Plate Convergence, Transcurrent Faults, and Internal Deformation

Home , Plate theory (volcanism), Sagami Trough

VOL. 77, NO. 23 JOURNAL OF GEOPHYSICAL RESEARCH AUGUST 10, 1972

Plate Convergence,Transcurrent Faults, and InternalDeformation Adiacentto SoutheastAsia and the WesternPacific

THOMAS J. FITCI-I 2

Lamont-Do.herty GeologicalObservatory o.• Columbia University,Palisades, New York 10964

A model for oblique convergencebetween plates of lithosphere is proposed in which at least a fraction of slip parallel to the plate margin results in transcurrent movements on a nearly vertical fault located on the continental side of a zone of plate consumption.In the extreme caseof complete decoupling,only the componentof slip normal to the plate margin can be inferred from underthrusting.Recent movementsin the western Sunda region provide the most convincingevidence for decouplingof slip, which in this region is thought to be oblique to the plate margin. A speculative model for convergencealong the margins of the Philippine Sea is constructedfrom an inferred direction of oblique slip in the Philippine region. This model requires that the triple point formed by the junction of the Japanese and Izu-Bonin trenches and the Nankai trough migrate along the Sagami trough. Absence of an offset between the trenches can be explained by rifting within the Izu-Bonin arc. Transcurrent movements in southwest.Japan and the Tai.wan region suggest that similar movements within well-developed island arcs are initiated during a nascent stage in the development of such plate margins and are often imposed on a pre-existing zone of weakness. The bulgelike configurationof the western margin of the Andaman basin is explained as a result of interarc extension within the western part of the basin. In the Indian Ocean plate south and west of this basin, strike-slip movements inferred from four mechanism solutions, two o.f which are reported in this work, suggest• distribution of stressin which axes of both maximum and minimum compressi.onare nearly horizontal, the stress maximum trending west of north. The larger recent earthquakes in the eastern Sunda region have resulted in deformation within the arc rather than shallow underthrusting at the inferred plate margin. A narrow region between the Celebes basin and the Philippine trench, including the Talaud Islands and a northern extensi.on of the Molucca basin, contains several zones of shallow thrusting. Mechanism solutions show that these movements trend E-W consistent with shortening normal to the arc-related structures in the region. Recent movements near the east coast of Luzon reveal a zone of underthrusting along the inner wall of a trough, marginal to the coast. There is an apparent hiatus in underthrusting between this zone and the one that follows the inner wall of the Philippine trench.

Several of the earth's most extensive zones sibly the Nicaragua trough [Maldo•ada-Koer- of transcurrentfaulting are locatedwithin and dell, 196.6].With the exceptionof the Atacama strike parallel to active island arcs and arclike [Arabasz and Allen, 1970], the other fault structures.Allen [1965] discussedevidence for zonesshow evidenceof recent strike-slipmove- movements on three such faults: the Atacama ments contemporarywith plate consumptionin in Chile, the Longitudinal in Taiwan, and the the sameregion. In the work presentedhere such Philippine. To this list can be added the a systemof contrastingmovements is explained Semangkofault in Sumatra [e.g., Katili, 1970], by a model of convergencein which slip that is pbssiblythe median tectonic line in southwest obliqueto the plate margin is at least partially Japan [Karneko,1966; Okada, 1971], and pos- decoupled between parallel zones of transcurrent faulting and underthrusting.A similar x Lamont-Doherty Geological Observatory con- model is proposedby Karig and Mammerickx tribution 1830. [1972] to explain an en echelon distribution 2 Now with Department of Geophysics and Geochemistry, The Australian National Univer- of oceanictroughs east of the New Hebrides sity, Canberra, Australia. arc south of 12øS. A schematic cross section of a decoupled Copyright • 1972 by the American Geophysical Union. plate margin is shown in Figure 1. The trans-

4432 •)LATE CONVERGENCEAND TRANSCURRENT FAULTS 4433

Asthenospher• • Asthenosphere

Fig. 1. Vertical sectionof a well-developedisland arc in a region of oblique convergence. Decoupling hypothesisis illustrated by transcurrent movement on a vertical fault adjacent to the center of active volcanism. current fault is near an axis of a chain of active either the vertical or the inclined fault are volcanoes,as the Philippine and Semangko obtained by equating horizontal shear stress faults are (Figure 2). One mechanismsolution to the product of the normal stress and the for a recent earthquake shows evidence for coefficientof friction •: strike-slipmovement [Mo.l•ar and Sykes,1969] alongthe Nicaragua trough, a volcanic gra.b•n Vertical fault thatis nearlyparallel to anadjacent section F(cosa- /zsin a) - /zNH (1) of the Middle America trench. Coexistence of activevolcanism and transcurrent. faulting is Inclinedfault extensionunderstandablecan alsoin bethat weaka zone in horizontalof weakness shear,in F(cos a -- /zsin a sin$)sin $ - /zNH (2) and vice versa. A zone of weakness of a dif- where N• is the contribution to normal stress ferent origin can be inferredfor right-lateral from the hydrostaticload. When (1) > (2), movementsalong the median tectonic line horizontalshear occurson the vertical fault [Kaneko, 1966; Okado, 1971]. These move- rather than on the inclinedfault. When a dip mentsparallel 'a zoneof underthrustingalong of 30ø is assumedfor the inclinedfault, the the inner wall of the Nankai trough [e.g., inequalitysimplifies to cosa • (3 • sin e•)/2, Fitch and Scholz, 1971]. In place of recent in whichcase angles of obliqueslip as largeas volcanism,of whichthere is nonein this part 45ø satisfythe inequal!ty,if it is assumed of Japan[e.g., Matsuda e't al., 1967],narrow that • is about 0.7. This argumentis only metamorphicbelts, considered relics of Meso- slightlyweakened by usingan expressionfor zoicplate margins [Miyashiro, 1967], provide obliqueshear rather than horizontalshear in a zone of weakness conducive to horizontal (2). shear. Of the four regions of oblique convergence Decouplingof obliqueslip in the manner revealedby modelsbased on the 'assumption proposedhere is favoredin manyreal situa- of rigid plates[e.g., Le Pichon,1968], one tions becausea nearly verticalsurface con- servesas a type exampleof a plate margin centrates horizontal shear more effectively suchas that illustratedin Figure1. The region than an inclinedsurface of equal or greater is the westernSunda, where horizontal shear strength.This argumentis only weaklyde- andunderthrusting occur concurrently on the pendenton the angleof obliqueslip. Consider Semangkofault and alongthe inner wall of the simplemodel in Figure3, in whichuniform the Java trench,respectively. Directions and horizontalforce per unit area F in the direc- rates of convergencein this regionare com- tionof the slipvector is appliedto a.plate mar- puted from a simplifiedmodel that doesnot gin. Simpleconditions for horizontalshear on includeslip betweenthe Chinaand Eurasian 4434 THOMAS J. FITCH nearly parallels the island arc [Gumper, 1971].

60N Similarly, underthrustingnorth of Puerto Rico parallelsthe plate margin in this regionmarked by the Puerto Rico trench. Near New Guinea, slip between the major plates is obscuredby movements between small intervening plates [Johnson and Molnar, 1972] and by possible internal deformation in western parts. of this region. A model for decoupled slip similar to that proposedfor the western Sunda arc is used to infer a direction of oblique convergencefor the Philippine arc. From this inference, although it is tenuous,a model is constructedfor o convergence along the eastern and western

