Vous êtes sur la page 1sur 31


B12, PAGES 28,295-28,326, DECEMBER 10, 2000

Neotectonics of the Sumatran fault, Indonesia

Kerry Sieh
Seismological Laboratory, California Institute of Technology, Pasadena

Danny Natawidjaja
Seoteknologi, Lembaga Ilmu Pengetahuan Indonesia, Bandung, Indonesia

Abstract. The1900-km-long, trench-parallel Sumatran fault accommodates a significant

amount of thefight-lateral component of oblique convergence between theEurasian and

Indian/Australian plates from10N to7S.Ourdetailed map ofthefault, compiled from topographic maps and stereographic aerial photographs, shows that unlike many other great strike-slip faults, theSumatran fault ishighly segmented. Cross-strike width ofstep overs
between the19major subaerial segments is commonly many kilometers. Theinfluence of these step overs onhistorical seismic source dimensions suggests thatthedimensions of

future events willalso beinfluenced byfault geometry. Geomorphic offsets along thefault
range ashighas-20 krnandmayrepresent thetotaloffset across thefault.If thisis so,other

structures must have accommodated much of thedextral component of oblique convergence during thepast fewmillionyears.Ouranalysis of stretching of theforearc region, near the southern tipof Sumatra, constrains thecombined dextral slipontheSumarran andMentawai faults tobenomore than100kmin thepast fewmillion years. Theshape andlocation of the Sumatran faultand theactive volcanic arcarehighly correlated withtheshape andcharacter
oftheunderlying subducting oceanic lithosphere. Nonetheless, activevolcanic centers of the Sumatran volcanic archavenotinfluenced noticeably thegeometry of theactive Sumatran fault.Onthebasis of its geologic history andpattern of deformation, we dividetheSumatran

plate margin intonorthern, central andsouthern domains. We support previous proposals that thegeometry andcharacter of thesubducting Investigator fracture zone areaffecting the
shape andevolution of the Sumatran fault system withinthecentral domain.The southern

domain isthemost regular.TheSumatran faultthere comprises sixfight-stepping segments. This pattern indicates thattheoverall trend of thefaultdeviates 4clockwise fromtheslip vector between thetwoblocks it separates. Theregularity of thissection andits association with theportion of thesubduction zonethatgenerated thegiant (Mw 9) earthquake of 1833 suggest thata geometrically simple subducting slab results in bothsimple strike-slip faulting and unusually large subduction earthquakes.
1. Introduction

roughly coincidentwith the active Sumarranvolcanic arc (Figure 1). On its northeastern side is the southeast Asian !.1. Plate Tectonic Environment plate, separated from the Eurasianplate only by the slow TheSumatran faultbelongs to a class of trench-parallel slipping Red River fault of Vietnam and southern China strike-slip faultsystems thatworkin concert withsubduction[Allen et al., 1984]. On its southwestern side is the Sumatran zones to accommodate obliquelyconvergent plate motion "forearc sliverplate"[Jarrard,1986],a 300-km-wide stripof [Yeats etal., 1997, Chapter 8]. Other strike-slip faults that lithosphere betweenthe Sumarranfault and the Sumatran

occur in similar settings include the left-lateral Philippine deformation

front. At its northwestern terminus the Sumarran

fault (parallel tothe Luzon and Philippine trenches), Japan'sfault transformsinto the spreadingcentersof the Andaman right-lateral Median Tectonic Line(parallel to theNankai Sea [Curray et al., 1979]. At its southeastern end, in the
trough), andChile's Atacama fault(parallel to the South
American trench).
Sunda Strait, the fault curves southward toward the deformationfront [Diament et al., 1992].

For itsentire 1900-km length theSumarran fault traverses The basic kinematic role of the Sumatran fault is rather thehanging wall blockof the Sumatran subduction zone, simple:It accommodates a significant amountof the strikefor 'Now at Division of Geological and Planetary Sciences, California

slip component of the oblique convergence betweenthe Australian/Indian and Eurasianplates. The pole of rotation
the relative motion between the Australian/Indian and

Institute ofTechnology, Pasadena.

Copyright 2000 by the American Geophysical Union.

Paper number 2000JB900120. 0148-0227/00/2000JB 900120509.00

Eurasian plates is in east Africa, ~50 west of Sumatra [Prawirodirdjo et al., thisissue, ?rawirodirdjo,2000; Larson

et al., 1997]. Northern Sumatra is closer to thispolethanis southern Sumatra.Thustheorientation andmagnitude of the relative-motion vectorvary significantly alongthe Sumatran











Sea South






Figure 1. Regional tectonic setting of the Sumatran fault. The Sumatran fault (SF) is a trench-parallel, right-lateral strike-slipfault that traverses the hangingwall blockof the Sumarran subduction zonefrom the SundaStraitto the spreading centers of the Andaman Sea. It separates a forearcsliverplatefrom the southeast Asianplate. Triangles are activevolcanoes of the Sundaarc. Arrow is relativeplatemotionvectors determined from GPS. Topography andbathymetryare from Smithand Sandwell[1997]. WAF is theWest Andamanfault. MF is the Mentawai fault.

portionof the plate boundary(Figure 1). At 6S, 102Eit is 60 mm/yr, N17E [Prawirodirdjo et al., this issue]. At 2N, 95E,it is 52 mm/yr, N10E. Furthermore, because the shape of the plate boundaryis arcuate,the natureof relativeplate motionchanges markedlyalong its strike. At the longitude of central Java the strike of the subductionzone is nearly orthogonalto the direction of relative plate motion, so any component of strike-slip motion need not be large [McCaffrey,1991]. At the latitudes of Sumatra, however, the strike-slip component of relative plate motion must be significantbecausethe direction of relative plate motion is substantially obliqueto thestrikeof thesubduction zone.

Fitch[1972] suggested thattheright-lateral component of this oblique convergence is the cause for the right-lateral
Sumarran fault. McCaffrey [1991, 1992] added more

substance to this hypothesis with his discovery thatslip vectors of moderate earthquakes along theSumatran portion of thesubduction zone arenearly perpendicular tothe strike of theplate boundary. He noted thatif these vector directions are representative of long-term slip trajectories along the subduction interface, thensubduction itself is onlyslightly oblique andmost of thedextral component of plate motion
must be accommodated elsewhere.

The Sumatranfault is the most obviouscandidate for



accommodation of theremaining component ofdextral slip. The Mentawai fault, discovered offshore by Diament et al.

localitiesto determinewhether or not the actual slip rates

conform to current kinematic models. Such rates would also

serve as a long-term average for the interpretation of geodetic [1992], complicates this slightly. This major, submarine, System(GPS) networksthat trench-parallel fault lies between theSumatran fault and the data from Global Positioning trench and may also have accommodated a significant amount now spanthe fault [Genrichet al., this issue]and historical

of the dextral component ofplate motion.

The combination of anarcuate plate boundary anda distant

triangulation data[?rawirodirdjoet al., this issue].

pole of rotation suggests that the rate ofdextral slip along the 2. A Modern Map of the Fault

Sumatran faultincreases northwestward [Huchon and Le To maptheSumatran faultefficiently andreliably, we have Pichon, 1984;McCaffrey, 1991]. Observations nearthe upon its geomorphic expression. Geomorphic northwestern andsoutheastern termini of theSumatran fault reliedprimarily
is especially reliablefor mappinghigh slip rate support this contention. Curray etal. [1979] suggested that expression commonlydevelopand are therateof opening acrossthe spreading centers of the faults,wheretectoniclandforms Andaman Sea (Figure 1) hasaveraged about 37 mrn/yr for the maintained at rates that exceed local rates of erosion or burial

past 11 Myr. They proposed that most ofthis motion has been
carried to thesoutheast by the Sumatran fault. Reanalysis of these data yields thesame rate;totalopening in thepast3.2

[Yeats et al., 1997, Chapter 8]. Examples of geomorphologically basedregionalmaps of active faults include activefaultmapsof Japan, Turkey,China,Tibet,and Group for ActiveFaults,1980;Saroglu Myr is~1 I8 km(J.Curray, written communication, 1999). Mongolia[Research and Molnar, 1977] as well as most The slip rate inferred fortheSumatran faultnear itssouthern et al., 1992; Tapponnier terminus, however, appears to be far lower than37 mrn/yr. mapsof submarine activefaults. Admittedly, the geomorphic expression of activefaults Bellier et aI. [1999] calculate a rate of ~6 mm/yr near the southern endof thefault from an offsetchannel incised intoa withsliprates thatarelower than or nearly equal tolocalrates dated Pleistocene tuff. of erosion or burialis likely to be obscure. This is especially likelyif thefaults areshort, havesmallcumulative offset, or 1.2. Motivation of This Work have no component of vertical motion. Becauseof our reliance on geomorphic expression, our map of the Sumarran Despite itsranking asoneof Earth's great strike-slip faults, fault undoubtedly excludes manyshort,low-rateactivefault

its high level ofhistorical seismic activity and itsmajor role in

theactivetectonicsand seismichazard of Southeast Asia, the Sumatranfault has not been well characterized. What


attention thefaulthasreceived hasbeenpredominantly froma

2.1. Resources and Methods

features of theSumatran faulthave longbeen great distance, mostly at plate tectonic scales. Untilrecently, Thegrossest of small-scale topographic andgeologic the geometry of thefaultwasknown onlyto first-order (see, knownfrom analysis small-scale mapsof the fault,based forexample, thesmall-scale maps of Fitch[1972],BelIieret maps. More detailed analysis of satellite imagery, havebeenproduced more al. [1997]or McCaffrey[ 1991 ]. More detailedstudies have upon been limited to local studies, suchas Tija's [1977] and Katili recently[Bellier et al., 1997; BeIlier and Sebrier,!994;
Detourbetet al., 1993]. The unavailabilityof stereographic and Hehuwat's [1967] work on exemplary offsetdrainages. however, limited theresolution andthereliability of The Sumatranfault has generatedmany historical imagery,

small-scale maps. Specifically, the lackof stereoscopic earthquakes withmagnitudes M> 7, butbecause most of these these precluded the recognition of importantsmall happened morethana half a century ago,theyhavenotbeen coverage landforms, unlessthey were favorablyilluminated. welldocumented. Reid [1913] used geodeticmeasurements tectonic inactive faults lacking small, latePleistocene and frombefore and after the 1892 Sumatran earthquakeas Conversely, support for his concept of elastic rebound.Berlage[1934] Holocenetectoniclandformsmay have been mappedas based uponthepresence of olderandlarger tectonic described theeffects of the 1933earthquake in south Sumatra. active, Visser [1927] described the effects of the 1926 landforms. Ourmapping of theSumatran faultis based primarily upon Padangpanjang earthquake in west Sumatra, andUntung et al. of l:50,000-scale topographic maps andl:100,000[1985] and Natawidjaja et al. [1995]recently reported dextral inspection aerial photographs. Where these werenotavailable or offsets formed during the nearby 1943 Alahanpanjangscale were of unsuitable quality, we utilized l:250,000-scale earthquake. maps andradarimagery. Figure2 displays the Thepaucity of detailed maps of thefault,thescarcity of geologic of materials thatweused. data onhistorical largeearthquakes, andthe lackof reliable coverage Figure 3 displays representative stereographic pairs of the estimates of sliprates areunfortunate. Theyseriously hamper aerialphotographs. Thesephotos display the attempts to forecast the seismic productivity of thefaultand 1:100,000-scale

that are efforts to understand quantitatively its role in the oblique fault at about0.3S,where it offsetsstreamchannels deeply incised intoa thickpyroclastic flow deposit.After convergence of theSumatran plateboundary. these and otherstereopairs, we compiled our Ourfirsttaskin thisstudy, then,hasbeento construct a interpreting onto l:50,000-scale topographic maps (or modem map of theactive components of theSumatran fault. interpretations To be of use in seismic hazard assessmentand in l:250,000-scale topographic maps,where the larger-scale Where stereographic aerial understanding theneotectonic roleof thefault,thescale of the maps were unavailable). wereunavailable, we interpreted activefault map needed to belarge enough to clearly discriminate major photographs andsense of slipdirectly fromthe l:50,000-scale fault strands andthe discontinuities andchanges in strike geometry

between strands.

Our second task, whichwill bedescribed in a future paper,

topographic maps. Thesedata were then digitizedand attributed, usingthe

Information System (GIS) software, Arc/Info. willbeto determine the sliprateof the faultat several Geographic







400 Kilometers

The Sumatran



Coverages of 1:250,000 topographic maps

Centersof aerial photographs

Coverages of 1:50,000 topographic maps

Figure 2. Dataupon which ourmap compilation isbased. Most of ourmapping is based oninspection of 1:50,000scale topographic maps produced by BAKOSURTANAL & JANTOP, thenational mapping agencies forIndonesia, andl:100,000-scale aerial photographs. Other data sources include smaller-scale geologic andtopographic maps.

of the Sumatran Fault The resultingG1S database includesfault geometry,sense of 2.3. Major Segments fault slip, and photo centers. Plate 1, constructed from the Superimposed upon the broad sinusoidalgeometryof the database,depicts all of the salient featuresof the Sumarran Sumarran fault are more than a dozendiscontinuities, ranging plate boundarythat we mappedand compiled. in width from -5 to 12 km (Plate 1). Major local changes in strikealsooccur. Most of the discontinuities are rightsteps in 2.2. Geometry of the Fault the fault trace and thts representdilatational step overs. The overall shapeof the Sumarranfault acrossSumatrais However, a few contractional bends also occur. and bendsin the faultare sinusoidal (Figure 1). The northern half of the fault is gently Theoretically,thesediscontinuities concave to the southwest, whereas the southern half of the large enoughto influencethe seismicbehaviorof the fault fault is concave to the northeast. Over the 1650-kin [Harris et al., 1991;Harris and Day, 1993]. The relationship ruptures to these geometrical segment boundaries subaerially exposed lengthof the fault, the "amplitude" of the of historical sinusoidal trace is -55 km. will be the subject of a futurepaper(D. Natawidjaja and K. in preparation,2000). Ornamentingthe broad, sinusoidalshapeof the Sumatran Sieh, manuscript We haveusedthesesecond-order geometric irregularities fault are numeroussmaller irregularities. Though smaller,
these have dimensions of the order of tens of kilometers and

to dividethe Sumarran fault into 19 segments (Figure 4 and

are thereforetectonicallyand seismologically significant. Table 1). Each segment bearsthe nameof a majorriveror The greatest of theseis a featurethat we call the Equatorial bay alongthe segment. In so namingthe varioussegments, B ifurcation (Figure 1 and Plate 1). This forceps-shaped we haveabandoned theusual practice of retaining names that feature is present between the equator and about 1.8N have precedence in the scientific literature. The latitude. It is characterized by the bifurcationof the Sumatran nomenclatural morass inherited Iom numerous earlier studies fault toward the southeastinto two principal active strands. includesmany fault names derived from nearbycities,
The two strands are distinct from each other even at their

districts,basins, and rivers. These include Banda Aceh Anu,

LamTeuba Baro,Reuengeuet Blangkejeren, Kla-Alas, UluAer, Batang-Gadis, Kepahiang-Makakau, Ketahun, Muara branch of the bifurcation does not rejoin the easternbranch Labuh, andSemangko [e.g.,seeKatiliandHehuwat, 1967;
pointof bifurcation (about 1.8N).The greatest separation of
km, near 0.7N. these two branches is -35 The western

farther south; instead, it dies out geomorphically at about


Cameron et al., 1983;Durham,1940]. Sincemanyof these

Other large irregularities include geomorphically expressed fault tracesat about5.5N, 4N, and 3.5S. The Batee fault, a right-lateralfault that may have displacedthe island's western shelf-150 km since the Oligocene[Karig et al., 1980], divergessouthward from the Sumatranfault at about4.6N. A 75-km-longfold-and-thrust belt, exhibiting clear geomorphicevidenceof youthfulness
lies about 40 km west of the Sumarran fault at about 1.3N. All of these features are described in section 2.3.

overlap ourgeometric segment boundaries or include only subparallel parts of oursegments, we have abandoned them in favor ofa
moresystematic and precise nomenclature.

