Earth and Planetary Science Letters 365 (2013) 108119Contents lists available at SciVerse ScienceDirectEarth and Planetary Science Letters0012-82

http://d

n Corr

E-mjournal homepage: www.elsevier.com/locate/epslSubduction system variability across the segment boundary of the2004/2005 Sumatra megathrust earthquakes

A. Shulgin a, H. Kopp a,n, D. Klaeschen a, C. Papenberg a, F. Tilmann b, E.R. Flueh a, D. Franke c,U. Barckhausen c, A. Krabbenhoeft a, Y. Djajadihardja d

a GEOMAR Helmholtz Centre for Ocean Research Kiel, Wischhofstr. 1-3, 24148 Kiel, Germanyb German Research Center for Geosciences (GFZ), Telegrafenberg, D-14473 Potsdam, Germanyc Federal Institute for Geosciences and Natural Resources (BGR), Stilleweg 2, D-30655 Hanover, Germanyd Agency for the Assessment and Application of Technology (BPPT), Jl.M.H. Thamrin No. 8, Jakarta 10340, Indonesiaa r t i c l e i n f o

Article history:

Received 23 December 2011

Received in revised form

19 December 2012

Accepted 20 December 2012Editor: P. Shearerthe Mw 8.6 March 28, 2005 Sumatra earthquakes. A set of collocated reflection and wide-angle seismicAvailable online 28 February 2013

Keywords:

subduction zone segmentation

seismic tomography

crustal structure

seismic slip barrier

Sumatra forearc1X/$ - see front matter & 2012 Elsevier B.V.

x.doi.org/10.1016/j.epsl.2012.12.032

esponding author. Tel.: 49 431 600 2334; fail address: hkopp@geomar.de (H. Kopp).a b s t r a c t

Subduction zone earthquakes are known to create segmented patches of co-seismic rupture along-

strike of a margin. Offshore Sumatra, repeated rupture occurred within segments bounded by

permanent barriers, whose origin however is still not fully understood. In this study we image the

structural variations across the rupture segment boundary between the Mw 9.1 December 26, 2004 and

profiles are available on both sides of the segment boundary, located offshore Simeulue Island.

We present the results of the seismic tomography modeling of wide-angle ocean bottom data,

enhanced with MCS data and gravity modeling for the southern 2005 segment of the margin and

compare it to the published model for the 2004 northern segment. Our study reveals principal

differences in the structure of the subduction system north and south of the segment boundary,

attributed to the subduction of 961E fracture zone. The key differences include a change in the crustalthickness of the oceanic plate, a decrease in the amount of sediment in the trench as well as variations

in the morphology and volume of the accretionary prism. These differences suggest that the 961Efracture zone acts as an efficient barrier in the trench parallel sediment transport, as well as a divider

between oceanic crustal blocks of different structure. The variability of seismic behavior is caused by

the distinct changes in the morphology of the subduction complex across the boundary related to the

difference in the sediment supply.

& 2012 Elsevier B.V. All rights reserved.1. Introduction

Since 2004, repeated large and great megathrust earthquakesalong the Sumatra subduction zone have ended a 40 yr period ofrelative moderate (Mwo8.5) global seismic activity. Studies of theMw9.1 SumatraAndaman earthquake in 2004 and followingevents of magnitudes Z8.4 in 2005 and 2007 confirmed previousobservations of segmented rupture areas of earthquakes alongsubduction zones (Ando, 1975; DeShon et al., 2005; Spence, 1977).Offshore Sumatra, rupture occurs in discrete patches (e.g. Chliehet al., 2007), bounded by distinct barriers along-strike. Detailedseismological studies reveal variations in the updip limit of the2004 and 2005 ruptures, respectively (Ammon et al., 2005; Ishiiet al., 2005). These observations imply across-strike heterogeneityAll rights reserved.

ax: 49 431 600 2922.in physical properties along the megathrust. The controllingphysical processes, however, are still not fully understood.

The tectonics around Northern Sumatra are predominantlycontrolled by the subduction of the oceanic Indo-Australian plateunderneath Eurasia. The current convergence rate offshore North-ern Sumatra is estimated at 51 mm/yr (Prawirodirdjo and Bock,2004). The increasing obliquity of the convergence northwardsfrom the Sunda Strait (McCaffrey, 2009; Moore and Curray, 1980)results in the formation and development of a number ofarc-parallel strike-slip fault systems (Fig. 1). The most significantare the Sumatra and the West Andaman Fault systems, accom-modating arc-parallel strain (Malod and Kemal, 1996; Mosheret al., 2008; Sieh and Natawidjaja, 2000) offshore central-southern Sumatra. For the Mentawai fault system (Berglar et al.,2010; Diament et al., 1992; Kopp et al., 2001; Lelgemann et al.,2000; Malod and Kemal, 1996), recent findings suggest deforma-tion dominated by backthrusting (Singh et al., 2010, 2011).

The seismic unzipping of the Sumatra margin since 2004offers the unique opportunity to investigate the relationship

www.elsevier.com/locate/epslwww.elsevier.com/locate/epslhttp://dx.doi.org/10.1016/j.epsl.2012.12.032http://dx.doi.org/10.1016/j.epsl.2012.12.032http://dx.doi.org/10.1016/j.epsl.2012.12.032http://crossmark.dyndns.org/dialog/?doi=10.1016/j.epsl.2012.12.032&domain=pdfhttp://crossmark.dyndns.org/dialog/?doi=10.1016/j.epsl.2012.12.032&domain=pdfhttp://crossmark.dyndns.org/dialog/?doi=10.1016/j.epsl.2012.12.032&domain=pdfmailto:hkopp@geomar.dehttp://dx.doi.org/10.1016/j.epsl.2012.12.032

Fig. 1. Map of the study area offshore Northern Sumatra. The location of the wide-angle (SeaCause) and the collocated MCS (BGR06-135) profiles described in this studyare shown by red line. The arrow shows the plate convergence vector. Major faults are shown by black solid and dashed lines (after Curray, 2005). Hypocenter locations of

MwZ7.0 earthquakes are shown by magenta stars. The local seismicity relocated in a pseudo 3D model (recorded for 3 month in 2006, see text) is shown by depth color-coded circles (rmso0.1 s, seismicity with poor depth constrains is shown by gray circles; after Tilmann et al., (2010). The proposed segment boundary between therupture areas of the 2004 and 2005 earthquakes is shown by thick gray line (Franke et al., 2008). Dashed white lines mark the extent of the Wharton Ridge in the vicinity of

the trench. Bold black dashed lines are the fracture zones on the oceanic plate identified from the bathymetric and magnetic data. Blue lineseismic profile SAGER shown

in Fig. 7 from Klingelhoefer et al. (2010). Dashed magenta linesMCS profiles described in Franke et al. (2008). (For interpretation of the references to color in this figure

legend, the reader is referred to the web version of this article.)

A. Shulgin et al. / Earth and Planetary Science Letters 365 (2013) 108119 109between interseismic coupling and the mechanical properties ofthe megathrust including the nature of asperities and barriers tomegathrust ruptures. The 2004 earthquake asymmetrically rup-tured to the north over a distance of more than 1200 km, whilethe 2005 event showed a bidirectional rupture. However, there isa clear delineation between these events, located underneath theisland of Simeulue (Ammon et al., 2005; Briggs et al., 2006;Gahalaut and Catherine, 2006; Subarya et al., 2006). Franke et al.(2008), based on bathymetry and multichannel seismic (MCS)data, proposed the segment boundary running NESW throughthe island of Simeulue, as discussed earlier by Briggs et al. (2006)(Fig. 1). The observed location of the pivot line (a proxy to thedowndip limit of the rupture area) of the pre-2005 earthquake netuplift on Simeulue (Meltzner et al., 2006) spatially correlates withthe proposed segment boundary.