lOS margins of the Philippine Sea. For the western margin, directions and rates of convergence comefrom recent movements (measureddirectly or inferred from seismic evidence or both) Fig. 2. Plate boundaries and major transcurrent faults in the Far East. Solid boundaries are within selected zones of shallow earthquakes. those that are well defined by recent shallow earth- The spreading history of the mid-oceansystem quakes [e.g., Bc•razangi•d Dorm•n, 1969]. Dashed of active ridgescannot be used,because no part curves extrapolate between well-defined bound- of this system passes through the Philippine aries. Dotted curves represent traces of known Sea. transcurrent faults of great length. MTL, median tectonic line; LF, Longitudinal fault; PF, Philip- No attempt is made to include specialregions pine fault; SF, Semangko fault; ST, Sagami such as interarc basins [Karig, 1970] in these trough. Plate margins east o.f New Guinea are models. Active extension within several of the from Johnso.n c•nd Molnar [ 1972]. basins marginal to the Philippine Sea and southeastAsia probably influencesthe rate and plates (Figure 2). In the absenceof a quan- direction of convergencealong adjacent frontal titative measure of this slip, large corrections arcs. The western part of the Andaman basin to slip vectorsin the Sundaregion are possible; (Figure 4) contains one or more centers.of however,such correctionsare unlikely because spreadingthat became active for the first time computed centers of rotation for the Indian Ocean and Eurasian (including China) plates are nearly 90 ø distant from this arc. The other regionsof oblique convergenceare the western Aleutians [e.g., McKenzie and Parker, 1967], northern New Guinea [e.g., Le Picho.n,1968; Morgan, 1972], and that part. of the Caribbean margin adjacent to Puerto Rico [Molnar and Sykes,1969]. The plate margin in each of these regions does not fit the model proposed here for the western Sunda arc without major modifications.Seismic evidence shows that at least part of the western Aleutian arc is boundedby active zonesof faulting, underthrusting seaward of the volcanic islands, and right-lateral shear on the continental side of these islands [Gumper, 1971]. However, un- Fig. 3. Schematic diagram of a zone of oblique like most shallowunderthrusting [Isacks et al., convergence. (Top) Map view. (Bottom) Vertical 1968], these movementsare in a direction that cross section. Large arrows are slip vectors. PLATE CONVERGENCEAND TRANSCURRENTFAULTS 4435 in the late Miocene [Rodol/o, 1969]. New sea shallow earthquakes are an essential part of floor is forming in the basin west of the Mar- this evidence.In additionto the new solutions, iana ridge and within the Izu-Bonin arc [Korig, those published by Katsumata and Sykes 1971b]. A great thickness of sedimentary fill [1969] for the Ryukyu, the Izu-Bonin, and suggeststhat the Okinawa trough, west of the the Mariana regionsand by Fitch [1970a, b] Ryuku ridge, is only marginallyactive [Karig, for the Indonesian-Philippineregion are used. 1971a]. Normal faults offsettingthese sediments Each new solution is based on a distribution of [Wageman et al., 1970], heat flow through the P-wave first motions and S-wave polarizations floor of the trough that is anomalously high shownin the appendix.A tabulation of param- [Watanabe et al., 1970], and a mechanismsolu- eters for the new solutions as well as those of tion for normal faulting near the northeast cor- Fitch [1970a] is also in the appendix.Detailed ner of this trough (solution32 in Figures 12 and seismicitymaps based on epicenterscomputed 13 of this paper) are consistentwith active ex- by the U.S. Coast and Geodetic Survey tension. (USCGS) are provided for some of the more The arguments presentedin this work rely complex seismiczones within this region. The strongly on recordedas well as.historic seismic model for plate convergencein the northwest evidence.Forty-six new mechanismsolutions for Pacific followspresentation of these data.

I00' E I10' E 120'E 130'E 140' E /

LUZON S.E. Philippine ASIA Sea Basin

IOøN IO'N

MINDANAO

Trench Celebes Talaud Basin Horst

BORNEO Geelvink 0ø

NEW rait ,/' JAVA Banda Basin

IO'S IOoS

Wharton Basin

..; AUSTRALIA

IO0øE IIO'E 120øE 130'E 140'E

Fig. 4. Major fault zones adjacent to southeast Asia. Shading and striping mark regions of known or suspected extension and shortening, respectively. Locations of transcurrent faults in Celebes and Java and the sense of motion on them are from Ka. tili [1970]. Saw-tooth curves mark zones of shallow underthrusting for which there is at least some seismicevidence. Smooth continuations of these curves represent zones of underthrusting along which at most only scattered concentrations of shallow earthquakes were recorded in recent years. Thin curves, dashed where there is little supporting evidence, mark fault zones that may or may not be part of a plate margin. Directions of relative motion between the major plates computed from rigid plate models are given by solid lines [Morgan, 1972] a.nd dashed lines [Le Pichon, 1968]. 4436 THOMAS J. FITCH PLATE CONVERGENCE AND these events, which was probably the largest, INTERNAL DEFORMATION horizontal offset at the fault, inferred from Western Sunda arc. By analogy with other geodetic evidence, was as large as 4 meters island arcs, a line of andesiticvolcanism in (Richter [1958], from earlier work by H. F. Sumatraand an adjacentdeep-sea trench result Reid). In addition, other right-lateral offsetsof from consumptionof oceaniclithosphere. The recent origin are known from numerouslocations Andaman-Nicobar ridge, which marks a north- throughout the fault zone [Katili, 1970]. A ward continuationof this arc, has stratigraphic comparisonwith seismicityin other parts of this and structural evidencefor a tectonic history in regionfrom lists of recordedearthquakes of large the Cenozoic similar to that of the nonvolcanic magnitude [Gutenberg a•d Richter, 1954; arc between Sumatra and the Java trench Rothe, 1969], as well as with recent seismicity [Rodolfo, 1969]. At shallow depths between [e.g., Barazangi and Do,rman, 1969], shows Sumatra and the nonvolcanic arc there are four that most shallow activity is seaward of the earthquakeswith mechanismsconsistent with island, whereas, in the Andaman region, earth- shallow underthrustingof the oceanic litho- quakes are concentrated along a submarine spherebeneath this arc (solutions7, $, 9, and 10 continuationof the Semangkofault. Qualitatively in Figure 5). A great earthquakeon June 26, this migration in activity toward the concave 1941, the largest recordedearthquake in the side of the arc is consistent with an increase in western Sunda region, may have been a mega- the ratio of parallel slip to normal slip toward thrust. It was located west of the Andaman the NW (Figure 4), provided that these com- Islands and appearsto have had a subcrustal ponentsof slip are decoupledin a manner simi- focus of about 50-60 km [Richter, 1958]. Solu- lar to that proposedhere. tions to eventsat intermediatedepths (solution Figure 5 showsmechanism solutions (solutions 1 in Figure5 and solutions2 and 3 in Fitch a•d 2, 3, 4, and 80) for four shallow earthquakes, Mol•r [1970]) are consistentwith a descending three of which are located near narrow troughs slab of lithospherebeneath Sumatra within (Figure 6) that mark a continuationof this fault which the axis of least compresslyestress is zone for a distance of approximately 500 km oriented in a down-dip direction. Although along the east side of the Andaman-Nicobar computeddepths for earthquakesbeneath the ridge. Fitch [1970a, b] interpreted solution 2 western arc are in general poorly controlled, as a shallowunderthrust; however,rightdateral recent evidence indicates a maximum depth of shear in this region,seismic evidence for which about 200 km for earthquakesbeneath Sumatra comesfrom solutions3, 4, and 80, suggeststhat and 100-200 km for earthquakes beneath the the nodal plane with a NW strike and a nearly Andaman-Nicobar ridge [e.g., Fitch, 1970a]. verticaldip is morelikely to representthe fault Parallelingthis zone of underthrustingis the plane. Motion on this plane was right-lateral Semangkofault in Sumatra (Figure 2) and its with a large componentof thrusting in the submarinecontinuation along the easternside of sense of the Andaman-Nicobar ridge over- the Andaman-Nicobar ridge [Rodolfo, 1969]. ridingthe Andamanbasin. Field evidenceshows This fault zone is seismically one of the more that the interpretationof solution3 given in active zones of transcurrent movements within Figure 5 is correct.This solutionis for the active island arcs. Offsets in the right-lateral Atjeh earthquakeof April 1964, after which senseare known in focal regions of the larger right-lateral offsetsas large as 0.5 meter were earthquakesrecorded in Sumatra [Katili, 1970]. observedalong a sectionof the Semangkofault These events included the 1892 Tapanuli, the in extreme northern Sumatra [Katili, 1970]. 1926 Padang-pandjang(listed by Gutenberg Internal deformationadjaceat to the western aad Richter [1954] as two shockswith an Ms of Suada arc. Contrasting distributionsof stress 6•/• and 63/i), the 1933 Liwa (given an Ms are inferred from mechanism solutions for shal- of 7•/• by Gutenbergand Richter [1954]), the low earthquakeson oppositesides of the western 1952 Tes (not listed by Gutenbergand Richter Sundaarc. A pair of thrust movementsalong [1954] and thereforeprobably having an Ms of the northern coast of Sumatra near the northern < 6•/•), and the 1964Atjeh (robof 6.7 computed tip of the island(solutions 5 and 6 in Figure5) by the USCGS) earthquakes.For the first of are similar to solutionsreported in recent years PLATE CONVERGENCE AND TRANSCURRENT FAULTS 4437 from the continental sides of other island arcs oceanic side of the Sunda arc is such that the [e.g.,Isacks et al., 1968].These movements show axesof both maximum and minimum compres- that, behindat least somearcs, compression is sion are nearly horizontal, the minimum stress largeenough to causecrustal shortening nearly being in a NE-ENE direction. This distribution normalto the adjacentplate margin.Although is inferred from nearly identical solutionsto four the originof suchstress is unknown,it is prob- widely separated earthquakes in the Indian ably associatedwith consumptionof oceanic Ocean plate. This unusual seismiczone is seen lithospherealong the frontal arc. as a broad band of scatteredactivity extending A tentative distribution of stress for the from Australia to India [Sykes, 1970]. Of the

