See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/303883853

Sensitivity analysis for probabilistic seismic hazard analysis (PSHA) in the

Aceh Fault Segment, Indonesia

Article in Geological Society London Special Publications · June 2016

DOI: 10.1144/SP441.5

CITATIONS READS

7 1,767

4 authors:

Amalfi Omang Phil R. Cummins

16 PUBLICATIONS 86 CITATIONS
Australian National University
156 PUBLICATIONS 4,088 CITATIONS
SEE PROFILE
SEE PROFILE

David Robinson Sri Hidayati

Geoscience Australia Badan Geologi
35 PUBLICATIONS 345 CITATIONS 22 PUBLICATIONS 324 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Field Survey: Ambient Noise Tomography (ANT) in Western Part of Java Region View project

Updating Indonesian Seismic Hazard Maps 2017 View project

All content following this page was uploaded by David Robinson on 16 August 2016.

The user has requested enhancement of the downloaded file.

 Downloaded from http://sp.lyellcollection.org/ at Australian National University on June 30, 2016

Sensitivity analysis for probabilistic seismic hazard analysis

(PSHA) in the Aceh Fault Segment, Indonesia
AMALFI OMANG1,2, PHIL CUMMINS2,3*, DAVID ROBINSON3 & SRI HIDAYATI1
1
Badan Geologi, Jalan Diponegoro No. 57, Jawa Barat 40122, Indonesia
2
Australian National University, Acton, ACT 2601, Australia
3
Geoscience Australia, Cnr Jerrabomberra Avenue and Hindmarsh Drive,
Symonston, ACT 2609, Australia
*Corresponding author (e-mail: [email protected])

Abstract: Slip rate, locking width and ground-motion prediction equation (GMPE) selection are
important in seismic hazard analysis because they are used to estimate earthquake recurrence, to
limit the maximum magnitude in an earthquake source and to estimate earthquake ground shaking.
In this study, we examine the sensitivity of probabilistic seismic hazard analysis (PSHA) to fault
slip rates, fault locking width and the selection of GMPEs for the Aceh Fault Segment, Indonesia.
The hazard level differences vary considerably owing to changes in these three parameters. There-
fore, careful consideration is needed in applying PSHA in areas of high fault parameter uncertainty.

The Sumatran Fault System (SFS) in Indonesia is a 1981). A more recent study by Ito et al. (2012) sug-
major right-lateral system consisting of 19 seg- gests a locking width of between 1 and 30 km, with a
ments. It runs parallel to the Sumatra Subduction slip rate of 20 + 6 mm a21.
Zone, where the Australian Plate subducts beneath This uncertainty in fault parameters is poten-
the Burma Microplate and Sunda Block (Fig. 1). tially important in seismic hazard analyses, both
Sumatra is a classical example of a partitioned sub- probabilistic and deterministic, which requires an
duction zone, where much of the trench-normal assessment of the future earthquake potential in a
plate motion is accommodated by the megathrust region. In this paper, we explore the sensitivity of
plate boundary, but much of the trench-parallel probabilistic seismic hazard analysis (PSHA) on
motion is accommodated by the SFS. According the Aceh segment to three key parameters: slip
to historical and instrumental data, most of the 19 rate; maximum magnitude as determined by locking
segments have ruptured within the last century. width; and the selection of ground-motion predic-
The largest recorded magnitudes for events on the tion equations (GMPEs: also known as attenuation
segments vary between Mw 6.5 and 7.7 (Natawidjaja functions). Slip rate is important in determining
& Triyoso 2007). seismic hazard because it is used to estimate earth-
Knowledge of the slip rate and locking depth quake recurrence (how often a fault is expected to
along the SFS is sparse, and has changed little host an earthquake). Earthquake hazard estimates
since the geological study of Sieh & Natawidjaja are sensitive to slip rate because slip rate controls
(2000) and the geodetic study of Genrich et al. the overall rate of earthquake activity on a fault
(2000), and has been summarized more recently (e.g. Youngs & Coppersmith 1985; Kramer 1996).
by Natawidjaja & Triyoso (2007). Geological stud- Estimating the maximum expected earthquake
ies generally suggest an increase in slip rate from on a specific fault or earthquake source is also
11 mm a21 below the equator to 27 mm a21 above it, important in PSHA (e.g. Beauval & Scotti 2004).
while geodetic estimates of slip vary between 19 and It is rare, however, that the largest possible earth-
26 mm a21, with no apparent trend along strike. quakes along individual faults have occurred during
Locking depths determined from GPS measure- the historical recording period. Therefore, maxi-
ments (Genrich et al. 2000) vary between 9 and mum magnitude is often evaluated from estimates
56 km. Knowledge of slip rate and locking depth of fault dimensions (e.g. Wells & Coppersmith
is particularly sparse for the northernmost tip of 1994; Leonard et al. 2007), and these dimensions
the SFS, where it splits into two fault segments: the are usually determined from a combination of geo-
Aceh and the Seulimeum faults (Fig. 1). Slip rates logical and geodetic observations.
for this segment vary from as low as 5 mm a21 The latest Indonesian seismic hazard map
(Genrich et al. 2000) to 38 mm a21 (Bennett et al. (Irsyam et al. 2010) assumed a constant locking

From: Cummins, P. R. & Meilano, I. (eds) Geohazards in Indonesia: Earth Science for Disaster Risk Reduction.
Geological Society, London, Special Publications, 441, http://doi.org/10.1144/SP441.5
# 2016 The Author(s). Published by The Geological Society of London. All rights reserved.
For permissions: http://www.geolsoc.org.uk/permissions. Publishing disclaimer: www.geolsoc.org.uk/pub_ethics
 Downloaded from http://sp.lyellcollection.org/ at Australian National University on June 30, 2016

A. OMANG ET AL.

Fig. 1. The Sumatran Fault System (SFS) is a major right-lateral system that runs parallel to the Sumatra
Subduction Zone, and is here indicated by its major segments plotted in alternating red and blue traces. The velocity
vectors of the Australian Plate and Sunda Block (in the ITRF2002 reference frame) are indicated, along with the
Burma Microlate, the velocity of which is omitted. The black rectangle indicates the area illustrated in Figure 2.

width of 17 km (depth of 3–20 km) for all SFS Sources of uncertainty in seismic
segments. However, Ito et al. (2012) used GPS hazard analysis
to study the locking width of the Aceh Fault Seg-
ment and found uncertainty in its size, which Seismic hazard may be analysed deterministically
ranged from a few kilometres up to 35 km, twice (e.g. when a particular earthquake scenario is
that used by Irsyam et al. (2010). Since earthquake assumed) or probabilistically, in which all possible
magnitude depends on the area of rupture, which earthquakes of various magnitudes, locations and
is the product of fault length and locking width, probabilities of occurrence are explicitly considered
the maximum magnitude on a fault is expected (Kramer 1996). We focus here on PSHA, which can
to vary with locking width. However, the uncer- be summarized in four steps (Reiter 1990):
tainty in fault locking width is rarely considered (1) Identification and characterization of earth-
in PSHA. quake sources. These include:
Another important aspect to consider in PSHA is (a) area source zones: in which an earth-
the selection of the GMPEs. There are no GMPEs quake can occur anywhere in a certain
derived for the Indonesian region. Consequently, area with a probability that is typically
Indonesia’s seismic hazard analyses adopt GMPEs determined from an analysis of earth-
from other regions. For example, Irsyam et al. quake catalogue data (i.e. a recurrence
(2010) used Boore & Atkinson (2008), Campbell relationship);
& Bozorgnia (2003) and Chiou & Youngs (2008) (b) fault sources: where earthquakes can
in equal weights to estimate ground motion from occur on particular faults with a specified
crustal fault sources. Sabetta et al. (2005) demon- probability typically determined from
strated that the selection of GMPEs has a greater either catalogue data, or from slip rate
impact than the expert judgement applied in assign- and locking width estimated using either
ing relative weights to the GMPEs. geological or geodetic observations.
 Downloaded from http://sp.lyellcollection.org/ at Australian National University on June 30, 2016

SUMATRA FAULT PSHA SENSITIVITY ANALYSIS

(2) Application of GMPEs to forecast the level of measurable uncertainties. In the case of fault slip,
shaking or propagation of ground motion. the uncertainties are estimated from the measure-
GMPEs are determined from analyses of ment error in offset length and in the dating, and
strong-motion data. the uncertainty in locking width is estimated
(3) Integration (or summation) over the uncertain- from the analysis of GPS velocities. We therefore
ties in earthquake location, earthquake size consider the uncertainty in fault parameters as alea-
and ground-motion parameters to obtain the tory, and attempt to formally include this as part
probability that the ground-motion para- of the PSHA in a way analogous to that used for
meter will be exceeded during a particular time GMPEs.
period.

