Sie sind auf Seite 1von 20

Geophysical Journal International

Geophys. J. Int. (2013) doi: 10.1093/gji/ggs020


G
J
I
G
e
o
d
y
n
a
m
i
c
s
a
n
d
t
e
c
t
o
n
i
c
s
Analysis of afterslip distribution following the 2007 September 12
southern Sumatra earthquake using poroelastic
and viscoelastic media
Ashar Muda Lubis,

Akinori Hashima and Toshinori Sato


Graduate School of Science, Chiba University, 133 Yayoi-cho, Inage-ku, Chiba 2638522, Japan. E-mail: asharmudalubis@yahoo.com
Accepted 2012 October 10. Received 2012 July 30; in original form 2011 July 05
SUMMARY
Most studies of afterslip distribution consider only elastic media. However, the effects of
poroelastic rebound in the upper crust and viscoelastic relaxation in the asthenosphere are part
of the observed post-seismic deformation. Therefore, these effects should be removed to give
a more reliable and correct afterslip distribution. We developed a method for calculating an
afterslip distribution in elastic, poroelastic and viscoelastic media, and we applied this method
to the case of the 2007 southern Sumatra earthquake (M
w
8.5). To estimate the coseismic slip
and time evolution of the afterslip distribution, we applied Akaikes Bayesian Information Cri-
terion (ABIC) inversion method of coseismic displacement, and analysed 15 months of GPS
post-seismic deformation data in 3-month observation periods. To calculate afterslip in each
period, we considered not only viscoelastic responses to coseismic slip but also viscoelastic
responses to afterslip in the preceding periods. We used viscoelastic model to compute post-
seismic deformation models every 3 months during the 15 months after the earthquake. The
viscosity value for the asthenosphere layer is a crucial unknown parameter. To overcome this
problem, we used a grid search method to determine the best-viscosity value, and we found
that the best viscosity for the Sumatra subduction zone was 2.5 10
18
Pas. After removing
the poroelastic and viscoelastic responses, we obtained maximum afterslip of 0.5 m during the
15-month investigation (the same as maximum afterslip estimated using the elastic medium
only), but the poroelastic and viscoelastic responses brought the afterslip distribution to a
shallower depth than the main coseismic rupture area. The results showed that the poroelastic
and viscoelastic responses added signicant corrections to the afterslip distribution. Compared
with the traditional method, this method improved the determination of the afterslip distribu-
tion. We conclude that consideration of poroelastic and viscoelastic behaviours is essential for
calculating the afterslip distribution. We propose that these parameters should be considered
to obtain more reliable and correct afterslip distribution models following earthquakes.
Key words: Inverse theory; Space geodetic surveys; Subduction zone processes; Dynamics
and mechanics of faulting; Rheology: crust and lithosphere; Rheology: mantle.
1 I NTRODUCTI ON
In the Sumatra region, the Indo-Australia Plate is subducting be-
neath the Sunda shelf at a velocity of 5 cmyr
1
(Fig. 1; Zachariasen
et al. 1999). The tectonic setting in this area is very complicated,
because of the oblique convergence of the Indo-Australian and
Eurasian plates in the Sumatra subduction zone. As a result, several
great plate interface earthquakes have occurred in the past on the
Sumatra subduction zone (e.g. Zachariasen et al. 1999; Natawidjaja

Permanent address: Department of Physics, Faculty of Mathematic and


Sciences, Bengkulu University, Indonesia. E-mail: asharmudalubis@
yahoo.com
et al. 2007). In the last 5 yr there have been three great megath-
rust earthquakes in this region. On 2004 December 26, a subduc-
tion earthquake of M
w
9.09.3 occurred in the northern Sumatra
and Andaman regions (Lay et al. 2005). This earthquake was the
largest earthquake in the past 40 yr. Three months following the
SumatraAndaman earthquake, on 2005 March 28, the 2005 Suma-
tra earthquake, also referred to as the Nias earthquake, of M
w
8.7,
occurred directly to the south of the rupture area of the Sumatra
Andaman earthquake. The rupture area of the Nias earthquake was
located in the region of the 1861 and 1907 events (e.g. Nalbant
et al. 2005; Briggs et al. 2006; Kanamori et al. 2010). About two
and a half years after the SumatraAndaman earthquake, on 2007
September 12, double earthquakes occurred in the region of the
1797 and 1833 historical earthquakes (Fig. 1). The rst earthquake
C
The Authors 2012. Published by Oxford University Press on behalf of the Royal Astronomical Society. 1
Geophysical Journal International Advance Access published November 19, 2012

a
t

K
y
o
t
o

U
n
i
v
e
r
s
i
t
y

o
n

M
a
y

1
4
,

2
0
1
4
h
t
t
p
:
/
/
g
j
i
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

2 A. M. Lubis, A. Hashima and T. Sato
Figure 1. Basic map of tectonic setting and historical great earthquakes along the convergent Sumatran Plate boundary. Rupture area of the 2004
SumatraAndaman and 2005 NiasSimeulue earthquakes is dotted in grey. Black and red rectangular boxes approximate the rupture area during the
great 1797 and 1833 earthquakes. Epicentres of the 2007 M
w
8.5 South Pagai and M
w
7.9 Pagai-Sipora earthquakes are shown by the two beachballs
(http://earthquake.usgs.gov/eqcenter/). Red arrows indicate the relative plate motion of the Indo-Australia Plate.
(M
w
8.5) occurred at 11:10:26 UTC. The epicentre was about 130
km southwest of Bengkulu on the southwest coast of Sumatra is-
land, at latitude 4.517

S and longitude 101.382

E. On the same day


(at 23:49:04 UTC), the second earthquake (M
w
7.9) occurred about
205 km northwest of Bengkulu and about 185 km southsoutheast
of Padang, in West Sumatra. The earthquakes devastated thousands
of houses in a narrow strip along the coastline from Bengkulu up
to Padang. The seismic sequence continued after these two large
earthquakes, with the biggest aftershock (M
w
7.1) occurring in the
next day. Hereafter, we refer to this 2007 September sequence as
the 2007 southern Sumatra earthquake.
Unlike the 1797 and 1833 events, which generated huge tsunamis
(Nalbant et al. 2005; Natawidjaja et al. 2006) on the western and
southern coasts of Sumatra, the 2007 southern Sumatra earthquake
produced only a tiny tsunami (Borrero et al. 2009). According to
Fujii & Satake (2008), the observed amplitude of the tsunami was
about 1 mat Padang, nearest Deep-ocean Assessment and Reporting
of Tsunami (DART) station to the source of the earthquake.
Alarge earthquake sequence such as this can cause stress changes
in the surrounding crust and lead to time-dependent post-seismic
deformation. Recently, the study of post-seismic deformation and
afterslip distribution has become a priority, because data on the

a
t

K
y
o
t
o

U
n
i
v
e
r
s
i
t
y

o
n

M
a
y

1
4
,

2
0
1
4
h
t
t
p
:
/
/
g
j
i
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

Afterslip of the 2007 Sep. Sumatra earthquake 3
afterslip distribution provide information on the stress changes,
frictional characteristics, and material properties along subduction
zones (e.g. Nishimura et al. 2000; Yagi et al. 2001); the afterslip
distribution can also affect the occurrence of future earthquakes
(Segall et al. 2000). Information on frictional characteristics along
subduction zones is important for simulations of earthquake gen-
eration. Once we obtain accurate and reliable information on the
afterslip distribution, which represents the frictional status of the
plate boundary, we may be able to establish more accurate models
of earthquake generation or earthquake cycles.
In general, in most studies the afterslip distribution is consid-
ered only in terms of elastic properties, through inversion of the
observed post-seismic displacement (e.g. Burgmann et al. 2002;
Miyazaki et al. 2004; Ozawa et al. 2004; Perfettini & Avouac 2004;
Takahashi et al. 2004; Baba et al. 2006). However, the observed
post-seismic displacement contains transient deformation due to
poroelastic rebound in the upper crust and viscoelastic relaxation in
the asthenosphere (J onsson et al. 2003). Therefore, we have to re-
move these two effects to obtain reliable and correct afterslip distri-
bution data following an earthquake. Wang et al. (2009) accounted
for the viscoelastic response to coseismic slip in their measurements
of afterslip distribution, but they neglected the viscoelastic response
to afterslip, and the poroelastic response caused by pore-pressure
changes in the crust. These two mechanisms can also contribute to
the observed post-seismic surface displacement, even though the
magnitudes of each contribution depend on the temporal and spa-
tial scales. Here, we have developed a method for calculating the
afterslip distribution following an earthquake using a model based
on media that is complete in terms of elastic, poroelastic and vis-
coelastic properties. We tested and applied our method to estimate
the afterslip distribution associated with the 2007 southern Sumatra
earthquake in Indonesia (Fig. 1), using data fromthe Sumatran GPS
Array (SuGAr) network.
Borrero et al. (2009) performed post-tsunami eld surveys and
numerical modelling to obtain the tsunami run-up heights due to
the main shock event. In addition, Konca et al. (2008) calculated
the rupture area of the main shock by using GPS, teleseismic, Inter-
ferometric Synthetic Aperture Radar (InSAR) and coral displace-
ment data. Lorito et al. (2008) performed inversion of the tsunami
waveforms to resolve the slip distribution and the average rupture
velocity. Lorito et al. (2007) estimated that the rupture propagated
to the northwest of the epicentre with a maximumslip of up to 15 m,
whereas Konca et al. (2008) obtained two slip peak zones, located
on South Pagai Island and around Mega Island, with a maximum
slip of 8 m. However, studies and modelling of afterslips following
the 2007 southern Sumatra earthquake have not yet been done. We
therefore took the opportunity to study these events and to test our
model by applying it to them.
By taking into account the poroelastic and viscoelastic responses
due to coseismic slip and afterslip, we aimed to create a better
assessment of the spatial and temporal distribution of the 2007
southern Sumatra earthquake, and to contribute a new approach for
investigating afterslip distribution following earthquakes.
2 METHODOLOGY AND ANALYTI CAL
APPROACH
2.1 GPS data analysis
We processed GPS data from the SuGAr network installed by the
Tectonics Observatory of the California Institute of Technology as
part of the Sumatra Plate Boundary Project (Fig. 2). In our anal-
ysis, we also included GPS data from the SAMP station located
in Sampali Medan in northern Sumatra province and from some
International GNSS Service (IGS) stations. To obtain data on the
coseismic and post-seismic deformation associated with the earth-
quake, we used GPS data spanning the period from 2007 January
1 to 2008 December 12. The GPS data were taken continuously,
with sampling every 30 s by the IGS station and every 120 s by
the SuGAr network. The entire data set was processed using the
GAMIT/GLOBK software packages version 10.35 (Herring 2000;
King & Bock 2004). The individual h-les, products of GAMIT,
were combined by using the GLOBK Kalman lter with regional
solutions (IGS1, IGS2, IGS3, IGS4 and IGS5) provided by the
Scripps Orbit and Permanent Array Center (SOPAC). The loosely
constrained daily solutions were transformed into International Ter-
restrial Reference Frame (ITRF) 2005 by using the GLORG mod-
ule of GLOBK (Herring 2000). We chose stations KARR, HYDE,
IISC, KUNM and LHAS, which lay a long way from the epicentre,
as stable sites. Finally, we estimated the daily north, east and up
components in the time-series from each station.
2.2 Modelling the slip distribution
To obtain the distributions of both coseismic and post-seismic slip,
we constructed the geometry of the plate boundary with a dip an-
gle of 12