I00ß 120ø • I '•,•.... ! i:.,•I 144 3.•_1.••'-•:•00 I

....ß '•' • 42 • 4945 52 -

• •'• •'"• 60 27 2 5

I I i I00 • 120ø 140• Fig. 5. Mechanismsolutions for recent shallowearthquakes in the Indonesian-Philippine region adapted from Fitch [1970a,Figure 10]. Open circle with oppositelydirected arrows gives epicenterand horizontal projectionof axis of minimum compression(the T axis) for recent normal faulting. A direction of slip is inferred from thrust-type solutionsin which one nodal plane is nearly vertical. This plane is assumedto be the auxiliary plane [e.g., McKenzie and Parker, 1967]. If both nodal planeshave nearly the same dip angle, the axis of maximum compressionis plotted. Closed circle and arrow gives epicenter and horizontal projectionof directionof slip. The senseof movementis' such that oceaniclithosphere moves in the direction of the arrow relative to an observerfixed to an adjacent island arc. Closed circle with opposingarrows gives epicenterand horizontal projection of axis of maximum compressivestress (the P axis) for recentthrust faulting. Strike-slip movements are illustrated by shearsymbols. The centralAndaman trough and rift valley [Rodoleo,1969] are represented by parallelsolid and dashedcurves. Regions enclosed by a solidcurve are prominentbathy- metric depressions(either deep-seatrenches or troughs). Solutions1-44 are from Fitch [1970a].Solutions 45-80, for whichdetails are givenin the appendix,are new solutionsmainly from 1968 to mid-19?0..Solutions9, 11, and 12 in Andaman region are from Fitch [1970b]. Solutions 1-5 in western New Guinea. and solution 20 near the north end of the Yap trench are from Johnsonand Molnar [1972] and Katsuma,ta and Sykes [1969], respectively.Solutions 5 and 6 in northern Sumatra and 31, 37, 39, 40, a.nd44 in the Philippine region were improved by using additional S-wave polarization angles and additional P-wave polarities. These improved solutionsare also given in detail in the appendix.Solutions I and 11 (solid squares) are for earthquakesat depths that are transitional from shallow to intermediate activity. With reference to the surface of the earth these solutions show normal faulting, one nodal plane having a steepdip. Such solutionsare consistentwith down-dip orientation of the axis of minimum stresswithin the descendinglithosphere in this region [Fitch and Molnar, 1970]. 4438 THOMAS J. F•Tc• An understandingof such a tectonicprovince o is soughtby analogy with interarc basinsthat have evolvedin specialregions in which there is a mechanism for the release of compressive 15N/ dß 15Nstress [Scholz et al., 1971]. In the western o Sunda region, compressionthat is proportional to the current rate of plate convergencewill reach a minimum value in the Andamanregion. Such a minimum is consistent with active extensionin the westernAndaman basin. Similarly, Scholzet al. [1971] suggestthat extensionin the ION ION Basin and Range province of the westernUnited Stateswas triggeredby a releaseof compressive stress that accompanied the formation of the San Andreas fault in the late Tertiary and the progressivedisappearance of a pre-existingisland arc off the coast of California [e.g., Atwater, 1970]. If this interpretation is applied to the SUMATRA western Andaman basin, the submarine con- 5N 5N tinuation of the Semangko fault is a counter-

91E IOOE Fig. 6. Recent shallow seismicity and bathy- merry of the Andaman region. Open and closed circles are epicenters computed by the USCGS from arrival times at more than 20 and 10-20 stations, respectively, for 1961-1970. Bathymetry (in meters) is from Rodoleo [1969, Plate 1]. Solid lines in northern Sumatra map fault traces within the Semangko fault zone [Katili, 1970].

events that yielded mechanismsolutions, one is W-- E located seaward of the northern end of the Java trench (solution65 in Figure 5), another is located on the Ninety-East ridge [Sykes, 1970], and a third is the Konya earthquake in western India [Sykes, 1970]. The remaining solution (Figure 7) is for an event located west of the Ninety-East ridge in an area covered by sediments marking the southernlimit of the Ganges cone [Heezen and Tharp, 1967]. OCT I0 1970 5.6S 86.2E 50 KM Behind the Andarnan-Nicobar ridge in the Fig. 7. Mechanism solution for shallow earth- western Andaman basin, geomorphic [Weeks quake in the Indian Ocean. Nodal planes are given e't al., 1967] and seismic [Fitch, 1970b; this by the heavy curves. Polarities of P-wave first paper, Figure 5] evidence shows that active motions are given by open circles for clear dilatation, closed circles for clear compression, open extension along the central Andaman trough triangle for weak dilatation, closed triangles for [Rodoleo, 1969] trends WNW-NW. However, weak co.mpression,and crossesfor weak arrival of the rate is not fast enoughto showthinning of uncertain polarity. Arrows give S-wave polariza- sedimentswithin the zoneof faulting that forms tion. Crosses enclosed in circles are poles to the the trough [Weeks et al., 1967]. An additional nodal, planes. T and P stand for axes of minimum and maximum compression,respectively. Within mechanism solution (solution 12) is consistent each set of parentheses is an azimuth and a with normal faulting on either a steep or a plunge. Data are plotted on an equal-area projec- shallowfault near the NW margin of this basin. tion of the lower hemisphere of the focal sphere. PT,ATE CONVERGENCE AND TRANSCURRENT FAULTS 4439 part of the San Andreas fault. However, this o o speculationimplies a greater rate of plate iI•, • • ... BORNEO' convergencein the past between Burma and Sumatra. No substantial evidence was found to supportor refute this implication. Eastern Sundaarc. Figure 4 showsthe Sunda Strait as a tectonic as well as a physiographic break in the arc. West of the strait the southern 0 end of the Semangkofault (Figure 8) shows evidence of recent movement in the right- lateral sense.Katili [1970] gives some evidence lOS o o * lOS for left-lateral movements on a fault in western Java that parallelsthe arc. Unlike the Semangko I I I I I I I I t I I I ! ! iOOE 115E fault, this fault is not part of an extensivezone of transcurrent movements. Fig. 8. •ecent •ha]]ow seismicity near 8und• A concentration of shallow earthquakes in 8[r•it. Ope• and c]o•ed circles are epicenter• computed by the USCGS from arrival times at more the vicinity of the strait (Figure 8) is further than 20. and 10:20 stations, respectively, for 1961- evidence for a tectonic break in this region. 1970. Heavy curves represent traces of major This activity is exceptional in that adjacent faults identified by Ka.tili [1970]. parts of the arc have not experiencedhigh levels of activity in recent years. Possiblythis activity results from extension on the seaward side and West of Sumba (Figure 4) the plate margin convergenceon the landward side of the strait follows the Java trench, along whoseinner wall as suggestedby the opposite senseof motion a zone of shallow earthquakes marks the fric- on faults adjacent to the strait (Figure 4). tional contact between the plates [e.g., Fitch, A steplikeincrease in the depth of the deepest 1970a]. East of Sumba this margin follows a earthquakes from 200 km west of the strait to narrow trough that makes a loop around the •600 km ca.stof it suggeststhat the rate of Banda basin [e.g., Chase and Menard, 1969, underthrustingin this regionhas a similar step- Chart HO 5990]. It passes north of Ceram like increase [Fitch and Molnar, 1970]. Fitch (Figure 4) and disappears approximately 200 [1970b] pointed out that an abrupt changein km farther west [Chase and Menard, 1969]. the orientation of the arc near the Sunda Strait Along the concave side of this trough, only (Figure 4) can accountfor such a changein the scattered concentrationsof shallow earthquakes rate of underthrustingand thus can accountin have been recorded. In comparisonwith deep- generalfor the configurationof the inclinedseis- sea trenches this trough is not a striking fea- mic zone beneath this part of the arc. ture; only locally doesit reach depths of •3000 Computedslip betweenthe Indian Oceanand meters. However, as occurswith active trenches, Eurasian plates in the entire Sunda region a strong gravity minimum parallels this trough (Figure 4) is nearly constantin magnitudeand [Gutenberg and Richter, 1954; unpublished direction because equators for the computed gravity profiles from Verna and Conrad 1971 centersof rotation for these plates passthrough cruises]. this region. However, a change in the orienta- The most striking bathymetric feature east tion of the arc near the Sunda Strait imposesa of the Java trench is the Weber deep (Figure 4). large changein the ratio of parallel to normal Fitch and Molnar [1970] and Fitch [1970a] slip. Along the western arc this ratio progres- erroneouslyinterpreted the Weber deep as part sively increasestoward the Andaman region, of the eastern continuation of the Java trench whereasalong the easternarc it remainsnearly (W. Hamilton, personalcommunication, 1971). constant at a low value. If the decouplinghy- The inner wall of this deep is seismicallyone of pothesisis correct, the absenceof an extensive the more active regions within this arc. For zone of transcurrent faulting along the eastern example,the only shallowearthquakes of large arc is explainedby the absenceof a large com- magnituderecorded in the easternSunda region ponent of parallel slip. are two events located near the inner wall of 4440 T•o•As J. F•TCH this deep: one on February 1, 1938, and the other on November 2, 1950 [Richter, 1958]. In fact, the larger recent movementsin this region,according to recordedseismicity, have occurred within the arc (e.g., Figure 5), not along the frictional contact between the convergent plates. The low level of shallowunderthrusting 0 ß 0 at the plate margins may be temporary, if 5 S •'""..... ' 5 $ computed rates of convergenceof 5-6 cm/yr [e.g., Le Pichon, 1968] are consistentwith a long recurrence time for major earthquakes, e.g., several hundred years. These rates are •' ....•'"%,,:;GUIN approximatelyhalf the ratesof convergencecomputed for arcs of the northwestern and the southwesternPacific. An alternative explana- lOS[, , I I I I I I I tlOS 130E 140E tion for the absenceof underthrustingis that the closeapproach of the Australian continent Fig. 10. Recent shallow earthquakesin western New Guinea and adjacent regions. Open and has slowedor stoppedplate consumptionin this closed circles are epicenters computed by the USCGS from arrival times at more than 20 and 10-20 stations, respectively, for 1961-1970. The heavy solid and dashed lines represent subareal I•.0" I$0" 140" I I 1 and inferred submarine traces of the Sorong fault ZO' ,. ,4• _ [Visser and Hermes, 1962].