Forthe entire group ofactive fault segments, from Aceh in

thenorth to theSunda Strait in thesouth, we have chosen the name"Sumarran fault,"first usedby Katili and Hehun'at

[1967]. Thisname represents bestthe dimension of the structure. Earlier names for the faultare"Semangko" and "Ulu-Aer," suggested by VanBemmelen [1949]and Durham
[1940]; but these refer to local features. "Great Sumatran



segments along the Sumatranfault, we suggest the particular namesin Figure4. In sections 2.3.1-2.3.19, we describe each segment,

beginningin the south. Each descriptionfocuseson the geomorphicexpression of the segmentand its terminations. Discussion of importanthistoricalearthquakes is minimal because the association of earthquakes with segments will be the focusof a futurepaper. Likewise, we do not focuson the slip ratesof the varioussegments because this alsowill be the principaltopicof a futurepaper. Plate 1 displaysthe fault at a scale that is appropriate for the detailed discussion that follows. (This plate and it's database are alsoavailable as postscript and GIS (ArcView) files at www.scecdc.scec.org/geologic/sumatra). 2.3.1. Sunda segment (6.75S to 5.9S). Bathymetric mapsof the SundaStrait, betweenJava and Sumatra,reveal that the southernmost portion of the Sumarranfault is
associated with two prominentsouth striking fault scarpson the seafloor [Nishimura et al., 1986; Zen et al., 1991; Pramumijoyo and Sebrier, 1991]. These scarps form a submarine graben,rangingin depth to 1800 below sea level (Figure 5). The large vertical displacements of the seafloor and the orientation and locationof the faults suggest that their sense of slip is normalanddextral. Focal mechanisms from a local seismicnetwork [Harjono et al., 1991] and from the Harvard Centroid Moment Tensor (CMT) cataloguesupport this interpretation. They show normal-fault mechanisms on the westernsideof the graben. Furthermore, faultsappearon both sides of the graben in three seismic reflection profiles
[Lassal et al., 1989].


The grabenwidenssouthward, toward the subduction zone, but losesbathymetricexpression ---130 km from the trench, near where one would expect it to intersectthe floor of the Sumatranand Javanforearcbasins(Figure 5). A belt of fault scarps andfoldsof the innertrenchslopecontinues across the southwardprojectionof the graben, but the outer-arcridge and forearc basin that are prominent in the offshore of Sumatraand Java are absentin this region. Instead,these features appear to converge upon each other and to be replaced by a narrow, 150-km-long plateau across the projection of the graben. The lessening of sliver-plate width occasioned by the absence of the forearc basinand outer-arc ridge appears to be accommodated by a landward deflection Figure 3. An example of the approximately l:100,000-scale of the trenchaxis (Figures 1 and 5). Huchon and Le Pichon [1984] were the first to propose aerial photographs we usedto compile mostof our mapof the Sumatran fault. These two setsof stereopairs showchannel that the disappearance of the outer-arcridge and the forearc offsets of *-720 m. The channels cut a late Pleistocene basin acrossthe southernprojectionof the Sumarranfault pyroclastic flow depositat about 0.3S. The flat upland indicates stretching parallelto the Sumatran fault. They also

surfaces aretheunincised topof theflow. These offsets yield

aaverage slip rateof- 11 mm/yr.

speculated that the subtlebending of the trenchtowardthe

Sunda Strait indicates arc-normal thinning of the region between the trailingedgeof the Sumarran forearcsliver plate
and the crust offshore from Java. This would be consistent

with the northwestwardtranslationof the forearc sliver plate

trough system" wasfirst usedby Westerveld [1953]. Since

"gmat"is not used for other faults of similar dimension, we

along the Sumarranfault. We attempt to quantify this

stretching in section 3. 2.3.2. Semangko segment (5.9S to 5.25S). From beneaththe watersof Semangko Bay at about5.9Sto a 6km-wide dilatationalstep over that has producedthe Suoh

suggest that it not be used for the Sumarranfault. In keeping

with convention generally accepted in California, where "San Andreas faultsystem" refers to theSanAndreas anditsmany auxiliary faults, we use"Sumarran faultsystem" (SFS)for the Valley at about5.25S,the principaltraceof the Sumarran
Sumarran fault and other structures that are related to the accommodation of strike slip along the Sumatranplate margin.These would includethe Batee fault, the Toru foldand-thrust belt, and the Mentawai and the West Andaman

fault runs almost linearly along the southwestern side of Semangko Bay andthe Semangko Valley (Plate 1 and Figure 4). The prominent northeast facingescarpment alongthe 65km lengthof thissegment attests to a significant component

sideup. An earthquake on July 26, faults inthe forearc region (Figure 1). For discrete, individual of dip slip, southwest

28 302














t Lake

Volcanic crater

Volcanic edifice

major valleys

Figure4. Map of 20 geometrically defined segments o1: theSumatran faultsystem andtheirspatial relationships to
activevolcanoes, major graben,and lakes.

1908, may have involved rupture of all or most of this segment [Berlage, 1934]. 2.3.3. Kumering segment(5.3S to 4.35S). This 150km-long segmentruns between the dilatationalstep over at Suoh Valley to a contractional jog at 4.35S. Near the center of this segment, the waters of Lake Ranau occupy a late
Pleistocenecaldera and conceal about 9 km of the trace (Plate

The northwesternmost 15 km of the Kumering segment deviateswestwardfrom the trend of the rest of the segment and is part of a 10-km-widecontractional jog. This portion of the segmentdisplaysa significantcomponent of reverse slip, as evidencedby a high escarpment and a mountainous anticline north of the fault trace. Aerial photography
available to us did not reveal the continuation of the fault

of 4.35S, through therestof thecontractional i and Figure 4). The southern part of the Kumeringsegment tracenorthwest traverses the drainagesof the Werkuk and upper Semangko bend. rivers. A less active southeastward continuation of this High intensities indicate ruptureof manytensof kilometers segment duringtheMs 7.5 Liwa earthquake segment may form the northeastern flank of the Semangko of theKumering 24, 1933[Berlage, 1934]. Deadlyphreatic explosions Valley [Pramumijoyo and Sebrier, 1991],but we did nothave of June occurred 2 weeks aftertheearthquake withintheSuoh Valley adequate materials to determine its activitythere. [Stehn, 1934]. North of Lake Ranau, a 40-km-long reach of the fault A geomorphically lessprominent subparallel strand of the traverses the headwatersof the Kumering River. The trunk of the principal active streamof this large river doesnot crossthe fault; instead, its fault exists2.5 km to the southwest south of LakeRanau [Natawidjaja, 1994;Widiwijayanti two majortributaries flow toward one another across the trace trace et al., 1996].The devastating Mw 6.8 Liwa earthquake of of the fault and flow northeastward away from the fault from 1994 wasgenerated by this lessprominent trace. Themost their confluence.This relationship of largestreamchannels to damage andtheaftershock region coincided with a25the fault is commonalong much of the Sumarran fault; not sev6re trace. uncommonly,the headwatersof a principal stream are near km reachof this secondary 2.3.4. Manna segment (4.35S to 3.8S). This85-km the fault, and none of the larger channelsof the drainage
network cross the fault trace. In these cases, dextral offsets of

segment deviates only a kilometer or two from being

recfilinearbut has rather obscureterminationson bothends

the stream channels are either ambiguous or small.





u"', o


u,'% o

',,ID u"',








deformation front

x, Java Sea

outer-arc ridge axis

---T-forearc basin axis

thrust '


inferred anticline


bathymetry contour
(in meters) depreasion/basin
a, active volt. dnO




Figure 5. Sumatranfault and relatedstructures near the SundaStrait and bathymetric map of the portionof the SundaStrait and surrounding seafloor.The Sunda segment of the Sumarran fault formsan 1800-m-deep graben that widenssouthward, towardthe deformation front. Northwestward movement of the forearcsilverplatealongthe Sumatran fault appears to have caused thinningof the regionbetweenthe trenchandthe strait. Bathymetryis Digital ElevationModel ETOPO02 and bathymetric surveys[Smithand Sandwell,1997].

heardeyewitness accounts of minorcrackingalongthefault whenwe visitedin 1993, but we saw no convincing evidence of tectonic surficialruptures from 1979. 2.3.6. Ketaun segment (3.35Sto 2.75S). This 85-kmlong segment consists of a linear trace with several discontinuitiesand stepovers of about a kilometerin dimension (Plate 1 and Figure 4). The segment's southern end is at a 6- to 8-km-widedilatational stepoverontothe Musi segment. An inactiveor lessactivecontinuation of the 1922]. Ketaunsegment naay extend beneath thestratovolcanic edifice 2.3.5. Musi segment (3.65S to 3.25S). This 70-kin of Bukit Kaba. This possibility is suggested by thepresence segmentof 'the Sumatranfault comprisesseveral highly of a geomorphically subdued fault, southeast of the volcano discontinuous fault segments (Figure4 andPlate 1). Despite and ---25 km eastof the centralMusi segment. The northern good coveragewith l:100,000-scaleaerial photography, we end of the Ketaun segmentis within a 6-km-wide couldnotidentifycleargeomorphic traces alongmuchof this contractional step over. Within this contractional stepoxer

(Figure 4). The Manna segmentappearsdiscontinuous on Plate I becausethe trace is obscurelocally on the aerial photographs and topographic maps.The southern end of the segmentabuts the contractionalbend mentionedabove. The northernend of the segmentis obscurebeyondabout3.8S but appearsto be within a geometricallycomplex right (dilatational)stepin the fault. Exceptionally clear 2.4 + 0.2 km dextral offsets of two large streams(Air Kanan and Air Kiri) exist on the dissected westernflank of an extinct volcanos,.utheast of Pajarbulan (Plate 2). We encountered surprisinglywe!l-preserved small tectonic landforms beneath the jungle canopy during an excursion in 'thedrainages of thesetwo streams. A destructiveearthquakeoccurred in the vicinity of this segmenton June 12, 1893. The area of greatestdamage coincided with the centralpart of the Manna segment [Visser,

stratovolcano, Bukit Kaba.

Stream channels cut into the

youngest flows there are offset -700 m. We have used these channels to determine the slip rate of 11 mm/yr for theMusi segment (D. Natawidjaja and K. Sieh, manuscriptin preparation, 2000).

The destructive, Ms 6.6 Kepahiang earthquake occurred

alongthis segment at about3.6Son December15, 1979. We

the topography rises severalhundredmetersabovethe The longest continuous trace that we were able to map surrounding landscape. traverses the southwesternflank of the large, active Twomajor rivers cross theKetaun segment, theKetaun in



terminus of the segment [Natawidjaja et aI., the south and theSeblat in thenorth. TheSeblat river valley southeastern
runsnorthwestward from the volcanic terraneof March 15,1952 (M 6.2,U.S. Geological Survey (USGS)), longsegment

valley may beoffset -23 km. A moderate earthquake on


appears tobe offset dextrally -17km, and the Ketaun river


2.3.10. Sumanisegment (1.0Sto 0.5S). This 60-kin-

which occupies a structural graben,rather than a volcanic caldera (Figure 4 andPlate1). Two opposing arcuate normal faultsformtopographic scarps thatrise400 m above predominantly linear, 60-km-long segment with several short,oblique of the lake (Plate 3). Ancientuplandsurfaces, obscure sections along its northern few kilometers (Plate1 the surface flowing away from the lake, are clearly and Figure 4). It shares a contractional stepoverwiththe with drainages by the steep scarps bounding the lake basinandthus Ketaun segment on its southeastern end. Its northwesterntruncated

Lake Diatasto the southwestern flank of Lake Singkarak, produced high intensities along the Ketaun segment [Kraeff, 2.3.7. Dikit segment(2.75Sto 2.3S). This is a

termination is at oneof thelarger dilatational step overs along appearto have been faulted down below the watersof the the Sumarran fault. On thesouthwestern flankof this11-km- lake. Failureof the Sumani segment produced the second of two wide step over, theDikit segment disappears intotheedifice (Ms 7.4 [Pachecoand Sykes,1992]) on of the small stratovolcano Kunyit.Thisis oneof thefewclear large earthquakes associations of a di!atational stepoveranda volcano alongthe June9, 1943 [Natawidjajaet al., 1995]. Shakingintensities
indicatethat the northwestern end of the fault rupturewas Eyewitnessaccountsled Untung et al. ~500 mby thefault. Justnorthof the smallcaldera lake,at [1985] to concludethat right-lateraloffsets of up to 2 m nearthe townof Solok,but Natawidjaja et aI. [1995] about 2.65S, the main trace appears to form an enigmatic occurred of ~1 m. Analysisof geodetic data d0gleg. TheDikitRiver Valleyfollows thefault for-20 km. couldonly verify offsets Wearenot convinced that this represents a dextraloffset of supports a meter or so of dextralslip [Prawirodirdjoet al., 20kin, because the construction of two largevolcanic edifices this issue]. High intensities in the vicinity of Lakes Dibawah and hasundoubtedly obscuredolder drainageson the block northeast of the fault. Diatassuggest thatthe entiresoutheastern part of the segment 23.8. Siulak segment (2.25S to 1.7S). Clear alsoruptured, andperhaps eventhe northwestern part of the dilatational stepoversdemarcate the terminations of this 70- Suliti segment. The first of two largeearthquakes on August4, 1926 was kin-long segment (Figure4 and Plate 1). The 11-kmwide stepover at thesoutheastern endis thewidest dilatational step most severein the narrow zone along the Sumanisegment. on October1, 1822, was most severe over along theSumatran fault,but ouraerialphotography did Anotherearthquake, notreveal its structural details. The northern terminus of the between theMarapiandTalangvolcanoes (Wichman, ascited may well have Siulak segment is a 4-km-widestepoveron the western flank by Visser [1927]). Thus this earthquake rupture of the Sumani segment.Genrichet al. [this of the great active stratovolcano Kerinci. West dipping involved duringthe early to midnormal faults cut lavas of Melenggokvolcanothere, and issue]showthat strainaccumulation with 23 + 5 mngyr of dextralslip on this appear to transfer slip from the Siulak segmentto its 1990sis consistent segment. northwestern neighbor. Along the Siulak segment'ssoutheastern reach, Lake 2.3.11. Sianok segment (0.7S to about 0.1N). This Kerinci andthe alluviumof a broadvalley obscure the fault predominantly straight and continuous segment runs-90 km trace for-30 km. Two largeearthquakes havecaused severe from the northeast shore of Lake Singkarak,along the damage alongthe Siulaksegment of the Sumatran fault. On southwest flankof thegreatstratovolcano Marapito a 10-kmJune 3, 1909,mostof the regiontraversed by this segment wide right stepover at the equator (Plate 1 andFigure4). Its was devastated by an earthquake judgedto havea magnitude southern 18 km, on the flank of Lake Singkarak, is arcuate ofabout Ms 7.7 [Abe, 1981]. The zone of greatest damage and must have a significant component of normalfaulting during theM 7.0 earthquake of October 6, 1995,waswithin downtowardthe lake. Geomorphic expression of thefault is thebroadvalley northwestof Lake Kerinci (Indonesian particularlyinteresting along the Sianok segment because it newspaper Kompas,October7, 1995). traverses the flank of Marapi volcanoand the young,200-m2.3.9.Suliti segment (1.75S to 1.0S). This95-km-long thick pyroclastic flow depositof Maninjou volcano. Stream segment hasa comparatively straight faulttrace,which channels flowing off Marapi displayclear dextra!offsets that terminates on both the northwest and southeast at di!atational rangefrom -120 to 600 m. The trunk channel of the Sianok step overswithin volcanicedifices(Figure 4 and Plate 1). River is incisedinto the ManinjouTuff and displayoffsetsof The northwestern stepover,at Lake DiatasandTalang ~700 m (Figure3). We havebeenableto usetheseoffsetsto volcano, is 4 km wide. The details'Of'the central reachesof determine a dextralrate of slip of- 11 mm/yr (D. Natawidjaja the segment areobscure because thefaulttraverses thenarrow andK. Sieh,manuscript in preparation,2000). valley of the Suliti River headwaters for more than 50 km. The second of two large earthquakes on August4, 1926, How much of thiscourse of thefaultalong theSuliti River wasmostsevere alongthe southeastern portionof the Sianok valley represents a dexraloffsetis unknown because the segment.This is consistent with Visser's[1927] observation trunk stream does notcross thefault. Alongthesouthernmost of fault rupturebetweenBukittinggiand Singkarak. Genrich
Sumatran fault.