A number of possible scenarios were suggested for the forma-tion of segment boundaries: these include tectonic structures inthe overriding plate (Collot et al., 2004; Ryan and Scholl, 1993),mechanical discontinuities in the subducting plate (Aki, 1979;Spence, 1977) and topographic relief on the oceanic plate such asseamounts, ridges or fracture zones (Bilek et al., 2003; Bilek, 2010.Offshore Sumatra, the oceanic plate is characterized by prominenttectonic features observed in the bathymetry and magnetic data(Sclater and Fisher, 1974): northsouth trending fracture zonesenter the trench around 961E and 971E (Deplus et al., 1998). Inaddition, the Wharton Ridge, representing the southwestnortheast trending segments of a fossil spreading axis, isapproaching the trench offshore Nias (Fig. 1). For northernSumatra, Subarya et al. (2006) suggested that the segmentboundary between the 20042005 patches is linked to thesubduction of the 961E fracture zone on the oceanic plate, basedon the coseismic slip modeling results. This fracture zone wasidentified from magnetic studies (Sclater and Fisher, 1974), aswell as from the bathymetry data seaward of the trench; in thevicinity of the trench no anomalous relief is observed, due toincrease of sediment fill. In an MCS profile presented by Frankeet al. (2008) (their Fig. 5) the fracture zone is manifested by anarea of low/absent reflectivity from the top of the oceanic crust, aswell as in the thinning of the sedimentary cover. It was alsoshown that this is a pre-existing feature and not subduction-related, based on the sediment onlap above the zone of shallowbasement (Dean et al., 2010).

A. Shulgin et al. / Earth and Planetary Science Letters 365 (2013) 108119110The rupture patterns of the 2004 and 2005 events have beendocumented in a number of studies (e.g. Ammon et al., 2005;Banerjee et al., 2007; Briggs et al., 2006; Chlieh et al., 2007;Gahalaut and Catherine, 2006; Ishii et al., 2005; Lay et al., 2005;Rhie et al., 2007; Sibuet et al., 2007) and both events range amongthe most intensely studied earthquakes globally. In addition tothe varying lateral extent of the rupture planes, the most strikingdifference is the variation in the updip limit of seismogenicrupture. The 2004 rupture area offshore Northern Sumatraexperienced co-seismic slip from the outer forearc basin to theouter plateau region, close to the trench (Henstock et al., 2006).Observations of co-seismic slip and afterslip of the 2004 earth-quake are based on GPS data (Banerjee et al., 2007; Chlieh et al.,2007) and are related to coral and satellite-derived uplift rates(Meltzner et al., 2006) and aftershock locations (Klingelhoeferet al., 2010), which support the updip limit of the seismogeniczone being closer to the trench compared to the 2005 event. The2005 rupture concentrated deeper in the system, underneath thetransition from the forearc high to the forearc basin (Tilmannet al., 2010) and is comparable to most megathrust eventsworldwide. The analysis of the aftershocks activity of the 2004and 2005 earthquakes (Tilmann et al., 2010) supports the changesin the ability to support seismic strain release across the proposedsegment boundary. Pesicek et al. (2010) found that the after-shocks just offshore the Banyak island group (Fig. 1) show an arctype distribution following variations in the trench geometry.They report a southward steepening of the dip of the downgoingoceanic plate in the distribution of the deep (450 km) aftershocks.The separation between the northern area with prevailing shallowseismicity from the southern part, where the deeper earthquakesare observed (Fig. 1), is roughly collocated with the active seismicprofile described in this study.

The shallow and deep structure of the southern portion of the2004 rupture was investigated using active seismic methods(Dean et al., 2010; Dessa et al., 2009; Fisher et al., 2007; Frankeet al., 2008; Gulick et al., 2011; Klingelhoefer et al., 2010; Mosheret al., 2008; Singh et al., 2008). Reflection seismic studies haverevealed the presence of a pre-decollement reflector in the 2004rupture area (Gulick et al., 2011), which terminates in the vicinityof the 20042005 rupture segment boundary (Dean et al., 2010)and is absent in the 2005 segment. These findings indicatecontrasting plate boundary shear zones and prism properties inthe two segments (Dean et al., 2010). The rupture area of the 2004earthquake shows a thick sedimentary cover (4 km offshorenorthern Sumatra) on the subducting Indo-Australian plate(Gulick et al., 2011); deformation is mostly restricted to the upperslope apron (Fisher et al., 2007) within a region of dominantlylandward-vergent thrust faults (Mosher et al., 2008). These obser-vations from the 2004 rupture area, together with the observationof a steep frontal slope and a broad marginal plateau, suggest thepresence of a strong inner structure (Gulick et al., 2011). Possibledewatering and lithification of the sediments during the earliersubduction history making them unusually strong, was suggestedby Gulick et al. (2011).

The area of the 2005 rupture is seismically not as intenselystudied; however, high-resolution bathymetry mapping has beenconducted in both segments in a multinational effort (e.g.Graindorge et al., 2008; Henstock et al., 2006; Kopp et al., 2008;Krabbenhoeft et al., 2010; Ladage et al., 2006). These studiesdocument a change in the frontal slope towards the 2005 rupturearea offshore Nias and a change in the morphology to a rougherappearance, suggesting lateral variations in material strength andmechanical properties. This lateral heterogeneity could be attrib-uted to changes in the sediment supply, related to the subductionof a basement ridge tracing the 961E fracture zone (Gulick et al.,2011; Klingelhoefer et al., 2010). Magnetic observations haverevealed variations in the age of the incoming oceanic plate by2 myr from the older 2004 segment to the younger 2005segment (Franke et al., 2008). Based on refraction seismic data,the upper plate forearc Moho has been reported at a depth of20 km in the northern domain (Dessa et al., 2009; Klingelhoeferet al., 2010) with possible deepening in the south to a depth of ca.2324 km (Simoes et al., 2004).

In this work we present the deep structural image of theSumatra margin south of Simeulue Island in the 2005 rupturearea (Fig. 1). We compare our results to seismic imagesacquired near the southern termination of the 2004 rupturearea north of Simeulue Island (Klingelhoefer et al., 2010) toquantify the crustal-scale changes across the proposed seg-ment boundary and possible effects on the variations in theseismogenic behavior.2. Data

The dataset used in this study consists of a multi-channelseismic (MCS) line and collocated wide-angle refraction/reflectionprofile (Fig. 1). The seismic data were collected within theSeaCause project during SO186 cruise of R/V Sonne offshoreNorthern Sumatra in 2006. The reflection profile data (MCS lineBGR06-135) were acquired to reveal and constrain the structureof the sedimentary fill as well as the fault deformation patterns.MCS processing included pre-stack processing and pre-stackdepth migration. A time-and-space-variant frequency filter wasapplied prior to a spherical divergence correction and a predictivedeconvolution. Normal moveout correction and velocity analysiswas followed by multiple suppression and common-depth-pointstacking. For the pre-stack depth migration, a scheme combiningcommon reflection point gathers and focusing analyses is used inassessing the seismic velocity field (Mackay and Abma, 1993).The pre-stack depth migrated data are shown in Fig. 2. In thisstudy it is further used in the interpretation of the shallowstructures, due to much higher resolution compared to wide-angle data, and to constrain the starting model for the seismictomography.

In total 24 ocean bottom seismometers (OBS) were deployedalong a 255 km long profile with an average spacing of 10 km,providing uniform coverage. The processing of the ocean bottomdata included the localization of the ocean bottom instrumentsusing the arrival information of the direct water wave and preciseshot geometry. A time-gated deconvolution was applied toremove the predictable air bubble reverberations in order toimprove the signal-to-noise ratio. Finally, a time and offset-variant Butterworth filter, in which the pass-band moves from0.87.56586 Hz at near offsets towards the lower frequencies(0.84.52030 Hz) as record time and offset increases, wasapplied to consider frequency changes caused by signal attenua-tion. Passband changes are discrete in four windows, which arelinearly ramped over 1 s travel time. Clear seismic phase arrivalswere recorded on all stations within 50 km offsets; for somestations clear travel times are available up to 100 km offset.Figs. 3 and 4 and Supplementary Figs. S1 and S2 provide examplesof the data quality and the recorded phases as well as pickedtravel times (reflection and refraction phases). Stations located onthe oceanic plate recorded clear arrivals of the oceanic sedimen-tary phases (Psed1/2) and crustal phases (Pg) (Fig. 3).The reflections from the Moho (PmP) and from the top of theoceanic crust (PtopP) were also recorded, permitting us to con-strain the crustal structure of the incoming oceanic plate in greatdetail (Fig. 3 and Supplementary Fig. S1). The stations located onthe forearc high and in the forearc basin show records of severalsedimentary phases (Psed1/2) as well as crustal phases (Pg) (Fig. 4

Fig. 2. Pre-stack depth-migrated seismic section (BGR06-135) collocated with the wide-angle profile. Top panel shows the entire section with the line-drawing interpretation overlain. Bottom panels are the zoom-ins on theselected areas of the profile: (A) oceanic plate, (B) forearc basin, (C) deep section of the contact with oceanic crust and (D) slope break area around forearc high.