region. Such a disruption is suggestedby an a.pparentreorientation of the islandsof the outer arc in the vicinity of Timor and by a break in the Java trench in this region. However, it is impossibleat present to decidewhich, if either, of these explanationsis more correct. Junction between[o'ur plates. Figure 4 illus- trates a complexarray of known and suspected fault zonesnear the junctionof four large plates of lithospherein the regionbetween the Philip- pine and Sundaarcs. Plate marginsin this region are hard to define with certainty from seismic evidence alone and probably are not stable over time intervals longer than a few million Fig. 9. Recent shallow seismicity, volcanism, years. McKenzie and Morgan [1969] point out and major faults in the Philippine and adjacent that junctionsbetween four or more plates are regions. Solid circles are epicenters of shallow generallyunstable. The metamorphichistory of earthquakes from 1961 to 1971 located by the the Celebesand other parts of this complex USCGS from 20 or more arrival times for P region showsthat several zonesof plate con- waves. Triangles give locations of active volcanoes, including the Solfutura fields, [Neuman vergencewithin this regionhave flourishedand Van Padang, 1953]. Heavy curves representtraces becomeextinct in the late Tertiary [Hamilton, of major transcurrent faults [e.g., Ka,tili, 1970]. 1970]. Dashed parts of these curves represent submarine Recent movementsin the Celebesregion (Fig- continuationsfor which there is seismicand bathy- ure 5) include extensionnear the south coast ot metric evidencereferenced in the text. Deep-sea trenchesand troughsare outlined with bathymetric the gulf that separatesthe two northern penin- contours from the charts of Chase and Menard sulas (solutions 27, 28, and 29) and thrust [1969]. movementsin a limited region along the west PLATE CONVERGENCE AND TRANSCURRENT FAULTS 4441 coast of this island (solution 61). A few foci at such a plate margin [Johnson and Molnar, intermediate depths are located beneath this 1972]. However, the low level of recent activity part of the island [e.g.,Barazangi and Dorman, in this seismic zone is inconsistent with the 1969]. A mechanismsolution for one of these large slip rates that would be necessaryfor such earthquakeshas a nearly vertical axis of mini- a plate margin (judging from computed rates mum compressionconsistent with extensional of approximately 10 cm/yr between the Pa- stresswithin a descendingpiece of lithosphere cific and Indian Ocean plates in this region [Fitch and Molnar, 1970]. [Le Pichon, 1968). A zone of left-lateral shear separates the Between northernmost Celebes and Halma- eastern and the western parts of the island, hera (Figure 4) is a double-arcstructure sepa- and the northern peninsulais traversed by a rated by a zone of E-W shortening.Figure 9 zone of right-lateral shear [Katili, 1970]. Solu- showstwo lines of active volcanismbelonging to tion 60 is consistent with left-lateral shear separate island arcs. Beneath the western arc within the former fault zone. Strike-slip (solu- and the Celebes basin is an inclined zone of tion 57) and normal faulting (solution 56) in earthquakesthat extends to depths of •600 the Makassar Strait between Borneo and Celebes km. In contrast, a. diffuse zone of activity is (Figure 4) are interpreted as evidence for a found at intermediatedepths beneath the eastern hinge for extensionin the northern gulf. arc [Fitch and Molnar, 1970]. This evidence According to recent distributionsof shallow suggests westward convergence beneath the earthquakes(e.g., Figure 9), a fault zonecrosses Celebes basin and eastward convergencebe- a spur of continentallithosphere (the Sulu spur) neath the west coast of Halmahera separated east of Celebes(Figure 4). This proposedfault by a remnant sea floor [Moores, 1970]. How- zone joins an active zone of left-lateral shear ever, the bathymetry of this sea floor and its (solutions21 and 22 in Figure 4) that in Fig- sediment cover do not reveal an active trench- ure 4 is drawn with great caution as a sub- like feature consistentwith plate consumption marine continuation of a left-lateral fault in along the western margin of Halmahera (un- extreme western New Guinea [Visser and publishedseismic data from Verna and Conrad Hermes, 1962]. Judging from epicentersof re- 1971 cruises). The best-developedtrenchlike cent earthquakesshown in Figure 10, the left- feature in this region is the southernend of the lateral fault is probably active. For solutions Philippine trench, which can be followed east 21 and 22 the question of which nodal plane of Halmahera to approximately0.5øN [Chase representsthe fault surfacewas resolvedby an and Menard, 1969]. The low level of seismic E-W trend in epicentersfrom events occurring activity on this side of Halmahera [e.g., Bara- within a coupleof monthsafter the main shock. zangi and Dotman, 1969] suggeststhat oceanic These events are assumed to be aftershocks. lithosphereis consumedat only a slow rate Figure 10 shows scattered concentrationsof along this part of the trench. Clearly plate recent activity along the margins of Geelvink consumptionadjacent to Halmahera and the Bay and in the vicinity of the Aroe Islands evolution of the double-arc structure are from south of New Guinea. This activity reveals a this evidencepoorly determined. major break in an E-W zone of shallowunder- Between the parallel volcanic chains and thrusting concentratedinland from the northern farther south in the Molucca basin, sediments coast of New Guinea. Directions of under- are so stronglydeformed that extensivereflecting thrusting inferred from mechanismsolutions 2, horizons are nonexistentwithin them (unpub- 4, and 23 (Figure 5) are nearly normal to the lished seismic data from Verna and Conrad 1971 coast,whereas directions of slip computedfrom cruises).Thrust faulting in the basementasso- rigid plate models are strongly oblique to the ciated with a closureof this basin appears to coast (Figure 4). This disparity is difficult to be consistent with the morphology of these resolve within the mechanicsof rigid plates sediments.An extreme gravity minimum in the unless another plate margin exists near• this Molucca basin is additional evidence for an region. A broad zone of shallow seismicityex- anomalousupper mantle in this region [Vening tending from the Mariana arc to the easternend Meinesz, 1964]. of the Bismarck Sea is a possiblecandidate for North of the double arc the zone of shorten- 4442 T'•oMAs J. F•Tc•