The small diamond-haped caldera ofDipatiampat isoffset beneaththe lake.

part of thissegment, tributaries of the Liki River are offset et aI. [this issue] show that strain accumulationacrossthis
several hundred meters.

segment in the early to mid-1990sis consistent with dextral

The firstof twolarge earthquakes of June 9, !943 (Ms7.1 slip of 23 + 3 mngyr.

[Pacheco and Sykes, 1992]), may have involved rupture ofthe

2.3.12. Sumpursegment (equatorto 0.3N). Dataalong

neighborare northern partof the Su!itisegment, judging fromserious this 30-km-long segmentand its northwestern

damage to Muaralabuh village,25 km northwest of scant.Our mapis based predominantly upon1:250,000-scale





geologic maps [Rock etal.,1983, Aspden etal.,!982] and the valleys of the Gadis and Angkola Rivers, between MalintangandLubukRayavolcanoes [Visser,1922]. poorly reproduced 1:50,000-scale topographic maps.
Bothtermini of the Sumpursegment are at large
2.3.15. Toru segment (1.2Nto 2.0N). Majorbends in
We definethe southern terminus to be at a regional bend of 15 at !.2N. The topographic high east dilatational step over. The northwestern step is associated thatthisis a contractional bend. with a high west facing escarpment andthe adjoining wide of thebendsuggests The northwestern termination of theTom segment occurs at a valley ofthe Sumpur-Rokan River. with a (0.3Nto 1.2N). This ! 15- 15 regionalbendin the fault, whichis coincident stepover.We canbe confident thatthisbend Ian-long segment is broadly concave toward the southwestkm dilatational because the segment to the northwest doesnot and forms mostof the northeastern leg of the Equatorial is dilatational across the fault andthe bend Bifurcation (Figure 4 andPlate1). Thesoutheastern 40 kmof displaynet verticaldeformation with theTamtungdepression. the Barumun segment formsthe boundary between a high coincides

dilatational steps. Thus, between theSianok andBarumun the fault trace delimit this segmentof the Sumatranfault

segments, thefaultexperiences a 35-km-wide, double-(Figure4 andPlate !).

west facing escarpment and the broaddepression of the

Northwest of Sibual-buali volcano, a 30-kin-wide caldera

of the caldera, southwest of the fault, must be concealed

of the fault is truncated by the fault. The otherhalf Sumpur River. We interpret thisescarpment andadjacent northeast

youngvolcanic deposits. The geomorphic expression extensional dip slip on thisportion of the Barumun segment. beneath
caldera is unusually We place the northwestern end of the Barumun segment of the fault in thevicinityof thetruncated

depression to be evidence of a significant component of

complex. Significant components of dip slip occuron faults that splay northwardfrom the main trace into the caldera. The Tom segmenthas not produced a major historical but right-lateral slip nearthe northern endof this inspect l:100,000-scale aerial photography. There thefault earthquake, did generate the Ms 6.4 PahaeJaheearthquake of traces display cleargeomorphic evidence of strikeslip.The segment channel of theBarumun River may be offsetabout20 km, but 1984. 2.3.16. Renun segment(2.0Nto 3.55N). This longest this offset is not compelling because the trunk stream doesnot cross the fault. segment of the Sumarran fault traverses the western flank of 2.3.14. Angkola segment (0.3N to 1.8N). The the 80-km-long Toba caldera, alleged to be the largest caldera on Earth [Chesner et al., 1991 ]. Much of southwestern branchof the Equatorial Bifurcation consists of Quaternary traverses thethickpyroclastic flow deposit acontinuous fault with an abrupt30 bend at about0.65N theRenunsegment eruption. The regional expression of (Figure 4 andPlate1). Geomorphic expression is particularly of that73,000-year-old clear between about 0.8N and 0.5N. Katili and Hehuwat this225-km-longsegment is linear,exceptfor a dogleg along 30 km, where the segmentforms the [1967] usedoffsetsof tributaries to the AngkolaRiver at its northwesternmost flank of the Alas Valley graben. This graben, about 0.55N to demonstrate right-lateral offsets ranging from southwestern graben along 200 to 1200m alongthis segment.The northern 30 km of the 45 km longand9 km wide, is oneof the largest Angkola segment consists of a set of discontinuous faultson the Sumarran fault. West of Lake Toba, the fault consistsof 30- to 40-km-longstrands, arranged en echelon, with the southwestern flank of the Sarullagraben. Althoughlarge- several separations of onlya kilometer or so. Although scale aerial photographs do show minor, discontinuous across-strike natureof the en echelon pattern suggests faulting at about 0.35N, the lack of through going the right-stepping geomorphic expression of the western branchsouthof 0.5N that the fault is experiencinga minor componentof in the uppercrust,the stepoversare associated shows thatthe fault is significantly less active there. The transtension not graben. western segment doesnot rejoin the northeastern strand just with horsts, The southeasternmost partof theRenunsegment exhibits a north of the Equator. Geologic mapping supports this well-defined 2-km offset of the 73,000-year-old Toba Tuff, interpretation, and suggests that total slip on the western to determine a 27 mm/yrslipratefor the branch cannotbe large [Rock et al., 1983]. Geodetic whichwe haveused measurements spanning the early to mid-1990ssuggest that fault [Sieh et al., 1991; D. Natawidjaja and K. Sieh, in preparation, 2000]. GPS measurements across modem strain ratesarehigher in the vicinityof theAngkola manuscript portion of thissegment suggest sliprates of 24 +_ segment thanon themainsegment farthereast[Genrich et al., thesouthern

somewhat arbitrarily at an abrupt15 bendin thetraceof the fault, near theheadwaters of theBarumun River. 0nlyalong its northernmost 35 km havewe been ableto

the northern portion of the this issue]. Combined slipat depth at a rateof 23 _+ 4 mm/yr I mm/yrbelow--9 km. Across Renunsegment, geodetic ratesappear to be 26 +_ 2 mm/yr satisfies thegeodetic measurements. The Angkola segment of theSumarran faultproduced the [Genrichet al., this issue]. The Renun segment was the sourceof three major famous earthquake of 1892,during theestablishment of the early in thetwentieth century. Accounts of these first primary triangulation network in theregion. Differencesearthquakes areverysparse, however, andthelimitsof therupture inangles measured just before and afterthe earthquakeevents from poorlyconstrained isoseismal enabled MiIler[ 1895]to calculate thatcoseismic fight-lateral can only be guessed dislocations of at least 2 m hadoccurred along a northwest contours. Visser [1922] reportsthat shakingduring the

22, 19!6, earthquake was very strong in the trending linecoincident withthat portion of thefault trace February valley andthattheradius of strong shaking was between 0.45ON and1.2N.These geodetic data, along with Tamtung
24, 1921, earthquake hada region of those fromthe 1906SanFrancisco earthquake and 1891 -200 km. TheJanuary shaking similar tothat oftheearthquake of 1916.The Mino-Owari earthquake inspired, Reid [1913] toformulate the severe of shaking for theearthquake of April !, 1921, was theory of elastic rebound [Yeatset al., 1997, Chapter 8]. radius 1922]. Prawirodirdjo et al. [thisissue] havereanalyzed theDutch twiceaslarge[Visser, 2.3.17.Tripa segment (3.2Nto 4.4N). Marked data and conclude that thedextral slipwas4.5+ 0.6m. The and curvature, mountainous terrain, and most serious damage reported in 1892 wasalong thefaultin irregularity



river canyons andstream offsets. Dextral separations of--25 offsetsof the Kuala Tripa and MeureuboRivers. Each of and20 km on theGeumpang andWoylaRiverchannels are. thesedeeplyentrenched rivers displays a clear offsetof-21 notcompelling evidence for offset, but theyaresimilar in km (Figures6 and 7 and Tables 2 and 3). magnitude to the size of clearoffsets of the Tripaand The segment's southeastern terminusis the northeastern MeureuboRivers farther southeast (Figure 7). The flank of the extensional Alas Valley graben.Its northwestern northwestern portion of theAceh segment traverses a region
limit is a 9-km-widerestraining bend,whichdisplays southside-upfaults with a significantcomponent of reverseslip. Onecouldarguethatan appreciable contractional jog at 4.0N and a major changein strike at 3.85Njustify dividingthis
segment further.

spectacular dextral offsets of major rivers characterizethis 180-kin-long segment (Figure4 andPlate 1). The location of the main traceof the fault is well constrained by spectacular

and Plate 1). The southeastern two thirdstraverse

mountainous terrain and are well expressed byaligned major

of low relief and is obscureon l:100,000-scale aerial photographs. Geomorphic expression of the lhult is subtle andstream offsets appear to be absent there. Although some published maps show the Sumarran faultrunning along the southwestern flankof theAcehValleyandcontinuing into the

Parallelto and ---15km northeast of the centralportionof

this segment (between 4.0 and 4.25N) is another active strike-slip fault. This 55-kin-long fault trace is also well

sea across thenorthwestern coast [Curray et al., 1979; Page

et al., 1979], we see no geomorphicevidence of active faultingwithin 25 km of the coastline.Therefore,we, arenot definedby alignedriver valleys and stream offsets. Stream convincedthat the fault is active northwestof about5.4N. patternssuggestthat this fault may convergewith the main Geomorphicevidence for inactivity is compatible with active trace at thenorthwestern terminus of theTripasegment. geodetic observations that strainis accumulating at nomore However, we could find no clear large-scale geomorphic than a few millimeters per year acrossthe fault [Genrich et evidenceof this, nor does the l:250,000-scalegeologic al., this issue]. mappingsuggest it [Cameron et al., 1983]. 2.3.19. Seulimeum segment (5.0N to 5.9N). This An earthquakeon September 19, 1936, occurredalong the segment represents the principal active trace of the Sumarran southeasternmost part of the Tripa segment (M., 7.2 [Newcomb fault throughnorthern Aceh province(Figure 4 andPlate1). and McCann, 1987]). A smaller, more recent shock(rob6.0, The activetraceis markedby sharpescarpments anddissected November 15, 1990) occurred near the middle of this young volcanic deposits on the southwesternflank of segment. SeulawahAgam volcano. Small tributariesof the Seulimeum 2.3.18. Aceh segment (4.4N to 5.4N). This 200-kin- River are clearly offset a few hundredmeters. Alongthe long segmentof the Sumarranfault has a smoothsinusoidal central part of this segment,young folds appearto be offset shape and lacksmajordiscontinuities or sharpbends(Figure4 -20 km (Figures6 and7 andTables2 and3).



7 'k s



part of Sumatra


.... .. d. kJ.




./" River

Select Large Offts

(e Table 3)

Small OffL 17 channel off2u t b ]Select (See Table 2) i Obmed Lines pointing ,o the exit


2122 23 19 20 '




3031 28

I.., ' E flF _.I '.;,:.'.'

Southern part of Sumatra
Figure 6. Map of small and largegeomorphic offsets alongthe Sumarran fault. See Tables2 and 3 for more information.The largest offsets indicate thattotalslipacross thefaultis at least20 km.





Andaman Sea
2!-----', /

active hult zone 21

offset ()

0 20 100 km

Anticline SynclineVleys

small river large river .,.,.-, Drainage Divide

centers and cones of active volcanoes

Figure 7. Twoof themost compelling large geomorphic offsets along theSumatran fault,the21-kmdextral offsets of theTripa and Meureubo Riversin northSumatra.The headwaters of the nearby Woyla River and folded

Quaternary sediments near 6Nalso appear tobeoffset bythis amount. These offsets appear to represent thetotal
dextral offset since initialupliftof thispartof theBarisan Mountains several million years ago.

Clearevidence of recentactivity along the southeastern 22 kmof this segment is absentfrom our aerial photos,but we infer thatthe fault continues through the long,narrowvalley oftheBaroRiver alongthisreachto an intersection with the Aceh segment at about 5N. Northwest from the coastline, bathymetry [Curray et al., 1979; J. Curray, written communication, 1999], focal mechanisms (Harvard CMT catalogue), geomorphic expression of faultingon Weh Island, and evidence on a seismicreflectionprofile [Peter et al.,

the geomorphic evidence for recentdextralslip alongthe Seulimeum segment, Gertrichet al. [this issue]showthat
strain accumulation acrossthis segmentin the early 1990s
could be nil.

2.4. Other Related Structures

2.4.1. Bateefault. The Bateefault is a majorright-lateral

strike-slip faultthat diverges from theSumarran faultatabout

with the Sumarran fault and 1966; Weeks et al., 1967] suggest that the fault continues 4.65N. Betweenits intersection the coastline, the fault traverses the 1000-m-high escarpment of theBarisan range.Kariget al. It is interestingthat dextral movementalong the southwestern that this structure continues ontothe Seulimeum segment hasproduced no deflection of theAceh [1980] haveshown shelfandoffsets theedgeof thecontinental shelf segment at their intersection. It is difficult to imagine how continental -150 km and the eastern edge of the outer-arc ridge -100 km. many kilometers of dextralslip on the Seulimeum segment
couldhave occurred without at least a broad deflection in the

One strandof the Batee fault terminates beforereaching the

northern partof NiasIsland (Plate1). Another strand runs along thenorthern coast of Niasandappears to offset the trench slope andouter-arc ridge (Plate1). Except very McCann, 1987; Soetardjo etal., 1985]) severely damaged the inner city of Banda Aceh,but the source of theeventis unknown. locally, theBatee faultdoes notappear to be active on the of Sumatra.Although several largeriverchannels An earthquake in 1964 (Ms 6.5, NationalEarthquake mainland dextral deflections of upto 10km,smaller ridge lines Information Center (NEIC))damaged Kmeng Rayamore display
Aceh segment. A largeearthquake in 1936 (M 7.1-7.3 [Newcomb and

severely thanBanda Aceh. SinceKrueng Rayais closer to

and channels exhibitno offset. We suspect that tl,se large

are,indeed, dextral offsets,but thelackof clear theSeulimeum segment, the Seulimeum segment of the deflections offsets suggests either noactivity in the past fewtens of Sumarran fault may have generated this event. In contrast to small






Table 2. Selected SmallOffsets AlongtheSumatran Fault(FromNorthto South)

River/Lake Name
a Aceh River

Offset, m


Toru River


offsetof several streams thatincised youngvolcanic deposits onthesouthwest flankof Seulawah Agamvolcano excellent offset of several streams deeply cutintothe73,000
year old Toba Tuff offsetsof a few streamson the northeastflank of the Sorik




AngkolaRiver (Ringkitbranch)
Sianok River


Merapivolcano offsets of several tributaries of the AngkolaRiver

excellent offsets ofseveral crossings of theSianok River, deeply

incised intothe 60,000 yearold ManinjauTuff

Anai River


offsets of several channels onthesouthwest flankof Merapi

offset of north sidewall of the caldera lake


Lake Dipatiampat
Musi River


excellentoffsetsof tributariesto the Musi River, on the

southwest flank of Kaba volcano

Manna River


offsetof Air Kiri andAir Kanan(Plate2) whichdrainsan eroded

volcanic edifice

WerkukRiver (Menjadi,Pisai,
Rebu branches)


offsets of threechannels thataredeeplyincised intothethick, Quaternary RanauTuff

thousands of years or activity at a ratemuch lowerthan along 3. Discussion, Interpretations,and Speculations

theSumatran fault. Thisinterpretation conflicts withthe 12 +_ In this paper, we have defined the geometry and 5 mm/yr estimate of dextralslip rate of BeIlier and Sebrier geomorphology of the Sumatran fault. There are now several [1995]. We question the validityof theirapproach, which that theserefinements allow us to address. These usesan empirical relationship of channel lengthand age to questions include the implications of the fault's historic behavior and derivean agefor a channel.This ageis thendividedintothe geometry for the evaluation of future seismic hazard and measured offset to determine a rate. about thetotal' offsetacross theSumatran fault and 2.4.2. Tom fold and thrust belt. Between about 1.0 and questions its role in oblique convergence during the past many millions 1.5N lies a geomorphologically remarkable set of active and foldsandfaultsthat strikeroughlyparallelto the Sumatran of years. Other questionsconcernthe geometric kinematic relationship of the Sumatran fault to the fault but lie 15 to 40 km farther southwest (Plate !). The neighboring subduction zone and the relationship of arc principal manifestations of this fold-and-thrust belt are a to strike-slip faulting. We address eachof these northwest striking anticline and sync!inc. The syncline volcanism in turn,below. underlies a 25-by-10-kin swamp, andtheanticline appears as fourquestions
a 30-by-15 km fold in Mio-Pliocene sediment. The Gadis

Riverandits tributaries meander across thesyncline andthen

traverse the anticline as an antecedent stream.