and a strike angle of 323

by considering the focal


mechanism of the Global Centroid Moment Tensor (CMT) solution
(http://earthquake.usgs.gov/). For afterslip estimation, we modelled
the geometry of the fault as rectangular, with dimensions of 528 km
in the strike direction and 288 km along the dip direction (Figs 2
and 3), but for the coseismic slip estimation we changed the length
of the fault to 480 km in the strike direction and the width of the
fault to 244 km in the dip direction.
We constructed a fault model with an elastic layer (060 km)
overlying a viscoelastic structure. The thickness of the elastic layer
was considered to be the same as in other post-seismic studies in
the Sumatra subduction zone, such as for the 2004 great Sumatra
Andaman earthquake (Pollitz et al. 2006, 2008; Panet et al. 2010).
To account for the poroelastic response, we divided the elastic layer
into two. The uppermost layer, 20 km in depth, contained uid
(water) in the rock. We therefore had a three-layered structure of
0 to 20 km (poroelastic), 2060 km (elastic) and more than 60 km
(viscoelastic). Details of the fault parameters are given in Table 1.
When a earthquake occurs, the coseismic pressure changes disturb
the equilibrium between uids in shallow crust and the host matrix.
These pressure changes drive uid ow from this initial undrained
toward nal drained conditions of rock and change the volume.
Such poroelastic relaxation can be modelled by modifying the ef-
fective Poissons ratio of the elastic solid, while keeping the shear
modulus constant (Rice & Cleary 1976). Therefore, we calculated
the poroelastic deformation fromthe surface deformation under two
conditions, drained and undrained, in terms of the elastic properties.
For this calculation, we subtracted the surface deformation obtained
using different Poissons ratios for the drained and undrained condi-
tions. We set a Poissons ratio () of 0.27 for drained condition and
a Poissons ratio (
a
) of 0.31 for undrained. This is in accordance
with the ndings of various studies of poroelastic post-seismic dis-
placement (e.g. Peltzer et al. 1998; J onsson et al. 2003; Fialko 2004;
Freed et al. 2006). Since pore pressure gradients dissipate as crustal
uids mobilize in the rst 12 months after the event for the case of
the 2004 SumatraAndaman earthquake (Hughes et al. 2010), we

a
t

K
y
o
t
o

U
n
i
v
e
r
s
i
t
y

o
n

M
a
y

1
4
,

2
0
1
4
h
t
t
p
:
/
/
g
j
i
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

4 A. M. Lubis, A. Hashima and T. Sato
Figure 2. GPS data stations used in this study from the SuGAr network. Also shown are observed GPS coseismic displacements (black and grey) and modelled
coseismic displacements (red and blue) estimated from the coseismic slip distribution on the plate boundary. Uncertainties of GPS data are 1.22.4 mm for the
horizontal component and 4.67.9 mm for the vertical component. Blue rectangle is fault dimensions on the plate boundary. Red star is epicentre of the 2007
M
w
8.5 South Pagai and pink star indicates epicentre of M
w
7.9 Pagai-Sipora earthquake.
consider that the poroelastic deformation only plays a signicant
role within the rst 3 months following the earthquake.
First, we attempted to remove the poroelastic effect fromthe post-
seismic displacement observed in the rst 3 months following the
main shock. We then calculated viscoelastic surface deformations
in the rst 3 months after the earthquake using the range of viscosity
values, and removed the viscoelastic effect estimated using various
viscosity values from the GPS post-seismic data. After removing
the poroelastic and viscoelastic displacements at various viscosities,
we inverted the residual GPS post-seismic data and tried to nd the
best viscosity for the Sumatra region by using the minimum ABIC
inversion value (Akaike 1980).
Because the viscoelastic post-seismic deformation depends on
the viscosity value, we searched for the best-viscosity value for
the asthenospheric layer by nding the minimum ABIC value of the
inversion, corrected for the post-seismic deformation. The viscosity
value was evaluated in the range of 5 10
17
to 5 10
22
Pas
by considering the viscosity values from various studies of post-
seismic deformation (e.g. Nishimura & Thatcher 2003; Gourmelen
& Amelung 2005; Pollitz et al. 2006, 2008; Ryder et al. 2007;
Hammond et al. 2009). For details of the viscosity values from
these studies of the Sumatra subduction zone see Table 2.
Once we had obtained the best-viscosity value for our model, we
were able to calculate the spatio-temporal distribution of viscoelas-
tic displacement. We used this viscoelastic displacement to remove
the viscoelastic response and to obtain the corrected GPS displace-
ments. In this case, we considered not only the effect of viscoelastic
responses to the coseismic slip, as proposed by Wang et al. (2009),
but also the viscoelastic responses to the afterslip.
To calculate the afterslip distribution, we inverted 15 months
of GPS post-seismic-observed displacements. In this inversion, we
divided the GPS data into ve groups, each group consisting of
3 months of observations. To obtain the slip distribution, we mod-
elled the coseismic and post-seismic displacements as linear equa-
tions. The observation equations are as follows:
y = Ax +g (1)

a
t

K
y
o
t
o

U
n
i
v
e
r
s
i
t
y

o
n

M
a
y

1
4
,

2
0
1
4
h
t
t
p
:
/
/
g
j
i
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

Afterslip of the 2007 Sep. Sumatra earthquake 5
Figure 3. Fault geometry for modelling afterslip of the 2007 southern Sumatra earthquake consists of a rectangular fault with dimensions of 528 km in the
strike direction and 288 km in the dip direction.
Table 1. Parameters used to monitor poroelastic and viscoelastic response following the 2007 southern Sumatra earthquake. Vp
and Vs are primary and secondary wave velocities.
Wave velocity (km s
1
) Poisson ratio
Undrain Drain Undrain Drain
Layer Thickness (km) Vp Vs Vp Vs Density (g cm
3
) Viscosity (Pas)
Lithosphere 20.00 6.90 3.65 6.46 3.65 0.31 0.27 2.9
40.00 7.56 4.00 7.56 4.00 0.31 0.31 3.3
Asthenosphere 8.30 4.40 8.30 4.40 0.31 0.31 3.5 510
17
510
22
y = (d(t
0
), d(t
0
+t ), d(t
0
+2t ), )
T
d(t ) = (d
1
(t ), d
2
(t ), d
3
(t ), d
4
(t ), ,)
T
x = (m(t
0
), m(t
0
+t ), m(t
0
+2t ), )
T
m(t ) = (m
1
(t ), m
2
(t ), m
3
(t ), m
4
(t ), , )
T
A =

H
u
V
1
+P H
d
0
V
2
V
1
H
d
V
3
V
2
V
1
H
d
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

, (2)
where t
0
denotes the time of origin of the earthquake; t is the time
step; d
i
(t) is the observation value at observation point i at time t,
m
i
(t) being the slip distribution at point i on the fault at time t; H
u
is the undrained elastic response matrix; H
d
is the drained elastic
response matrix; Pis the poroelastic response matrix (P=H
d
H
u
);
V
k
is the viscoelastic response matrix at kt time after the event
and g is the error vector. We assume g to be Gaussian with zero
mean and a covariance of
2
E. We use the undrained response for
the coseismic period, because the coseismic slip occurred suddenly,
and the drained response for the post-seismic period, because the
afterslip occurred gradually.
By combining eqs (1) and (2), we obtain the observation equation,
as follows:

d(t
0
)
d(t
0
+t )
d(t
0
+2t )
d(t
0
+3t )
.
.
.

H
u
V
1
+P H
d
0
V
2
V
1
H
d
V
3
V
2
V
1
H
d
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

m(t
0
)
m(t
0
+t )
m(t
0
+2t )
m(t
0
+3t )
.
.
.

+g. (3)

a
t

K
y
o
t
o

U
n
i
v
e
r
s
i
t
y

o
n

M
a
y

1
4
,

2
0
1
4
h
t
t
p
:
/
/
g
j
i
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