CO•VEaCE•CE A•.O•C MAac•s or PHiLiPPiNE SEA Convergencein the Philippine region has evolved one of the earth's more complexisland arcs. Active volcanism extends from southern •N 5N Mindanao to northern Luzon (Figure 9); however, activity in this region is not confinedto a central rift, as it is in some well-developed

o island arcs such as the Sunda arc. Seismic O© activity extends to depths of about 200 km beneath the lesser Philippine Islands adjacent to the northern end of the Philippine trench [Fitch and Molnar, 1970; Fitch, 1970a]. Farther cO0 ß south, adjacent to Mindanao and the eastern margin of the CelebesSea, this inclinedseismic zone reachesdepths of >600 km. Inland from and parallel to the Philippine trench is a. zone of transcurrent faulting showing evidence of ee 0 ß ß 0 o Quaternaryto Recentmovements [Allen, 1962]. ß 0 0 At a latitude of approximately12øN, wherethe 2N •N 126E 129E trench ends, the fault zone bends gently west- wa.rd and finally ends in central Luzon as a Fig. 11. Recent shallo.wearthquakes and bathy- broad zone containingseveral distinct branches. merry near the Talaud Islands. Open and closed circles are epicenters computed by the USCGS Figure 9 shows epicenters of recent earth- from arrival times at more than 20 and 10-20 quakes of shallow focus, a few of which are stations, respectively, for 1961-1970. B•thymetry locatedon or near mapped traces of this fault. (in fathoms) is from Krause [1966, Plate 1]. Only one of these events was large enoughto yield a reliable mechanismsolution (solution41 in Figure 12), and that solution is consistent ing splitsinto an easternand • westernbranch with geologicevidence of recent transcurrent (Figure 9). The western bra.nch follows the movementsin the left-lateral sense[Allen, 1962]. westernmargin of the Talaud ridge marked by Of the recordedearthquakes, only one of large the island of that name shown in Figure 11. magnitudewas located near the central part The eastern zone lies between the eastern mar- of the fault trace. That event occurred in 1901 gin of the Snelliusridge and the southern end and was given a magnitude of 7.8 by Richter of the Philippine trench. Much of the scatter [1958]. The only other recordedearthquake of in the distributionof theseepicenters may result large magnitude near this fault occurred in from an absence of data from local sta.tions 1924 [Allen, 1962] and was located near a and from near-sourceerrors causedby propaga- submarinecontinuation of the fault that appears tion through inclined slabs of high-velocity on bathymetricmaps as a narrow trough in the lithosphere.At least in recent years the Talaud sea floor [Krause, 1966] south of Mindanao. trough dividing these two ridges has been Tectonic movementsmarginal to Luzon in- seismicallyquiet. Every solutionto earthquakes clude possibleunderthrusting along the west in both branchesis consistentwith some type coast of the island. Consumption of oceanic of thrust movement. Thus the Talaud and lithospherebeneath this coastcan explainearth- Snelliusridges are active horstlikestructures, as quakes at intermediatedepths, one of which was previously suggestedby Krause [1966] yielded a mechanismsolution consistentwith a from bathymetric data. East of Mindanao the nearly vertical axis of minimum compression thrust zonesthat parallelthe sidesof thesehorst- [Fitch and Molnar, 1970]. Also suggestiveof like structures join to form a zone of shallow underthrustingis the west Luzon trench [Hayes thrusting that continuesalong the inner wall of and Ludwig, 1967], which parallelspart of the the Philippine trench (Figure 9). coast. PLATE CONVERGENCEAND TRANSCURRENTFAULTS 4443 Shallow underthrustingalong the east coast [Allen, 1962; Chasea•d Menard, 1969]. The of the island is well documentedby a seriesof part of the Philippinefault nearestthis activity movementsthat began in August 1968 (Fig- strikesapproximately 322 ø . Alongthe innerwall ures 9 and 12). A mean azimuth of underthrust- of the Philippinetrench an averageazimuth of ing of 283ø inferred from mechanismsolutions underthrustinginferred from solutions40, 39, 44, 47, and 51 (Figure 12) is nearly normal 45, 35, and36 (Figure12) is 264ø. A comparison to the coast and the adjacent deep-seatrough with underthrusting farther north reveals a,

30*N

0 •, - o \ øø o o ß 2½ PLATE 0 O- "• 0 0 ,•N 24 • LUZON 0 o oøo 44 0 o o 5147' 0 o o o o 0 o 41 \ o •.. •. \ 0 o o oo•9

o 37 o I(• o o•

,1•• NEWGUINEA o oooooo 2 --

IIIIIIIIIIIIIIIIIIIIII I 120*E 13o* 14o' •50 o Fig. 12. Mechanismsolutions for recentshallow earthquakes along margins of thePhilippine Sea(adapted from Katsumata and Sykes [1969]). Mechanism solutions are represented by the symbolsused in Figure5. Symbolsin key: 1, trenchaxis; 2, islandchains and submarine ridges; 3, major fault zones.Solutions with numbersless than 27 are from Katsumata and Sykes[1969]. Solution 52 and the solutionsin the Philippineregion are the sameas those shownin Figure5. Solutions27-35 are newsolutions for 1968to mid-1970;details are shown in Figure 13. The slip vectorfor the Kanto earthquakewas determineduniquely from seismicand geodeticevidence reported by Kanamori [1971a]. 4444 THOMAS J. FITCIt rotation of approximately19 ø in the direction slip to normalslip observedas a changein the of underthrustingwithin a distanceof a few direction of underthrusting.In the extreme case hundred kilometers.This apparent rotation is of completedecoupling, a directionof oblique largerthan the uncertaintyin the determination convergencefor this region is given by the of the averagedirection of slip and takes place strike of the section of the Philippine fault in too short a distance to be consistent with between the two zones of underthrusting.With simple convergencebetween two plates unless this greatly simplifiedmodel for slip in the the centerof rotation for theseplates lies within Philippineregion and an additionalassumption this region. Such a rapid changein direction of rigid plates,a modelis constructedfor con- can be explainedby the addition of a small vergencealong both sidesof the PhilippineSea platethat includesmost of the lesserPhilippine plate. Someof the grossseismic and geologic Islands.This plate wasproposed by Katsumata observationsfor this plate marginare explained, and Sykes[1969] and Fitch [1970a] to explain such as the sense of horizontal shear on interior a largechange in the directionof underthrusting faults of major proportions(Figure 2). How- betweenthe Philippineand Ryukyu arcs.Some ever,no attemptis madeto accountfor localized of the largestearthquakes recorded in thePhilip- complicationssuch as interarcspreading within pine regionhave foci near one of the lesser the west Mariana basin [Karig, 1971b]. PhilippineIslands [e.g., Allen, 1962], and The next step in constructingthis modelis recent shallow seismicityin this region [e.g., to selectdirections of slip from other parts of Barazangiand Dotman, 1969] is concentrated the westernmargin of this plate that give, in in a narrowzone suggestive of a plate boundary. addition to the inferred direction of slip in the However,this evidencemay be misleading.In Philippineregion, the directionsof total slip parts of westernNew Guineaand elsewhere, between the plates. Plate movementsin the narrow zonesof shallowearthquakes may mark Taiwan region (Figure 12) are excludedfrom zonesof internal deformation rather than plate the modelcomputations because of the obvious margins. tectonic complexityin this region.This com- In the lesserPhilippine region, mechanism plexityis revealedby a superpositionof shallow solutions to three shallow earthquakes (solu- activity that has been dominatedin recent tions43, 42, and54 in Figure12) revealstrike- yearsby transcurrentmovements and a nearly slipmovements that are difficultto incorporate vertical zone of intermediate earthquakes con- into a modelof slipalong a simpleplate margin. centratedat depthsof <100 km [Katsumata Such movementscan be explainedas internal andSykes, 1969; Wu, 1970].The onlyremain- deformationcaused by stress resulting from ing directionsof slip are thosefrom shallow plate convergence.For example,in western earthquakesalong the Ryukyuarc and south- Japan,mechanism solutions for shallowearth- westJapan, including the Kantoregion (Fig- quakes[Ichikawa, 1961, 1965] and recentoff- ure 12). Exceptfor the directionin the Kanto setson mappedfaults [Allenet al., 1970]show region,the directionsare inferredfrom mecha- a preferredorientation of compressivestress nism solutionspreviously reported by Katsu- that is nearly normalto the Japanesetrench. mata and Sykes[1969] and from newsolutions With a similarexplanation in mind for earth- for 1968-1970 (Figure 13). quakesin the lesserPhilippine region, solutions In this paperthe Sagamitrough (Figure 12) 43 and 54 (Figure 12) are drawn as conjugate is taken to be the northern boundary between movementssuggestive of compressionalong a the Eurasianand PhilippineSea plates. A direc- NE-SW trend. tion of slip alongthis trough (Figure12) was Speculativemodel [or plate convergencein inferred from the mechanismof the great Kanto western Pacific. The decoupling hypothesis earthquakeof 1923,determined from seismic offersanother explanation for the changein the [Kanamori,1971a] and geodeticevidence directionof underthrustingnorth of the Philip- [Ando,1971]. The mechanism was right-lateral pinetrench and the apparenthia.tus in under- shearand had a strongcomponent of thrusting. thrustingin this part of the arc. From this Figure12 showsthat the directionof slipfor hypothesis,the bend in the Philippinefault the Kanto earthquakeis nearly identicalto correspondsto a changein the ratio of parallel that inferredfrom underthrustingnear the coast PLATE CONVERGENCEAND TRANSbURRENTFAULTS 4445 of southwestJapan. To this evidence can be movementspredate the currentepisode of plate added directionsof slip inferred from mecha- tectonicsthat begain in Japan during the nism solutionsfor two historicmegathrusts in Miocene[Kaneko, 1966; Matsuda et al., 1967]. this region,the Nankaido and Tonankai earth- Geomorphicevidence of Quaternarymovements quakesof 1946 and 1944, respectively[Kana- in the right-lateralsense [Kaneko, 1966; Okada, mori, 1972a]. This evidencesuggests that the 1971] showsthat at least part of the rejuve- slip vectoris at mostonly partiallydecoupled nated activity along this fault zone is contem- in southwestJapan. porary with underthrustingalong the Nankai Inland fromthe zoneof underthrustingis the trough. Seismicevidence for the onset of under- median tectonic line, along which earliest thrusting1-2 m.y.ago [Fitch and Scholz, 1971;