A.Sh

ulgin

etal./Earth

andPlaneta

ryScien

ceLetters

365(2013)108119

111

0

10

20

30

40

0 20 40 60 80 100 120 140 160 180 200 220 240

2

3

4

5

6

7

8

Tim

e - D

ist/6

[sec

]

90 100

2

3

4

5

6

7

8

SW NE

Distance, [km]

Dep

th, [

km]

2

3

4

5

6

7

8

Tim

e - D

ist/6

[sec

]

-6 -50 -4 -30 -20 0 0 -10 0 10 20 30 40 50 60 70 80

-6 -50 -4 -30 -20 0 0 -10 0 10 20 30 40 50 60 70 80

100Offset [km]

2

3

4

5

6

7

8

PwPw

Pg

Pg

PgPmP

PmP

PtopPPtopP

Psed

Pn

Fig. 3. Data example for ocean bottom station located on the outer accretionary prism and the corresponding seismic rays obtained with forward modeling approach,location is marked in Fig. 5. Top plot shows the schematic final model with the OBS marked by red triangles, seismic rays are shown by color code: redrefractions,

bluePmP mantle reflections, greenPtopP reflection from the top of the oceanic crust, dotted box shows the lateral extent of the data sections. Middle panel shows the

data sections plotted with a reduction velocity of 6 km/s. Bottom panel shows the picked travel times (gray lines, thickness corresponds to the pick uncertainty), color lines

are the forward predicted travel times for selected phases computed through the final tomography model. The phases: Pwdirect water wave, Pgcrustal refraction,

Pnmantle refraction, Psed1/2refracted waves in the sedimentary layers, PmPMoho reflection, PtopPreflection from the top of the oceanic crust, PfbPreflection

from the top of forearc basement, PmfcPreflection from forearc upper/lower crust boundary. (For interpretation of the references to color in this figure legend, the reader

is referred to the web version of this article.)

A. Shulgin et al. / Earth and Planetary Science Letters 365 (2013) 108119112and Supplementary Fig. S2). A number of reflected phases wereidentified and picked from these stations, including reflectionsfrom the oceanic Moho (PmP), top of the oceanic crust (PtopP),and the forearc basement reflections (PfbP) (Fig. 4). Pn mantlephases were recorded on five stations (Supplementary Fig. S2),providing the opportunity to define the velocity in the uppermost

0

10

20

30

40

0 20 40 60 80 100 120 140 160 180 200 220 240

0

1

2

3

4

5

6

7

Tim

e - D

ist/6

[sec

]

-90 -80 -70 -60 -50 -40 -30 -20 -10 0 10 20 30 40

0

1

2

3

4

5

6

7

0

1

2

3

4

5

6

7

Tim

e - D

ist/6

[sec

]

-90 -80 -70 -60 -50 -40 -30 -20 -10 0 10 20 40Offset [km]

0

1

2

3

4

5

6

7

PwPw

Pg

PgPg

Pn

PmP

PtopP

Psed2

Psed2

Psed1Psed1

PmfcP

Distance, [km]

Fig. 4. Data example for ocean bottom station above the forearc basement high and the corresponding seismic rays obtained with forward modeling approach, location ismarked in Fig. 5. For detailed description please see Fig. 3 caption.

A. Shulgin et al. / Earth and Planetary Science Letters 365 (2013) 108119 113mantle. The final wide-angle dataset included 15,000 first arrivaltravel times and 4000 reflection travel times. Pick errors assignedto picked travel times range from 20 ms for the clear short-offsetsto a maximum of 100 ms for the weak large-offset reflections.

0

10

20

30

400 20 40 60 80 100 120 140 160 180 200 220 240

0 20 40 60 80 100 120 140 160 180 200 220 2400

10

20

30

400 20 40 60 80 100 120 140 160 180 200 220 240

0 20 40 60 80 100 120 140 160 180 200 220 240

0

10

20

30

400 20 40 60 80 100 120 140 160 180 200 220 240

0

10

20

30

40

0

10

20

30

40

0

10

20

30

40

0

10

20

30

40

0

10

20

30

400 20 40 60 80 100 120 140 160 180 200 220 240

1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5

500 1000 1500 2000 2500 3000

Profile Distance [km]

Dep

th [k

m]

Dep

th [k

m]

Vp, km/s

Ray path density

VE=2:1

VE=2:1

SW NE

3 4

4

55

67

7

46

Profile Distance [km]

LVZ: remnant of input model

Fig. 5. Tomographic inversion offshore northern Sumatra. Top panel: Vp distribution along the profile, with the black lines showing the position of the reflectors (dashedlinesreflectors which are poorly constrained). Bottom panel shows the final ray path coverage during the tomographic inversion. Yellow triangles on middle panel show

the location of the ocean bottom stations described in Figs. 3 and 4 and Supplementary Figs. 1 and 2. The low-velocity layer on top of the oceanic crust is a remnant of the

input model and not required by the data. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

A. Shulgin et al. / Earth and Planetary Science Letters 365 (2013) 1081191143. Modeling

3.1. Seismic tomography

The crustal structure was modeled by the joint refraction/reflection 2D tomography code Tomo2D (Korenaga et al.,2000). We applied a top-to-bottom approach with a simplelayered starting model. We chose a grid spacing of 250 m in thehorizontal direction. Vertical spacing increased with depth from50 m at the seafloor to 330 m at 40 km depth. The correlationlength parameters were defined as 1 km0.2 km (horizon-tal vertical) at the top and linearly increasing to 8 km4 kmat 40 km depth. Initially, the model was constrained only for thenear offsets; then the depth extent of the rays was increased toinclude the deeper sections of the model space. Simultaneously,the positions of the reflectors were obtained, based on theinversion of the available reflection phases. The structure andthe geometry of the sediments for the starting model wereadopted from the collocated pre-stack depth migrated MCSprofile, thus constraining the upper section of the model indetail. In addition, the position of the top of the oceanic crust inthe tomographic inversion was controlled by the MCS depthmigrated image (Fig. 2) (consistency of the two datasets isdiscussed in Section 3.2 and shown in Supplementary Fig. S3d).The generalized procedure during the inversion was as follows:constraining the interface geometry based on the correspondingMCS depth section; inverting the near offset refracted phases toverify the consistency between datasets; fixing the upper sectionof the model with a weight of 1000 compared to thedeeper sections. Then, we ran the inversion for the velocityand reflector geometry for the next depth layer. Subsequently,we fixed the obtained layer structure and repeated the previousstep for the following deeper layer. Finally, we kept the entirecrustal structure fixed and inverted for the velocity in the uppermantle. For the final model we have obtained RMS of 52 ms witha w22.13 during the inversion. The final Vp model along theprofile is shown in Fig. 5.

3.2. Resolution tests

Several tests were performed to analyze the resolution of theobtained intermediate and final models. Initially, we used aforward ray shooting method to compute the travel times forthe first arrival and reflected phases through our final models. Thecomparison of the seismic sections and the calculated travel timesfor several stations is shown in Figs. 3 and 4 and SupplementaryFigs. S1 and S2. The calculated travel times, as predicted by thetomographic Vp velocity model, are in a good agreement with therecorded seismic sections. All the refracted phases, including Pg,Pn, and sedimentary phases Psed fit the observed arrivals. Thereflection phases (PmP, PtopP, PfbP, and PmfcP, see above, orFig. 3 for annotation) also match the recorded data.