6 A. M. Lubis, A. Hashima and T. Sato
Table 2. Viscosity estimates derived from geodetic measurements of post-loading deformation.
Case 2004 SumatraAndaman earthquakes (M
w
9.2) Viscosity (10
18
Ps) Reference
Kelvin element 0.5 Pollitz et al. (2006)
Maxwell element 10
Transient 0.5 Han et al. (2008)
Steady state 510
Maxwell 350 Reddy et al. (2010)
Maxwell 0.051000, (best: 0.5) Paul et al. (2007)
Maxwell 0.1 Panet et al. (2007)
Case 2000 June, Sumatra earthquake (M
w
7.9) 11000 Selva et al. (2004)
Because matrix A is similar to the lower triangular matrix, we
can solve the observation equation step by step, as follows:
First, we solve the coseismic slip distribution
d(t
0
) = H
u
m(t
0
) +g (t
0
) . (4)
Next, we try to solve the following equation:
d(t
0
+t ) = (V
1
+P) m(t
0
) +H
d
m(t
0
+t ) +g (t
0
+t ) .
(5)
In this equation, we already know the slip distribution m(t
0
).
Then,
D(t
0
+t ) = H
d
m(t
0
+t ) +g (t
0
+t ) , (6)
where D(t
0
+t ) = d(t
0
+t ) (V
1
+P) m(t
0
) .
We can solve this equation (for afterslip distributions) in the
same manner as shown above, and we can use this equation with the
constraint of a smooth slip distribution using the ABIC inversion
method (Yabuki &Matsuura 1992). To estimate the elastic response
only, we set V
k
= 0.
The kernel matrix H gives the theoretical relationship between
the surface displacement and a vector of slip on the rectangular fault,
which describes the function of the fault geometry parameters (e.g.
strike, dip, rake and fault dimensions). The matrix kernel H and V
k
are calculated by the method proposed by Matsuura &Sato (1989),
and we use a bicubic B-spline function to smooth the grid of the
kernel matrix on the plate boundary.
2.3 ABIC inversion
To estimate the coseismic and post-seismic slip distributions, we
performed an inversion analysis of the geodetic data, incorporating
the ABIC method (Yabuki & Matsuura 1992). Inversion by using
the minimumABICmethod has been successfully adopted by many
authors (e.g. Yoshioka et al. 1993; Fukahata et al. 2004; Matsuura
et al. 2007; Fukahata & Wright 2008).
To obtain high resolution in the slip distribution, discretization of
the fault may require a large number of slip patches, and this num-
ber may exceed the number of observations. In our case, the GPS
data for the Sumatra region were obtained from fewer than 30 sites,
whereas the number of unknowns in regard to slip was more than
1250. To stabilize the problem, we use an a priori constraint, namely
a smooth variation of slip along the fault surface, based on prob-
abilistic or Bayesian theory, as proposed by Yabuki & Matsuura
(1992). Using a measurement of the roughness of the slip distribu-
tion, we conducted an ABIC minimum inversion in the same way
as used by Yabuki and Matsuura, except for the following point:
their proposed ABIC inversion uses a non-negativity constraint to
force the slip direction to be positive, but the ABIC method cannot
be used with this kind of constraint, because, when the positivity
constrain is applied, the marginal likelihood could not be computed
analytically, so that it is impossible to determine the hyperparame-
ter using ABIC method (Fukuda & Johnson 2008). Therefore, we
did not use the non-negativity constraint in our modelling, and we
allowed slip in all directions.
3 RESULTS
3.1 GPS Coseismic deformation and post-seismic
deformation
Fig. 2 shows the coseismic GPS displacements. The estimated er-
rors of the coseismic offsets in the GPS data are 1.22.4 mm for the
horizontal component and 4.67.9 mm for the vertical component.
The average error for the north component was 1.54 mm, and for the
east component, 2.05 mm. The average error for the updown com-
ponent was 6.04 mm. We observe large coseismic displacements
to the south of Pagai Island. Two sites, namely BSAT and PRKB,
experienced coseismic horizontal displacements of 1.82 and 1.5 m,
respectively, towards the southwest. Substantial deformation was
also observed at sites LNNG, MKMK and LAIS, with a horizon-
tal component magnitude of 6283 cm. In particular, the coseismic
deformations south of Pagai Island were comparable not only to
those from InSAR and intensity matching of SAR images (Lubis
2011), but also to coral reef eld observations (Konca et al. 2008;
Sieh et al. 2008). Extreme crustal displacement in the area of South
Pagai Island gave us a clue that the deformation source might have
been very close to the area south of Pagai Island.
The cumulative temporal horizontal post-earthquake GPS dis-
placements in the 15 months following the 2007 southern Suma-
tra earthquake increased in magnitude with time (Fig. 4). A large
horizontal displacement of 13 cm (about 8 per cent of the coseis-
mic displacement) was observed at station BSAT during the rst
3 months of observation. However, during the second 3 months
of observation, the horizontal displacement decreased to 7 cm. We
observed a cumulative post-seismic displacement of about 22 cm
at BSAT during the rst 15 months of observation. We plotted the
horizontal post-seismic deformation for every time step of obser-
vation (Figs S1S5). Most of the errors in the post-seismic dis-
placements were less than 5 mm. The horizontal GPS displace-
ments for every stage moved trenchward, and the magnitudes of
the horizontal component declined over the 15-month observation
period. With respect to post-seismic displacement, on Sumatra Is-
land the InSARpost-seismic deformations (Lubis 2011) also exhibit
good agreement with the GPS observations and give more evidence
that substantial post-seismic displacement occurred following the
2007 southern Sumatra earthquake. The GPS data show that post-
seismic surface deformation occurred, especially at two sites on
Pagai Island (BSAT and PRKB), and that movement continued over
the 15 months following the main shock. We examined the daily

a
t

K
y
o
t
o

U
n
i
v
e
r
s
i
t
y

o
n

M
a
y

1
4
,

2
0
1
4
h
t
t
p
:
/
/
g
j
i
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

Afterslip of the 2007 Sep. Sumatra earthquake 7
Figure 4. Cumulative temporal horizontal post-seismic GPS displacement over the 15-month observation period. Black rectangle is fault dimensions for
afterslip model on the plate boundary. Red and pink stars represent the epicentres for earthquake of M
w
8.5 and 7.8 on 2007 September 12.
positions of vertical post-seismic displacement (Fig. S6). The post-
seismic displacements actually include both aseismic deformation
and displacements due to large aftershocks. We therefore checked
the sub-daily GPS observations to see whether there was any sub-
stantial deformation associated with large aftershocks in the rst 3
days following the main shock, but we could not see any substantial
jumps in the subdaily position data. In addition, we estimated the
cumulative seismic moments for 15 months of aftershocks from the
USGS catalogue (NEIC) at depths less than 30 km (M < 7). We
found that the cumulative aftershock moment was 0.6 per cent of
the main shock moment and 5 per cent of total afterslip moment
that will be shown in the Sections 4.2 and 4.6. We therefore inferred
that the inuence of the remaining aftershocks on the post-seismic
measurements was small.
3.2 Coseismic slip distribution
The inferred maximum coseismic slip distribution obtained by in-
version of the surface coseismic GPS deformation for the horizontal
and vertical components was less than 6 m (Fig. S7); this value was
smaller than that of Konca et al. (2008). Therefore, we investigated
the coseismic slip distribution from the joint inversion of coseismic
GPS deformation for the horizontal and vertical components and the
coral data displacement obtained by Konca et al. (2008) and Sieh
et al. (2008). We excluded the GPSdata fromstation LNNGfromthe
inversion data, because the displacement at this site showed a large
discrepancy compared with that at station MKMK, even though the
two sites were very close. Moreover, when we included the inver-
sion data from this site we found that the coseismic slip directions
around the site were opposite to the main coseismic slip direction
and inconsistent with the earthquake mechanism. The result of the
coseismic slip distribution obtained with the joint inversion data
is given in Fig. 5, and the estimated error is shown in Fig. S8. In
general, our slip distribution was similar to the coseismic slip dis-
tribution estimated by Konca et al. (2008). We compared modelled
displacements with the observed displacements from both the GPS
data and the coral data (Figs 2 and 6, respectively). The rms of the
mist between the modelled data and the coseismic displacement
observed using GPS was 2, 3 and 6 mmin the east, north and vertical
components, respectively; against the coral data it was 8.2 cm. The
mean mist was less than that obtained with the model of Konca
et al. (2008).
The coseismic slip of the main shock had three major asperities
located to the north of the epicentre of the rst earthquake, and a
maximum slip of 7 m was located southwest of Pagai Island. The
seismic moment of the main shock event was 7.71 10
21
Nm
(M
w
= 8.5) if we assume a rigidity of = 40 GPa for the crust.
This seismic moment is comparable to the Harvard CMT (6.71
10
21
Nm) and the results of Konca et al. (2008) (7.3 10
21
Nm).
To examine the resolution of our model, we used the checkerboard
method and inputted 100 100-km rectangular slip patches of 7,
3, 1, 0.3, 0.1 and 0.05 m. The resolution was generally good for
all input slip distributions, particularly in areas where there were
substantial asperities on the plate boundary (Fig. S9).

a
t

K
y
o
t
o

U
n
i
v
e
r
s
i
t
y

o
n

M
a
y

1
4
,

2
0
1
4
h
t
t
p
:
/
/
g
j
i
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

8 A. M. Lubis, A. Hashima and T. Sato
Figure 5. Coseismic slip distributions on plate boundary, estimated from
joint ABIC inversion of GPS and coral data. Black arrows indicate slip
amounts and directions. Green rectangle is fault dimensions on the plate
boundary. Red and pink stars represent the epicentres for earthquake of M
w
8.5 and 7.8 on 2007 September 12.
3.3 Afterslip distribution using elastic properties only
We then modelled the afterslip using an elastic medium only: the
results of the inversion using post-seismic GPS data for the ve
time periods are shown in Figs 7(a)(e), and their error is shown
in Fig. S10. The temporal afterslip distribution displayed a clear,
gradual drop fromthe 0- to 3-month observation period to the 12- to
15-month period. Afterslip during the rst 3 months occurred with
a maximum slip of 28 cm. The slip distribution during the period of
36 months following the main shock was concentrated in the centre
of the main slip area of the rst 3 months, with a maximum slip
of 10.5 cm. The slip in the third time period also had one area, and
it differed in magnitude from that during the 3- to 6-month-period
inversion. The afterslip distribution for the period 912 months was
located in the same place as the afterslip distribution for the 6- to
9-month period, but the maximumslip and area of slip were less than
in the previous time periods. The slip in the last observation period
was much smaller than those in the previous periods, although it
had the same pattern as in the two previous periods; a substantial
part of the afterslip occurred in an area deeper than 30 km.
3.4 Afterslip distribution corrected by poroelastic
response
We calculated surface poroelastic displacements using an elastic
formulation with the different Poisson ratios under both undrained
and drained conditions. Fig. 8 shows the post-seismic poroelastic
displacements within the rst 3 months, which is difference between
Figure 6. Coral data used for ABIC inversion (grey) (from Konca et al.
2008 and Sieh et al. 2008), and modelled ground deformation (blue). Red
and pink stars represent the epicentres for earthquake of M
w
8.5 and 7.8 on
2007 September 12.
undrained and drained conditions. We found substantial poroelastic
horizontal displacement in the shallow region of the plate bound-
ary (e.g. at stations BSAT, PRKB and NGNG), with magnitudes of
24 cm. The subsidence and uplift of the surface poroelastic dis-
placement contrasted with the vertical GPS coseismic displacement
(Fig. 2). The maximum subsidence (magnitude of up to 2 cm) was
located mainly to the west of the Pagai island, whereas the maximum
uplift occurred on the eastern part of Pagai Island and in particular
on Sumatra Island.
We used this poroelastic displacement to correct the observed
GPS post-seismic displacement in the rst 3 months of observa-
tion. Again, we inverted corrected post-seismic GPS displacement
to obtain the afterslip distribution. We obtained a maximum slip
distribution of 29 cm; it was concentrated deeper, and was larger,
than the maximum afterslip obtained using elastic properties only
(Figs 9b and c). Comparison of the afterslips that do and do not
account for effect of poroelasticity revealed that the poroelastic
response added a signicant correction at shallow depth. Between
Pagai Island and Sipora Island, slip amounts of the poroelasic model
are much smaller than those of the elastic only.
3.5 Viscosity effect on surface deformation, and selection
of viscosity value
We calculated viscoelastic post-seismic surface deformations using
a range of viscosity values from 5 10
17
to 5 10
21
Pas. The vis-
cosity value range was chosen based on the viscosity values found in

a
t

K
y
o
t
o

U
n
i
v
e
r
s
i
t
y

o
n

M
a
y

1
4
,

2
0
1
4
h
t
t
p
:
/
/
g
j
i
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

Afterslip of the 2007 Sep. Sumatra earthquake 9
Figure 7. Changes in afterslip distribution. (ae) Results of inversion of
post-seismic GPS data for ve time periods using the elastic model. (fj)
Results using the poroelastic and viscoelastic model for the same ve time
periods. Black arrows indicate slip amounts and directions.
various investigations of post-seismic deformation (e.g. Nishimura
& Thatcher 2003; Gourmelen & Amelung 2005; Pollitz et al. 2006,
2008; Ryder et al. 2007; Hammond et al. 2009). We examined
the viscoelastic post-seismic deformation at various viscosities in
the rst 3 months following the earthquake (Fig. 10). The results
show that viscoelastic post-seismic deformation depends strongly
on the viscosity value. After removing the poroelastic response, we
searched for the best-viscosity value in the asthenospheric layer by
nding the minimum ABIC value for each given viscosity value.
The best viscosity for the southern Sumatra region was 2.5 10
18
Pas (Fig. 11). In general, this value was in the range of viscosity
values recently proposed for the asthenosphere layer by some au-
thors for Sumatra region (e.g. Pollitz et al. 2006, 2008; Ogawa &
Heki 2007; Panet et al. 2007, 2010; Cannelli et al. 2008; Han et al.
2008).
3.6 Afterslip distribution corrected by consideration of
poroelastic and viscoelastic responses to coseismic slip and
afterslip distribution
We determined the cummulative spatio-temporal distributions of
the viscoelastic response due to coseismic slip distribution for 3,
6, 9, 12 and 15 months after the earthquake (Fig. 12). In each ob-
servation period, the largest viscoelastic deformation was located
above the fault plane around the north of Pagai Island (stations
Figure 8. Poroelastic displacements using the estimated coseismic slip dis-
tributions (Fig. 5). Black, grey and red arrows indicate the observed hori-
zontal, the observed vertical and the poroelastic horizontal displacements,
respectively. Colour image indicates the poroelastic vertical displacements.
Blue rectangle is fault dimensions on the plate boundary. Red and pink
stars represent the epicentres for earthquake of M
w
8.5 and 7.8 on 2007
September 12.
BSAT and PRKT). The largest surface viscoelastic deformation
was about 1112 cm, at stations BSAT and PRKB, located on
Pagai Island, far off the coast of Sumatra Island, within 15 months
after the earthquake. In addition, we obtained small horizontal sur-
face deformations of up to 4 cm on the coast of Sumatra Island,
around stations such LAIS, MKMK and LNNG, whereas substan-
tial viscoelastic vertical deformations were observed in a subsidence
pattern and reached up to 8 cm after a year or more of observation.
On Pagai Island, viscoelastic models predict at stations BSAT and
PRKBshould move away fromthe trench following the main shock;
the predicted post-seismic horizontal velocities at almost all sites
are anticorrelated with the observed GPS velocities. Conversely,
the viscoelastic horizontal deformations on Sumatra Island were
trenchward, particularly at station LAIS.
In the rst 3-month period we plotted the results of our inversions
for afterslip distribution corrected by poroelastic response and vis-
coelastic response, or both, for coseismic slip (Fig. 9). The effects
of the poroelastic and viscoelastic responses on the afterslip distri-
bution were clear. By inverting the GPS data, without any correc-
tions, we estimated maximum afterslip of 28 cm, but it decreased
to 26 cm if we took into account the effects of the viscoelastic

a
t

K
y
o
t
o

U
n
i
v
e
r
s
i
t
y

o
n

M
a
y

1
4
,

2
0
1
4
h
t
t
p
:
/
/
g
j
i
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

10 A. M. Lubis, A. Hashima and T. Sato
Figure 9. Results of inversion of GPS data for the 3 months after the earthquake by using the elastic, poroelastic and viscoelastic models and their combinations.
(a) Coseismic slip distribution; (b) afterslip distribution using elastic medium only; (c) afterslip distribution using elastic and poroelastic media; (d) afterslip
distribution using elastic and viscoelastic media; and (e) afterslip distribution in complete media (elastic, poroelastic and viscoelastic responses). Black arrows
indicate slip amounts and directions.
response. The magnitude of the afterslip was 29 cm if we included
both the poroelastic and the viscoelastic response.
We plotted the afterslip distributions corrected by the poroelastic
and viscoelastic responses over the ve time periods (Figs 7fj).
In the rst 3 months of observation the trend was similar to
that of the afterslip distribution loaded by elastic properties only.
After the rst 3 months, the afterslip distribution obtained by ap-
plying the poroelastic and viscoelastic models contrasted with that

a
t

K
y
o
t
o

U
n
i
v
e
r
s
i
t
y

o
n

M
a
y

1
4
,

2
0
1
4
h
t
t
p
:
/
/
g
j
i
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

Afterslip of the 2007 Sep. Sumatra earthquake 11
Figure 10. Post-seismic viscoelastic deformation with various viscosity values in the rst 3 months following the earthquake.
obtained by using elastic modelling only (contrast Figs 7ae with
fj). The results clearly differed not only in magnitude but also
in location of the afterslip distribution. With the complete me-
dia, the afterslip distribution started to move from the deeper part
to a shallow depth within 6 months after the earthquake: after
6 months, the afterslip was still moving to a shallow area located
up-dip of the coseismic rupture area, west of Pagai Island. The
maximum magnitude of the afterslip distribution decreased with
increasing length of the observation period. At 69 months, the
peak of the afterslip distribution was about 9.5 cm; at 912 months
it had changed slightly to 9.2 cm. In the last observation period
there was maximum afterslip of 8.2 cm. At 6 months the slip
distribution in the deeper part was removed successfully by ac-
counting for viscoelastic effects, but at 15 months the afterslip
distribution at shallow depth was still substantial. The post-seismic
slip distribution in the rst 3 months was dominated by afterslip and
poroelastic effects, whereas accounting for the viscoelastic compo-
nents of post-seismic deformation resulted in substantial corrections
to the afterslip distribution from 6 months after the earthquake.
We plotted the cumulative 15-month afterslip distributions as-
sociated with the 2007 southern Sumatra earthquake by using the
various models (Fig. 13). The peak cumulative afterslip using the
elastic model only was concentrated in the area southeast of Pagai
Island and between the southern part of the M
w
7.9 event and the
northern part of the M
w
8.5 event. Using the elastic model only,
the maximum cumulative afterslip distribution in the 15 months
following the main shock was 49 cm, whereas if we included the
poroelastic effect the magnitude of the afterslip distribution in-
creased to 53 cm. If we considered a combination of the elastic and
the viscoelastic effects, we observed an increase in maximum after-
slip to 62 cm; the peak of the afterslip distribution was located in
the up-dip region of the coseismic rupture area. Combination of the
elastic, poroelastic and viscoelastic effects gave maximum afterslip
of 50 cm. The most important differences among the models were as
follows: (1) When we considered the elastic and poroelastic effects,
the afterslip distribution was located mainly in the deeper region
on the plate boundary, with substantial overlap with the coseismic
rupture. (2) When we considered the viscoelastic component, the
afterslip distribution occurred mainly in the shallow area up-dip of
the main coseismic slip distribution on the plate; this area did not
slip during the 2007 main shock event. (3) When we considered the
elastic, poroelastic and viscoelastic effects, we obtained a pattern

a
t

K
y
o
t
o

U
n
i
v
e
r
s
i
t
y

o
n

M
a
y

1
4
,

2
0
1
4
h
t
t
p
:
/
/
g
j
i
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

12 A. M. Lubis, A. Hashima and T. Sato
Figure 11. Best-tting viscosity for the Sumatra region. A viscosity of 2.5 10
18
Pas gave the minimum value of ABIC inversion.
Table 3. Comparison of moment magnitudes in four cases of afterslip modelling.
Case Max. slip (cm) Moment (10
21
Nm) Magnitude (M
w
)
Elastic 49 1.030 7.942
Elastic and poroelastic 53 1.003 7.934
Elastic and viscoelastic 62 1.071 7.953
Elastic, poroelastic and viscoelastic 50 0.994 7.931
similar to that with afterslip distribution loaded by viscoelastic re-
sponse only, but the magnitude of the maximum afterslip decreased
by about 12 cm. The moment release due to the cumulative afterslip
distribution using the elastic, poroelastic and viscoelastic effects
was 0.994 10
21
Nm (M
w
7.9; Table 3); this was smaller than the
moment released due to afterslip using the elastic effect only, and
was about 15 per cent of the coseismic moment of the main shock.
We examined both the observed and the predicted horizontal dis-
placements due to cumulative post-seismic slip on the fault over the
15-month GPS observation period (Fig. 14). The cumulative post-
seismic GPS displacements obtained were comparable to the cumu-
lative post-seismic displacements modelled by using the poroelastic
and viscoelastic responses. The observed GPS displacements (black
arrow) tted well with the cumulative afterslip distributions pre-
dicted using the elastic, poroelastic and viscoelastic effects (green
arrow) in Fig. 14b. The mist for the northsouth component and
eastwest component was about 1 cm; the mist for the vertical
component was about 4 cm.
4 DI SCUSSI ON
The vertical displacements estimated using the poroelastic model
(Fig. 8) had opposite sign to the observed coseismic (Fig. 2) and
post-seismic displacements (Fig. 8). This pattern has also been re-
ported for other earthquakes (e.g. J onsson et al. 2003; Feigl &
Thatcher 2006; Hughes et al. 2010). The modelled vertical sur-
face displacements from viscoelastic relaxation due to coseismic
slip (Fig. 12) were generally similar to the observed post-seismic
displacements. The majority of the subsidence area was located on
Sumatra Island. Both the vertical surface displacement with co-
seismic and post-seismic deformation observed by using GPS data
and the vertical post-seismic viscoelastic relaxation were similar on
Sumatra Island. In contrast, the post-seismic horizontal components
of viscoelastic relaxation in the main area of the fault were anticor-
related with the observed coseismic and post-seismic displacements
(Figs 12 and S1S5). An opposite polarity of the observed horizon-
tal coseismic displacement and horizontal displacements expected
of viscoelastic relaxation has been suggested previously by theo-
retical viscoelastic modelling (Matsu ura et al. 1981). Such patterns
were also reported by Pollitz et al. (2006, 2008) in the case of the
2004 M
w
9.2 great SumatraAndaman earthquake and the 2005
M
w
8.7 Nias earthquake; within the area above the rupture patch,
the observed post-seismic displacements had opposite signs to the
modelled viscoelastic displacements. Our result is also consistent
with that of an investigation of long-term viscoelastic deformation
in the great 1960 Chile earthquake (Hu et al. 2004).
Understanding the discrepancies between afterslip due to elastic
effect only (Fig. 9b) and afterslip corrected by poroelastic rebound
(Fig. 9c) during the rst 3-month observation period is made easier
by an examination of the GPS data for the horizontal and vertical
displacement components. In shallowarea, during the rst 3 months,
the horizontal displacements in the poroelastic model tended to have

a
t

K
y
o
t
o

U
n
i
v
e
r
s
i
t
y

o
n

M
a
y

1
4
,

2
0
1
4
h
t
t
p
:
/
/
g
j
i
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

Afterslip of the 2007 Sep. Sumatra earthquake 13
Figure 12. Spatio-temporal distribution of cumulative surface viscoelastic deformation using the estimated coseismic slip distributions (Fig. 5). The number
at the top of each panel denotes time elapsed (in months) after the earthquake. Red arrows indicate the horizontal viscoelastic displacement and black and grey
arrows indicate the observed horizontal and vertical displacements of GPS data, respectively. Colour image indicates the vertical viscoelastic displacements.
Black rectangle indicates the horizontal projection of the fault plane. Red and pink stars in (a) represent the epicentres for earthquake of M
w
8.5 and 7.8 on
2007 September 12.
spatial patterns and directions (Fig. 8) similar to those of the post-
seismic observed GPS data. Since the horizontal displacements due
to afterslip have similar trend in the shallow area, afterslip using the
poroelastic model was reduced in the shallow part (Fig. 9c) com-
pared with that using the only elastic model. In contrast, the vertical
components had the opposite sign in deeper parts (observed; down,
poroelastic response; up), particularly in the Sumatra region, such
as at stations LAIS, LNNG and MKMK (Fig. 8). To compensate
this anticorrelated poroelastic response, afterslip using the poroe-
lastic model was increased in the deeper part (Fig. 9c), because
the vertical displacements due to afterslip are down in the deeper
parts.
In the shallow parts, afterslip was still present at the end of
the 15 months of observation (Figs 7hj). The later slip was
substantially affected by the contribution of viscoelastic deforma-
tion, which was anticorrelated with the post-seismic GPS observa-
tions in the shallow parts (Figs 12 and S1S5). After 3 months,
the horizontal and vertical displacements of the viscoelastic re-
sponse at the shallows part are landward and up directions (Fig. 12).
These directions are opposite to the observed horizontal and vertical
displacements (Fig. 12). Since afterslip at the shallow parts gives
trenchward and subsidence displacements, the large afterslips were
estimated using the viscoelastic model to compensate the anticorre-
lated viscoelastic response. In contrast, the viscoelastic response at

a
t

K
y
o
t
o

U
n
i
v
e
r
s
i
t
y

o
n

M
a
y

1
4
,

2
0
1
4
h
t
t
p
:
/
/
g
j
i
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

14 A. M. Lubis, A. Hashima and T. Sato
Figure 13. Cumulative 15-month afterslip distribution. (a) elastic model only; (b) elastic and poroelastic properties; (c) elastic and viscoelastic properties;
and (d) afterslip corrected by poroelastic and viscoelastic modelling. Red and pink stars represent the epicentres for earthquake of M
w
8.5 and 7.8 on 2007
September 12.
the deeper parts has similar trend with the observed displacements
(Fig. 12).
Our condence in the location of afterslip distribution is driven
primarily by the observations of seismicity distributions (Collings
et al. 2012). 27 temporary seismic arrays installed from 2007
December to 2008 October recorded that the majority of after-
shocks were located immediately up-dip of the main rupture of the
2007 southern Sumatra earthquake (Fig. 15). In addition, a large
proportion of aftershocks activity appears in the region between the
north part of Pagai Island and Sipora Island, which does not rupture
during the 2007 southern Sumatra earthquake (Fig. 5), while after-
slip concentrates surrounding this area. Overall, we found evidence
that the afterslip distribution corresponds to location of aftershock
distributions (Fig. 15). Moreover, we tried to investigate whether
the afterslip is related to static coseismic Coulomb stress change
(Lin & Stein 2004). We found that the afterslip coincides clearly
with the area of increased coseismic Coulomb stress change, in
the up-dip region of the main rupture patch of the 2007 southern
Sumatra earthquake (Fig. 15). We suggest that increased Coulomb
stress change may promote afterslip around the western part of Pagai
Island. In addition, we also nd that aftershock locations are consis-
tent with areas of increased static coseismic Coulomb stress change
following the main shock. This is not surprising since numer-
ous studies have found good qualitative agreement between static
Coulomb stress changes (in areas experiencing positive stress
changes) and aftershock distributions (

Arnad ottir et al. 2003; Lin


& Stein 2004; Steacy et al. 2005).
We compared the afterslip distribution corrected only by vis-
coelastic relaxation due to coseismic slip using the method proposed
by Wang et al. (2009) with the afterslip distribution corresponding
to the poroelastic and viscoelastic responses to coseismic slip and
afterslip in each preceding observation period (Fig. S11). This re-
vealed the direct importance of viscoelastic responses to afterslip
distributions in the period preceding each observation. First, we
found that the difference in magnitude of afterslip distributions
between these two models was subtantial in the rst 3 months of
observation. Thereafter, we found that there were discrepancies of
up to 1 cm. We then evaluated all of the viscoelastic responses.
We found that the contribution of the vertical component of the
viscoelastic response to afterslip should be considered (Fig. 16).

a
t

K
y
o
t
o

U
n
i
v
e
r
s
i
t
y

o
n

M
a
y

1
4
,

2
0
1
4
h
t
t
p
:
/
/
g
j
i
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

Afterslip of the 2007 Sep. Sumatra earthquake 15
Figure 14. Cumulative 15-month GPS observations and synthetic displacements (a) for elastic medium; and (b) with complete media. Black arrows indicate
the observed horizontal displacements. Red, blue and green arrows indicate estimated surface displacements due to afterslips, poroelastic and viscoelastic
responses, and summed with afterslips, poroelastic and viscoelastic responses, respectively.
Because the cumulative viscoelastic response to afterslips in each
preceding observation period was approximately 10 per cent of the
viscoelastic response to coseismic slip, we considered that these re-
sponses contributed to the correction of the observed post-seismic
deformation data. The afterslip distribution has greater amplitude
when viscoelastic-corrections include the relaxation of afterslip it-
self, in addition to the relaxation of the coseismic slip (Fig. S11).
Post-seismic displacement depends strongly on viscosity values,
and most studies of post-seismic displacement show that it differs
among places and as a function of tectonic evolution and the envi-
ronment of the region of interest (e.g. Pollitz et al. 2000; Nishimura
& Thatcher 2003; Gourmelen & Amelung 2005; Ryder et al. 2007;
Hammond et al. 2009; Johnson et al. 2009). In addition, labora-
tory experiments have demonstrated that the viscosity of rock in the
lithosphere or asthenosphere depends on grain size, hydrous condi-
tion (water saturation), strain rate, temperature and lithosphere or
asthenosphere strength (when the materials are subjected to stress)
(e.g. Sheu &Shieh 2004; Burgmann &Dressen 2008). We therefore
needed to consider the viscosity characteristics. A grid search for
the best-t viscosity value for the asthenosphere yielded a viscosity
of 2.5 10
18
Pas (Fig. 11) for the Sumatra region of Indonesia.
Our results are broadly consistent with those of other studies infer-
ring the lithospheric rheology in the Sumatra subduction zone in
relation to the great 2004 SumatraAndaman earthquake and the
2005 Nias Sumatra earthquake. For example, Pollitz et al. (2006,
2008) used GPS post-seismic deformation to explain viscoelastic
behaviour associated with to the great 2004 SumatraAndaman
earthquake and the 2005 Nias Sumatra earthquake. They obtained
best-tting values of 10
19
Pas for steady-state asthenosphere vis-
cosity and 5 10
17
Pas for transient asthenosphere viscosity.
A viscosity of 5 10
17
Pas was also observed by Paul et al.
(2007) in an investigation of post-seismic GPS deformation in the
Andaman Islands. In addition, the post-seismic signature of Grav-
ity Recovery and Climate Experiment (GRACE) data associated
with the great 2004 SumatraAndaman earthquake gave a best-
tting viscosity between 5 10
17
and 1 10
18
Pas (e.g. Ogawa &
Heki 2007; Panet et al. 2007; Cannelli et al. 2008; Han et al. 2008).
Acombination of GPSand GRACEdata gave a best-tting viscosity
of 8 10
18
Pas (Panet et al. 2010). Comparison of the best-tting
viscosity for the 2007 southern Sumatra earthquake with that for the
2004 great SumatraAndaman earthquake reveals that the viscosity
for the former event is within the range of viscosity values derived
for the latter.
In this study, we use the most common viscoelastic model;
Maxwell viscoelastic rheology in the asthenosphere (e.g. Piersanti
et al. 1997; Piersanti 1999; Khazaradze et al. 2002; Hu et al. 2004;
Tanaka et al. 2009; Kogan et al. 2011). Other models for viscoelas-
tic model are Kelvin solid and Standard linear solid (e.g. Pollitz
et al. 2008; Segal 2010), but the pattern of viscoelastic relaxation
is similar for many models. The main difference is generally time
dependence. Low viscosity leads to large displacements in the ear-
lier stage. Therefore, we estimated the viscosity value using the
ABIC method. In addition, calculation of poroelastic deformation
by taking the difference in coseismic predicted surface displace-
ment between drained (fully relaxation) and undrained conditions
is also a common approach (e.g. Peltzer et al. 1998; J onsson et al.

a
t

K
y
o
t
o

U
n
i
v
e
r
s
i
t
y

o
n

M
a
y

1
4
,

2
0
1
4
h
t
t
p
:
/
/
g
j
i
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