N N N

27 28 2g

W E W E W E

APRI 1968 5,?..5N152.2E 50 KM AUG:5 1968 25.6N128.4E 19 KM NOV12 1968 27.5N128.4E 30 KM

N N N 30 $1 32

w E w E W E

APR21 1969 :52.1N 151.8E 50 KM SEP17 1969 SUN I$1.$E 8 KM DEC.$1 1969 28.5N 129.1E 50 KM

N N N 33 34 $õ

W 0 + E W E W E ,P

MAR4 1970 12.IN143.7E 50 KM JUL25 197052.1N 151.6E 50 KM JUL26 197032JN 151.8E 30 •KM Fig. 13. Additionalmechanism solutions for shallowearthquakes in Ryukyu,southwest Japan,and Marianaregions. Locations were computed by the USCGS.Nodal planesare givenby heavycurves. Polarities of P-wavefirst motions are givenby opencircles for clear dilatation,open triangles for weakdilatation, closed circles for clearcompression, closed tri- anglesfor weakcompression, and crosses for weakarrival of indeterminatepolarity. Arrows giveS-wave polarization. Crosses enclosed in circlesgive the locationsof polesto oneof the nodalplanes. Locations of P (axisof maximumcompression) and T (axis of minimumcom- pression)axes are given by small closedcircles. 4446 THoM^s J. FITCH Kanamori, 1972a] coincideswith geologicevi- margin of this plate restrict centersof rotation dence for post-Plioceneto Recent 'upwarping for the Pacific and Philippine Sea plates to a and block movements'in this part of Japan region S-WSW of the Mariana arc [Kats'umata [Matsuda et al., 1967]. and Sykes, 1969]. Because the instantaneous The most active section of the median tec- vectors of rotation can be summed, such a tonic line, accordingto geomorphicevidence, is vectorfor the Eurasianand PhilippineSea plates the middle section, where recent movements and another for the Pacificand Eurasianplates, yield an average rate of right-lateral slip of when they are summed,yield a rotation vector about 5 mm/yr [Okada, 1971]. This rate is for the Pacificand PhilippineSea plates.When lower than that of underthrustingalong the this sum is computed,there is an additional as- Nankai trough by more than 1 order of mag- sumption that slip between the Eurasian and nilude. A recordof large-magnitudeearthquakes China plates can be neglectedin comparison in weslern Japan that is thought to be com- with the other relative plate motions in the plete for about the last 600 years [e.g., Ima- region. Given instantaneous rotations for the mura, 1928] revealsno major eventsalong this Pacificand Eurasian platescomputed by others fault zone. Thus large movements have not (solutions5 and 6 in Table 1), it is apparent occurred on this fault in recent years unless that, if the summed vectors are to yield a they occurredaseismically. centerof rotation for the Pacificand Philippine Minor movementsin the right-lateral sense Sea platesthat satisfiesconstraints imposed by along a fault zone of this strike can also be recent seismicityalong this plate margin, the explainedas internal deformationresulting from center of rotation for the Eurasian and Philip- E-W compressioninferred from recent move- pine Sea plates must be in the northeastPacific. menis on other faults in western Japan [e.g., Of the two centers of rotation for the Pacific Ichikawa, 1961,1965]. In any case,the assump- and Philippine Sea plates,the preferredone is tion that slip is coupledin southwestJapan is in east-centralAustralia. (solution 7 in Table 1). probablyas goodas the assumptionthat slip is This preferencecannot be judgedwith certainty completelydecoupled in the Philippine region. from availablemechanism solutions (Figure 12) After directionsof slip that are thought to becauseof large uncertainty in the determina- approximatedirections of total slip betweenthe tions of the nodal planes (e.g., solutions 20 plates are selected,centers of rotation are fitted and 14 of Katsumata and Sykes [1969]) and to these data. Two solutionsgiving nearly equal possible complicationsresulting from interarc fits are given in Table 1 (solutions1 and 4). spreadingin the westernMariana trough (solu- The center of rotation near Alaska is preferred tions 18 and 19 of Katsumata and Sykes [1969]. becausean anticipated center of rotation for Implied directions of convergence(Figure 14) the eastern margin of the Philippine plate re- that are normal in the Izu-Bonin and Mariana quires a center of rotation for the western arcs are consistentwith a large component of margin of this plate in the northeast Pacific. A horizontal shear along latitudinal fracture zones large distancebetween this center of rotation that separate the Palau and Yap trenchesand and southwestJapan is satisfyingbecause it is the Yap and Mariana trenches (Figure 12). large enoughto keep computedrates of slip in From arguments in favor of the center of the Philippine region no larger than the largest rotaiion in Australia it would be hazardous to spreadingrates at creslsof active ridges (about conclude that the center of rotation for the 10 cm/yr) [e.g., Le Pichon, 1968]. Rates that Eurasian and Pacific plates computed by are much larger would not violate plate theory Le Pichon [1968] is less accurate than that [e.g., McKenzie and Parker, 1967] but would, computedby usingthe centerof rotation for the in view of the speculativenature of this model, American and Pacific plates from McKenzie require additional justification. a•d Parker [1967]. The latter center is some- The configuration of inclined seismic zones what more satisfyingbecause it is a fit not only beneath the Izu-Bonin and Mariana arcs (Fig- to the strike of transform faults within and ure 2) and a relative absenceof activity at all marginal to the Pacific Ocean but also to slip depths in the region of the Palau and Yap vectors from mechanism solutions for under- trenchesand along other parts of the southern thrusts along the convex side of the Aleutian PLATE CONVERGENCEAND TRANSCURRENTFAULTS 4447

:• o • o o •:%oz 4448 THOMAS J. FITCH

Sagami A standard deviation of 18ø is larger than the Trough uncertainty in many of the inferred directionsof underthrusting.Solution 3 is completely un- acceptable because it implies opening along either the Ryukyu or the Philippine arc. These negativeresults can be interpretedas additional support for the decouplinghypothesis. LF CONCLUSION The senseof parallel slip along the western Philippine margin of the Philippine Sea inferred from the preferred center of rotation for the Eurasian and Philippine Sea plates agreeswith the sense Pl=te of recent offsetson the median tectonicline, the Longitudinal fault, and the Philippine fault PF• • (Figure 2). However, computedratios of parallel slip to normal slip in southwest Japan, Taiwan, and, to a lesserextent, in the Philip- pines (Table 2) conflict with seismicevidence from these regions. The largest estimates of Fig. 14. Convergencealong the margin of the right-lateral shearalong the mediantectonic line Philippine Sea plate. Plate margin is in several (about 5 mm/yr [Okada, 1971]) are less than regions greatly simplified (for details see text). Large arrows are inferred relative motion be- the right-lateral shear required for complete tween plates. MTL, median. tectonic line; LF, decouplingby about an order of magnitude.If Longitudinal fault; PF, Philippine fault. continueddevelopment of an islandarc in southwest Japan resultsin completedecoupling, the arc. In any case,the least well determinedrota- ratio of parallel slip to normalslip inferred from tion vector in the sum is the one for the Eura- solution1 (Table 1) will be approximately1:2. sianand PhilippineSea plates. This speculation assumesthat the preferred The remaining two solution (solutions2 and center of rotation will remain in its present 3) in Table I contradict the decouplinghy- positionnear southwestAlaska. Such an assump- pothesis by assuming that the directions of tion is difiqcultto defendfor a plate margin such underthrustingalong the eastside of the Philip- as this one, parts of which are in early stages pine Islands give the directionof total slip in of development.For example, imagine within this region.Solution 2 is a poor fit to the data. a plate of lithospherea nascentisland arc ex-

TABLE 2. Rates of Slip

Decoupled Slip Ratesõ Azimuth Azimuth of Fault, of Slip,t Parallel Normal Fault deg deg (Strike Slip) (Dip Slip)

Median tectonic line (southwest Japan) 249 310 4 7 Longitudinal fault (Taiwan) 22 307 2 9 Philippine fault adjacent to hiatus in 321 305 10 1 underthrusting* Philippine fault adjacent to Philip- 346 304 7 7 pine trench