3.1. Historical and Future Seismicity

In thepreceding discussion, wehave described very briefly

what is knownaboutlargeearthquakes alongthe Sumamn

In addition, severalsmaller northwest striking reverse faults appearto break the anticline(Plate 1). The anticline alsois cutby smallnorthstriking strike-slip faults. However, these faultsare so smalland closelyspaced that theydo not
appear on Plate 1.

fault. Eventhese highlyabbreviated accounts suggest that geometric segmentation influences seismic rupture of the
Sumatran fault. In contrast to the San Andreas fault in

California[Lawsonet al., 1908; Sieh, 1978], the Sumam

Table 3. Proposed LargeOffsetsAcross the Sumarran Fault

Features Offset, km Quality
offsetof a few fold axeswhich deformed Pliocene,

Quaternary folds
Meureubo River



Miocene, andOligocene strata dextral offset is clearly indicated by thedeflection of the

trunk channel

Tripa River



dextral offset is clearly indicated by thedeflection of the

trunk channel

Singkarak graben



based onaninterpretation of thegraben opening


Seblat River
Ketahun River



clearly shown by a sharp deflection of themainchannel

clearly shown by a sharp deflection of themainchannel









fault appears to produce earthquakes with rupturelengths no

greaterthan a hundredkilometersor so. We speculate that thiscontrast in behavior results from the contrast in continuity of the two fault systems:The SanAndreas fault hasonly one stepoverdiscontinuity with a cross-strike width greater thana kilometer(near San GorgonioPass[Allen, !957]), whereas the Sumarran fault hasat least 12. The SanAndreas hasonly two large bends (near Monterey Bay and at Tejon Pass) [Jenningsand Saucedo, 1994], whereas the Sumatranfault hasabouteight. A morepreciseand detailedevaluation of the relationship of these irregularitiesand their relationshipto historical ruptures is warrantedbut is beyondthe scopeof this paper. We havebeguna thorough analysis of the historical accounts

Thetwomajor offsets between 5 and5.5N provide the

most compelling evidence from stream channels forlarge

andhavelong,straight courses along the faulttrace.The neighboring WoylaRiverdrainage alsoappears to beoffset -21 km,butthisoffset isless certain because thematch across the fault is of trunkchannel to tributary channels. The drainage dividebetween the WoylaandGeumpang Rivers
alsoappears to be offsetby ~20 km.

offset along theSumatran fault(Figure 7). Thedeeply incised

trunk channels ofboth streams cross the fault atahigh angle

entrenched channels of the TripaandMeureubo Rivers, but

One could propose 40- to 50-kin offsets forthe deeply

topography. Our proposed20- to 30-kin offset of an pairat about 6.4N, which is based upon a andhopeto interesta seismologist in studying instrumental anticline/syncline offset of folded Pliocene, Miocene, andOligocene records in order to assess more fully the role of geometric plausible the segmentation in controlling ruptureparameters.Until this rocks [Bennettet al., 1981] (Figure 7), supports futureworkis completed, onecanobtain a crudesense of the interpreted20-21 km offset of the Tripa and Meureub0 influence of fault segments on historical ruptures by Rivers. Anotherlarge offsetthat we will consider in moredetail is comparing Kati!i and Hehuwat's[1967] compilation of the one we can infer from the geometry of the normal faults along felt regionsof historicalearthquakes with our map of the graben at about1.4S.Thisis morespeculative fault. Bellier et al. [1997] have redrawnKati!i andHehuwat's theSingkarak offsets described above. In most cases, map and include a few more recentearthquakes in their thanthe geomorphic the length of a pull-apart graben along a strike-slip fault compilation of historical felt areas.
probably doesnot represent the total slip across thefaultzone
3.2. Offsets Across the Sumatran Fault and the Evolution

this would leave implausible mismatches in thesurrounding

(for example, the 7-km-long step over mappedby

Zachariasen and Sieh [1995] between two faults in California

of Dextral Slip Along the Sumatran Margin 3.2.1. Exemplary small to moderate offsets. Noneof the geomorphic offsetsacross the Sumatran fault are greater than ~20 km, and most are far smaller(Figure 6 and Tables2 and 3). The smallest knownoffsetsalongthe Sumarran fault are thoseassociated with particular historic fault ruptures.These include offsets of a meter or two on the Sumani segment (0.75S), during the 1943 earthquake[Untungeta!., 1985] and up to about 4.5 m during the 1892 earthquake, on the Angkola segment(about0.6N) [Miiller, 1895; Reid, 1913; Prawirodirdjoet al., thisissue]. Our bestexamples of dextra! offsets in the range of hundredsof meters to a couple kilometersare on or near the flanks of young volcanoes:
Channelson the southwest flank of Kaba volcano (3.6S) are offset~700 m. The walls of Dipatiampatcaldera(2.65S)are
offset -500 m. Stream channels incised into the southwest

hasonly300 m of totaloffsetacross it). The particular nature of the faultsbounding the Singkarakgrabensuggests that it may be an exception. Althoughthe dextralfault segments cominginto thestep over from the northwest and southeast are misaligned byonly ~3.5 km, the normalfaultsbounding the lake are separated by
as much as 7.5 km (Plates 1 and 3). Becauseof their saladtong geometry,we surmisethat the normal faults represent collapseof shallow crust into the expandingrectangular region thatis beingproduced by dextralslipon themisaligned
lateral faults.

The predominance of volcanicrocks of Plio-Pleistocene age on the flanksof the grabenindicatethat the graben is no morethana few millionyearsold. Bellierand Sebrier [1994] proposed that the Singkarakbasin is an extinct pull-apart graben,inactivatedwhen the trace of the Sumatranfaultcut

flank of Marapi volcanodisplayoffsetsrangingfrom 120 to 600 m. The Maninjou Tuff (0.4S) has been offset 700 m (Figure 3), and channelscut into the Toba Tuff (2.2N) are
offset about 2 km (Table 2). We have usedthree of theseto

acrossthe lake.

The very steep scarpsand youthful

topography associated with the graben-bounding normal faults strongly suggest, however, that accommodation space continues to be created by dextralslip on the en echelon

determine the modem slip rate of the Sumatran fault, but full documentation of theseratesis the subject of a manuscript in preparation. As one would expect,highly dissected volcaniclandforms are offset more than their youngerneighbors are. The two offsetstreams cuttinga dissected volcanicedificeat 4.2Sare goodexamples of this. They are offsetabout 2.5 km (Plate2). 3.2.2. Largest geomorphicoffsets. The largest plausible geomorphic offsets along the Sumatranfault are ~20 km (Table 3, Figures6 and7, and Plate 1). Theseincluderightlateral deflection of the channels of the Ketaun River channel

Sumani andSianok segments. Furthermore, the location of the 1943rupture is inconsistent with a competing model for
theevolution of thefaultby BellierandSebrier[1994]. We hypothesize that the normalfaults shouldonlybe

active adjacent to foundering crust within theaccommodation space generated bydextral slipalong theenechelon faults. A hypothetical evolution of these normal faults asthestrike-slip displacement grew is depicted in Figure8. Therefore, we propose, thatthe totaloffset on these two misaligned strikeslipsegments is -23 km,thelength of thearcuate normal fault
zones on either side of the lake.

at 3.2S,the SeblatRiver channelat 2.9S,andthe Tripa and

Thisis,of course, nottheonlyplausible evolution forthe

Meureubu River coursesat 4.1 and 4.4N (Plate 1). Late Singkarak pull-apart graben, but it is onethatis consistent Cenozoic folds at 5.25N may also be offset-20 km. with -20 km of totaloffset along theSumatran fault. One Furthermore, we speculate below that the Singkarakgraben could, forexample, accept ourinference thatthelengths of

(at 0.6S) hasdeveloped in response to 23 km of offset.

thenormal faults reflect thefault-parallel length of actively





Smgkarak ,'

13km offset

18 km offset

23 km offset

0 et
Figure 8. Hypothetical evolution of theSingkarak graben and bounding normal faults showing howthelength of the normal-oblique faultsmightrepresent thetotaloffset on theSianok andSumani segments. ProfileE shows the
current geometry of the graben.

only a few tensof kilometers of offset. An example foundering crust but hypothesize that the length of the accrued North Anatolianfault, whichhasa foundedng regionhasremained unchanged at --23 km since is Turkey's 1500-km-long in a the faults initiated. This wouldimply that the lengthof the totaloffsetof only 85 km [Annijo et al., 1999]. Second,

foundering regionhas no bearingon the amount of total strict sense, the Sumatra fault is not one fault; rather, it is a offset. We favorour hypothesis because it is consistent with fault zone that consistsof many segments,which range in otherevidence for---20 km of total offset. lengthfrom 60 to 220 km. Many strike-slipfaults with as shortas thesehaveaccrued only a few kilometers 3.2.3. Total offset. Why are the largestgeomorphic lengths offsets no greater than ---20km? Is it possible that these to a few tens of kilometersof offset (for example, the San represent totalstrike-slip offsetalongthe Sumatran fault? Or Jacinto fault in California is a zone with 24 km of dextral isthere a limitto thesizeof geomorphic offsets related tothe offset that consistsof many disjunct segments,tens of
susceptibility of landformsto erosionand burial? We will

give reasons below why20 km might well bethetotaloffset

across the fault, but we will also show that a total offset as
greatas -,-100 km can not be ruled out at this time.

kilometerslong). Anotherreasonto suspect that total slip would be >20 km is the transformation of the Sumarranfault into the spreading centersof the Andaman Sea [Curray et al., 1979]. This

thatoffsetcouldequalthe 460 km of spreading that Indirect arguments for offsetmuch greater than20 km are suggests as follows:One mightexpect thatthe greatlength of the hasoccurredtherein the past 10 Myr. But we will seebelow Sumarran faultrequires substantially greater total offsets than that much of this offset has been carried by faults that splay acouple tens of kilometers. It is certainly truethatmanyvery
into the forearc, west of the Sumarranfault zone.

Regardlessof plausible analogues and the fault's long strike-slip faults, such astheAlpine (NewZealand) and San Andreas (California) and manyoceanic ridge-ridgeconnectionto the spreadingcentersof the Andaman Sea,

transform faultsdisplaygeologic offsetsof hundreds of

reasons. First, manyothervery long strike-slip faultshave

directgeologic evidence for total oftket across the Sumatran

fault is sparse and equivocal. McCarthy and Elders [1997] kilometers [Yeats etal., 1997, Chapter 8]. suggest 150 km of dextralslip,on the basis of similarities in But this is nota strong argument for large offset, for two
isolated outcrops of crystalline basement on bothsides of the




0.SN '


Ocean. The flow of the Sumpur River,between about 0.10 and0.75N, hasalso been strongly influenced by subsidence alongthe fault; majortributaries flow into andacross the Sumput Valleybefore flowing eastward toward theJava Sea
from their confluence at the fault.

A thirdlimitto thesizeof geomorphic offsets isimposed

by thespacing of major drainage channels thatcross thefault.

Cumulative offsets are unlikely tobegreater than the spacing

80 km

between majorriver channels because piracyoccurs astrunk channels of one drainagesystemare offset to positions

upstream from neighboring trunk channels [see, e.g., Prentice,

Young Volcanic Cover

1988; Allen etal.,1984; Yeats etal.,1997, Chapter 8]. Along

onlya smallpercentage of the Sumarran faultaremajor
stream channels spaced morethana coupletensof kilometers



apart (Plate1). Piracy of the headwaters of theAlasijani

River by the Manna River, for example,may haveoccurred at about 4.1S (Plate 1). Furthermore,where the Sumarran drainage divide is within just a couple kilometersof the Sumarranfault, large trunk streamchannelsdo not cross the fault trace. In theseplaces,the Sumarranfault traverses only ' ,....... Geologicaloffsets ...... about 20 to 25 km smaller tributary drainages. Because tributaries are more closely spaced, geomorphic interference will result where Plate 4. Geologicmapsof offset bedrock unitsalongthree offsetsexceeda few kilometers. Thus only abouthalfof the sections of the Sumarran fault suggesting that the total offset


'4" ''C9/._tt " "- Granites



fault mightbe expected to express offsets greater across the fault is only -20 km. Reproduced from Katili and Sumarran
than a few kilometers.

Hehuwat [ 1967].

Nonetheless, thereis reason to favorthe hypothesis that the largestgeomorphicoffsetsare, in fact, the total offsetacross the fault. The 20- to 21-km offsetsof deeply incised channels fault in central Sumatra. Katili and Hehuwat [1967], in northern Sumatra probably record total offset sincethe however, infer that total dextral offset at three localities (near initiation of uplift of the Barisan mountain range in this the equator,3S, and 4S) is only 20 to 25 km on the basisof region, and that uplift is quite old. The age of initiation of regional-scale mapsof late Paleozoicto early Cenozoicrocks uplift is poorly constrained, but sedimentation historyof the (Plate 4). Cameron et al. [1983] suggesta 20-kin dextral forearcbasin suggests that Sumarransedimentsources began offset of Oligocenebedsat about4.1N. Neither the larger erodingin late mid-Miocene time (about 10 Ma) [Karig etal., nor thesmalleroffsetsare adequately defended by sufficiently 1979; Harbury and Kalagher, 1991], and Cameronet al. detailedmapping. [1980] document major activity of a range-bounding fault The geologic setting of the Sumarranfault supports the about 10 Ma. If this is true, then incision of the Tripa and notion that geomorphic offsetsmight be limited to lessthan a MeureuboRivers would also have begunabout 10 Ma, and few tens of kilometers and that these values could be the 21-km offsets would necessarilyreflect total offsetsince significantly less than the total offset. The abundanceof that time. The nearby 20-km offset of an Oligocene young volcanic cover, the spacing of major river channels, sedimentary unit proposed by Cameron et al. [1983] suggests and the length of individual fault segmentsall limit the thatthismaybe thetotaloffsetsince the Oligocene aswell. accumulation of geomorphically evident offset. Let us Our analysis of theSingkarak graben alsosuggests that23km consider volcaniccoverfirst. More thana quarter(-450 km) is the total offset across the Sumarran fault since formation of

require discovery of an older fault, hidden beneaththe younger sediments of theregion. Even if the Sumatran fault carriedall the dextralcomponent 3.2.4. Evidenceof stretchingnear the SundaStrait. A of the relativeplate motionvector(-30 mm/yr), no morethan simple structuralanalysisof the forearc region nearthe
-15 km of offset could have accumulated since their

of the 1650-km-long Sumatran fault traverses youngvolcanic edificesand their thick pyroclasticdeposits. Most or all of thesevolcanicconstructions are probablyfar lessthan a half million years old, given their generally undissected nature.

the two bounding faults,the Sumani andSianoksegments. If the total offset acrossthe fault were greater,proofwould

southern terminus of the Sumarran fault provides support for

---100 km of total offset across the Sumarran fault system.

deposition. Burial of older offsets would have obscuredor eliminated their cleargeomorphic expression. Clear geomorphic offsetsare also limited by the lengthof individualfault segments, which range in lengthfrom-35 to 220 km (Table 1). Since the majority of the fault segments are right-stepping, grabenare commonalongthe fault. These grabenform intramontane valleys that occupyabout-350 km
of the fault. As these basins form, streams divert into them.

However,as we will show,not all of this,nor evena majority

of it, need be associated with the Sumarran fault.

Two earlier papersdiscuss stretching of the forearc near

the southern terminus of the fault. Huchon and Le Pichon

[1984] werethefirstto suggest thatarc-parallel stretching of

theforearc region neartheSunda Straitis related to strike slip along the Sumatran fault. They hypothesized thatthe
landward bend in the subduction deformation front and the

The Alas graben, between 3.1 and 3.9N, has probably

enabled such a diversion. The Alas River drains a 130-km

absence of an outer-arc ridgeandforearc basin south of the

reachof thefault into the50-km-longgraben beforebreaching Sunda Strait (Figure 5) indicate arc-parallel stretching ofthe the graben wall and flowing southwestward to the Indian forearc region. However, theydid not usethisto calculate



they accepted sparse and equivocal evidence for-100 kmof

Sunda Strait. These twofeatures disappear nearthestrait, and plausible amounts ofoffset along the Sumatran fault. Instead, thedeformation frontbowslandward. Following Huchon and

thisas an indication of faulttotal offset along thefaultand attempted to demonstrate that Le Pichon[1984],we interpret and fault-normal neckingof the forearc this offset is consistent withreasonable estimates of arc- parallelstretching
informationfrom the forearc and outer-arc assessment of theimplications of theforearc geometry on stratigraphic regions north of theequator show thatthepaired forearc basin total offset along theSumatran fault.
region. Extensiveseismic reflectionstudiesand structural and parallel stretching. Theydid notattempt a rigorous

LassaI et al. [1989] also attemptedto quantify the

andouter-arc ridgedeveloped throughout theMiocene epoch

rapidly during the Plioceneepoch stretching of the forearc regionsouth of the Sunda Strait. but grew particularly
et aI., 1997;Samuel and Harbury, 1996]. Thuswe They show three seismic reflection lines froma 80x 50 km [Samuel

area inand ontheflanksof thegraben at thewestern entrance infer that the deformation of these features has occurred tothestrait. They annotate thesewith five stratigraphic withinjustthepast few millionyears. We beginwith an estimate of the boundaries of the volume boundaries, whose geometryand ages they defend by reference to unpublishedwork. They claim (without that has been stretched.The concavityof the deformation discussion or argument)that an allegedly upper Miocene front and merging of the outer-arc ridge and forearcbasin that the current lengthof the deformed region,L, is stratal package contains reefdeposits (anindicator of shallow suggest depthson or near the water). Theyassume an uppermost Miocene (5 Ma) agefor ~356 km (Figure9). Hypocentral the reefs andthenusethe depthof this packetof sediment to subduction interface constrain the northeast dippingbaseof calculate the "stretching factor" since5 Ma. This factor is the deformed forearcwedge. The deformation front and the slopedefinethe seaward andlandward described by Le Pichon and Sibuet [1981], who apply a baseof the continental of thedeformed region. stretching modelof McKenzie[1978] to passive continental boundaries Using these boundaries, we calculatethat the deformed margins. Theuseof thismodel seems whollyinappropriate to wedge has a volume V, of about 1.01 x 106 km 3. We ussince the parameters neededto calculatestretching are crustal that this volumeis equalto the original,unreformed mostly unknown for the SundaStrait. LassaIet al [1989], assume thatthe cross-sectional areas conclude by asserting, withoutany discussion or calculation, volumeVo. By furtherassuming and northwestern edgesof the that thisstretching factor "probably explainsthe opening of of the current southeastern the strait since 5 Ma ago,with a maximum displacement of 50 deformed region, A and B, have not changed since began,we can calculate the originalarc-parallel to70km alongthe centralSumatra fault." Their paperis, in deformation region. A and B are currently 2870 fact, sosparse on dataanddocumentation thatits conclusions lengthof the deformed and4970 km2: areleft undefended. We propose a simple measureof extension acrossthe Lo = 2*V / (A+B) =258km. (1) graben of the Sundasegment, whichestablishes a minimum mount of dextralslip on the Sumatran fault. If we assume The total amountof northwest-southeast stretching is: that thefaults bounding the graben dip 60, we cancalculate
the Sumatran fault.

the horizontal extension across the faults in the direction of We calculate a 6.5-km lower bound on

A L = L - Lo= 356 km - 258 km = 98 km.


fault forms the northeastern boundary of extension of the grabenparallelto the Sumatran fault if we Sincethe Sumatran the forearc sliver block, we are tempted to conclude that this assume thatthe 2-kin heightof the scarprepresents vertical of stretching of the forearc equalsthe amountof throw across the faults. This assumption is manifestly an estimate

right-lateral slipalong the Sumatran fault. However, in fact, of deposits within the grabenare clear on the seismic this 100 km is only an upperboundon offsetof the pastfew structure in the forearc reflection cross sections.Thus 6.5 km is probably several millionyearssincethereis another kilometers less than the actual amount of extension acrossthe regionthat could also have accommodated someof this Sunda graben.Several morekilometers of dextral slipcould stretching. The Mentawai fault [Diament et al., 1992], probably also beadded to totalslipalong theSumarran faultif located between the forearc basin and the outer-arcridge 1 and8 andPlate1), couldalsohaveaccommodated the geometry andtimingof faulting farthereastwithinthe (Figures
underestimate of total vertical throw, sincehundreds of meters
strait and buffed beneath >2000 m of volcanic debris

some of this motion. The linearity of this large structure

(summarized by Huchon andLe Pichon[1984])wereknown better. In summary, extension of theSunda graben andfilled graben farther east is consistent with dextral slipof theorder of10kmalong theSumatran fault. However, more detailed stratigraphic andstructural datawill be necessary to calculate extension across thegraben more precisely.

suggests a significant component of strike-slip motion, butthe magnitude of strike-slipmotion, if any, has not been


Plausible evolution of dextral slip along the

Sumatranmargin. Although knowledge of the geology of

the Sumatran fault and other faults of the Sumatran fault

is incomplete, enough information exists to attempt a Letusnowattempt a quantitative analysis of stretching of system of the system's deformational history overthe the forearc region, to provide a maximum limittodextral slip reconstruction fewmillion years (Figure 10). Theprincipal constraints on the Sumatran faultduring thepast fewmillion years. This past and timingof the analysis simplycarriesthe geometrical observations of on this historyare: (1) the magnitude discrepancy between spreading in the Andaman Sea and Huchon andLe Pichon [1984]to theirlogical conclusion. near theSunda Strait; (2) a range of plausible total From simple volumetric balancing of theforearc wedge, we stretching calculate ~!00 kmof stretching of theforearc parallel to the offsetsfor the Sumatranfault; (3) the timing, style, and Sumatran fault. magnitude ofdeformation intheSumatr. forearc region; and decrease in thecurrent rates of slipalong Aswe discussed in section 2.3.1.(Figure 5), theforearc (4) a southeastward

basin and outer-arc ridge areattenuated in the region ofthe

the Sumatranfault.

These constraintssuggestthat the





-., , SUMA'I RA

X 'X

,, --;..- ,.._ , ' ' ""



! .

\ Java Sea

outer-arc ridge axis

(in meters)

deformation front


___.i.__ forearc basin axis % bathymetry contour

' active volcano

' 5 .... O-kin isoba/h

,,r -.:

* ; JVA K RX'Xa;X,,

" depression/basin

56 km -



Figure 9. Stretchingof the forearc sliver plate near the SundaStrait, which appearsto have thinned the forearc wedgeperpendicular to the deformation front. By volumetric balancing, we calculate that ---I00 km of stretching of the forearcsilver hasoccurredparallelto the Sumatran fault sinceformationof the outer-arcridge and forearcbasin. This would be a maximumvaluefor northwestward translation of the part of the torearcsilver plate that is south of the equator.

Sumarranfault systemhas evolved significantlyin the past discrepancybetween deformation in the Andaman seaand severalmillion years and that the currentconfiguration of SundaStrait during the past 3 Myr may be very smallor deformation is not representative of pre-Quaternary nonexistent. deformation. Nonetheless, the currentrate of slip on the Sumatran fault One hundred kilometers of motion near the Sunda Strait appears to diminishsignificantlyfrom northwest to southeast. contrasts markedlywith the 460 km of openingsuggested by Although new geodeticevidence suggeststhat there is no et Curray et al. [1979] for the Andaman spreadingcenters significantdecreasebetweenabout lS and 2N [Gertrich (Figure 1). The contrast disappears,if one compares al., thisissue],geologicslip ratesacross this section suggest a Andamanextensionand Sundanoffset for similar periodsof muchlargerdecrease in rate, from 27 mm/yr (near2.2N)to time. Only about 118 km of Andamanextension may have 11 mm/yr (near 0.4S) [Sieh et al., 1991, 1994; D. accumulated in the past 3 Myr (J. Curray, written Natawidjaja and K. Sieh, manuscript in preparation, 2000]. communication, 1999). This does not differ greatly fi'omthe Bellier and Sebrier's[ 1995]estimations of slip ratealong the 100 km of stretching of the forearc near the SundaStrait for fault,based uponcorrelations of stream length withage, also aboutthe sameperiodof time (i.e. sincethe rapid riseof the decrease from northwest to southeast. Sumatran outer-arcridge in the early Pliocene). Hence the If thetotaloffset along theSumatran faultis only-20 km

Figure10. (opposite) A plausible (butnonunique) history of deformation along theobliquely convergent Sumatran platemargin,based uponour work and consistent with GPS results andthe timingof deformation in the forearc
region. (a) By about 4 Ma, the outer-arc ridge has formed. The former deformationfront and the Mentawai

homocline provide a set of reference features for assessing laterdeformations. From4 to 2 Ma, partitioning of
obliqueplate convergence occurs only northof theequator. Dextral-slip faultson the northeast flank of the forearc sliverplate parallelthe trenchin northernSumatrabut swingsouthanddisarticulate the forearcbasinand outer-arc ridgenorthof the equator. (b) Slip partitioning begins southof the equatorabout2 Ma, with the creationof the

Mentawai andSumarran faults.Transtension continues in theforearc north of theequator. (c) In perhaps justthe
past100 yr, the Mentawaifault hasbecome inactive, and the rateof slip on the Sumatran fault northof 2N has morethandoubled. This differencein slip rate may be accommodated by a new zoneof transtension between the Sumarran fault andthedeformation frontin theforearc andouter-arc regions.





Z' Ma %,........_... -


(76 km)

/ ..... _l

' 42myr
. / %

mv '47 7m y
/ z

40mnVyr J
(57 kin)
, I

'mmvr 48

, '
o 7Ma

-,.._ -, N..

Zone oftranstension

, 37mm/yr













800 1000 !ometers






andsliprateshavebeenconstant, thenthenorthern partof the faultzonewouldbe lessthana millionyears old. South of the Equatorial Bifurcation, where late Quaternary slip rates appear to be --10 mrn/yr, 20 km of slipmighthaveaccrued in
--2 Myr. Our calculation of-100 km of fault-parallel stretching of

that the strand ofthe Batee fault northwest ofNias offsets the ancient deformation front ~50km,from northwest ofNias to a position west of Nias (Plate 1). Farther south onthe inner
trenchslope,between Nias and Siberut Islands, the deformation front may beoffset byabout anadditional 50 km
alonganother northstrikingfault.

the forearc nearthe Sunda Straitsuggests thateitherthetotal offsetalongthe Sumatran fault is muchlargerthan20 km or

the Bateefault is -150 km [Kariget al., 1980]. From that another structure in the Sumatran fault system has paleonto!ogically constrained seismic stratigraphy, Matson accommodated -80 km of the stretching. The only plausible andMoore[1992]show thattheBatee faultwasactive from
othercandidate for dextral slipwouldbe theMentawai fault,
well constrained from seismic reflection data to run between

Dextral offset ofthe eastern edge ofthe forearc basin by

thirty kilometers of dextral slipappear tohave occurred on the Singkel fault in the late Miocene epoch.Thus it is the outer-arcridge and the forearc basin [Diamentet al., nearby to suggest thatthe firstfew tensof kilometers of 1992]. The linearity of the feature suggests that it is reasonable portions of the principally a strike-slip feature. Diamentet al. [1992] also the 150-kmdextraloffseton the northern faultaccrued in thelateMiocene.However, thebulk of arguethat the structure of the fault zone indicates that its Batee
the sense is primarilystrike-slip. In our opinion, the structural the slip must be late Plioceneand youngerbecause homocline of Nias Island is offset -100 kin. This argumentis a less compelling one becausewe are not Pliocene convinced that the Mentawai fault zone exhibits the "flower" offsetmusthaveaccrued overat least1.5 Myr, since a shorter would requireratesof dextralslip in excess of the structurecharacteristic of strike-slipfaulting. In fact, the duration position of the fault,on thenortheastern flankof theouter-arc rateof relativeplatemotion. Plate1 alsoshows a disruption of theouter-arc ridge and ridge,is consistent with the fault beinga backthrust, along and whichthe outer-arc ridgehasrisen. The existence of a large innertrenchslopesouthof Nias Island,at the Pini basin homocline in the samepositionrelativeto the forearcbasin between Tanabala and Siberut Islands. The Pini basin rapid subsidence beginning about4 Ma. This andouter-arc ridgenorthof the equator [Karig et al., 1980] experienced is probably contemporaneous with activity of (Plate 1) supports this interpretation. So it is with some subsidence
reluctance that, in the evolutionary modelbelow, we usethe Mentawaifault as a strike-slip element of the Sumarran fault
system. A final constraint on the evolution of the Sumatran fault

thelate Miocene through thePleistocene epochs. Twenty to

north striking faults thatbound thebasin [Matson and Moore, 1992]andwithminor north striking dextral-slip faults onNias [Samuel andHarbury,1996]. A disruption in theinner trench slopefarthersouth,alongstrikeof the Pini basin, may

systemis the Mio-Pliocenehistoryof the forearcand outer- representa 40- to 50-km dextral offset of the sameancient arc regions. The Andamanspreading centerswere actively deformation front mentioned above. Figures 10a-10c depict a plausibleevolutionof the spreading at ~40 mm/yr during this period,yet we have no faultandother structures of theplateboundary that evidence of contemporaneous dextral deformationof the Sumatran forearcsliverplate southof the equator. How andwhere,in is consistentwith available geologic, geodetic,and Pliocene and late Miocene time (about 2 to 10 Ma), was the seismographic data. Variationsof this historyare als0 dextral component of oblique convergence accommodated? possible;our principal intention is to show that the fault evolvedsignificantly in the past few millionyears. Matson and Moore [1992] suggest that some of this system of this speculative history areas discrepancycan be accommodated by the dextral-normal The main characteristics (1) thecurrent 15mm/yrdifference in Sumarran fault faultsof the forearcregionnear Nias Island(Figure2 and follows: Plate 1). We consider thispossibilitybelow. slipratenorth andsouth of theequator is veryyoung (perhaps Stratigraphic and structural studies by Samuelet al. [1997] onlyl00,000 years old), and (2) active normal- anddextralfaulting in the forearc and outerarc and Samueland Harbury [1996] showthat broadening and slip (transtensional) current) uplift of the outer-arcridge occurredearly in the Pliocene betweenlS and 2N is an ancient(and perhaps to the stretching at the southern endof theSumatran epochthroughout the Sumatran forearcregion. This is critical analogue to reconstructing deformationof the forearc sliver plate fault. Figure10a shows the geometry of the regionat about 4 becausethe early Pliocene growth of the outer-arcridge basin produced an elongatefeaturethat has beendeformed in the Ma. Justprior to this time, relief betweenthe forearc subsequent severalmillion years. The ridge is clearin the and the outer-arc ridge increasedgreatly acrossthe ridge bathymetry of Plate1. South of about1S,it is regular and60 homoclinalfold betweenthe forearcbasin and outer-arc to 80 km wide. Its northeastern boundaryis the Pliocene [Karig et al., 1980; Samuelet al., 1997;SamuelandHatbury, that as the outer-arc ridgegrew, the Mentawaihomoclinal flexure. On the southwest the ridgeis 1996]. We speculate bounded by a plateauthat sits at a depthof-2400 m. We subduction deformation frontjumped southwestward to its

location, from a deformation frontstill visible in the speculate that this plateau was formerly a part of the present Australian plate and that its northeastern edgeis the former bathymetry, closer to theouter-arc ridge (Plate1). From 4 to
deformation front of the subduction zone. Similar features are