Checkerboard resolution tests provide information on thespatial and amplitude resolution, which is dependent on the givenray geometry and the velocity distribution. Small perturbations(75%) with a checkerboard pattern are added to the final Vpmodel. Using the same source-receiver geometry as in the tomo-graphic inversion, synthetic travel times are computed through theperturbed medium. Next, in order to recover the initial perturba-tion pattern, the tomography is recomputed based on the synthetictravel times, with small (7 pick uncertainty) random noise added.The results are presented in Supplementary Fig. S3. Three differentanomaly sizes were used in the tests: 52.5 km, 105 km and2010 km (horizontal vertical size). For all of them perturbationvalues of 75% are used. For the small size perturbation we are ableto recover the amplitude and size of the anomaly for the upper5 km below sea floor (bsf), which roughly corresponds to thethickness of the sedimentary cover except across the consolidated

0

10

20

30

40

0 20 40 60 80 100 120 140 160 180 200 220 240

0

10

20

30

40

0

10

20

30

40

0

10

20

30

40

Dep

th, k

m

Dep

th, k

m

Distance, km

Oceanic crust

Upper mantle

Sumatra

block

Accretionary prismFrontal prism

2 3 4 5 6 7 8

Vp, km/s

Accretionary prism

Upper mantle

Oceanic crustSumatra

block

12.26.2004Earthquake

2 3 4 5 6 7 8

Vp, km/s

0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300

Fig. 7. Comparison of the crustal structure along two adjacent segments offshore Sumatra. Identical scale and color-code are used for both models. Top panel: SAGERprofile north of the segment boundary in the 2004 rupture area. Background model and line-drawing modified after Klingelhoefer et al. (2010) (location is shown by blue

line in Fig. 1). Bottom panel represents the model obtained in this study: profile located south of the segment boundary in the 2005 rupture area. Black lines are the line-

drawing from the MCS profile, the dashed lines are the reflectors constrained by the tomography and gravity modeling. Circles are the projection of the seismicity onto the

line of the profile within a 710 km corridor (seismicity of Tilmann et al. (2010) relocated in a pseudo 3D model based on the refraction model, see text). Blue circles arewell-constrained hypocenters. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

20

30

0

10

40

0

10

20

30

40

20 40 60 80 100 120 140 160 180 200 220 240

0 20 40 60 80 100 160 180 200 220 240

100

0

100

0

120 140

1.5 2.0 2.5 3.0 3.5

Distance, [km]

Gra

vity

, [m

Gal

]D

epth

, [km

]

3.35

3.35

3.35

3.30

2.4

2.42.4

2.3

2.65

2.65

2.652.85 2.85

2.9

2.9

2.9

2.9

Density, [g/cm]

Fig. 6. Forward gravity model based on tomographic image. The final preferred velocity model is converted to density and the gravity response is computed. The top panelshows the ship gravity data (black), and the model response (solid red). The bottom panel shows the density model, values are in g/cm3. A pure 2D approach is used in the

gravity modeling. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

A. Shulgin et al. / Earth and Planetary Science Letters 365 (2013) 108119 115accretionary prism. For the forearc basin, along the north-easternportion of the profile, it is possible to recover the perturbations upto 8 km bsf, corresponding to the upper crust of the Sumatra block.Despite the presence of the high resolution MCS data, this test isessential for confirming that the given ray geometry is adequatelyresolving the shallow structures and does not produce extra errorsin the deeper sections of the model. For the middle size anomalythe major part of accretionary prism (except the deepest part,where smearing occurs) and Sumatra basement is recovered, butnot the downgoing oceanic crust. Finally, the entire model spacewas only recovered by the largest size perturbation. In addition tothe checkerboard test, a comparison of the refraction tomographywith the MCS-derived velocity field was performed (SupplementaryFig. S3d). For the shallow portion of the profile where the MCS dataprovide higher resolution and precision, the differences between thetwo methods yield comparable Vp distributions. For the major partof the profile the difference between the models are on average5070 m/s. The maximum variations are observed at the struc-tural interfaces, as the tomography tends to smooth the Vp field andthe resolution is incomparably weaker. The maximum variations of

A. Shulgin et al. / Earth and Planetary Science Letters 365 (2013) 108119116150 m/s account for the limited quantity of the wide-angle recordsections, the defined grid spacing of the tomography model, and thepicking precision of the travel times.

3.3. Gravity modeling

Forward 2D gravity modeling was performed based on thefinal preferred velocity model, to verify that the structural modeldoes not contradict gravity observations. The seismic velocities ofthe tomographic model were converted to densities using experi-mental relationships (Carlson and Herrick, 1990; Christensen andMooney, 1995). The model space was extended laterally by100 km and down to a depth of 70 km, where the deep seismicity(global seismicity catalog used) marks the position of the down-going oceanic plate. The mantle was modeled with a constantdensity of 3.35 g/cm3 with the exception of the mantle wedgewhere a density of 3.30 g/cm3 provided the best fit (Fig. 6).The study area covers a wide range in bathymetry (600070 mwater depth) as well as being located across a chain of forearcislands, so strong 3D effects are to be expected. However, as theaim of the gravity modeling is to confirm the tomography resultsand not to create a stand-alone model, pure 2D modeling is used.Despite these simplifications, we obtain a reasonable fit to thedata, where the misfit does not exceed 2025 mGal at the edgesof the profile (Fig. 6). Within the given above simplifications, thekey observation is that the forearc Moho should be located at20 km depth at the plate interface, and dip towards Sumatra toat least 25 km on the edge of the profile. A shallower Mohogeometry would require unrealistically low mantle wedgedensities.

3.4. Local seismicity relocation

To relocate the aftershocks of the 2005 earthquake a pseudo3D model was constructed based on the refraction model byextruding the model in the profile-perpendicular direction butfollowing the curved geometry of the trench; additionally theactual topography is respected in the construction of the 3Dmodel. The seismicity catalog of Tilmann et al. (2010) is relocatedwithin this model using a nonlinear search method for themaximum likelihood location (NonLinLoc, Lomax et al., 2000)using P waves only, or assuming Vp/Vs values between 1.70 and1.82. For well-located events (colored circles in Figs. 1 and 7), thedifferences in depth and horizontal location between the P-onlyrelocations and the relocations with various Vp/Vs ratios wereminor and in general smaller than the difference between the 3Drelocation and the original location in the minimum 1D model.The locations shown in the figures assume a Vp/Vs ratio of 1.74,which gave close to the minimum rms.4. Results and discussion

Our interpretation is based on the results of the tomographicinversion and the pre-stack depth migrated reflection imageshown in Figs. 2 and 5, which will be discussed starting fromthe south-west towards Sumatra in the north-east. In thesouth-west portion of the profile, 8 km thick oceanic crust over-lain by a sedimentary cover enters the trench. The interpretedthickness of the oceanic crust may be dominated by the crustalstructure of the fracture zone. However, as we observe a similarthickness of the downdoing crust along the profile, we speculatethat the increased thickness of the oceanic crust occupies abroader area to the south, contrary to the narrow localizedridge-related feature. Sediment thickness ranges from ca. 1 kmon the outer rise and increases to up to 3.5 km at the deformationfront (offset 50 km in Fig. 2). The observed sedimentary coverthickness is in good agreement within both our datasets, and alsofits the observed values on MCS lines BGR06-118 and BGR06-119located offshore Simeulue (Franke et al., 2008) (Fig. 1). The MCSdata show horizontally well-stratified sediments cut by normalfaults (Fig. 2a). The oceanic Moho is dipping from 13 km depth atthe southwestern termination of the profile to 15 km depth belowthe deformation front (Fig. 5). The Vp distribution within theoceanic crust corresponds to a velocity profile characteristic formature oceanic crust (White et al., 1992): increasing from 4.7 km/s at the basement to 5.96.0 km/s at 2 km depth below thebasement, corresponding to the upperlower crust transition,and finally increasing to 6.97.0 km/s at the Moho, although itis 11.5 km thicker than typical. The structure of the oceanic crustdoes not vary significantly along the profile.