*Large ratio of oblique slip to normal slip in this region is a consequenceof con- straints placed on solution 1 (see text for detailed argument). õRates are considered no better than first-order approximations because of uncertainty in the rotation vectors. ñComputed from solution 1 in Table 1. PLATE CONVERGENCEAND TRANSCURRENTFAULTS 4449 tending itself in only one direction. As such a simplemodel for obliqueconvergence presented convergentplate boundary continuesto grow, in the introductionsuggests that decoupling it will causemigration of an initial center of can occur at angles as large as 45ø , even when rotation in a direction that in a simple caseis all fault zoneshave equal strength. given by the direction in which the arc is ex- One of the more interesting results of the tending itself. preferred plate model for the northwest Pacific In contrast to southwest Japan, the region is that the Sagami trough marks the path of between the southern end of the Ryukyu arc an unstabletriple point formed by the junction and Luzon hasyet to developa well-definedzone of three major plates near Japan (Figure 2). of underthrusting.Recorded shallow activity in The computationsillustrated in Figure 15 with this region strongly suggeststhat the Longi- diagrams proposedby McKenzie and Morgan tudinal fault is at least as active and probably [1969] reveal a large componentof right-lateral more active than other faults in the region,with shearalong the Sagamitrough at a rate of about the possibleexception of the zone of under- 70 km/m.y. Absence of an offset between the thrusting alongthe southernend of the Ryukyu Izu-Bonin and Japanesetrenches requires ex- arc [Katsumata and Sykes, 1969]. Thus a ratio tension between the Philippine Sea and Pacific of parallel slip to normal slip of < •, computed from the inferred direction of oblique slip for this region,is inconsistentwith recentseismicity of the regon. RATE STRIKE AZ Earthquakesat depthsas great as 150 km that 292 are found beneath eastern Taiwan and in de- 312 creasingconcentrations south of the islandsug- 268 gest plate consumption;however, the manner in which this activity manifests itself at the surfaceis still a puzzle.The spatialconfiguration of the seismiczone can be explainedby a curled edgeon the part of the plate adjacentto Taiwan. The assumptionof completedecoupling in the Philippineregion results in ratiosof parallel slip to normal slip of 1:1 for the part of the Fig. 15. Migration of a triple point along the Sagami trough. Regions A, B, and C represent arc adjacent to the Philippine trench and 10:1 pieces of Eurasian, Pacific, and Philippine Sea in the region of the hiatus in underthrusting plates near Japan. The lines that join at the north of the trench. Although a weak argument triple point represent a possibleconfiguration for can be made for a concentration of recorded boundaries between these plates at the onset of underthrusting adjacent to southwest Japan; the seismicityon the part of the fault adjacent to lines between A and B, C and B, and A and C this hiatus, the level of activity is much less represent the Japanese trench, the Izu-Bonin than that in the zonesof underthrusting.Thus trench, and the Nankai trough, respectively. The it is difficult to justify the assumptionof com- approximate time lapse since the onset of under- plete decouplingwithout appealing to geo- thrusting along the Nankai trough is 1 m.y. [Fitch and $cholz, 1971]. By using this time inter- morphic evidence that at least qualitatively val and the rates of plate convergencecomputed appears to be more substantial than the seismic from solutions 1 and 7 in Table 1, it is possible evidencefor slip [Allen, 1962]. to estimate the configuration of consumed litho- As is shownin Figure 13, anglesformed by sphere(dashed lines) and new positionsfor plate margins that move in responseto migration of the the western margin of the Philippine Sea and triple point. The triple point migratesin the direc- the inferred directionsof obliqueslip vary by tion of the heavy arrow, which nearly coincides as much as 60ø . In the Taiwan region, where with the strike of the oceanic part of the Sagami decoupledslip is hypothesized,this angle is trough. The rate is about 70 km/m.y. The shaded >70 ø. Such large anglesof oblique slip-seem- region representsthe new oceanic lithosphere resulting from plate separationalong the Izu-Bonin ingly require a zone of horizontal shear that arc. This rifting is inferred from the absenceof has little strength relative to the strength of an offset between the Japaneseand the Izu-Bonin the zone of plate consumption.However, a trenches. 4450 THOMAS J. F•TC• plates in the vicinity of the triple junction. vince are tectonicallysimilar in that each region Normal faulting is known in this region from is boundedon one side by an extensivezone of a mechanism solution to the Boso-Oki earth- transcurrent faulting and on the other side by quakeof November26, 1953 [Kanarnori, 1972b]. a semistablecontinental margin. Scholz et al. Karig [1971b] speculated that the elongated [1971] interpreted the Basin and Range pro- troughs within the Izu-Bonin arc result from vince as a latent interarc basin. Similarly, the a nascentstage in the developmentof an inter- western Andaman basin can be interpreted as arc basin. Extensionwithin this arc is supported an interarc basin. Previous interpretations of by a mechanismsolution (solution 13 in Figure this basin as a 'rhombochasm'[Carey, 1958; 12) for normal faulting [Katsumata and Sykes, Rodol)•o, 1969] require lateral movements of 1969] and high values of heat flow [Watanabe adjacent parts of southeastAsia and the Sunda et al., 1970]. If the Izu-Bonin trench has re- arc, whereasthe interpretation proposedin this mained stationary, the argument for extension work requiresmovement of only the Andaman- near the triple junction can be applied to the Nicobar ridge to form a bulge that is apparent remainder of the Izu-Bonin arc. From this from physiographicmaps of the region [e.g., argument, extensionwithin this arc is a conse- Rodoleo,1969, Figure 4]. quence of the mechanicsof rigid plates, and Some closecomparisons can be made between thus only passive upwelling of magma is re- the physiographicexpressions and the tectonic quired to form new sea floor. However, this settingsof the Alpine and Philippine faults. For speculationis hazardousbecause the bulgelike example,the Alpine fault is confinedto a narrow configurationof the Mariana ridge (Figure 2) zone in the southernpart of South Island and strongly suggeststhat interarc extension can broadensnorthward into a seriesof subparallel generate enough pressureto push the frontal branches.Although active volcanism is absent arc into an adjacentocean basin [Karig, 1971a]. in South Island, seismicevidence suggests that a Recordedseismicity along the Izu-Bonin arc zone of convergencefollows a greater part of the is conspicuouslylacking in shallowearthquakes west coast of the island. Two submarine earth- of large magnitude [e.g., Kanarnori, 1972a]. quakes north of Milford Sound yielded mech- A cessationof plate consumptionalong this anism solutionsconsistent with thrust faulting platemargin was proposed by Kanamori[1971b, [Johnson and Molnar, 1972]. A possiblesub- 1972a] to account for this absence.Interarc marine continuation of this fault south of Mil- extension provides an alternative method by ford Sound parallels a zone of earthquakesat which compressivestress normal to the margin intermediate depths [Smith, 1971] and a deep- of the Pacific plate is reducedto a value that is sea trench [Hayes et al., 1971]. From this evi- low enough for consumptionof the Pacific dence it seemsprobable that the plate margin plate to take place without large-magnitude in this regionis not a zone of pure horizontal earthquakes. Such a mechanismfor stress re- shear. Consequently,computed centers of rota- duction again implies passive upwelling of tion for the Indian Ocean and Pacific plates magma, unlike that occurringwithin the west that are constrainedto give plate motion in Mariana basin. Wu [1971] interprets the low New Zealandthat is pure horizontalshear along level of shallow seismicity along the entire the Alpine fault will be in error proportionalto eastern margin of the Philippine Sea as evi- the rate of convergencein this region. dencefor plate separation. In this work, concurrenthorizontal shear and Within tectonically active margins of south- plate consumptionin the sameregion are ex- east Asia and the western Pacific,the Philippine fault and the western Andaman basin are of plainedby the decouplingof obliqueslip. In special interest becauseeach appears to have some regions,such as the westernSunda arc an analog in another part of the world. The and eastern parts of the Philippine arc, de- Philippinefault is in someways similar to the coupledslip occursalong parallel zones of hori- Alpine fault in New Zealand, and the western zontal shear and shallow underthrusting. The Andamanbasin is similarto the Basin and Range lithospherebetween these fault zonesis a long province in the United States. The western narrow plate that in an ideal situationwill Andaman basin and the Basin and Range pro- movein sucha way as to causeextension at one PLATE CONVERGENCEAND TRANSCURRENT FAULTS 4451 end of the plate and compressionat the other. APPENDIX In general,these simple end effectsare obscured by complicationssuch as a changing ratio of parallel slip to normal slip along the arc and internal deformation resulting from volcanism. In addition, the narrownessof sucha plate may result in earthquake sequenceson one side of the plate influencing those on the other side. If this were the case, a rigid plate approxima- tion would be of doubtful validity. Zones of convergencein continentalregions, such as the one that extendsfrom easternTurkey to central Asia southwestof Lake Baikal, may be extreme casesof decoupledslip in which the boundary between the major plates is composedof numerous plates so small that their motion is de- pendent on the motion of the surrounding plates, even on a time scale as short as a few years. The width of the contact zone between the major plates, which in placesexceeds 100 km (judging from the width of the seismiczones in the region [e.g., Nowroozi, 1971]), can be consideredsupport for such a plate model. In regionsof plate convergencewithout a well- developedisland arc, such as southwest Japan and Taiwan, transcurrent movements appear to be part of an initial stagein the lateral growth of an adjacent zone of convergence.Thus transcurrent faults within active island arcs may predate the formation of the arc in that region. Furthermore, such faults can inherit not only their traces but also their sense of movement from a pre-existingfault.

Theseideas can be appliedto Cenozoicfault- 00•00o00000 ing on the San Andreassystem in California. There is evidencefor slip on the San Andreas

that predatesthe time when this fault became o part of the boundarybetween the American and Pacific plates [e.g., Atwater, 1970]. Some of these early movementsmay have occurred within an active island arc associated with a mid-Cenozoic trench that existed off the coast of western North America [Atwater, 1970]. As o the plate marginwas transformed progressively from a convergentmargin betweenthe Pacific and the Farallon plates [e.g., McKenzie and o Morgan, 1969], the pre-existing transcurrent fault inheritedthe presentplate marginin the region. This speculationexplains why the lo- cationof the pre-existingzone of underthrusting was not transformed into the present plate margin. 4452 T•oMAs J. FITCH

o o o o o o o o o o o o o o o o o o o o o o o o o o o

o o o o o o o o o o o o o o o o o o o o o o o PI•AT]• CONVERGENCE AND TRANSCURRENT FAULTS 4453

Ot O0 ('NI •'•'1 ,--• ,--• •,•'1 ,--I (",1 ,---• 'el' I'• •4D •'•'t tO (",1 •D t•"l O0 .el' t•"l I'• .el- .el' t•"l ('NI •D

o o o o o o o o o o o o o o o o o o o o o o o o o o o

(D 4-• •J ..-• ß ß 4454 T•o•s J. F•c•

o o o o o o o o o o o o o o o o o o o o o o

...... PLAT• CONVERGENCE AND TRANSCURRENT FAULTS 4455 4456 THOMAS J. FXTC•

z • z

c) •AT•_ •ONYERGENCEAND TRANSCURRENTFAULTS 4457

z • z • 4458 T•o•As J. FXTC• PLATE CONVERGENCE AND TRANSCURRENT FAULTS 4459

Acknowledgments. A lecture on tectonic move- tinental tectonics in Central Asia, J. Geophys. ments in southwest Japan given by Professor Res., 75, 2699-2709, 1970b. Arata Sugimura helped convince me that the Fitch, T. J., and P. Molnar, Focal mechanisms Sagami trough marked a plate boundary in this along inclined earthquake zones in the Indo- region. Dr. W. Pitmann III gave me computer nesia-Philippine region, J. Geophys. Res., 75, programsfor determining instantaneouscenters of 1431-1444, 1970. rotation between rigid plates. Dr. Warren Hamil- Fitch, T. J., and C. H. Scholz, A mechanism for ton pointed out locations of the more prominent underthrusting in southwest Japan: A model for zones of convergencein the eastern Sunda and the convergent plate interaction, J. Geophys. Res., Celebes regions. Dr. Charles Windisch showed me 77, 7260-7292, 1971. unpublished seismic records of sediment cover in Gumper, F. J., Tectonics of the western Aleutians this complex region. Discussions with Professor determined from focal mechanisms and seismic- L. R. Sykes, Dr. C. H. Scholz, Dr. P. Molnar, and ity (abstract), Eos Trans. AGU, 52, 279, 1971. T. Johnson were helpful in constructing the argu- Gutenberg, B., and C. F. Richter, Seismicity o! ments presented in this work. Professor.Sykes and the Earth and AssociatedPhenomena, p. 310, Dr. Scholz read the manuscript critically and sug- Princeton University Press, Princeton, N.J., gested numerous improvements. 1954. This work is supported by NASA under grant Hamilton, W., Tectonic map of Indonesia--A NGR-33-008-146. progressreport, Proj. Rep. Indones. Invest. (IR) IND-5, pp. 1-29, U.S. Geol. Surv., Washington, REFERENCES D.C., 1970. Hayes, D. E., and W. J. Ludwig, The Manila Allen, C. R., Circum-Pacific faulting in the Philip- trench and west Luzon trough gravity and mag- pines-Taiwan region, Y. 6eophys. Res., 67, 4795- netics measurements, Deep Sea: Res., 14, 545- 4812, 1962. 560; 1967. Allen, C. R., Transcurrent faults in continental Hayes, D. E., M. Talwani, and D. A. Christoffel, areas--A symposium on continental drift, Phil. The Macquarie ridge complex (abstract), Eos Trans. Roy. Soc. London, Ser. A, 258, 82-89, 1965. Trans. A GU, 52, 237, 1971. Allen, C. R., T. Matsuda, and A. Okada, Relation- Heezen, B.C., and M. Tharp, Indian Ocean floor ship between seismicity and geologic structure (map), Nat. Geogr. Soc., Washington, D.C., 1967. in western Japan (•bstract), Bull. Geol. Soc. Amer., 2, 481-482, 1970. Ichikawa, M., On the mechanism of the earth- Ando, M., A fault-origin model of the great Kanto quakes in and near Japan during the period earthquake of 19'23 as deduced from geodetic from 1950 to 1957, Geophysics,30, 355-403, 1961. data, Bull. Earthquake Res. Inst. Tokyo Univ., Ichikawa, M., The mechanism of earthquakes 49, 19-32, 1971. occurring in central and southwestern Japan Arabasz, W. J., and C. R. Allen, Tectonics of and some related problems, Palp. Meteorol. northern Chile as reflected by the Atacama Geophys., 56, 104-156, 1965. fault system,Res. Rep. 1970,p. 77,/)iv. of Geol. Imamura, A., On the seismic activity of central and Planet. Sci., Calif. Inst. of Technol.,Pasa- Japan, Jap. J. Astron. Geophys.,6, 119-137,1928. dena, Calif., 1970. Isacks, B., J. Oliver, and L. R. Sykes, Seismicity Atwater, T., Implications of plate tectonics for and the new global tectonics, J. Geophys. Res., the Cenozoic tec•,onic evolution of western 73, 5855-5899, 1968. North America, Geol. Soc.Amer. Bull., 81, 3513- Johnson, T., and P. Molnar, Focal mechanisms 3536, 1970. and plate tectonics of the southwest Pacific, Barazangi,M., and J. Dorman, World seismicity J. Geophys. Res., 77, in press, 1972. map of ESSA Coast and GeodeticSurvey epi- Kanamori, H., Faulting of the great Kanto earth- center data from 1961-1967,Bull. Seismol.$oc. quake of 1923as revealedby seismologicaldata, Amer., 59, 369-380, 1969. Bull. Earthquake Res. Inst. Tokyo Univ., 49, Carey, S. W., A tectonic approachto continental 13-18, 1971a. drift, in Continental Drift: A Symposium,p. Kanamori, H., Seismologicalevidence for a litho- 363,University of Tasmania,Hobart, Tasmania, spheric normal faulting--The Sanriku earth- 1958. quake of 1933,Phys. Earth Planet. Interiors, 4, Chase, T. E., and H. W. Menard, Bathymetric 289-300, 1971b. Atlas o! the North-Western Pacific Ocean, Kanamori, H., Tectonic implicationsof the 1944 Hydrogr. O]fice Publ. 1301, 50 charts,U.S. Nav. Tonankai and the 1946 Nankaido earthquakes, Oceanoar.Office, Washington, D.C., 1969. Phys. Earth Planet. Interiors, 5, 129-139, 1972a. Fitch, T. J., Earthquake mechanisms and island Kanamori, H., Relations among tectonic stress arc tectonics in the Indonesian-Philippine re- and earthquakesswarm, Tectonophysics,13, in gion, Bull. Seismol. Soc. Amer., 60, 565-591, press, 1972b. 1970a. Kaneko, S., Transcurrent displacementalong the Fitch, T. J., Earthquake mechanismsin the Hima- medianline, south-westernJapan, N. Z. J. Geol. layan, Burmese,and Andaman regionsand con- Geophys., 2, 45-59, 1966. 4460 THOMAS J. FITCH

Karig, D. E., Ridges and basins of the T'onga- Islands and Cochin China, International Vol- Kermadec island-arc system, J. Geop,hys. Res., canological Association, Naples, 1953. 75, 239-254, 1970. Nowroozi, A. A., