Why earthquake hazard map often fail and what to do about it”,
Tectonophysics. 562-563:1-25.
Wilson J.L., Lam N.T.K. & Gad E.F. (2015). “Hazard Identification and Behaviour of Reinforced Concrete
Framed Buildings in Regions of Lower Seismicity”, Proceedings of the 10th Pacific Conference on
Earthquake Engineering, 6 - 8 November 2015, Sydney, Australia.