The outer prism extends from the deformation front (offset51 km) to 85 km offset and is characterized by thrust faultingresulting from active accretion of trench sediment (Fig. 2). Theouter prism is bounded by the inner accretionary prism locatedlandwards (Fig. 2). The boundary between the outer prism andthe inner prism is positioned based on the area where theinterpreted imbricate thrusts disappear. The outer prism showsVp velocities ranging from 2 km/s below the seafloor to 4.2 km/sat the top of the oceanic plate with a general increase of the Vpaway from the trench, due to compaction. The inner prism ischaracterized by higher Vp values with less lateral variability,varying from 2 km/s at the top to 5.4 km/s, suggesting highlycompacted, lithified and possibly metamorphosed material nearthe plate interface.

Bathymetry data shows much smoother morphology on theinner accretionary prism compared to the lower slope of the outerprism and additionally constrains the surface boundary betweenthe outer prism and tectonically less active inner prism (Koppet al., 2008). The spatial correlation of the updip limit of theseismogenic zone with the changes in the morphology of amargin was shown by the dynamic Coulomb model of Wangand Hu (2006). Offshore Sumatra, based on morphology observa-tions, the updip limit is located at around 110 km offset at a depthof 14 km (Fig. 7), which correlates with the location of the slopebreak, also suggested by the bathymetry analysis of Krabbenhoeftet al. (2010). The location of the updip limit of the seismogeniczone is additionally supported by the relocation of seismicity inthe area (Fig. 7).

A well-developed forearc basin underlain by Sumatra base-ment dominates the morphology northeast of offset 160170 km(Fig. 2b). The sedimentary basin fill reaches a thickness of 3.5 kmabove the depocenter. The deep pre-Neogene sedimentary strata(Berglar et al., 2010) are tilted trenchward. An unconformityseparates the pre-Neogene strata from more recent deposits,suggesting a complex episodic basin formation (Berglar et al.,2010). The geometry of the basin is controlled by the Sumatrabasement and bounded by the accretionary prism at around175 km offset. This boundary also correlates with the locationbetween the surface traces of the West Andaman and Mentawaifault zones, resolved in the bathymetry and MCS datasets (Berglaret al., 2010). Interpreted seismic Vp velocities are increasingwithin the sedimentary layers from 1.7 km/s near the seafloorto 3.5 km/s at the top of the Sumatra basement (Fig. 5).Vp velocities range from 5.0 km/s at the basement down to 6.06.1 km/s at a depth of 6 km below the basement, interpreted asthe transition to the lower crust of the Sumatra block. The lowercrust shows Vp velocities increasing from 6.2 km/s at the top to7.0 km/ at the plate interface (Fig. 5). The geometry and the depthof the forearc Moho are not recovered in our seismic dataset.However, based on the gravity modeling discussed above theMoho likely is found at a depth of ca. 2027 km (Fig. 6).

A. Shulgin et al. / Earth and Planetary Science Letters 365 (2013) 108119 117We compare the obtained crustal structure model south of thesegment boundary of the 20042005 rupture areas (Franke et al.,2008) to the SAGER profile published by Klingelhoefer et al.(2010), located to the north of the fracture zone (see Fig. 1 forlocations; see Fig. 7 for the profile comparison). The comparedprofiles were modeled using different techniques, based on differentdata acquisition setups. In addition, within our dataset only few Pnarrivals are available, which limits seismic sampling of the uppermantle on our profile compared to the SAGER data. This differencecauses difficulties in one-to-one comparison of the small features.

Principal changes are observed in the structure of the incom-ing Indo-Australian plate as well as in the sediment supply at thetrench. The lateral extent of the southern domain is unclear, andthe interpreted structures could be region-specific, as was sug-gested by Dean et al. (2010). In the north, 5 km thick oceanic crustis subducting, posing a large contrast to the 8 km thick crustmodeled in the southern domain. Furthermore, in the 2004rupture area the oceanic crust is overlain by ca. 33.5 km ofsediment, reaching 5 km at the trench. In the southern domain,the sedimentary cover is much thinner: 3 km thick in the trenchand less than 1 km on the incoming oceanic plate. Furthermore,the velocity structure of the sediments in the trench fill isdifferent, with lower values in the southern domain (Dean et al.,2010); also inferred from our tomography model. The observedsediment structure is consistent with the results of Dean et al.(2010), showing the shallowing of the oceanic basement from thenorth towards the subducted fracture zone. Farther south-east atthe latitude of the 2005 earthquake rupture zone Dean et al.(2010) suggested the oceanic basement to deepen further.However, due to limited number of interpreted MCS lines thisobservation is still uncertain. Our observations suggest that the961E fracture zone is not only separating oceanic crustal blocks ofdifferent thickness, but also acts as a barrier in the sedimenttransport along the trench, confirming previous results of Deanet al. (2010) and Gulick et al. (2011). In addition, we seedifferences in the size and structure of the accretionary prismand possibly of the Sumatra block. Along the southern profile inthe 2005 rupture area, the outer prism and accretionary wedgereach a total width of ca. 130 km, while the SAGER profile to thenorth in the 2004 domain documents a lateral extent of theaccretionary complex exceeding 170 km. The velocity structurewithin the accretionary complexes is comparable in the range ofvalues, but an outer or frontal prism in the SAGER model withinthe 2004 rupture area could not be marked out. The outer prismnorthwest of the segment boundary has well-developed land-ward-vergent fold ridges up to 80 km long, contrasting to2530 km long mixed-vergence folds to the southeast (Deanet al., 2010). The geometry of the accretionary complexes alsovaries between the profiles. The distinct differences in accretion-ary prism geometry documented by the two seismic profilesshown in Fig. 7 as well as variations in material properties asdiscussed by Dean et al. (2010) are suggested to be linked to thechanges in updip limit of seismicity as described by Tilmann et al.(2010). Furthermore, as discussed by Gulick et al. (2011),the presence of strong incoming sedimentary fill in the area ofthe 2004 rupture can explain the shift of the updip limit closer tothe trench and increased tsunamigenesis, enhanced by ruptureoccurring beneath more seaward parts of the margin for the 2004event and thus greater water displacement.

The SAGER profile north of the segment boundary shows amaximum thickness of the accretionary prism of 21 km, while onour profile located south of the segment boundary it does notexceed 17 km. The dip angle along the profiles is similar, althoughin the trench region and below the outer prism the southerndomain shows a steeper dip, also consistent with the observationsof Dean et al. (2010). However, the dip angle comparison iscomplicated by the profile geometry being not exactly perpendi-cular to the trench. The lateral increase of Vp values (around 200220 km profile distance) in the accretionary prism on the north-ern SAGER profile towards the Sumatra block is also not observedon our profile located south of the fracture zone, where theseismic velocities slightly decrease when approaching the Suma-tra basement. This observation could be linked to the presenceand activity of the West Andaman and/or Mentawai fault zones.In the northern domain, the West-Andaman fault is locatedaround 260270 km (SAGER profile Fig. 7), showing only minorVp decrease around it. Possible extra fluid inflow into theaccretionary complex along the fault zone in the southern seg-ment could be one of the reasons for a decrease of seismicvelocities, assuming the fault zone is more active in the south.The structure of the Sumatra crust shows a possible change acrossthe segment boundary: the northern segment is characterized bya relatively thin crust with a flat Moho located at around 20 kmdepth, while the southern segment shows a much thicker crustwith the Moho dipping towards Sumatra (in the southern seg-ment the forearc Moho is not resolved in seismic data, and thethickness constraints are based on gravity modeling discussedabove). The increase in crustal thickness southwards from Simeu-lue might suggest that the location of the segment boundary isdetermined not only by the features present on the oceanic plate,but might also spatially correlate to the changes in the crustalthickness of the overriding plate. This observation might alsoexplain the increase in the dip of the oceanic slab from north tosouth, which is observed in the deep seismicity (Pesicek et al.,2010) and which is also visible on the seismic profiles.

The downdip limit of 2004 seismogenic rupture is reported tobe at 48 km depth and hence located deeper than the forearcMoho (2223 km) (Hsu et al., 2006), which is commonly assumedto define the downdip limit of a seismogenic zone. Theseobservations suggest that the 2004 rupture extended along thecontact zone of the underthrusting oceanic plate and the forearcmantle, implying that the forearc mantle wedge is not serpenti-nized. For the area of 2005 earthquake we do not recover theseismic velocity structure of the mantle wedge, however based onthe gravity modeling (see details above) we speculate that themantle wedge might be partially serpentinized.5. Conclusions

The wide-angle deep sampling tomographic model in the 2005Sumatra earthquake rupture area, additionally constrained bygravity and MCS data reveals the crustal scale structure of theSunda margin in the 2005 earthquake rupture area. The majorresults obtained are: the oceanic plate carries a thin sedimentarycover (o1 km), except for the trench fill. The thickness of theoceanic crust is about 8 km and uniform along the profile. Theactive outer accretionary prism of a width of ca. 35 km is boundedby the ca. 85 km wide inner prism. The relocated seismicity at thedecollement coincides with the suggested position of the updiplimit of the seismogenic zone ca. 65 km away from the trench.

The distinct differences in incoming oceanic plate structureand the accretionary complex structure observed in the tworefraction seismic profiles located north and south of SimeulueIsland support the existence of a segment boundary between the2004 and 2005 rupture areas. The segment boundary is attributedto the subduction of the 961E fracture zone. Furthermore, thefracture zone presents a natural efficient barrier to the trenchparallel sediment transport. Such an obstacle results in thedifference of the amount of sediment present on the oceanicplate (3.5 km vs. o1 km) and in the trench (5 km vs. 3 km), atleast around the fracture zone. This incoming sediment thickness

A. Shulgin et al. / Earth and Planetary Science Letters 365 (2013) 108119118affects the width of the accretionary complex which is muchnarrower in the 2005 rupture segment, as well as being 4 kmthinner, resulting in ca. 35% volume difference. The comparison ofsubduction zone structure in the 2005 Sumatra earthquakerupture area to the southern domain of the 2004 rupture areareveals a marked variation in forearc geometry reflected indifferences in the regional morphology as well as deep marginstructure. These tectonic changes are linked to the slip termina-tion of the 2004 and 2005 seismic events, however, the physicalprocesses related to the termination of slip propagation requirefurther studies to fully comprehend earthquake physics in futureruptures.Acknowledgments

We would like to thank the crew of R/V Sonne and theSeaCause Working group for their enormous help in collectingand processing of the data. We express great gratitude toJun Korenaga for the discussion of seismic tomography and theTomo2D code. We would like to thank reviewers T. Henstock andS. Gulick for their comments and suggestions on improving themanuscript and editor P. Shearer for his careful editorial guidance.The SeaCause project is funded by the German Federal Ministry ofEducation and Research (BMBF) under grant 03G0186B. A.S.wishes to thank the Petersen-Stiftung for their funding to com-plete this study.Appendix A. Supporting information

Supplementary data associated with this article can be foundin the online version at http://dx.doi.org/http://dx.doi.org/10.1016/j.epsl.2012.12.032.References

Aki, K., 1979. Characterization of barriers on an earthquake fault. J. Geophys. Res.84, 61406148.

Ammon, C.J., Ji, C., Thio, H.K., et al., 2005. Rupture process of the 2004 SumatraAndaman earthquake. Science 308, 11331139, http://dx.doi.org/10.1126/science.1112260.

Ando, M., 1975. Source mechanisms and tectonic significance of historical earth-quakes along the Nankai Trough, Japan. Tectonophysics 27, 119140.

Banerjee, P., Pollitz, F., Nagarajan, B., Burgmann, R., 2007. Coseismic slip distribu-tions of the 26 December 2004 SumatraAndaman and 28 March 2005Nias earthquakes from GPS static offsets. Bull. Seismol. Soc. Am. 97, S86S102,http://dx.doi.org/10.1785/0120050609.

Berglar, K., Gaedicke, C., Franke, D., Ladage, S., Klingelhoefer, F., Djajadihardja, Y.S.,2010. Structural evolution and strike-slip tectonics off north-western Sumatra.Tectonophysics 480, 119132.

Bilek, S., Schwartz, S., DeShon, H., 2003. Control of seafloor roughness on earth-quake rupture behavior. Geology 31 (5), 455458.

Bilek, S., 2010. The role of subduction erosion on seismicity. Geology 38 (5),479480, http://dx.doi.org/10.1130/focus052010.1.

Briggs, R.W., Sieh, K., Meltzner, A.J., et al., 2006. Deformation and slip along theSunda megathrust in the great 2005 NiasSimeulue earthquake. Science 311,18971901.

Carlson, R.L., Herrick, C.N., 1990. Densities and porosities in the oceanic crust andtheir variations with depth and age. J. Geophys. Res. 95, 91539170.

Chlieh, M., Avouac, J.-P., Hjorleifsdottir, V., et al., 2007. Coseismic slip and afterslipof the great Mw 9.15 SumatraAndaman earthquake of 2004. Bull. Seismol.Soc. Am. 97, S152S173, http://dx.doi.org/10.1785/0120050631.

Christensen, N.I., Mooney, W.D., 1995. Seismic velocity structure and compositionof the continental crust: a global view. J. Geophys. Res. 100, 97619788.

Collot, J.-Y., Marcaillou, B., Sage, F., Michaud, F., Agudelo, W., Charvis, P., Graindorge, D.,Gutscher, M.-A., Spence, G., 2004. Are rupture zone limits of great subductionearthquakes controlled by upper plate structures? Evidence from multichannelseismic reflection data acquired across the northern Ecuadorsouthwest Colombiamargin. J. Geophys. Res. 109, B11103, http://dx.doi.org/10.1029/2004JB003060.

Curray, J.R., 2005. Tectonics and history of the Andaman Sea region. J. Asian EarthSci. 25, 187232.

Dean, S.M., McNeill, L.C., Henstock, T.J., Bull, J.M., Gulick, S.P.S., Austin, J.A., Bangs,N.L.B., Djajadihardja, Y.S., Permana, H., 2010. Contrasting decollement andprism properties over the Sumatra 20042005 earthquake rupture boundary.Science 329 (5988), 207210.

Deplus, C., Diament, M., Hebert, H., Bertrand, G., Dominguez, S., Dubois, J., Malod,J., Patriat, P., Pontoise, B., Sibilla, J.-J., 1998. Direct evidence of activedeformation in the eastern Indian oceanic plate. Geology 26, 131134.

DeShon, H.R., Engdahl, E.R., Thurber, C.H., Brudzinski, M., 2005. Constraining theboundary between the Sunda and Andaman subduction systems: evidencefrom the 2002 M-w 7.3 Northern Sumatra earthquake and aftershock reloca-tions of the 2004 and 2005 great earthquakes. Geophys. Res. Lett. 32, 24, http://dx.doi.org/10.1029/2005GL024188.

Dessa, J.-X., Klingelhoefer, F., Graindorge, D., Andre, C., Permana, H., Gutscher, M.-A., Chauhan, A., Singh, S., 2009. Megathrust earthquakes can nucleate in theforearc mantle: evidence from the 2004 Sumatra event. Geology 37, 659662,http://dx.doi.org/10.1130/G25653A.1.

Diament, M., Harjono, H., Karta, K., Deplus, C., Dahrin, D., Zen Jr., M.T., Gerard, M.,Lassal, O., Martin, A., Malod, J., 1992. Mentawai fault zone off Sumatra: a newkey to the geodynamics of western Indonesia. Geology 20 (3), 259262.

Fisher, D., et al., 2007. Active deformation across the Sumatran forearc over theDecember 2004 Mw 9.2 rupture. Geology 35, 99102.

Franke, D., Schnabel, M., Ladage, S., Tappin, D.R., Neben, S., Djajadihardja, Y.,Mueller, C., Kopp, H., Gaedicke, C., 2008. The great SumatraAndamanearthquakesimaging the boundary between the ruptures of the great 2004and 2005 earthquakes. Earth Planet. Sci. Lett. 269, 12, http://dx.doi.org/10.1016/j.epsl.2008.01.047.