Steps (1a), (b) and (2) all introduce uncertainty into

Methodology for sensitivity analysis
the PSHA analysis. The GMPEs of step (2) are each The modelling in this study utilizes the Earthquake
associated with an aleatory uncertainty that reflects Risk Model (EQRM), which is a computer package
the observed variability in the ground-motion mea- for estimating earthquake hazard and earthquake
surements and is formally included in the PSHA. In risk (Robinson et al. 2005, 2006). The EQRM is
addition, the epistemic uncertainty associated with developed by Geoscience Australia and can be
the lack of knowledge about which GMPE best freely downloaded from http:/code.google.com/
describes the actual ground motion is usually p/eqrm/source/checkout
accounted for via a logic tree. The EQRM uses an event-based approach. This
However, none of the parameters describing the approach differs from the traditional ‘site-based’
recurrence relationships of step (1a) and the fault approach to PSHA integration over the magnitude
behaviour in step (1b) are known perfectly, but and distance distributions on a site-by-site basis.
their uncertainly is rarely considered (except for The traditional approach was introduced by Cornell
maximum magnitude, for which alternative val- (1968) and was summarized by McGuire & Ara-
ues are sometimes considered in a logic tree). In basz (1990). In contrast, an event-based analysis
this study, we explore the sensitivity of the PSHA with the EQRM begins with the generation of a sin-
results to the uncertainty in fault parameters, namely gle simulated event catalogue, which is, in turn,
fault slip rate and maximum magnitude (as deter- used to generate ground-motion fields at all sites.
mined by locking width). Although PSHA results Both the ‘site-based’ and ‘event-based’ approaches
are also affected by uncertainty in recurrence solve the same equations – they differ only in
parameters for aerial source zones (step 1a), an numerical mechanics. The generation of the syn-
assessment of this sensitivity is beyond the scope thetic event catalogue relies on a model for the
of this paper. seismicity in the region. Typically, the model of
In order to account for uncertainty in fault seismicity comes from an interpretation of histori-
parameters in PSHA, we need to consider whether cal earthquakes, geology and neotectonics (Robin-
this uncertainty is best described as aleatory or epi- son et al. 2005). In this paper, the seismicity
stemic. Aleatory is often defined as the natural var- model consists of the geometry and slip rate of
iability of a quantity that cannot be reduced by the Aceh Fault Segment.
making additional measurements (McGuire 2004). We focus on the Aceh Fault Segment, located in
The uncertainty in GMPEs is normally regarded as northern Sumatra, which typically hosts events with
an aleatory uncertainty. Epistemic uncertainty, a strike-slip mechanism. Its length is approximately
however, is usually regarded as uncertainty due to 200 km (Fig. 2). In the Irsyam et al. (2010) seismic
a lack of knowledge and can, in principle, be hazard map, the slip of the Aceh segment is
reduced if additional data become available. Uncer- 2 mm a21, the locking width is 17 km and the haz-
tainty in maximum magnitude, and that represented ard is derived using three GMPEs, all with equal
by choice of GMPE, are often regarded as episte- weights. The GMPEs used are Boore & Atkinson
mic uncertainties that are accounted for via the (2008), Campbell & Bozorgnia (2003) and Chiou
use of logic trees. We prefer the more practical de- & Youngs (2008) (Irsyam et al. 2010).
finitions of Bommer (2003), in which aleatory We study the sensitivity of fault slip rates, uncer-
uncertainty is defined as uncertainty that can be mea- tainty in locking width and different GMPEs for
sured, while epistemic uncertainty is that which PSHA by comparing and analysing hazard curves
must be judged. at two different sites. Both sites are located near
The estimations of the fault parameters slip the middle of the fault trace, with Site 1 located
rate and locking width used in this paper are based on top of the fault line (0 km) and Site 2 located
on measurements of offsets in dated geological 20 km perpendicularly from the segment mid-point
formations and geodetic strain rates, respectively. (see Fig. 2). We also compare the hazard maps for
Both types of observations are associated with Aceh using different input parameters.
 Downloaded from http://sp.lyellcollection.org/ at Australian National University on June 30, 2016

A. OMANG ET AL.

Fig. 2. The location of GPS measurements made by Ito et al. (2012), along with surface traces of the Aceh,
Seulimeum and Tripa segments of the SFS. The two sites used for PSHA calculation are denoted by orange boxes
marked 1 and 2.