16 A. M. Lubis, A. Hashima and T. Sato
Figure 15. Static Coulomb stress change due to main shock with friction
coefcient of 0.4 at depth of 20 km. Grey circles are aftershocks from
3 to 15 months, data obtained from Collings et al. (2012). Green contour
indicates coseismic slip distribution. Contour interval is 0.5 m. Black contour
is afterslip distribution from 3 to 15 months following the 2007 southern
Sumatra earthquake. Contour interval is 0.05 cm.
2003; Fialko 2004; Freed et al. 2006). Investigation of the time
dependence of poroelastic deformation at full relaxation using the
nite-difference method shows a similar result of poroelastic de-
formation using a drained--undrained solution (Barbot & Fialko
2010). Extraction of Vp, Vs and density from CRUST2 (A New
Global Crustal Model) (http://igppweb.ucsd.edu/gabi/rem.html0)
for our study area shows that the average of Poissons ratio (
a
)
is 0.31. For modelling the drained condition, we reduce Poissons
ratio from the undrained condition of 13 per cent to a depth of
20 km. Previous analysis of poroelastic deformation showed that
by taking a reduction in Poissons ratio of the undrained condition
of 13 per cent an observed component of post-seismic surface de-
formation could be explained by poroelastic models (Peltzer et al.
1998; J onsson et al. 2003). Therefore, we have condence in our
chosen parameters in this model, although to understand more de-
tails the rock properties, laboratory experiments are needed (Rice
& Cleary 1976).
It is well known that the crust is composed of rock containing
water-lled pore space, while the mantle is composed of viscous
material. Therefore, in this research, we demonstrate analysis of
afterslip distribution by comparing afterslip using an elastic only
model with simple poroelastic and viscoelastic models. Since the
magnitude of porelastic and viscoelastic effects is model depen-
dent, we determine the optimal viscosity by minimizing the ABIC
value during inversion. For the poroelastic model, we use the com-
mon model and parameters described above. In future studies, it is
very important to explore the amount of poroelastic and viscoelas-
Figure 16. Illustration of vertical displacement at 912 months (4th step
investigation). A is vertical displacement at 912 months due to viscoelas-
tic response to coseismic slip, B, C and D are vertical displacements at
912 months due to viscoelastic response to 03, 36 and 69 months after-
slips, respectively. E is vertical displacement due to afterslip in the previous
observation periods where E = B + C + D. F is total vertical viscoelastic
response (A + E) at 912 months. G is observed vertical displacement.
tic effects in detail by using different models as well as different
parameters.
One important result of the analysis of afterslip distribution is that
post-seismic slip surrounded the region of the greatest coseismic slip
during the 2007 southern Sumatra earthquake. Such a pattern was
mentioned by Hsu et al. (2007) in their investigation of post-seismic
slip in relation to the 1999 Chi-Chi earthquake in Taiwan. The
same pattern of afterslip distribution was found following the 2003
Tokachi-oki earthquake (M
w
8) in northern Japan (e.g. Miyazaki
et al. 2004; Baba et al. 2006). An example very close to ours in
terms of migration of the afterslip distribution to the up-dip area
occurred following the 2005 Nias earthquake (M
w
8.7) in Sumatra
subduction zone (Hsu et al. 2006). The cumulative afterslip over
the 15-month observation period after correction for poroelastic
rebound and viscoelastic relaxation was located predominantly in
one patch (Fig. 17a). The main slip (maximum slip 50 cm) occurred
in the up-dip region of coseismic rupture on Pagai Island. The
cumulative afterslip distribution was expanded to a shallowarea and
concentrated at depths of 1520 km; this area did not slip during
the 2007 September earthquakes. However, if we had neglected the
poroelastic and viscoelastic responses (see Fig. 13a), then we would
not have obtained an afterslip distribution pattern like that of the
2005 event, and the majority of the afterslip distribution would have

a
t

K
y
o
t
o

U
n
i
v
e
r
s
i
t
y

o
n

M
a
y

1
4
,

2
0
1
4
h
t
t
p
:
/
/
g
j
i
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

Afterslip of the 2007 Sep. Sumatra earthquake 17
Figure 17. (a) Comparison of coseismic slip distribution and cumulative 15-month afterslip following the 2007 southern Sumatra earthquake, cor-
rected by consideration of poroelastic and viscoelastic responses. Color image is afterslip distribution and black arrows indicate slip amounts
and directions of afterslip. Red and pink stars represent the epicentres for earthquake of M
w
8.5 and 7.8 on 2007 September 12. Pink con-
tour indicates coseismic slip distribution. Blue contours indicates coseismic slip distribution of the 2010 October Mentawai earthquake (M
w
7.7;
http://www.geol.ucsb.edu/faculty/ji/big_earthquakes/2010/10/25/sumatra_update.html). Grey circles are aftershocks distribution (M > 4.0) during 15 months
after the main event. Black lines are plate boundary interface for interval of 20 km and black rectangle is fault dimensions for afterslip model on the plate
boundary. Epicentre for the M
w
7.6 2009 September Padang earthquake is shown by green star. (b). Cross-sectional prole of aftershock distribution projected
N45
o
E vs. depth 15 months after the main shock on 2007 September 12.

a
t

K
y
o
t
o

U
n
i
v
e
r
s
i
t
y

o
n

M
a
y

1
4
,

2
0
1
4
h
t
t
p
:
/
/
g
j
i
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

18 A. M. Lubis, A. Hashima and T. Sato
overlapped with the coseismic rupture area; this would not have been
consistent with the theoretical post-seismic simulation in subduction
zones by Kato &Hirasawa (1997). Therefore, we consider ours to be
a relevant example of the importance of poroelastic and viscoelastic
responses in measuring afterslip distribution.
The main shock of the 2007 event was following by substantial
afterslip in the shallower path, and the afterslip may have triggered
a large aftershock. One important outcome of our results, if we
include the effects of poroelastic and viscoelastic responses on the
estimation of afterslip distribution, is that the sequence of afterslip
in the 2007 southern Sumatra earthquake may have been correlated
with, and may have triggered, the large shock (M
w
7.7) in 2010
October in the shallower area (510 km depth) of the main after-
slip region, 3 yr after the 2007 event (see Fig. 17). The occurrence
of the 2010 earthquake completed the partial rupture of the 2007
events; thus the remaining stress accumulated since the 1833 event
was released probably by the 2010 event, but no slip occurred at the
plate boundary in the Mentawai segment area (particularly, around
Siberut Island and the north part of Sipora Island), where the earth-
quake (M
w
about 8.58.9) occurred in 1797. This indicates that
the accumulation of stress on Siberut Island and the north part of
Sipora Island was not released during the coseismic slip and cu-
mulative afterslip distributions in the 15- month observation period
following the 2007 southern Sumatra earthquake or in the rupture
that occurred in the large aftershock that constituted the M
w
7.7
earthquake in 2010. According to our results, and those of Chlieh
et al. (2008), this region remains locked, leaving a seismic gap area
that will probably be released in the coming decades.
5 CONCLUSI ONS
We successfully modelled post-seismic deformation following the
2007 southern Sumatra earthquake using poroelastic rebound and
viscoelastic relaxation modelling. Using elastic modelling only, the
afterslip area in the 15 months following the 2007 southern Suma-
tra earthquake overlapped with the coseismic rupture area to the
southeast of Pagai Islandabout 100 km from the epicentre of the
rst earthquake. After correction for poroelastic and viscoelastic
responses, the afterslip mostly migrated from the main coseismic
rupture area southeast of Pagai Island to the west of Pagai Island at
shallow depths, showing a consistency with aftershock distribution
The total moment released during the 15-month observation period
was 0.99 10
21
Nm (M
w
7.9)about 15 per cent of the coseismic
moment of the main shock. The cumulative afterslip corrected by
using poroelastic and viscoelastic modelling was 0.50 m over the
15 months (almost the same as for the afterslip obtained with elastic
modelling only), but consideration of the poroelastic and viscoelas-
tic responses in calculating afterslip distribution provided a more
appropriate explanation of the afterslip process.
Consideration of poroelastic response led to a substantial cor-
rection in the afterslip distribution in the rst 3 months after the
earthquake. After the rst 3 months, viscoelastic responses led to
a substantial correction. Therefore, we concluded that the afterslip
and poroelastic response in the early period of the post-seismic
event should be combined with viscoelastic relaxation during all
periods of the post-seismic event in order to obtain the best and
most reliable afterslip distribution.
We note that the viscosity value of the asthenosphere layer inu-
ences the afterslip distribution. Modelling of post-seismic deforma-
tion using a model of the structure of the Maxwellian viscoelastic
asthenosphere underlying two layers of elastic lithosphere depends
strongly on the viscosity value. The viscosity value for the astheno-
sphere layer is a crucial unknown parameter. In this research, we
estimated a viscosity of 2.5 10
18
Pas for the Sumatra subduction
zone by using a grid search method for the minimum ABIC value.
ACKNOWLEDGMENTS
We thank the Tectonic Observatory at the California Institute of
Technology (Caltech) and Earth Observatory of Singapore (EOS)
for operating the SuGAr array. The authors are grateful to editor,
Massimo Cocco, and anonymous reviewers for thoughtful reviews.
Careful reading of the manuscript by Dr. Emma Hill is gratefully
acknowledged. We thank Sylvian Barbot for discussion about poroe-
lasticity. We particularly thank Thomas Herring and Robert King
of Massachusetts Institute of Technology (MIT) for kindly guiding
us in the GPS data processing. Some of the gures were created
using the GMT software (Wessel & Smith 1998). This work was
supported by JSPS KAKENHI Grant Number 20540404.
REFERENCES
Akaike, H., 1980. Likelihood and the Bayes procedure, in Bayesian Statis-
tics, pp. 143166, eds Bernardo, J.M., DeGroot, M.H., Lindley, D.V. &
Smith, A.F.M., University Press, Valencia.

Arnad ottir, T., J onsson, S., Pedersen, R. & Gudmundsson, G.B., 2003.
Coulomb stress changes in the South Iceland Seismic Zone due to
two large earthquakes in June 2000, Geophys. Res. Lett., 30(5), 1205,
doi:10.1029/2002GL016495.
Baba, T., Hirata, K., Hori, T. & Sakaguchi, H., 2006. Offshore geodetic
data conducive to the estimation of the afterslip distribution following the
2003 Tokachi-oki earthquake, Earth planet. Sci. Lett., 241(12), 281292,
doi:10.1016/j.epsl.2005.10.019.
Barbot, S. & Fialko, Y., 2010. A unied continuum representation of post-
seismic relaxation mechanisms: semi-analytic models of afterslip, poroe-
lastic rebound and viscoelastic ow, Geophys. J. Int., 182(3), 11241140,
doi:10.1111/j.1365-246X.2010.04678.x.
Borrero, J.C., Weiss, R., Okal, E., Hidayat, R., Suranto, Arcas, D. & Titov,
V.V., 2009. The Tsunami of September 12, 2007, Bengkulu Province,
Sumatra, Indonesia: post-tsunami eld survey and numerical modeling,
Geophys. J. Int., 178, 180194, doi:10.1111/j.1365-246X.2008.04058.x.
Burgmann, R., Ergintav, S., Segall, P., Hearn, E.H., McClusky, S.C.,
Reilinger, R.E., Woith, H. &Zschau, J., 2002. Time-dependent distributed
afterslip on and deep belowthe Izmit earthquake rupture, Bull. seism. Soc.
Am., 92(1), 126137, doi:10.1785/0120000833.
Burgmann, R. & Dressen, G., 2008. Rheology of the lower crust
and upper mantle: evidence from rock mechanics, geodesy, and
eld observations, Annu. Rev. Earth planet. Sci., 36, 53167,
doi:10.1146/annurev.earth.36.031207.124326.
Briggs, R.W. et al., 2006. Deformation and slip along the Sunda megathrust
in the great 2005 Nias-Simeulue earthquake, Science, 311(5769), 1897
1901, doi:10.1126/science.1122602.
Cannelli, V., Melini, D., Piersanti, A. & Boschi, E., 2008. Postseismic sig-
nature of the 2004 Sumatra earthquake on low-degree gravity harmonics,
J. geophys. Res., 113, B12414, doi:10.1029/2007JB005296.
Chlieh, M., Avouac, J.P., Sieh, K., Natawidjaja, D.H. & Galetzka, J.,
2008. Heterogeneous coupling on the Sumatra megathrust constrained
from geodetic and paleogeodetic measurements, J. geophys. Res., 113,
doi:10.1029/2007JB004981.
Collings, R., Lange, D., Rietbrock, A., Tilmann, F., Natawidjaja, D., Suwar-
gadi, B., Miller, M. &Saul, J., 2012. Structure and seismogenic properties
of the Mentawai segment of the Sumatra subduction zone revealed by lo-
cal earthquake traveltime tomography, J. geophys. Res., 117, B01312,
doi:10.1029/2011JB008469.