2 Ma, dextral slipon theAcehsegment andtheBatee fault

innertrench slope wereoffset37 km by a curved southern

also presentbetweenabout 1.5N and 3N, near Simeulue occurred at37mrn/yr, and thehomocline, outer-arc ridge, and

of theBatee fault,off thenorth coast of Nias Island, Between1.5Nand2S,the outer-arc ridge,the homocline, extension the Pini basin. This is consistent with and the ancientdeformation front and plateauare markedly and37 km moreacross disarticulated. Karig et al. [!980] observed that the Pliocene the stratigraphy of Matson and Moore[1992]. Several homocline on the eastsideof Nias is dextral!y offset- 100km kilometers of arc-parallel elongation of Nias Island along by two strands of the Bateefault. We infer from bathymetry north striking dextral-slip faults andconjugate sinistra!-slip








Plate $. Geometric and structural details ofthe Sumatran fault, the forearc basin, outer-arc ridge, and volcanic arc,

outer arc, forearc, and Sumatran fault geometries are inthe southern domain. The coincidence ofthis structural domain with the source region ofthe giant (Mw 9) subducton earthquake of 1833 suggests that geometrical

suggesting the division ofthe Sumatran plate margin into northern, central, and southern domains. The simplest

simplicity encourages large ruptures. The central domain appears to have been the source region ofthe great (Mw 8.4) subduction earthquake of 1861. Fragmentation ofthe central domain appears tohave been caused by
interval is 200 m.

subduction ofthe Investigator fracture zone during the past 5Myr. The locus ofimpingement ofthe fracture zone

onthe deformation front was calculated byassuming the current relative plate motion vector and the forearc

deformation history ofFigure 10.Contours inred are the top ofthe Benioff-Wadati zone. Bathymetric contour



lhultsalsooccurred duringthis period[Samuel and Harbury, 1996]. Subruction southof the equator was parallelto the relative plate motion vector and highly oblique to the deformation front. Subructionnorth of the equator was mostlyor wholly dip slip because mostor all of the dextral component of plate motion was occurring alongthe Batee-

3.3. Tectonic ModeloftheSumatran PlateMargin

Transtensional necking of theforearc region between lOS and 2Nduring thepast 4 Myr has hada profound effect on all of themajor elements of theplate margin there. The inner trench slope,outer-arc ridge,andforearc basin have been fragmented by this process. Even the shapes of the Aceh fault. subduction interface,the active volcanicarc, andthe About2 Ma (Figure 10b), both the Mentawaifault andthe Sumatran faultappear to havebeen affected. In fact, wecan Sumatran fault formed. From 2 Ma to 100 ka, they carried

divide the Sumatran plate boundaryinto three tectonic -40 mm/yrof the dextralcomponent of oblique convergence domains, based upontheirrelationship to thisPlio-Pleistocene south of the equator, and the subruction interface transtension (Plate 5). The southern domain,whichwe accommodated only the dip-slip component. North of the suggest hasbeenpartof theforearc sliverplateonlyforthe equator, 40 mm/yr of dextralslip was accommodated by the past 2 Myr,isthemost simple geometrically and stmcturally. Sumatran fault (10 mm/yr)andAceh-Batee fault (30 mm/yr). The central domain, which comprises all the transtensionally Figure 10c depicts our suggestionfor the current fragmented pieces,is themostcomplex. neotectonic partitioningof deformation. The Aceh-Batee The southern domain has thefollowing characteristics: (1) fault is no longer active or is only minimally so. The the Sumatranfault displays a right-stepping en echelon Sumatran fault is slipping~15 mm/yr fasternorth of about pattern andcourses above the 100- to 135-kmisobaths ofthe 2N than south. The massbalanceproblemcaused by this subduction interface, (2) the locus of volcanism is discrepancy is beingtakenup by a nascent belt of deformation predominantly northeast of or near the fault, (3) theforearc
that crosses the outer-arc ridge at the equator. This deformation belt is superjacent to Fauzi et al. 's [1996] swath of exceptionallyhigh seismic activity in the down going oceanic slab. It alsoencompasses the activeTom foldsof the mainlandcoast,two youngfaults on and southof Nias and north-south grabenthat bathymetry suggest may existon the inner trenchslope (Plate 1). Figure 10c is consistent with recentmeasurements of geologically measured Sumatran fault slip ratesbut is inconsistent with the ratesof geodetic strain measured by GPS south of the equator. If the Sumatran fault is carryingonly ~ 10 mm/yrof dextral slip southof the equator[Sieh eta!., 1994; Bellier et al., 1999], the remainder of the dextralcomponent of slip mustbe taken up along either the subduction interfaceor by a fault within the forearc sliver. The GPS data show no sharp gradientsin shear in the forearc region, so the remaining dextralcomponent is probablyaccommodated by slip on the subductioninterface [McCaffrey et al., this issue]. This portionof the dextralcomponent, x, wouldbe -27 mm/yr(x = 58 mm/yr*sin 41 - 10 mm/yr, where 58 mm/yr is the magnitudeof relative plate motion and 41 is the angle between the plate motionvectorand the trenchnormaland 10 mm/yr is the slip rate on the Sumatran fault). Slip vectors for earthquakes on the subduction interfacedeviate from the trenchnormal by -20 , on average. These suggest that the dextralcomponent on the interfacewouldbe a bit lessthan ourmodel predicts, only-16 mrn/yr.

basin is remarkably simple, ~2 km deepandunbroken by

majorfaults,(4) theouter-arc ridgeis relativelynarrow, forms a singleantiformal high,andis geometrically simple, (5) the Mentawaifault and homocline, which separate the basin and ridge,are unbroken and relativelystraight, and (6) theinner trenchslopeis relativelyuniformandpossesses a prominent plateauabouthalf way betweenthe activedeformation front and the outer-arcridge. The sourceof the giant(Mw9) subduction earthquake of 1833 was the subduction interface
beneathmuch of this domain [Newcomband McCann, 1987; Zachariasenet al., 1999]. Strainsmeasuredby GPS in the

earlyto mid-1990s showthattheouter-arc islands aremoving parallel to the relative plate motion vector and that the subduction interfacebeneath the southern domainis currently fully locked[Prawirodirdjoet al., 1997;McCaffreyet aI., this
issue]. The Sumatran fault appears to be slippingat a rateof about 10 mm/yr in the Southerndomain [Sieh et al., 1991, 1994; Belllet et al., 1999].

Thenorthern domain is characterized by these features: (1)

a geometrically irregularSumatran fault, with bothreleasing and restraining bends,which residesabovethe 125- to 140km subduction isobaths, (2) a volcanicarc on andnorthof the Sumatran fault, (3) a 1- to 2-km-deep forearc basin, (4) a very

broad, structurally and bathymetrically complexouter-are ridge,(5) a homocline alongits southernmost few hundred
kilometers that is similar to the Mentawai structure of the

The historydepicted in Figure 10 is consistent with the domain is distinguished by thesefeatures: (1)a timingof activity on faults bothoffshore andonshore Nias The central 350-km-long section of the Sumatran fault that is markedly [Kariget al., 1980;Matson andMoore,1992;Samuel and with the subduction isobaths, (2) a volcanic arc Harbury, 1996]. It also incorporates ourobservation that the discordant
Bateefault is not currently activealongmostof its exposed that cuts dramaticallyacrossthe Sumatranfault, (3)a
trace but retains clear evidence of 5-km dextral offsetsof a

southern domain,and(6) a very narrowinnertrench slope.

topographically shallow (0.2-0.6km deep)forearc basin, into severalblocks during few of thelargest channels thatcross it (Plate 1). Restoration which has been fragmented faulting, (4) a fragmented outerarc,(5)a of ~80 km of slipon thefaults between lSand2Nin the oblique-normal homocline between theouter-arc ridge and forearc offshore regioneliminates the dimplein the subductionfragmented and (6) a fragmented inner trench slope. Thegiant (Mw deformation front west of Nias and Simeulue, just as basin, earthquake of 1861andnumerous other large restoration of-80 km of slip on the combined Sumatran and 8.5) subduction subduction earthquakes originated within this domain Mentawaifaults nearlyeliminates the dimplewest of the historic andMcCann, 1987]. Strains measured byGPS in Sunda Strait. Thuswe suggest that the concavities of the [Newcomb
deformation front west of Nias and west of the SundaStrait are features inheritedfrom Plio-Pleistocene dextra!strike-slip motionin the forearcregion.

theearly to mid-1990s indicate thatthehanging wallblock across thecentral domain is currently moving parallel tothe subduction deformation front[Prawirodirdjo et al., 1997;



McCaffrey etal.,this issue]. Thegeologic rate of slipof the

Sumatran fault increases markedlyfrom southeast to northwest across thecentral domain, from-11 mm/yr to -27

A more logical proposition may be that transtensional neckingof the centraldomainhas led to bendingof the subducting slab Trench-orthogonal thinningof the forearc
appears to have drawn the deformation front and trench

rnm/yr [Sieh etal.,1991 ].

We suspect that transtensional fragmentationhas northeastward, tensof kilometerscloser to the mainlandcoast. dominated the central domain becausethe Investigator If this process had not alsodrawnthe deeperpartsof the fracture zone hasbeensubducting beneath thecentral domain subducting slabnortheastward, the dip of the interface in the for the past several million years. Itslocus of impingement on forearc andouterarc wouldbe steeper thanin the southern the deformation front hasmigrated from the northern to the domain.The isobaths showthe contrary, thatthe subduction southern margin of thecentral domain during thepast5 Myr zone beneaththe centraldomainhas a very similarcrossprofile to that beneath the southern domain. One (Plate 5). This may besignificant because fault activity inthe sectional hanging wallblock of theforearc region appears tohave been test of this hypothesis would be to determineif the active restricted during thisperiod to thecentral domain (Figure 10). volcanic arcin thecentral domain is substantially northeast of Furthermore, the orientations of faults in the centraldomain thelateMiocene andPliocene arc. If so,it wouldsuggest that are predominantly north-south, parallel to thetopographic and thesubduction isobaths havemovednortheastward in thepast structural grainof the underlying Investigator fracturezone. few million years. Wehypothesize therefore thatthe topographic heterogeneity
of the SumatranFault has led to disruption of the forearcand outer-arc regions. 3.4. Relationship to the Modern Volcanic Arc Currently, the Investigator fracturezone is also associated with a band of intense seismicity withinthe downgoingslab Manyhavenoted theproximity of theSumatran faultto the in themiddleof the central domain (Plate 5) [Fauzi et al., volcanic arc andhavesuggested that it formedtherebecause 1996] andan abrupt change in the azimuthof GPSvectors on of theeffectof magmatism on thelithosphere [e.g.,Fauziet the outer-arc ridge[Prawirodirdjo et al., 1997,McCaffrey et al., 1996; Tikoff, 1998]. Sumatra aside for the moment, al.,thisissue]. however, most trench-parallel strike-slip faults are not Thesubduction interface curves broadlyacross theCentral coincident with their volcanic arcs. The Median Tectonic
domain (Plate5) [Fauzi et al., 1996]. The closeassociation of
this curve with the other elements of the central domain

oftheInvestigator fracturezone beneaththe centraldomain

Line (Japan) doesnot have an associated arc; the Denali fault

(Alaska) lies much farther from the trench than the Alaskan arc volcanoes;the Atacama fault (Chile) lies between the

suggests cause and effect or at leasta shared cause. Could flexure of thedowngoing slabhavebeen produced by necking of the hangingwall block? Or did deformation within the downgoing slab lead to transtension in the forearc sliver plate? We suggest the former. The existenceof the 1500-km-wide boundarybetween Indian andAustralian platesoffshore western Sumatra andthe Andaman Islands givesreason to suspect thatthedowngoing slab west of theInvestigator fracture zoneis deforming. This broad region of deformationabuts all of the central and
northern domains.Gordon et al. [1990] calculatethatthe two

trenchand volcanicarc; and the Philippinefault is tens of kilometers from the major Philippinearc volcanoes [Yeatset al., 1997]. Furthermore, mostvolcanicarcsalongobliquely convergent marginsdo not sportlarge strike-slipfaults. This general lack of association suggests thatthe alignment of the

Sumatran volcanic arc and the Sumatran fault is purelya

coincidence. In fact, McCaffreyet al. [this issue]haveused

finite elementmodeling of stresses acrossthe obliquely

convergent Sumatran plate boundary to showthat formation

of the trench-parallel Sumatran fault did not requirethe

oceanic plates areconverging north-south at an angular rateof presence of themagmatic arc. Nonetheless, Tikoff[1998] has that faults suchas the Sumatranfault form above 0.3/Myr about a poleof rotation in thecentral IndianOcean. suggested At the Sumatran deformation front this translates into a the locus of greateststrain gradient in the lower crust or by the magmatism of the volcanicarc. nominal 13-kmnorth-south shortening of'thedowngoing slab mantle,occasioned

andancient releasing stepovers alongthe Sumarran fault. We can test directly whether or not magmatism has influenced the locationof the fault or, conversely, whether or 1998]. To accommodate north-southcontraction,these not faulting has influencedthe location of volcanismand structures wouldneedto be rotating clockwise, domino-like, magmatism. Plate 1 allows us to searchfor a relationship toenable eastward extrusion of lithosphere [Gordon et aI., between the volcanic arc and the Sumatran fault, since it not only the mostprominent tracesof the Sumatran 1990]. Theprecise lociof such deformation is unknown, and displays so its impact ontheoverriding central andnorthern domains is fault but also the youngestvolcanoes. We mappedthese usingthe samesources we usedto mapthe hard toassess. Nonetheless, it isplausible that the contrast in volcanicfeatures nature of the southern and northern hanging wall domains fault (Figure 2). We limited our mappingto thosefeatures minimalerosion, sincehighlyeroded, older could havearisen,at least in part, from subduction of thathavesuffered deforming oceanic lithosphere beneath thenorthern domain. volcanic constructs are harder to recognize andmapping wouldhaverequired a more It ishard to imagine, however, howdextrat transtension on geomorphologically effort. The features we mappedexhibitvery little north striking faults withinthecentral domain could berelated substantial

in the past 3 Myr. The actual nature of lithospheric deformation westof the deformation frontis quite uncertain, however. Simplenorth-south buckling is unlikely. Focal mechanisms andstructure indicate a predominance of northsouth left-lateral slip on north-south faults [Deplus et al.,

BetIier and $ebrier [1994] have claimed that numeroussmall and large volcaniccones and calderasoccur at both current

tosinistral slipandclockwise rotation onnorth striking faults







Many have been active historically. Those that have been in thesubjacent subducting lithosphere, unless eastward

datedradiometrically are typically<100,000yearsold (e.g., extrusion of theoceanic lithosphere has ledtonorthwestward extrusion oftheforearc sliver plate, asplate collision has done Toba caldera,73 ka [Chesneret al., !991], and Maninjou inTurkey and Tibet. caldera, 60-90 ka [Nishimura, 1980]). In addition to mapping









jV.olcano diameter
large (> 15kmj



20 ............ , ........... i -' ' 5 ' j '





Figure 11. Plot of the distanceof volcaniccentersfrom the Sumatranfault showingthat the volcanicarc hasnot
influenced the location of the fault. However, 9 of the 50 volcanic centers are within 2 km of the fault. Most of

theseare associated with extensional (right) stepoversin the fault. Large (15-km diameter)volcanicedificesare listedalong the horizontalaxis. Smallervolcanoes mentioned in the text are named.

craters and calderas, which are indicators of volcanic source

The local centerline of the volcanicarc variesalong the

strike of the Sumatran fault. It is a few kilometers northeast

vents, we also mapped the edges of the volcanic cones in

orderto displaya crudemeasure of the outputof individual of thefaultbetween 5.5Sand0.4S,swings southwest ofthe sources. Sumatran fault between 0.4Sandabout2N, andthenswings At first glance,the moststrikingrelationships between the to a position-25 km northeastof the Sumarranfault between Sumatran fault and the youngarc volcanoes are that: (1) the 2N and5.5N. This broaddisparity between the localcenter averagecenterline of the active arc is decidedly landward line of the volcanic arc and the Sumatran fault is another (northeast)of the Sumatranfault and (2) the local centerline indicationthat modernarc magmatism has not guided the of the young volcanicarc switchesback and forth across the formation of the fault. traceof theSumatran faultasit traverses the 1650-km length It also doesnot appearthat individualvolcanic conduits of Sumatra. Figure 11 shows these relationships. The 10-km have influencedthe locationof particularfault segments. separations northeastfrom the $umatran fault are common, Onlyrarely do individual segments of thefaultbisect volcanic 25-km distances are not rare, and a few volcanoes are even centers or bendin their vicinity (counterexamples areKaba farther northeast. Only two volcanoesare more than 10 km and Dipatiampat). However,we would not expectsuch an southwest of the Sumatranfault. From Figure 11 one can association, sincethe volcanoes thatwe havemapped are far estimate that the averagedcenter line of the largestvolcanic youngerthan the age of initiation of the mapped fault edifices is --10 km northeast of the Sumarran fault. This segments.We suspect that mostof the uneroded edifices are skewed distribution of volcanoes relative to the Sumatran less than100,000 years old,whereas we havemade a case that fault suggests that the modernmagmaticarc has not created a thefault planes wehave mapped areprobably -2 Myrold.If
weak crustal zone that has favored the concentration of shear.