Slip-rate sensitivity analysis fault, also indicates a similar trend (Fig. 3b). At a
500 year return period, the hazard level for a higher
Reported slip rates for the Sumatran Fault System slip rate reaches 0.31g, while the hazard level for a
(SFS) are based on geological and geodetic studies. lower slip reaches 0.09g. This represents a 300%
The latest seismic hazard map of Indonesia (Irsyam variation. As expected, the hazard level at the sec-
et al. 2010) assigned slip of 2 mm a21 to the Aceh ond site illustrates that hazard decreases as we
Fault. However, recent work by Ito et al. (2012) esti- move away from the fault. These hazard curve
mated the slip rate in the middle of the Aceh seg- comparisons show that seismic hazard analysis is
ment (Profile A in Fig. 2) to be about 20 + 6 sensitive to fault slip rate (in fact, it is linear in
mm a21. Since slip-rate estimates are used to deter- slip rate) and demonstrate that the distance from
mine the recurrence interval, it is important to study an earthquake source plays an important role in
how hazard varies with slip rate. Our study focuses the hazard level.
on the sensitivity of PSHA to fault slip rate. We do We also compare the hazard maps derived using
not attempt to determine which slip-rate estimates these two slip rates for the Aceh Fault Segment. Fig-
are most suitable for the Aceh segment. ure 4 shows a comparison for the Aceh hazard maps
Figure 3a, b illustrates a comparison between the at a 500 year return period. Figure 4a is the hazard
hazard curves for two different slip rates at sites 1 map using 2 mm a21, while Figure 4b is the hazard
and 2. The green and red dashed lines are the hazard map using 20 mm a21. Both maps demonstrate that
curves for 2 and 20 mm a21 slip, respectively. the highest hazard level is located along the fault
Figure 3a shows the hazard curves comparison at a trace, and that hazard decreases as we move further
site located in the middle of the fault (Site 1). The from the fault. The main differences between these
hazard level is significantly higher for larger slip two maps are the maximum hazard level and the
rate at all return periods. At a 500 year return period, area impacted. A higher slip rate clearly leads to a
which is commonly used in PSHA, the hazard dis- greater maximum of ground shaking and a larger
crepancy is significant: 0.83g for 20 mm a21 slip area impacted by future earthquakes. Consequently,
and 0.18g for 2 mm a21 slip. This represents a var- the recent work by Ito et al. (2012) suggests that the
iation in hazard of approximately 400%. The com- hazard level in the Aceh area (Irsyam et al. 2010)
parison at the second site, located 20 km from the is underestimated.
 Downloaded from http://sp.lyellcollection.org/ at Australian National University on June 30, 2016

SUMATRA FAULT PSHA SENSITIVITY ANALYSIS

Fig. 3. Hazard curves comparison for different parameters in this study. (a) shows the comparison between hazard
curves for slip rates of 2 and 20 mm a21 at Site 1, while (b) shows the comparison at Site 2. (c) & (d) show the
comparison of hazard curves comparison for different maximum magnitudes. The grey shaded hazard curves on (c)
& (d) show the hazard curves derived from 1000 samples of the magnitude PDF at sites 1 and 2. (e) & (f) show the
comparison of hazard curves using different GMPEs at sites 1 and 2.
 Downloaded from http://sp.lyellcollection.org/ at Australian National University on June 30, 2016

A. OMANG ET AL.

Fig. 4. Comparison of hazard maps for the Aceh Fault Segment at a return period of 500 years. (a) shows the
hazard map for a slip rate of 2 mm a21, while (b) shows the hazard map for the 20 mm a21 slip rate.

Locking width (maximum magnitude) To study how broadly the hazard level might
sensitivity analysis vary due to aleatory uncertainty in the locking width,
we undertook a detailed sensitivity study by sam-
Another key parameter in PSHA is the maximum pling the maximum magnitude PDF tightly (1000
magnitude, which can be related to the locking samples) and then plotted all the hazard curves
width. Ito et al. (2012) studied GPS constraints on (the grey shaded hazard curves in Fig. 3c, d). The
locking width for two cross-sections on the Aceh hazard curve from the largest maximum magnitude
Fault Segment (Fig. 2). Their results are presented in the PDF is presented as the dashed red line, while
as locking width probability density functions the smallest maximum magnitude is presented as
(PDFs) for two profiles in Figure 5. We use the lock- the magenta dashed line. These hazard curves indi-
ing width data from Ito et al. (2012) to estimate a cate a large variation in hazard at all return periods.
PDF for the maximum magnitude using the formula However, the largest discrepancies occur at longer
from Wells & Coppersmith (1994) (Table 1) that return periods.
relate rupture width to magnitude. For the purpose The maximum magnitudes based on the Ito et al.
of this exercise, we assume a mapped length of (2012) locking widths are also plotted as a maxi-
200 km for the Aceh Fault Segment. mum magnitude PDF (Fig. 6). The mean value for

Fig. 5. Probability density functions for (a) the locking width, from the study of Ito et al. (2012), and their
conversion to (b) the magnitude PDF (blue) and cumulative PDF (green curve).
 Downloaded from http://sp.lyellcollection.org/ at Australian National University on June 30, 2016

SUMATRA FAULT PSHA SENSITIVITY ANALYSIS

Table 1. Regressions of rupture (RW) width and moment magnitude (M) (Wells & Coppersmith 1994)

Coefficients

Equation Slip type a b Magnitude range

M = a + b∗ log (RW) Strike-slip 3.80 2.59 4.8– 8.1

Reverse 4.37 1.95 4.8– 7.6
Normal 4.04 2.11 5.2– 7.3
log (RW) = a + b∗ M Strike-slip 20.76 0.27 4.8– 8.1
Reverse 21.61 0.41 4.8– 7.6
Normal 21.14 0.35 5.2– 7.3

The regressions are derived using a worldwide database for 421 historical earthquakes.