a
t

K
y
o
t
o

U
n
i
v
e
r
s
i
t
y

o
n

M
a
y

1
4
,

2
0
1
4
h
t
t
p
:
/
/
g
j
i
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

Afterslip of the 2007 Sep. Sumatra earthquake 19
Feigl, K.L. & Thatcher, W., 2006. Geodetic observations of post-seismic
transients in the context of the earthquake deformation cycle, Compt.
Rend. Geosci., 338(1415), 10121028, doi:10.1016/j.crte.2006.06.006.
Fialko, Y., 2004. Evidence of uid-lled upper crust from observations of
post-seismic deformation due to the 1992 M
w
7.3 Landers earthquake, J.
geophys. Res., 109, B08401, doi:10.1029/2004JB002985.
Freed, A.M., B urgmann, R., Calais, E., Freymueller, J. & Hreinsd ottir, S.,
2006. Implications of deformation following the 2002 Denali, Alaska
earthquake for postseismic relaxation processes and lithospheric rheol-
ogy, J. geophys. Res., 111, B01401, doi:10.1029/2005JB003894.
Fujii, Y. & Satake, K., 2008. Tsunami waveform inversion of the 2007
Bengkulu, southern Sumatra, earthquake, Earth Planet Space, 60(9), 993
998.
Fukahata, Y. &Wright, T.J., 2008. Anon-linear geodetic data inversion using
ABICfor slip distribution on a fault with an unknown dip angle, Geophys.
J. Int., 173(2), 353364, doi:10.1111/j.1365-246X.2007.03713.x.
Fukahata, Y., Nishitani, A. & Matsuura, M., 2004. Geodetic data inver-
sion using ABIC to estimate slip history during one earthquake cycle
with viscoelastic slip-response functions, Geophys. J. Int., 156, 140153,
doi:10.1111/j.1365-246X.2004.02122.x.
Fukuda, J. & Johnson, K.M., 2008. A Fully Bayesian Inversion for Spatial
Distribution of Fault Slip with Objective Smoothing, Bul. seism. Soc. Am.,
98(3), 11281146, doi:10.1785/0120070194.
Gourmelen, N. & Amelung, F., 2005. Postseismic mantle relaxation
in the Central Nevada seismic belt, Science, 310(5753), 14731476,
doi:10.1126/science.1119798.
Hammond, W.C., Kreemer, C. & Blewitt, G., 2009. Geodetic constraints
on contemporary deformation in the northern Walker Lane: 3, Central
Nevada Seismic Belt postseismic relaxation, Geol. Soc. Am. Spec. Pap.,
447, 3354, doi:10.1130/2009.2447(03).
Han, S.C., Sauber, J., Luthcke, S. B., Ji, C., Pollitz, F. F., 2008. Implications
of postseismic gravity change following the great 2004 Sumatra-Andaman
earthquake from the regional harmonic analysis of GRACE intersatellite
tracking data, J. Geophys. Res., 113, B11413, doi:10.1029/2008JB005705.
Herring, T., 2000. Global Kalman Filter VLBI and GPS Analysis Program,
Version 5.0. Massachusetts Institute of Technology, Cambridge, MA.
Hu, Y., Wang, K., He, J., Klotz, J. & Khazaradze, G., 2004. Three-
dimensional viscoelastic nite element model for postseismic deforma-
tion of the great 1960 Chile earthquake, J. geophys. Res., 109, B12403,
doi:10.1029/2004JB003163.
Hsu, Y.J. et al., 2006. Frictional afterslip following the M
w
8.7, 2005
NiasSimeulue earthquake, Sumatra, Science, 312(5782), 19211926,
doi:10.1126/science.1126960.
Hsu, Y.J., Segall, P., Yu, S.B., Kuo, L.C. & Williams, C.A., 2007. Tem-
poral and spatial variations of post-seismic deformation following the
1999 Chi-Chi, Taiwan earthquake, Geophys. J. Int., 169(2), 367379,
doi:10.1111/j.1365-246X.2006.03310.x.
Hughes, K.L.H., Masterlark, T. & Mooney, W.D., 2010. Poroelastic stress-
triggering of the 2005 M8.7 Nias earthquake by the 2004 M
w
9.2
SumatraAndaman earthquake, Earth planet. Sci. Lett., 293, 289299,
doi:10.1016/j.epsl.2010.02.043.
Johnson, K.M., B urgmann, R. & Freymueller, J.T., 2009. Coupled af-
terslip and viscoelastic ow following the 2002 Denali Fault, Alaska
earthquake, Geophys. J. Int., 176(3), 670682, doi:10.1111/j.1365-
246X.2008.04029.x.
J onsson, S., Segall, P., Pedersen, R. & Jornsson, G., 2003. Post-earthquake
ground movements correlated to pore pressure transients. Nature, 4242,
179183.
Lay, T. et al., 2005. The great Sumatra-Andaman earthquake of 26 December
2004, Science, 308(1127), 11271133, doi:10.1126/science.1112250.
Lin, J. & Stein, R.S., 2004. Stress triggering in thrust and subduction
earthquakes and stress interaction between the southern San Andreas
and nearby thrust and strike-slip faults, J. geophys. Res., 109, B02303,
doi:10.1029/2003JB002607.
Lorito, S., Romano, F., Piatanesi, A. & Boschi, E., 2008. Source process
of the September 12, 2007, M
w
8.4 southern Sumatra earthquake from
tsunami tide gauge record inversion, Geophys. Res. Lett., 35, L02310,
doi:10.1029/2007GL032661.
Lubis, A.M., 2011. Modeling afterslip distribution with elastic, poroelastic
and viscoelastic responses investigated from Global Positioning System
(GPS) observations, PhD thesis, Chiba University, Japan.
Kanamori, H., Rivera, L. & Lee, W.H.K., 2010. Historical seismograms for
unravelling a mysterious earthquake: the 1907 Sumatra Earthquake, Geo-
phys. J. Int., 183(1), 358374, doi:10.1111/j.1365-246X.2010.04731.x.
Kato, N. &Hirasawa, T., 1997. Anumerical study on seismic coupling along
subduction zones using a laboratory-derived friction law, Phys. Earth
planet. Inter., 102(12), 5168, doi:10.1016/S0031-9201(96)03264-5.
Khazaradze, G., Wang, K., Klotz, J., Hu, Y. & He, J., 2002. Prolonged
post-seismic deformation of the 1960 great Chile earthquake and im-
plications for mantle rheology, Geophys. Res. Lett., 29(22), 2050,
doi:10.1029/2002GL015986.
King, R. & Bock, Y., 2004. Documentation for the GAMIT GPS Analysis
Software, Release 10.2. Massachusetts Institute of Technology, Cam-
bridge, MA, and Scripps Institute of Oceanography, La Jolla, CA.
Kogan, M.G., Vasilenko, N.F., Frolov, D.I., Freymueller, J.T., Steblov, G.M.,
Levin, B.W. & Prytkov, A.S., 2011. The mechanism of postseismic de-
formation triggered by the 20062007 great Kuril earthquakes, Geophys.
Res. Lett., 38, L06304, doi:10.1029/2011GL046855.
Konca, A.O. et al., 2008. Partial rupture of a locked patch of the Sumatra
megathrust during the 2007 earthquake sequence, Nature, 456, 631635.
Matsu ura, M., Tanimoto, T. & Iwasaki, T., 1981. Quasi-static displacements
due to faulting in a layered half-space with an intervenient viscoelastic
layer, J. Phys. Earth, 29, 2354.
Matsuura, M. & Sato, T., 1989. A dislocation model for the earthquake
cycle at convergent plate boundaries, Geophys. J. Int., 96(1), 2332,
doi:10.1111/j.1365-246X.1989.tb05247.x.
Matsuura, M., Noda, A. & Fukahata, Y., 2007. Geodetic data inver-
sion based on Bayesian formulation with direct and indirect prior in-
formation, Geophys. J. Int., 171(3), 13421351, doi:10.1111/j.1365-
246X.2007.03578.x.
Miyazaki, S., Segall, P., Fukuda, J. & Kato, T., 2004. Space time distribution
of afterslip following the 2003 Tokachi-oki earthquake: implications for
variations in fault zone frictional properties, Geophys. Res. Lett., 31,
L06623, doi:10.1029/2003GL019410.
Nalbant, S.S., Steacy, S., Sieh, K., Natawidjaja, D.H. &McCloskey, J., 2005.
Seismology: earthquake risk on the Sunda trench, Nature, 435, 756757,
doi:10.1038/nature435756a.
Natawidjaja, D.H. et al., 2006. Source parameters of the great Sumatran
megathrust earthquakes of 1797 and 1833 inferred fromcoral microatolls,
J. geophys. Res., 111, B06403, doi:10.01029/02005JB004025.
Natawidjaja, D.H., Sieh, K., Galetzka, J., Suwargadi, B. W., Cheng, H., Ed-
wards, R. L., Chlieh, M., 2007. Interseismic deformation above the Sunda
Megathrust recorded in coral microatolls of the Mentawai islands, West
Sumatra, J. Geophys. Res., 112, B02404, doi:10.1029/2006JB004450.
Nishimura, T. & Thatcher, W., 2003. Rheology of the lithosphere inferred
from postseismic uplift following the 1959 Hebgen Lake earthquake, J.
geophys. Res., 108(B8), 2389, doi:10.1029/2002JB002191.
Nishimura, T. et al., 2000. Distribution of seismic coupling on the subduct-
ing plate boundary in northeast Japan inferred from GPS observations,
Tectonophysics, 323, 217238.
Ogawa, R. & Heki, K., 2007. Slow postseismic recovery of geoid depres-
sion formed by the 2004 Sumatra-Andaman Earthquake by mantle water
diffusion, Geophys. Res. Lett., 34, L06313, doi:10.1029/2007GL029340.
Ozawa, S., Kaidzu, M., Murakami, M., Imakiire, T. & Hatanaka, Y., 2004.
Coseismic and postseismic crustal deformation after the M
w
8 Tokachi-
oki earthquake in Japan, Earth Planets Space, 56, 675680.
Panet, I. et al., 2007. Co-seismic and post-seismic signatures of the
Sumatra December 2004 and March 2005 earthquakes in GRACE
satellite gravity, Geophys. J. Int., 171(1), 177190, doi:10.1111/j.1365-
246X.2007.03525.x.
Panet, I., Pollitz, F.F., Mikhailov, V., Diament, M., Banerjee, P. & Grijalva,
K., 2010. Upper mantle rheology from GRACE and GPS postseismic
deformation after the 2004 Sumatra-Andaman earthquake, Geochem.,
Geophys., Geosyst., 11, Q06008, doi:10.1029/2009GC002905.
Paul, J., Lowry, A.R., Bilham, R., Sen, S. &Smalley, R. Jr., 2007. Postseismic
deformation of the Andaman Islands following the 26 December, 2004