Perhapsthe active volcanic arc has failed to influence the locusof faultingbecause the volcanicconduits "soften"only a small percentageof the length of the arc. Alternatively, perhaps magmatic plumbing beneath the Sumatran fault, associatedwith an unmapped, extinct volcanic arc, did
influence the location of the fault.

thelocus of faulting wereinfluenced by magmatic softening of the crust,the magmatic plumbing that led to the
concentrationof strains beneath the Sumatran fault would

have formed longbefore theyoung volcanoes onPlate 1. To test the hypothesis that magmatic concentration of shear
stresses led to the formation of the fault within the arc, one

would need to map the Pliocene and early Pleistocene



volcanic centers. Wemay attempt this ata future date, but it

is beyond the scope ofour current efforts.

by Engdahl et al. [1998]) andas determined by Fauzi et al. [1996] in their local seismic surveyin the regionof Lake

Despite thelackof influence of active magmatism on Toba. tectonism, tectonism is influencing magmatism, butonly toa From about 6S to the equator,the relationship is minor extent. Thisconclusion contrasts withthatof Bellier particularly regular;the subduction interfacelies 100 to 135 and Sebrier [1994], whoproposed that extensional pullaparts km belowtheSumarran fault,except along thesouthemmost segment (Plate1 andTable 1). Between about3.5N along theSumatran faulthave affected thelocation of the (Sunda) volcanoes. In fact, ourmap shows that only9 of the50young and 6.0N the subduction interface is 125 to 140 km below
volcanic vents shown onPlate1 arelocated within 2 kmof a
fromsoutheast to northwest,Suoh, Seminung,Kaba,

the Sumatran fault, exceptbeneath the northern (possibly

are, on average,~20 km greaterthan depthssouthof the

partof theAcehsegment. These depths in thenorth mapped trace of theSumatran fault (Figure 11). These are, inactive)
equator. The relationship of subduction isobaths to the Dipatiampat, Kunyit,Melenggok, Talang,Sibual-buali,

Seulawah Agam,andPulauWeb. Kaba,Kunvit,Meleggok, Sumatran faultis markedly aberrant between theequator and

Talang, Sibual-buali, Seulawah Agam,andPulau Web are

stratovolcanoes greaterthan about 10 km in diameter and, thus, embody the mostsubstantial volumes. Suoh,Kaba,

about 3.5N.

There the traces of the Sumarran fault and the

subduction isobaths aremarkedly discordant; the depth of the interface beneath theSumatran faultranges from-100 to 175

Kunyit, Melenggok, Talang,and Sibual-buali are located

within dilatational stepovers or on oneof the bounding faults of a dilatationa! step over. One of these(Suoh) is a large

Because of thewell-behaved relationship of Sumatran fault to isobaths in thenorthern andsouthern domains, we propose

phreatic explosion crater thatformed 15 days after thelarge that the Sumatranfault formedfirst in thosetwo domains,as
Semangko segment ruptureof 1933 [Stehn,1934], most two separatestructures. As displacement on the faults has convincingly in association withtectonic activity.Bellierand grown,they haveformeda linkageacross the centraldomain Sebrier [1994]proposed thatTobaandRanau calderas also andwill onedaybecome a single structure.

and are based solelyon the use of SPOT imageryto map more 4. Summary, Conclusions, ancient fault strandsin the vicinity of these two calderas. Remaining Questions Although linearions may exist along theseallegedancient We have used stereographic aerial photographyand faults, theirdocumentation of the lineations is scant,and they topography to map 1650 km of the Sumatran fault (Figures 2 present no geologic mapping to confirm theirexistence or to and 3). The resultingmap showsthat the fault comprises quantify thestyle, age,or amount of shear along them. numerous segments separated by dilatational and We suspect thatthe association of just 9 of the 50 young contractional stepoversand abruptchanges in trend(Plate 1 volcanoes with the Sumatran fault is a random occurrence. If andFigure4). This segmentation appears to haveinfluenced one peppered an elongate rectangle (with the 1700-by-50 km the rupturedimensions of historicallarge earthquakes and dimensions of the volcanicarc) with a randomdistribution of limitedtheirmagnitudes to ~7.5. 50points andthenranstraight linesrandomly through its long The largest geomorphically evident offsets along the dimension, several pointswouldtypicallybe within2 km of Sumatran fault are between17 and 23 km (Plate 3, Figures7, eachline. Thus the close association of several volcanoes and 9 and Table 3). Theseare predominantly deeplyincised with theSumatran fault zonedoesnot, by itself, demonstrate a fiver channels, but one apparent offset of a fold pair and the genetic relationship. The closeassociation of sixof theeight accumulated offset acrossa major step over also fall within close encounters with dilatationalstep overs does,however, this range. A lack of detailedand complete mapping along suggest thattectonic stepoversare influencing the locations the fault precludes confidentmatchingof geologicunits of a few of the arc's volcanic centers. across the fault, but rock offsetssuggested by Katili and Hehuwat [1967] and Cameron et al. [1983] supportthe 3.5.Relationship of the Sumatran Fault contention that the 20-km geomorphic offsetsrepresent the
to the Subduction Zone

formed at extinct extensional step overs along the Sumatran fault zone,but these hypotheses are not well founded.They

Thegeneral shape of theSumatran faultmimics thatof the Straitsuggests ~ 100 km of arc-parallel stretching of the deformation front offshoreso faithfully that one wonders Sunda sliverplatesince theearlyPliocene (Figures 5 and8). about a genetic relationship between thesubduction interface forearc that20 km of thiswasaccommodated by dextral and thestrike-slip fault(Plate5). Northof theequator, both We propose structures are concave toward the southwest. South of the slipontheSumatran faultandthattheMentawai fault,a long, within the forearcregion,accommodated the equator, both are broadly concavetoward the northeast. linear structure

total offset across the fault. The distention of forearc structures and the trench near the

Along theentire length of the Sumatran faulton land,its

horizontal distance from the deformationfront variesno more

remaining dextralslip.

than ~10% from290km(Table1 and Plate 5).

A similarcoincidence existsbetweenthe shapeof the Sumatran fault andthatof the subduction interface downdip distinct tectonichistories(Plate 5). The southerndomain andmayhavebeenaccreted from itstrace. Thisis clearfromPlates1 and5, whichshow (from7Sto 1S)is the simplest the50-, 100-, and 200-km isobaths of the subduction to the forearcsliverplate only about2 Myr ago by the

Oursynthesis of datafromtheSumatran fault,thevolcanic arc, and the forearc regionshows that the Sumarran forearc sliver plate consists of three tectonicdomains with very

interface. Thecontours aredrawn onthetopofthe Wadaft- creation of the Sumatran and Mentawai faults. The northern andits northern part Benioff zone, as defined by hypocentral locations in the domain(northof 2N) is morecomplex,
hasbeenexperiencing arc-parallel translation for at leastthe International Seismological Center (ISC) catalog (as relocated



past10 Myr. The central domain is themost complex of the three and has been a regionof transtension between the
northernand southern domainssinceat least4 Myr ago.

Our map of theSumatran fault can serve asajumping-off point forcareful analysis of theseismic hazard posed bythis major structure. To what degree does thehistorical record of
largeearthquakes along the Sumarran faultdemonstrate that

Geodetic measurements suggestthat slip acrossthe

Sumatran fault between about 0.8 S and 2.7N is nearly

large structural irregularities constrain rupture lengths?

records help constrain the uniform at about 25 mm/yr[Genrich et aI., thisissue].These Would primitiveinstrumental parameters of these largeevents of thefirsthalfofthe rates areincompatible withthe27 andl1 mm/yrgeologic slip source ratesthat we have determinedat 2.2N and 0.3S [Siehet al., twentieth century? Whetheror not segmentation of the 1991, 1994; D. Natawidjajaand K. Sieh, manuscript in Sumatran fault has markedly influenced ruptures, answering could profoundly affect our general preparation, 2000). We propose thatthegeologic difference these questions in rateshas arisenin just the past 100 ka or so, because understanding of theimportance of structural geometry on
structural evidence for accommodationof the 15 mrn/yr

seismic ruptureprocesses.

difference is obscure. We suggest that a belt of auxiliary,

transtensional deformation between the Sumatran fault and

the trench is the nascentmanifestation of this rate change Yehuda Bock and his colleaguesat the IndonesianNational

Acknowledgments. Thisworkbegan withan invitation by

Coordination Agency for Surveying and Mapping (Figure10). This beltincludes the western (Angkola) branch (BAKOSURTANAL) to visit Sumatra. They were interested in of the Equatorial Bifurcation, the Tom fold-and-thrust belt establishing a better geologic context fortheirGPSgeodetic studies, underNSF grants EAR-8817067andEAR-9004376. Our along themainland coast, andsubmarine faults in theforearc supported
basin,outer-arc ridge,andinnertrenchslope. Although the Sumatran fault andtheSumarran volcanic arc
affected the location of the other. Rather than being
initial reconnaissance, in 1991, supported by donations fromthe Caltech Associates, convincedus that geologicalwork aimed at

share the same jungle,neither appears to havefundamentally In 1992, we initiated our own NSF sponsored project(EARcoincident, the fault and the arc intertwine(Figure 11). The

understanding the activetectonics of thefaultcouldbe veryfruitful.

9205591) to map and characterize the Sumatran fault. BAKOSURTANAL, through Bock, supplied most of our topographic coverage of the fault. We are gratefulto Suparka and averaged center line of thevolcanic arcis distinctly northeast Hery Harjono at the IndonesianInstitute of Sciences(LIPI)for

of, not at, the Sumatran fault. Nevertheless, the few volcanic helping usacquire manyimportant aerialphotographs andarranging centersthat are on or very near the Sumatranfault are fieldwork. Rudy Bachruddin at the Volcanological Survey of
and Development Centers(GRDC) also helpedus by providing essential aerial photographs alongportions of the Sumarran fault. dramatic bend in the modem volcanic arc between 0.7N and Various private companies also allowed us accessto topographic 2.5Nis mostprobably theresultof transtensional thinning of maps andaerialphotographs. PaulTapponnier (IPG Paris) shared the forearcsliverplatein the past4 Myr. We can notrule out topographic mapsfrom the Singkarak region. We alsobenefited with Rob McCaffrey and Bob Kieckhefer and the possiblilitythat the Pliocene and Miocene volcanicarc from conversations of our data by Carolyn White. were lesssinuous andcloserto the locusof later strike-slip from an initial GIS compilation Reviews of anearlymanuscript by Yehuda Bock, Joseph Curray, and faulting. RobMcCaffrey wereveryhelpful.Finally, we greatly appreciate the

predominantly at major extensional stepovers,whichmay well haveattracted a smallpercent of thearcvolcanism. The

Indonesia (VSI) and GunawanBurhan at the Geological Research

The broadsimilarityin shapeof the Sumatran fault and thorough, thoughtful, andcareful reviews of Jeff Freymuller, Eli subduction interfacesuggests a genetic relationship.The Silver, and Paul Tapponnier. This is CaltechSeismological contribution 8625. broad, low-amplitudesinusoidalshape of the subduction Laboratory interface is mimicked by the Sumatran fault,andalong most
of its trace the Sumatran fault lies above the 110- to 140-km


isobaths of the subduction interface.These relationships are Abe, K., Magnitudes of largeshallow earthquakes from 1904 to
particularly regular northof 3.5Nandsouth of theequator, in the northern and southern domains. We suggest that the Sumatran fault first formedas two separate faultsin these two
!980: Phys.Earth Planet.Inter. 27, 72-92, !981.

domains, andare in theprocess of linking together through

the central domain and across the volcanic arc. We ascribe

Allen, C., A. Gillespie, H. Yuan, K. Sieh,Z. Buchun, andZ. Chengnan, Red Riverandassociated faults, Yunnan Province, China: Quaternary geology, sliprateandseismic hazard, Geol.
Soc.Am. Bull., 95, 686-700, 1984.

Allen,C.R., SanAndreas faultzonein SanGorgonio Pass, southern

fracture zonethroughout thepast5 Myr. kinematics, Geology, 27, 267-270, 1999. Not unexpectedly, this work has generatedas many Aspden, J.A., W. Kartawa, D.T. Aldiss,A. Djunuddin, R. Whandoyo, D. Diatma, M.C.G. Clarke, andH. Harahap, The questions as answers:What are the detailsof the creation and
evolution of the three tectonic domains of the forearc sliver

California, Geol. Soc.Am. Bull., 68, 315-350, 1957. the disrupted nature of the central domain's outer-arc ridge R., B. Meyer, and A. Hubert, Westward propagation ofthe and forearcbasinsto its locationabovethe Investigator Armijo, North Anatolian fault into the northern Aegean: Timing and

plate? How, for example, did deformation in the 1982. transtensional centraldomainevolvethrough the pastseveral Barka, A.,TheNorth Anatolian fault zone, Ann. Tectonicae, 6,164195, 1992. millionyears?Why did the Sumarran fault form whereit did,
290 km from the subduction deformation front and 100 to 150

geology of the Padangsidempuan and Sibolga Quadrangle, Sumatra, report Geol. Res. and Dev. Centr., Bandung, Indonesia,

Bellier, O., andM. Sebrier, Relationship between tectonism and

image analyses, Tectonophysics, 233,215-231, 1994.
Res.Lett., 22 1969-1972, 1995.

km above the subductioninterface? Would careful, detailed

volcanism along the Great Sumatran fault zone deduced by SPOT

mapping confirm totalSumatran faultoffsets of only-20 km?

When did the contrast in slip ratesalongthe Sumatran fault begin? Why is this gradientin rates not apparent in the

Bellier, O., andM. Sebrier, Is theslipratevariation onthe Great Sumatran fault accommodated by fore-arc stretching?: Geophys.

Bellier, O.,M. Sebrier, S.Pramumidjojo, T. Beaudouin, H.Harjono, geodetic data? Is it plausible that the Mentawai faulthasa I. Bahar, and O. Fomi, Palcoseismicity and seismic hazard along strike-slip component as largeas 80 km? Did the two faults theGreat Sumatran fault (Indonesia), J. Geodyn., 24,169-183,
originate a mere2 Myr ago?




Bellier, O.,H.Bellon, M. Sebrier, Sutanto, and R.Maury, K-Ar age

faultsin Sumatra, Indonesia, J. Geosc, Osaka City Univ., 10, 517, 1967.

Kraeff,A., The earthquake of theTesregion, Benkulen residency on eraplacement andslip-partitioning in Sumatra (Indonesia), March 15th,BeritaGunung Berapi,1, 40-56, 1952. Tectonophysics, 312,347-359, 1999. Larson,K., J. T. Freymueller, and S. Philipsen,(31obal plate Bennett, J.D.,etal.,Geologic map ofthe Banda Aceh quadrangle, velocities from the GlobalPositioning System, J. Geophys. Res., Sumatra, Geol. Res. and Dev.Cent., Bandung, Indonesia, 1981.

of the RanauTuffs: Implications for the Ranaucaldera

Bedage, H.,Jr., Deaardbeving in zuid Sumatra van 25Juni 1933:

Ned. Indie, 94, 15-36,1934.

102, 9961-9981, 1997.

dansle Waarremingen in her epicentrale gebied, Natuurwet. Tijdschr. Lassal,O., P. Huchon,and H. Harjono,Extensioncrustale

Cameron, N., M. Clarke, D. Aldiss, J. Aspden, and A. Djunuddin,

Reportof the StateEarthquake Investigation Commission, 451 pp.,Carnegie Ins.of Washington, Washington, D.C., 1908. Cameron, N., et al.,Geology of theTakengon quadrangle, Sumatra, Le Pichon,X., and J. C. Sibuet,Passivemargins:A model of formation, J. Geophys. Res.,86, 3708-3720,1981. report Geol. Res. and Dev. Cent., Bandung, Indonesia, 1983. Chesner, C.A.,W. I. Rose, A. Deino, R. Drake,andJ. A. Westgate, Matson, R., and G. Moore, Su'ucturalinfluences on Neogene

The geo.logical evolution ofnorthern Sumatra, paper presented at

Jakarta, 1980.

detroitde la Sonde (Indonesie): Donnees de la sismique reflexion (campagne Krakatau), Geophysics, 309(2), 205-212, 1989. Lawson,A., et al., The CaliforniaEarthquake of April 18, 1906,

Proc. 9"Indonesian Petroleum Association Annual Convention,

Eruptive history of Earth's largest Quaternary caldera (Toba,

Indonesia) clarified, Geology, 19,200-203,199 !.

subsidence in the centralSumatrafore-are basin,in Geologyand

Curray, J.,D. Moore, L. Lawyer, F. Emmel, R. Raitt, M. Henry, and

R.Kieckhefer, Tectonics of theAndaman SeaandBurma, AAPG Mem., 29, ! 89-198,1979.

Geophysics of Continental Margins, edited by J. Watkins,F. Zhiqiang, and K. McMillen,pp. 157-181, Am. Assoc. of Pet.
Geol., Tulsa, Okla., 1992.

McCaffrey, R., Slipvectors andstretching of theSumarran forearc,

Geology, 19, 881-884,1991.

Deplus, C., et al.,Direct evidence of active deformation in the

eastern Indian oceanic plate,Geology, 26, ! 31-134,!998. Detourbet, C., O. Bellier,andM. Sebrier, La caldera volcaniqiae de

W. Stevens, S.S. O. Puntodewo,and C. Subarya, Strain SPOT, C.R.Acad. Sci.,Ser.Ii, 316, 1439-1445, 1993. partitioning during oblique plate convergence in northern Diament, M., H. Harjono, K. Karta,C. Deplus, D. Dahrin, M. T. Zen Sumatra: Geodetic and seismologic constraints and numerical Jr.,M. Gerard, O. Lassal, A. Martin,andJ. Malod,Merttawai modeling, J. Geophys. Res.,thisissue. fault zoneoff Sumatra:A newkey to thegeodynamics of western McCarthy, A., and C. Elders,Cenozoic deformation in Sumatra: Indonesia, Geology, 20, 259-262,1992. Oblique subduction andthedevelopment of theSumatran fault Durham, J., OeloeAer fault zone,Sumatra, Bull. Am. Assoc. Pet system, in Petroleum Geology of Southeast Asia,edited by A. J. Geol., 24, 359-362, 1940. Engdahl, E., R. vanderHilst,andR. Buland, Global teleseismic Fraser,S. J. Matthews,and R. W. Murphy, Geol. Soc. Spec.

Toba et le systeme de faille de Sumatra (Indonesia) vuepar

McCaffrey, R., Oblique plateconvergence, slip vectors, andforearc deformation, J. Geophys. Res., 97, 8905-8915,1992. McCaffrey, R., P. Zwick,Y. Book, L. Prawirodirdjo, I. Genrich, C.

earthquake relocation withimproved travel times and procedures Publ., 126, 355-363, 1997. of sedimentary fordepth determination: $eisrnol. Soc.Am.Bull.88, 722-743, McKenzie,D., Someremarkson the development

basins, Earth Planet.Sci.Lett.,40, 25-32, 1978.

P., andP. Tapponnier, Cenozoic tectonics of Asia: Effects Fauzi, R. McCaffrey, D. Wark, Sunaryo, and P. Y-PrihHaryadi, Molnar,
Lateral variation in slab orientation beneath Toba caldera,

of a continental collision, Science, 189, 419-426, 1975.

Muller, J. J.A., De verplaasting vaneenige traiangulatie pilaren in de northern Sumatra, Geophys. Res. Lett.,23, 443-446,1996. residenile Tapanuli (Sumatra) tengevolge de aardbeving van 17 Fitch, T. J., Plate convergence, transcurrent faults,and internal Mei ! 892,Natuurwet. Tijdscht. Ned.Indie.,54, 299-307, 1895. deformation adjacent to southeast Asiaandthewestern Pacific, J. Natawidjaja, D., Quantitative geological assessments of Liwa Geophys, Res.,77,4432-4462,1972. earthquake 1994, paper presented at AnnualConvention, Genrich, J.,Y. Bock, R. McCaffrey, L. Prawirodirdjo, C. Stevens, S. Indonesian Assoc.of Geophys. (HAG!), Bandung, Indonesia, S.O.Puntodewo, C. Subarya andS. Wdowinski, Distribution of 1994.

slip atthenorthern Sumatra faultsystem, J. Geophys. Res., this


Natawidjaja, D., Y. Kumoro, and J.Suprijanto, Gernpa bumi tektonik

di daerah Bukittinggi- Muaralabuh: hubungan segmentasi sesar aktifdengan gempa bumitahun 1926dan!943,paper presented at AnnualConvention, Geoteknologi-LIPI, Bandung, Indonesia,

Gordon, R., C. Demets, and D. Argus,Kinematic constraints on

distributed lithospheric deformation in theequatorial Indian ocean from present motion between the Australian andIndian plates:
Tectonics 9, 409-422, 1990.

K. R, and W. R. McCann, Seismichistory and Harbury, N., andH. Kalagher, The sunda outer-arc ridge,north Newcomb, seismotectonics of the Sunda Arc, J. Geophys. Res.,92, 421-439, Sumatra, Indonesia, J. Southeast AsianEarthSci.,6(3/4),4631987.
476, 1991.

Harjono, H., M. Diament, J.Dubois, and M. Lame, Seismicity of the

Nishimura, S., Re-examination of the fission-track ages of volcanic

Indonesian IslandArcs, editedby S. Nishimura, pp. 148-153, Kyoto Univ.., Kyoto, Japan, 1980. Harris, R.,and S.Day,Dynamics offault interaction: Parallel strike- Nishimura,S;, J. Nisida, T. Yokoyama, and F. Hehuwat, slip faults, J. Geophys. Res., 98,4461-4472, 1993. Neotectonis of the Strait of Sunda,Indonesia,J. Southeast Asia

Sunda Strait: Evidencefor crustalextensionand volcanological implicationa, Tectonics, 10, 17-30,1991.

ashes and ignimbrites in Sumatra, in Physical Geology of

Harris, R., R. Archuleta, andS. Day,Faultsteps andthedynamic Earth Sci., !, 81-91, 1986. rupture process: 2-D numerical simulations of a spontaneously Pacheco, J.,and L. Sykes, Seismic moment catalog of large, shallow propagating shear fracture, Geophys. Res.Lett.,18, 893-896,

earthquakes, 1900-1989, Bull, Seismol. Soc. Am..,82,1306-1349,

Jeffcry, W. Keats, and J. Thaib, A review of themain structural and magmatic features ofnorthern Sumatra, J. Geol. Soc. London,

1992. Ituchon, P., and X. Le Pichon,SundaStrait and central Sumatra Page, B. (3.N.,J.D.Bennett, N. R. Cameron, D. M. Bridge, D. H. fault, Geology, 12,668-672,1984.

Jarrard, R.,Terrane motion by strike-slip faulting of fore-arc slivers,

Geology, 14, 780-783,1986.

136, 569-579, 1979. Jennings, C., andG. Saucedo, Faultactivity mapof California and G., L. Weeks, andR. Bums, A reconnaissance geophysical adjacent areaswith locations and agesof recentvolcanic Peter, survey in the Andaman Sea and across the Andaman-Nicobar eruptions, Calif.Div. of MinesandGeol.,Sacramento, 1994. Island area, J. Geophys. Res., 71,495-509, 1966. Karig, D., S. Suparka, G. Moore, and P. Hehanussa, Structure and Cenozoic evolution of the Sunda arc in the central Sumatra region, AAPGMere.,29, 223-237, 1979.

Pramumijoyo, S., andM. Sebder, Neogene andQuaternary fault

AsianEarth Sci.,6, 137-145, 1991.

kinematics around the Sunda Straitarea,Indonesia, J. Southeast

Karig, D., M. Lawrence, G. Moore, and J. Curray,Structural L., A geodetic study of Sumatra and theIndonesian framework of the fore-arc basin, NW Sumatra, J. Geol.Soc.of Prawirodirdjo, region: Kinematics and crustal deformation fromGPS and London, 137,77-91,!980. Ph.D. Thesis, Univ. ofCalif., San Diego, 2000. Katili, J.,andF. Hehuwat, On theoccurence of large transcurrent triangulation,





Prawrodirdjo,L. et al., Geodetic observations of interseismicstrain

segmentation at the Sumatra subduction zone: Geophys. Res.Lett.

24, 2601-2604, 1997.

Tregoning, P., F. K. Brunner, Y. Bock,S.S. O. Puntodewo, It. McCaffrey, J.F. Genrich, E. Calais, J.Rais, and C. Subarya, First

Prawirodirdjo, L., Y. Bock,J. Genrich,S.S. O. Puntodewo, J. Rais, C. Subarya,and S. Sutisna, Onecentury of tectonic deformation

geodetic measurement of convergence across theJava trench,

Geophys. Res.Lett.,21, 2135-2138, 1994.

System surveys, J. Geophys. Res.,thisissue. Prentice, C. Earthquake Geology of theNorthern SanAndreas Fault
near Point Arena, California,Ph.D. thesis,CaliforniaInstituteof Technology, Pasadena, 1988. Reid, H., Sudden earth movements in Sumatra in 1892, Bull.

Untung, M., N. Buyung, E. Kertapati, Undang, andC. Allen, along theSumatra faultfromtriangulation andGlobal Positioning Rupture along theGreat Sumatran fault,Indonesia, during the
earthquakes of 1926and1943,Bull.Seismol. Soc. Am.,75,313317, 1985.

Van Bemmelen, R., TheGeology of Indonesia, 732 pp.,Gov. Print.

Off., The Hague,Netherlands, 1949.

Visser,S., Inland and submarine epicentra of Sumatra andJava

Seismol. Soc.ofAm.Bulletin, 3, 72-79, 1913. earthquakes: Koninklijk Magnetischen Meteorologisch Observatorium te Batavia, 9, 1-14, 1922. Research Groupfor ActiveFaults, Main Active Faultsin andAround Visser, S., De aardbevingen in de Padangse Bovalanden, Natuurwet. Japan, University of TokyoPress, Tokyo,1980. Tijdschr. Ned. Indie.,87, 36-71, 1927. Rock, N.M. S., D. T. Aldiss,J. A. Aspden, M. C. G. Clarke,A. Weeks,L., R. Harbison, and G. Peter,Island arc system in the Djunuddin, W. Kartawa, Miswar, S. J. Thompson,and R. Andaman Sea, Am.Assoc. Pet. Geol.Bull.,51, 1803-1815,1967. Whandoyo,The geologyof the Lubuksikaping quadrangle, J.,Eruptions of acidpumice tuffsandrelated phenomena Sumatra, report, 60 pp. and map, Geol. Res. and Der. Cent., Westerveld, Bandung, Indonesia, 1983. along theGreat Sumatran fault-trough system, in Proceedings of Samuel, M. A., and N. A. Harbury,The Mentawaifault zoneand thePacific Science Congress, vol. 2, Pacific Science Congress,
deformationof the Sumarranforearc in the Nias area, in Tectonic
New Zealand, 1953.

Evolution of Southeast Asia, editedby R. Hall and D. J. Blundell, Geol.Soc.Spec.Publ. 106, 337-351, 1996. Samuel, M., N. Harbury,A. Bakri, F. Banner, andL. Hartoho, A new stratigraphy for the islands of the Sumarran forearc,Indonesia, J.
Asian Earth Sci., 15, 339-380, 1997.

Widiwijayanti, C., J. Deverchere, R. Louat, H. Harjono, M. Diament,

andD. Hidayat,Analysis of the aftershock sequence of theMw 6.8 Liwa earthquake, Indonesia, Geophys. Res. Lett., 23, 30513054, 1996.

Saroglu, F., O. Entre,andI. Kuscu,Activefault mapof Turkey,Gen. Dir. of Miner. Res.andExplor.(MTA), Ankara, Turkey,1992. Sieh,K., Slip alongthe SanAndreas fault associated with thegreat 1857earthquake, Bull. Seisrnol. Soc.Am., 68, 1421-1428,1978. Sieh,K., Y. Book,andJ. Rais,Neotectonic andpalcoseismic studies in west and north Sumatra,Eos Trans. AGU 72(44), Fall Meet. Suppl., 460, 1991. Sieh, K., J. Zachariasen, Y. Bock, L. Edwards, F. Taylor, and P. Gans, Active tectonicsof Sumatra, Geol. Soc. Am. Abstr. Programs, 26, A-382, 1994. Smith, W. H. F., and D. T. Sandwell, Global seafloortopography from satellitealtimetryand ship depthsoundings, Science, 277,
1956-1962, 1997.

Yeats,R., K. Sieh,andC. Allen, The Geology of Earthquakes, 568 pp., OxfordUniv. Press, New York, 1997. Zachariasen, J., and K. Sieh,The transfer of slip between twoen echelon strike-slip faults: A casestudyfrom the 1992Landers earthquake, southern California, J. Geophys. Res.,100, 15,28115,301, 1995.

Soetarjo, U., E. Arnold,R. Soetadi, S. Ismail,andE. Kertapati, Ser. Seismol., vol. 5, 199 pp., Indonesia, Southeast AsianAssoc. of Seismol.and Earthquake Eng., (SEASEE), SeriesEditor and Program Coordinator: E.P.Arnold, USGSDenver, Colo.,1985. Stehn, C., Explosionen desPematang Batain der Suoh-senke (sud Sumatra) im jahre 1933,Natuurwet. Tijdschr. Ned.Indie,94, 4668, 1934.

Zachariasen, J., K. Sieh, F. W. Taylor, R. L. Edwards,andW. S. Hahtoro,Submergence and uplift associated with the giant1833 Sumatran subduction earthquake: Evidence from coral microatolls, J. Geophys. Res.,104, 895-919, 1999. Zen, M., Jr., D. Dahrin, M. Diament, H. Harjono, K. Karta,C. Deplus, M. Gerard, O. Lassal, J. Malod, and A. Martin, Mantawai-90 cruiseresult: The Sumatra obliquesubduction and strike slip fault zones,in Geodynamic Processes in the Forearc SIiver Plate and General Topics,editedby H. Prasetyo, p. 46, !ndonesian Assoc.of Geophys., Bandung, 1991.

D. Natawidjaja andK. Sieh,Seismological Laboratory, California Institute of Technology, Pasadena, CA 91125. (danny @gps.caltech.edu; sieh @gps.caltech.edu)

Tapponnier, P., and P. Molnar,Active faultingand tectonics in

China, J. Geophys. Res.,82, 2905-2930,1977. Tikoff, B., Sunda-style tectonics and magmatic are processes, Eos Trans. AGU, 79(45),Fall Meet. Suppl., F222, 1998.

Tjia, H., Tectonic depression alongthe transcurrent Sumatra fault

zone,Geol. Indonesia, 4(1), 13-27, 1977.

(Received April 21, 1999;revisedMarch 3, 2000; accepted April 7, 2000.)