the maximum magnitude from this PDF is Mw 6.13, As described above, the five median (i.e. from
which is considerably smaller than the Mw 7.7 used each quantile) maximum magnitudes were used sep-
by Team-9 (Irsyam et al. 2010). arately in the hazard calculations and used to derive
Standard PSHA tools, such as the EQRM we use the weighted hazard curve. The weighted hazard
here, do not provide for the incorporation of aleatory curve is the weighted sum of the five hazard curves
uncertainty in fault parameters, but allow only for using equal weights of one-fifth (see Table 2). We
epistemic uncertainty to be included using logic also compared the difference between the weighted
trees. Therefore, in order to incorporate the maxi- hazard curves using three, five and seven quantiles.
mum magnitude aleatory uncertainty into our PSHA, As the latter two results were similar, we reasoned
we had to borrow the tools of epistemic uncertainty that five quantiles are sufficient to represent the
and then apply them to an aleatory uncertainty: that aleatory uncertainty.
is, we used a logic tree approach with parameters To assess the differences between hazard curves
chosen in such a way as to account for the aleatory utilizing different maximum magnitude estimates
uncertainty in the maximum magnitude. To achieve and the hazard curve that accounts for aleatory
this, we sampled the maximum magnitude PDF by uncertainty by combining the five quantiles, we
dividing it into five different quantiles with an compared the hazard curves for two different loca-
equal one-fifth area below the PDF (Fig. 6). The tions in the Aceh Fault Segment (Fig. 1). The com-
median values for each of the five quantiles can be parison between the hazard curves for the five
seen in Table 2. median maximum magnitudes and also the corre-
sponding weighted hazard curve derived from
these five median hazard curves at sites 1 and 2
can be seen in Figure 3c, d.
Figure 3c shows a comparison for Site 1, located
in the middle of the Aceh Fault. The hazard curves
for different maximum magnitudes differ widely
at most return periods. At shorter return periods,
the hazard level for lower maximum magnitude is
higher than the hazard curve with the largest maxi-
mum magnitude. For instance, a comparison
between the Mw 4.8 and Mw 7.1 hazard curves at a
100 year return period show that the peak ground

Table 2. The median and weight values used

to sample the maximum magnitude PDF

Division Median Weight

Fig. 6. Magnitude PDF derived from Ito et al. (2012) Five quantiles 4.8 1/5
data. The solid curve is the probability density, while 6.0 1/5
the dashed curve is the cumulative density. The PDF is 6.4 1/5
divided into five quantiles, each of which has a 6.7 1/5
probability of one-fifth. The grey line indicates the 7.1 1/5
median value for each quantile.
 Downloaded from http://sp.lyellcollection.org/ at Australian National University on June 30, 2016

A. OMANG ET AL.

acceleration (PGA) for the lower magnitude is the maximum magnitude PDF derived from the
0.65g, and 0.56g for the higher magnitude. How- results of Ito et al. (2012), show how wide the
ever, at a longer return period (100 kyr), the hazard uncertainty in the hazard level might be due to
level for the lower magnitude reaches 0.9g, while the uncertainty in the locking width from the geo-
the hazard level for a higher magnitude is 1.84g. detic observations in northern Sumatra. The maxi-
This result is consistent with that of Youngs & Cop- mum hazard level from these data varies between
persmith (1985), who studied a similar variation 0.7g and 1.8g. These hazed grey plots also show a
in hazard curve changes with maximum magnitude similar trend. The hazard level for a lower maxi-
given a constant moment rate. They found that mum magnitude at shorter return period would be
lowering the maximum magnitude with constant higher than the hazard level for higher magnitude
moment rate forces the activity rates of smaller at a shorter return period.
earthquakes to increase, so that hazard levels The second comparison at Site 2 (Fig. 3d) shows
increase for short return periods. The approach that a similar trend with the comparison at Site 1
we used does not assume a constant moment rate (Fig. 3c). However, the probability for exceeding
because, for each different maximum magnitude, a certain ground-motion level (e.g. 0.5g) occurred
we also change the moment rate based on the lock- at a longer return period. The difference varies
ing width for each magnitude. Nevertheless, we between 0.04g and 0.06g at a return period of
observe a similar effect. 500 years. At a return period of 1000 years, the dif-
The hazard curve at Site 2, located 20 km from ference is about 0.1g: while at a 2500 year return
the fault, is shown in Figure 3d. The hazard curves period, it reaches 0.16g. The overall hazard
show a similar trend to those of the previous com- level has decreased, which is simply owing to the
parison. The main difference is that the hazard increased distance from the fault.
level decreases significantly at each maximum mag- Since the hazard map consists of many sites at a
nitude. This is simply due to the increased distance single return period, it is important to note that the
from the fault. hazard level for lower maximum magnitude is
We are also interested in understanding the higher than the hazard level for higher maximum
implication of uncertainty in the maximum magni- magnitude at shorter period. Based on this result, a
tude on our understanding of hazard, as illustrated careful consideration is needed before we assign a
in the latest Indonesian seismic hazard map. To certain maximum magnitude for PSHA work.
explore this, we considered the hazard contribution
from the Aceh Fault Segment using the maximum
magnitudes of the Aceh Fault corresponding to the Ground-motion prediction equation
Mw 7.7 used by Irsyam et al. (2010), the mean of (GMPE) sensitivity analysis
the maximum magnitude PDF of Mw 6.13 of Ito
et al. (2012), the weighted hazard curve using the The other aspect that we study is the impact from the
five quantile maximum magnitudes (as given in selection of GMPEs. Many studies mention the
Table 2) and results from 1000 maximum magni- importance of the GMPE selection for PSHA (e.g.
tudes sampled from the maximum magnitude PDF Lombardi et al. 2005; Sabetta et al. 2005; Bommer
of Ito et al. (2012). These results for sites 1 and 2 & Abrahamson 2006). The problem of GMPE selec-
(Fig. 2) can be seen in Figure 3c, d. tion is further complicated in Indonesia as there are
The first hazard curve comparison for Site 1 no GMPEs derived from Indonesian earthquakes.
(Fig. 3c) shows that the hazard level for higher max- The GMPEs used in the most recent hazard map
imum magnitudes is smaller at shorter return peri- of Irsyam et al. (2010) are Boore & Atkinson
ods. At a 500 year return period, the hazard level (2008), Campbell & Bozorgnia (2003) and Chiou
for Mw 7.7 is 0.85g, while the weighted magnitude & Youngs (2008). The seismic hazard map of
and the Mw 6.13 are 0.99g and 1.08g. The difference Irsyam et al. (2010) used an equal weight for each
is significant, reaching up to 0.23g. At a 1000 year of these GMPEs. We examine the contribution
return period, the difference in hazard level between from these three GMPEs at sites 1 and 2, and com-
these three different magnitudes becomes smaller, pare the results with the weighted results. We also
about 0.03g –0.17g. At a 2500 year return period, compare the hazard curves and hazard maps using
the hazard differences are much smaller, at between each of the GMPEs used in the Irsyam et al.
0.01g and 0.07g. However, at very long return (2010) seismic hazard map. As with Irsyam et al.
periods (100 kyr), the hazard level from the largest (2010), we use equal weights when combining the
maximum magnitude (Mw 7.7) reaches 1.83g, GMPEs.
which is larger than the weighted magnitude and Figure 3e, f show the hazard curve comparisons
Mw 6.13, which have hazard levels of 1.59g and at sites 1 and 2. The first comparison at Site 1 (0 km)
1.66g, respectively. The hazed grey lines, which shows that Boore & Atkinson (2008) (BA08) and
represent the sampled maximum magnitudes from Chiou & Youngs (2008) (CY08) lead to larger
 Downloaded from http://sp.lyellcollection.org/ at Australian National University on June 30, 2016

SUMATRA FAULT PSHA SENSITIVITY ANALYSIS

hazard estimates than Campbell & Bozorgnia magnitude (a shorter return period) is higher than
(2003) (CB08). The result for the weighted combi- the hazard level for higher magnitude at shorter
nation of GMPEs sits in the middle. The hazard return periods. While this has been found to be the
level for the weighted combination is about 23% case when moment rate is held constant (Youngs
larger than CB08. However, compared to BA08 & Coppersmith 1985), we find that it is still true for
and CY08, the weighted combination hazard level the case of the Aceh Fault, even when the moment
is about 13 and 11% smaller, respectively. The haz- rate increases as the locking width increases.
ard level for BA08 is 5% smaller than CY08 GMPE selection, however, shows a very signifi-
for shorter return periods (,600 years), and larger cant difference in the hazard level, and we found
by approximately 10% for longer return period that the difference may vary at difference distances.
(.800 years). A GMPE that results in the highest hazard level at a
The second comparison at Site 2 (20 km) shows certain distance might have smaller hazard level at a
a different result (Fig. 3f). The hazard level for different distance. However, given the fact that there
CY08 is larger than BA08 and CB08 for all return is no single GMPE derived for Indonesia, a further
periods. This shows that the hazard level depen- study is needed in order to select the most reliable
dence on GMPE depends on distance, highlight- GMPE(s) for use in seismic hazard analysis in
ing the important role of GMPE selection on Indonesia.
seismic hazard analysis. The weighted combination
GMPEs in this case also sits in the middle between
CB08, BA08 and CY08. The main difference com- Conclusions
pared to the first site is that the overall hazard level
decreases owing to the increased distance from In this study, we present a sensitivity analysis for
the fault. three parameters used in probabilistic seismic haz-
ard analysis. We analysed hazard curves and maps
for two different sites in the Aceh Fault Segment,
Discussion Indonesia. The parameters that we studied were
slip rates, locking widths and GMPEs.
The sensitivity analysis for probabilistic seismic The comparison between different slip rates
hazard analysis (PSHA) for two different sites shows a significant hazard difference between
near the Aceh Fault Segment, Indonesia, shows a smaller (2 mm a21) and larger (20 mm a21) slip
significant sensitivity in hazard level to the parame- rates. The study at Site 1 shows that at a 500 year
ters that we analysed: slip rates, locking widths return period, the hazard level from smaller slip is
(Mmax) and GMPEs. 0.06g, while the larger slip hazard level is 0.55g
The hazard curves comparison for the Aceh seg- (Fig. 3a), resulting in a difference of 0.49g, which
ment using two different slip rates show significant is significant. Another comparison at a 2500 year
differences in hazard level. The previous slip rate return period shows that the hazard level for smaller
used for this fault segment (Irsyam et al. 2010) is slip reaches 0.45g, while the larger slip produces
only 10% of that suggested by more recent work 1.13g, which still gives a difference of approxi-
(Ito et al. 2012). We clearly show that seismic haz- mately 0.68g. A similar result also appears at Site
ard analysis is sensitive to such a large difference in 2 (Fig. 3b): however, owing to the increased dis-
fault slip rates, with a higher slip rate leading to a tance from the fault, the hazard level is smaller.
higher maximum hazard, and a larger area subject Conversely, the maximum magnitude (as deter-
to high hazard levels. In general, geodetic studies mined from the fault locking width) also plays an
in the northern part of Sumatra suffer from limited important role in determining the hazard level.
measurements in the far field because of the narrow The comparison between Mw 4.6 (magenta dashed
land area available for measurements. Therefore, it line) and Mw 7.7 (red line) at Site 1 (Fig. 3c)
seems important to account for uncertainty in geo- shows two important points. The first point is that
detic measurements of slip rate there. the hazard difference is 0.08g at a return period
Fault geometry, particularly the locking width, of 500 years. The second point is that the hazard
influences the maximum potential magnitude of a level from a smaller maximum magnitude results
fault and is, hence, important for hazard. It is com- in a higher hazard level at a return period of
mon to use geological data to estimate maximum ,700 years. At longer return periods, the hazard
magnitude: however, recent developments in geod- level from the higher maximum magnitude shows
esy (Ito et al. 2012) can provide researchers with a a significant difference. For example, at a 2500
PDF for locking width and, hence, a new option year return period, a Mw 7.7 maximum magnitude
for determining the maximum magnitude on faults. results in a 1.06g exceedance in ground shaking, at
The analysis for different maximum magnitudes Mw 4.6, however, the maximum magnitude results
show that the hazard level for lower maximum in only 0.68g. The comparison at Site 2 (Fig. 3d)
 Downloaded from http://sp.lyellcollection.org/ at Australian National University on June 30, 2016

A. OMANG ET AL.

also shows a similar result, but the hazard level is horizontal and vertical components of peak ground
smaller owing to the increased distance. acceleration and acceleration response spectra. Bulle-
The study of different GMPEs also shows the tin of the Seismological Society of America, 93,
importance of the correct selection of the GMPE. 314–331.
Chiou, B.S.-J. & Youngs, R.R. 2008. An NGA model
At smaller return periods, the difference is relatively for the average horizontal component of peak ground
small, while at longer return periods the difference motion and response spectra. Earthquake Spectra, 24,
is larger, as expected. At a 500 year return period 173–215.
(Fig. 3e), CY08 resulted in a similar hazard level Cornell, C. 1968. Engineering seismic risk analysis. Bul-
to BA08: for CB08, however, the hazard level is letin of the Seismological Society of America, 58,
smaller by 0.21g. At a 2500 year return period, the 1583– 1606.
hazard level difference with CY08 and BA08 was Genrich, J.F., Bock, Y. et al. 2000. Slip distribution
still small, but the difference with CB08 increased at the northern Sumatra Fault System. Journal of
to about 0.41g. However, at a return period of Geophysical Research, 105, 28327–28341.
Irsyam, M., Sengara, W. et al. 2010. Development
.2500 years, the hazard difference between CY08 of Seismic Hazard Maps of Indonesia for Revision of
and BA08 is larger. The comparison at Site 2 Seimic Hazard Map in SNI 03-1726-2002. Reseach
(Fig. 3f) for the GMPE selection shows a different report submitted to the Ministry of Public Works by
trend. The hazard level between the three GMPEs the Team for Revision of Seismic Hazard Maps of
can be seen clearly at both short and long periods. Indonesia.
The main difference compared to Site 1 is that the Ito, T., Gunawan, E. et al. 2012. Isolating along-strike
hazard level is smaller owing to the increased dis- variations in the depth extent of shallow creep and
tance from the fault. fault locking on the northern Great Sumatran Fault.
The results show that each of the parameters Journal of Geophysical Research, 117, B06409,
http://doi.org/10.1029/2011JB008940
studied are equally important and affect the hazard Kramer, S. 1996. Geotechnical Earthquake Engineering.
level significantly. Therefore, careful consideration Prentice-Hall, Englewood Cliffs, NJ.
is needed as we work on seismic hazard analysis in Leonard, M., Robinson, D., Allen, T., Schneider, J.,
areas where there is significant uncertainty in fault Clark, D., Dhu, T. & Burbidge, D. 2007. Toward
parameters and GMPEs. However, it is important a Better Model of Earthquake Hazard in Australia.
to note that PSHA also depends on other factors, Geological Society of America, Special Papers,
such as earthquake catalogue data and source zona- 425-17.
tion, that have not been considered in this study. Lombardi, A.M., Akinci, A., Malagnini, L. & Muel-
ler, C.S. 2005. Uncertainty analysis for seismic hazard
This work was supported by the Australian Department of in Northern and Central Italy. Annals of Geophysics,
Foreign Affairs Australian Aid Program and Australian 48, 853– 865.
Research Council Linkage Project LP11010525. McGuire, R.K. 2004. Seismic Hazard and Risk Analysis.
Earthquake Engineering Research Institute, Oakland,
CA.
References McGuire, R.K. & Arabasz, W.J. 1990. An introduction
to probabilistic seismic hazard analysis. In: Ward,
Beauval, C. & O.& Scotti, 2004. Quantifying sensitivi- S.H. (ed.) Geotechnical and Environmental Geophys-
ties of PSHA for France to earthquake catalog uncer- ics, Volume III. Society of Exploration Geophysics,
tainties, truncation of ground-motion variability, and Tulsa, OK, 333–353.
magnitude limits. Bulletin of the Seismological Society Natawidjaja, D.H. & Triyoso, W. 2007. The Sumatran
of America, 94, 1579–1594. fault zone – from source to hazard. Journal of Earth-
Bennett, J.D., Bridge, D.McC. et al. 1981. Geologic quake and Tsunami, 1, 21–47.
map of the Banda Aceh Quadrangle, North Sumatra. Reiter, L. 1990. Earthquake Hazard Analysis – Issues
1:250,000. Geological Research and Development and Insights. Columbia University Press, New York.
Centre, Bandung, Indonesia. Robinson, D., Fulford, G. & Dhu, T. 2005. EQRM:
Bommer, J.J. 2003. Uncertainty about the uncertainty in Geoscience Australia’s Earthquake Risk Model.
seismic hazard analysis. Engineering Geology, 70, Geoscience Australia, Record, 2005/0 1.
165– 168. Robinson, D., Dhu, T. & Schneider, J. 2006. Practi-
Bommer, J.J. & Abrahamson, N.A. 2006. Why do mod- cal probabilistic seismic risk analysis: a demonstra-
ern probabilistic seismic-hazard analyses often lead tion of capability. Seismological Research Letters,
to increased hazard estimates? Bulletin of the Seismo- 77, 453– 459.
logical Society of America, 96, 1967– 1977. Sabetta, F., Lucantoni, A., Bungum, H. & Bommer,
Boore, D.M. & Atkinson, G.M. 2008. Ground-motion J.J. 2005. Sensitivity of PSHA results to ground
prediction equations for the average horizontal compo- motion prediction relations and logic-tree weights.
nent of PGA, PGV, and 5%-damped PSA at spectral Soil Dynamics and Earthquake Engineering, 25,
periods between 0.01 s and 10.0 s. Earthquake Spec- 317–329.
tra, 24, 99– 138. Sieh, K. & Natawidjaja, D.H. 2000. Neotectonics of the
Campbell, K.W. & Bozorgnia, Y. 2003. Updated near- Sumatran fault, Indonesia. Journal of Geophysical
source ground-motion (attenuation) relations for the Research, 105, 28295– 28326.
 Downloaded from http://sp.lyellcollection.org/ at Australian National University on June 30, 2016

SUMATRA FAULT PSHA SENSITIVITY ANALYSIS

Wells, D.L. & Coppersmith, K.J. 1994. New empirical Youngs, R.R. & Coppersmith, K.J. 1985. Implications
relationships among magnitude, rupture length, rupture on fault slip rates and earthquake recurrence mod-
width, rupture area and surface displacement. Bulle- els to probabilistic seismic hazard estimates. Bul-
tin of the Seismological Society of America, 84, letin of the Seismological Society of America, 75,
974–1002. 939– 964.

View publication stats