a
t

K
y
o
t
o

U
n
i
v
e
r
s
i
t
y

o
n

M
a
y

1
4
,

2
0
1
4
h
t
t
p
:
/
/
g
j
i
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

20 A. M. Lubis, A. Hashima and T. Sato
Great Sumatra-Andaman earthquake, Geophys. Res. Lett., 34, L19309,
doi:10.1029/2007GL031024.
Perfettini, H. & Avouac, J.P., 2004. Postseismic relaxation driven by brittle
creep: a possible mechanism to reconcile geodetic measurements and the
decay rate of aftershocks, application to the Chi-Chi earthquake, Taiwan,
J. geophys. Res., 109, B02304, doi:10.1029/2003JB002488.
Peltzer, G., Rosen, P., Rogez, F. & Hudnut, K., 1998. Poroelastic rebound
along the Landers 1992 earthquake surface rupture, J. geophys. Res.,
103(B12), 30 13130 145.
Piersanti, A., Spada, G. & Sabadini, R., 1997. Global postseismic re-
bound of a viscoelastic Earth: theory for nite faults and application
to the 1964 Alaska earthquake, J. geophys. Res., 102(B1), 477492,
doi:10.1029/96JB01909.
Piersanti, A., 1999. Postseismic deformation in Chile: constraints on
the asthenospheric viscosity, Geophys. Res. Lett., 26(20), 31573160,
doi:10.1029/1999GL005375.
Pollitz, F.F., Peltzer, G. & B urgmann, R., 2000. Mobility of continen-
tal mantle: evidence from postseismic geodetic observations following
the 1992 Landers earthquake, J. geophys. Res., 105(B4), 80358054,
doi:10.1029/1999JB900380.
Pollitz, F.F., B urgmann, R. & Banerjee, P., 2006. Post-seismic relax-
ation following the great 2004 Sumatra-Andaman earthquake on a
compressible self-gravitating Earth, Geophys. J. Int., 167(1), 397420,
doi:10.1111/j.1365-246X.2006.03018.x.
Pollitz, F.F., Banerjee, P., Grijalva, K., Nagarajan, B. & B urgmann, R., 2008.
Effect of 3-D viscoelastic structure on post-seismic relaxation from the
2004 M
w
= 9.2 Sumatra earthquake, Geophys. J. Int., 173(1), 189204,
doi:10.1111/j.1365-246X.2007.03666.x.
Reddy, C.D., Prajapati, S.K. & Sunil, P.S., 2010. Co and post-
seismic characteristics of Indian subcontinent in response to the
2004 Sumatra earthquake, J. Asia Earth Sci., 39(6), 620626,
doi:10.1016/j.jseaes.2010.04.019.
Rice, J.R. & Cleary, M.P., 1976. Some basic stress diffusion solutions for
uid-saturated elastic porous media with compressible constituents, Rev.
Geophys., 14(2), 227241, doi:10.1029/RG014i002p00227.
Ryder, I., Parsons, B., Wright, T.J. & Funning, G.J., 2007. Post-seismic
motion following the 1997 Manyi (Tibet) earthquake: InSARobservations
and modeling, Geophys. J. Int., 169(3), 10091027, doi:10.1111/j.1365-
246X.2006.03312.x.
Segall, P., B urgmann, R. & Matthews, M., 2000. Time-dependent triggered
afterslip following the 1989 Loma-Prieta earthquake, J. geophys. Res.,
105(B3), 56155634, doi:10.1029/1999JB900352.
Segall, P., 2010. Earthquake and Volcano Deformation, Princeton University
Press, Princeton, 246 pp. ISBN 978-0691133027.
Selva, J., Marzocchi, W., Zencher, F., Casarotti, E., Piersanti, A. & Boschi,
E., 2004. A forward test for interaction between remote earthquakes
and volcanic eruptions: the case of Sumatra (June 2000) and Denali
(November 2002) earthquakes, Earth planet. Sci. Lett., 226(34), 383
395, doi:10.1016/j.epsl.2004.08.006.
Sheu, S.Y. & Shieh, C.F., 2004. Viscoelasticafterslip concurrence: a pos-
sible mechanism in the early post-seismic deformation of the M
w
7.6,
1999 Chi-Chi (Taiwan) earthquake, Geophys. J. Int., 159(3), 11121124,
doi:10.1111/j.1365-246X.2004.02437.x.
Sieh, K. et al., 2008. Earthquake supercycles inferred fromsea-level changes
recorded in the corals of West Sumatra, Science, 322(5908), 16741678,
doi:10.1126/science.1163589.
Steacy, S., Gomberg, J. & Cocco, M., 2005. Introduction to special section:
stress transfer, earthquake triggering, and time-dependent seismic hazard,
J. geophys. Res., 110, B05S01, doi:10.1029/2005JB003692.
Tanaka, Y., Klemann, V., Fleming, K. & Martinec, Z., 2009. Spectral
nite element approach to postseismic deformation in a viscoelas-
tic self-gravitating spherical Earth, Geophys. J. Int, 176, 715739,
doi:10.1111/j.1365-246X.2008.04015.x.
Takahashi, H. et al., 2004. GPS observation of the rst month of postseismic
crustal deformation associated with the 2003 Tokachi-oki earthquake
(MJMA 8.0), off southeastern Hokkaido, Japan, Earth Planets Space, 56,
377382.
Wang, L., Wang, R., Roth, F., Enescu, B., Hainzl, S. & Ergintav, S., 2009.
Afterslip and viscoelastic relaxation following the 1999 M
w
7.4

Izmit
earthquake from GPS measurements, Geophys. J. Int., 178(3), 1220
1237.
Wessel, P. & Smith, W.H.F., 1998. New, improved version of the Generic
Mapping Tools Released, EOS, Trans. Am. geophys. Un., 79, 579.
Yagi, Y., Kikuchi, M. & Sagiya, T., 2001. Co-seismic slip, post-seismic
slip, and aftershocks associated with two large earthquakes in 1996 in
Hyuga-nada, Japan, Earth Planet Space, 53, 793803.
Yabuki, T. & Matsuura, M., 1992. Geodetic data inversion using a Bayesian
information criterion for spatial distribution of fault slip, Geophys. J. Int.,
109(2), 363375, doi:10.1111/j.1365-246X.1992.tb00102.x.
Yoshioka, S., Yabuki, T., Sagiya, T., Tada, T. & Matsuura, M., 1993. In-
terplate coupling and relative plate motion in the Tokai district, central
Japan, deduced from geodetic data inversion using ABIC, Geophys. J.
Int., 113(3), 607621, doi:10.1111/j.1365-246X.1993.tb04655.x.
Zachariasen, J., Sieh, K., Taylor, F., Edwards, R. & Hantoro, W., 1999. Sub-
mergence and uplift associated with the giant 1833 Sumatran subduction
earthquake: evidence from coral microatolls, J. geophys. Res., 104(B1),
895919.
SUPPORTI NG I NFORMATI ON
Additional Supporting Information may be found in the online ver-
sion of this article:
Figures S1S5. Horizontal post-seismic GPS displacements in each
3-month observation period.
Figure S6. Daily positions of vertical post-seismic displacements
at several GPS stations following the 2007 southern Sumatra earth-
quake.
Figure S7. Coseismic slip distribution on the plate boundary, esti-
mated fromABICinversion of 3-DGPS data. Black arrows indicate
slip amounts and directions. Green rectangle is the fault dimensions
on the plate boundary. Red and pink stars represent the epicentres
for earthquake of M
w
8.5 and 7.8 on 2007 September 12.
Figure S8. Estimated error for coseismic slip distribution from
joint inversion of the GPS and coral data. Green rectangle is fault
dimensions on the plate boundary. Red and pink stars represent the
epicentres for earthquake of M
w
8.5 and 7.8 on 2007 September 12.
Figure S9. Test of resolution of coseismic modelling by using the
checkerboard method, with a default spacing of 100 km 100 km
for input slips of (a) 7 m, (b) 3 m, (c) 1 m, (d) 0.3 m, (e) 0.1 m and
(f) 0.05 m.
Figure S10. Estimated error of afterslip distribution (ae) errors
of inversion of post-seismic GPS data for ve observation periods
using the elastic model only; (fj) errors using the poroelastic and
viscoelastic model in the ve observation periods.
Figure S11. Differences between afterslip distribution corrected
by considering the viscoelastic response due to coseismic slip
only (ae), and afterslip distribution corrected by considering the
poroelastic response and viscoelastic responses due to coseis-
mic slip and afterslip in the preceding observation periods (fj).
(http://gji.oxfordjournals.org/lookup/suppl/doi:10.1093/gji/ggs020/-/
DC1)
Please note: Oxford University Press are not responsible for the
content or functionality of any supporting materials supplied by
the authors. Any queries (other than missing material) should be
directed to the corresponding author for the article.

a
t

K
y
o
t
o

U
n
i
v
e
r
s
i
t
y

o
n

M
a
y

1
4
,

2
0
1
4
h
t
t
p
:
/
/
g
j
i
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

Das könnte Ihnen auch gefallen