Gahalaut, V.K., Catherine, J.K., 2006. Rupture characteristics of 28 March 2005Sumatra earthquake from GPS measurements and its implications for tsunamigeneration. Earth Planet. Sci. Lett. 249, 3946.

Graindorge, D., Klingelhoefer, F., Sibuet, J.-C., et al., 2008. Impact of the lower plateon upper plate deformation at the NW Sumatran convergent margin fromseafloor morphology. Earth Planet. Sci. Lett. 275, 201210, http://dx.doi.org/10.1016/j.epsl.2008.04.053.

Gulick, S.P.S., Austin, J.A., McNeill, L.C., Bangs, N.L.B, Martin, K.M., Henstock, T.J.,Bull, J.M., Dean, S., Djajadihardja, Y.S., Permana, H., 2011. Updip rupture of the2004 Sumatra earthquake extended by thick indurated sediments. Nat. Geosci.4, 453456, http://dx.doi.org/10.1038/ngeo1176.

Henstock, T.J., McNeill, L.C., Tappin, D., 2006. Seafloor morphology of the Sumatransubduction zone: surface rupture during megathrust earthquakes? Geology34, 485488.

Hsu, Ya-Ju, Simons, M., Avouac, J.-P., et al., 2006. Frictional afterslip following the2005 NiasSimeulue earthquake, Sumatra. Science 312 (5782), 19211926,http://dx.doi.org/10.1126/science.1126960.

Ishii, M., Shearer, P.M., Houston, H., Vidale, J.E., 2005. Rupture extent, duration, andspeed of the 2004 SumatraAndaman earthquake imaged by the Hi-Net array.Nature 435, 933936.

Klingelhoefer, F., Gutscher, M.-A., Ladage, S., Dessa, J.-X., Graindorge, D., Franke, D.,Andre, C.H., Permana, T.Y., Chauhan, A., 2010. Limits of the seismogenic zone in theepicentral region of the 26 December 2004 great SumatraAndaman earthquake:results from seismic refraction and wide-angle reflection surveys and thermalmodeling. J. Geophys. Res. 115, B01304, http://dx.doi.org/10.1029/2009JB006569.

Kopp, H., Flueh, E.R., Klaeschen, D., Bialas, J., Reichert, C., 2001. Crustal structure ofthe central Sunda margin at the onset of oblique subduction. Geophys. J. Int.147, 449474.

Kopp, H., Weinrebe, W., Ladage, S., et al., 2008. Lower slope morphology of theSumatra trench system. Basin Res. 20, 519529.

Korenaga, J., Holbrook, S., Kent, G., Kelemen, P., Detrick, R.S., Larsen, H.-C., Hopper,J.R., Dahl-Jensen, T., 2000. Crustal structure of the southeast Greenland marginfrom joint refraction and reflection seismic tomography. J. Geophys. Res. 105,2159121614, http://dx.doi.org/10.1029/2000JB900188.

Krabbenhoeft, A., Weinrebe, R.W., Kopp, H., Flueh, E.R., Ladage, S., Papenberg, C.,Planert, L., Djajadihardja, Y., 2010. Bathymetry of the Indonesian Sundamargin-relating morphological features of the upper plate slopes to thelocation and extent of the seismogenic zone. Nat. Hazards Earth Syst. Sci. 10,18991911, http://dx.doi.org/10.5194/nhess-10-1899-2010.

Ladage, S., Gaedicke, C., Barckhausen, U., Heyde, I., Weinrebe, W., Flueh, E.R.,Krabbenhoeft, A., Kopp, H., Fajar, S., Djajadihardja, Y., 2006. Bathymetricsurvey images structure off Sumatra. EOS Trans. AGU 87 (17), 165.

Lay, T., et al., 2005. The great SumatraAndaman earthquake of 26 December2004. Science 308, 11271133.

Lelgemann, H., Gutscher, M.-A., Bialas, J., Flueh, E.R., Weinrebe, W., Reichert, C.,2000. Transtensional basins in the Western Sunda Strait. Geophys. Res. Lett.27, 21, http://dx.doi.org/10.1029/2000GL011635.

Lomax, A., Virieux A.J.,. Volant P., Berge C., Thurber C.H.. Rabinowitz N., (Eds.), 2000.Probabilistic earthquake 510 location in 3D and layered models: Introduction of aMetropolis-Gibbs method and comparison with linear locations. 511 Advances inSeismic Event Location, Kluwer Amsterdam, pp. 101134.

Mackay, S., Abma, R., 1993. Depth-focusing analysis using a wave-front-curvaturecriterion. Geophysics 58 (8), 11481156.

Malod, J.-A., Kemal, B.M., 1996. The Sumatra Margin: oblique subduction andlateral displacement of the accretionary prism. Geol. Soc. Spec. Publ. 106,1928, http://dx.doi.org/10.1144/GSL.SP.1996.106.01.03.

McCaffrey, R., 2009. The tectonic framework of the Sumatran subduction zone.Annu. Rev. Earth Planet. Sci. 37, 345366.

Meltzner, A.J., Sieh, K., Abrams, M., Agnew, D.C., Hudnut, K.W., Avouac, J.,Natawidjaja, D.H., 2006. Uplift and subsidence associated with the greatAcehAndaman earthquake of 2004. J. Geophys. Res. 111, B02407, http://dx.doi.org/10.1029/2005JB003891.

http://dx.doi.org/10.1016/j.epsl.2012.12.032http://dx.doi.org/10.1016/j.epsl.2012.12.032http://dx.doi.org/10.1016/j.epsl.2012.12.032http://dx.doi.org/10.1016/j.epsl.2012.12.032http://dx.doi.org/10.1126/science.1112260http://dx.doi.org/10.1126/science.1112260http://dx.doi.org/10.1126/science.1112260http://dx.doi.org/10.1126/science.1112260http://dx.doi.org/10.1126/science.1112260http://dx.doi.org/10.1126/science.1112260http://dx.doi.org/10.1126/science.1112260http://dx.doi.org/10.1130/focus052010.1http://dx.doi.org/10.1130/focus052010.1http://dx.doi.org/10.1130/focus052010.1http://dx.doi.org/10.1785/0120050631http://dx.doi.org/10.1785/0120050631http://dx.doi.org/10.1785/0120050631http://dx.doi.org/10.1029/2004JB003060http://dx.doi.org/10.1029/2004JB003060http://dx.doi.org/10.1029/2004JB003060http://dx.doi.org/10.1029/2005GL024188http://dx.doi.org/10.1029/2005GL024188http://dx.doi.org/10.1029/2005GL024188http://dx.doi.org/10.1029/2005GL024188http://dx.doi.org/10.1130/G25653A.1http://dx.doi.org/10.1130/G25653A.1http://dx.doi.org/10.1130/G25653A.1http://dx.doi.org/10.1016/j.epsl.2008.01.047http://dx.doi.org/10.1016/j.epsl.2008.01.047http://dx.doi.org/10.1016/j.epsl.2008.01.047http://dx.doi.org/10.1016/j.epsl.2008.01.047http://dx.doi.org/10.1016/j.epsl.2008.04.053http://dx.doi.org/10.1016/j.epsl.2008.04.053http://dx.doi.org/10.1016/j.epsl.2008.04.053http://dx.doi.org/10.1016/j.epsl.2008.04.053http://dx.doi.org/10.1038/NGEO1176http://dx.doi.org/10.1038/NGEO1176http://dx.doi.org/10.1038/NGEO1176http://dx.doi.org/10.1126/science.1126960http://dx.doi.org/10.1126/science.1126960http://dx.doi.org/10.1126/science.1126960http://dx.doi.org/10.1029/2009JB006569http://dx.doi.org/10.1029/2009JB006569http://dx.doi.org/10.1029/2009JB006569http://dx.doi.org/10.1029/2000JB900188http://dx.doi.org/10.1029/2000JB900188http://dx.doi.org/10.1029/2000JB900188http://dx.doi.org/10.5194/nhess-10-1899-2010http://dx.doi.org/10.5194/nhess-10-1899-2010http://dx.doi.org/10.5194/nhess-10-1899-2010http://dx.doi.org/10.1029/2000GL011635http://dx.doi.org/10.1029/2000GL011635http://dx.doi.org/10.1029/2000GL011635http://dx.doi.org/10.1144/GSL.SP.1996.106.01.03http://dx.doi.org/10.1144/GSL.SP.1996.106.01.03http://dx.doi.org/10.1144/GSL.SP.1996.106.01.03http://dx.doi.org/10.1029/2005JB003891http://dx.doi.org/10.1029/2005JB003891http://dx.doi.org/10.1029/2005JB003891http://dx.doi.org/10.1029/2005JB003891

A. Shulgin et al. / Earth and Planetary Science Letters 365 (2013) 108119 119Moore, G.F., Curray, J., 1980. Structure of the Sunda trench lower slope off Sumatrafrom multichannel seismic reflection data. Mar. Geophys. Res. 4, 319340.

Mosher, D.C., Austin Jr., J.A., Fisher, D., Gulick, S.P., 2008. Deformation of thenorthern Sumatra accretionary prism: evidence for strain partitioning fromhigh-resolution seismic reflection profiles and ROV observations. Mar. Geol.252, 8999.

Pesicek, J.D., Thurber, C.H., Widiyantoro, S., Zhang, H., DeShon, H.R., Engdahl, E.R.,2010. Sharpening the tomographic image of the subducting slab belowSumatra, the Andaman Islands and Burma. Geophys. J. Int. 182, 433453,http://dx.doi.org/10.1111/j.1365-246X.2010.04630.x.

Prawirodirdjo, L., Bock, Y., 2004. Instantaneous global plate motion model for 12years of continuous GPS observations. J. Geophys. Res. 109, B08405, http://dx.doi.org/10.1029/2003JB002944.

Rhie, J., Dreger, D., Burgmann, R., Romanowicz, B., 2007. Slip of the 2004 SumatraAndaman earthquake from joint inversion of long-period global seismicwaveforms and GPS static offsets. Bull. Seismol. Soc. Am. 97, S115S127.

Ryan, H.F., Scholl, D.W., 1993. Geologic implications of great interpolate earth-quakes along the Aleutian arc. J. Geophys. Res. 98, 2213522146.

Sclater, J.G., Fisher, R.L., 1974. Evolution of the east-central Indian Ocean, withemphasis on the tectonic setting of the Ninetyeast Ridge. Geol. Soc. Am. Bull.85, 683702.

Sibuet, J.-C., Rangin, C., Le Pichon, X., et al., 2007. 26th December 2004 greatSumatraAndaman earthquake: co-seismic and post-seismic motions innorthern Sumatra. Earth Planet. Sci. Lett. 263, 88103, http://dx.doi.org/10.1016/j.epsl.2007.09.005.

Sieh, K., Natawidjaja, D., 2000. Neotectonics of the Sumatran fault, Indonesia. J.Geophys. Res. 105 (B12), 2829528326, http://dx.doi.org/10.1029/2000JB900120.Simoes, M., Avouac, J.P., Cattin, R., Henry, P., 2004. The Sumatra subduction zone: acase for a locked fault zone extending into the mantle. J. Geophys. Res. 109,B10402, http://dx.doi.org/10.1029/2003JB002958.

Singh, S.C., Hananto, N.D., Chauhan, A.P.S., 2011. Enhanced reflectivity of back-thrusts in the recent great Sumatran earthquake rupture zones. Geophys. Res.Lett. 38, http://dx.doi.org/10.1029/2010GL046227.

Singh, S.C., Carton, H., Tapponnier, P., et al., 2008. Seismic evidence for brokenoceanic crust in the 2004 Sumatra earthquake epicentral region. Nat. Geosci. 1(11), 777781, http://dx.doi.org/10.1038/ngeo336.

Singh, S.C., Hananto, N.D., Chauhan, A.P.S., Permana, H., Denolle, M., Hendriyana,A., Natawidjaja, D., 2010. Evidence of active backthrusting at the NE Margin ofMentawai Islands, SW Sumatra. Geophys. J. Int. 180 (2), 703714, http://dx.doi.org/10.1111/j.1365-246X.2009.04458.x.

Spence, W., 1977. The Aleutian are: tectonic blocks, episodic subduction, straindiffusion, and magma generation. J. Geophys. Res. 82, 213230.

Subarya, C., Chlieh, M., Prawirodirdjo, L., et al., 2006. Plate-boundary deformationassociated with the great SumatraAndaman earthquake. Nature 440, 4651.

Tilmann, F.J., Craig, T.J., Grevemeyer, I., Suwargadi, B., Kopp, H., Flueh, E.R., 2010.The updip seismic/aseismic transition of the Sumatra megathrust illuminatedby aftershocks of the 2004 AcehAndaman and 2005 Nias events. Geophys. J.Int. 181, 12611274, http://dx.doi.org/10.1111/j.1365-246X.2010.04597.x.

Wang, K., Hu, Y., 2006. Accretionary prisms in subduction earthquake cycles: thetheory of dynamic Coulomb wedge. J. Geophys. Res. 111 (B6), http://dx.doi.org/10.1029/2005JB004094.

White, R.S., McKenzie, D., ONions, R., 1992. Oceanic crustal thickness from seismicmeasurements and rare earth element inversions. J. Geophys. Res. 97,19,68319,715, http://dx.doi.org/10.1029/92JB01749.

http://dx.doi.org/10.1111/j.1365-246X.2010.04630.xhttp://dx.doi.org/10.1111/j.1365-246X.2010.04630.xhttp://dx.doi.org/10.1111/j.1365-246X.2010.04630.xhttp://dx.doi.org/10.1029/2003JB002944http://dx.doi.org/10.1029/2003JB002944http://dx.doi.org/10.1029/2003JB002944http://dx.doi.org/10.1029/2003JB002944http://dx.doi.org/10.1016/j.epsl.2007.09.005http://dx.doi.org/10.1016/j.epsl.2007.09.005http://dx.doi.org/10.1016/j.epsl.2007.09.005http://dx.doi.org/10.1016/j.epsl.2007.09.005http://dx.doi.org/10.1029/2000JB900120http://dx.doi.org/10.1029/2000JB900120http://dx.doi.org/10.1029/2000JB900120http://dx.doi.org/10.1029/2003JB002958http://dx.doi.org/10.1029/2003JB002958http://dx.doi.org/10.1029/2003JB002958http://dx.doi.org/10.1029/2010GL046227http://dx.doi.org/10.1029/2010GL046227http://dx.doi.org/10.1029/2010GL046227http://dx.doi.org/10.1038/ngeo336http://dx.doi.org/10.1038/ngeo336http://dx.doi.org/10.1038/ngeo336http://dx.doi.org/10.1111/j.1365-246X.2009.04458.xhttp://dx.doi.org/10.1111/j.1365-246X.2009.04458.xhttp://dx.doi.org/10.1111/j.1365-246X.2009.04458.xhttp://dx.doi.org/10.1111/j.1365-246X.2009.04458.xhttp://dx.doi.org/10.1111/j.1365-246X.2010.04597.xhttp://dx.doi.org/10.1111/j.1365-246X.2010.04597.xhttp://dx.doi.org/10.1111/j.1365-246X.2010.04597.xhttp://dx.doi.org/10.1029/2005JB004094http://dx.doi.org/10.1029/2005JB004094http://dx.doi.org/10.1029/2005JB004094http://dx.doi.org/10.1029/2005JB004094http://dx.doi.org/10.1029/92JB01749http://dx.doi.org/10.1029/92JB01749http://dx.doi.org/10.1029/92JB01749

Subduction system variability across the segment boundary of the 2004/2005 Sumatra megathrust earthquakesIntroductionDataModelingSeismic tomographyResolution testsGravity modelingLocal seismicity relocation

Results and discussionConclusionsAcknowledgmentsSupporting informationReferences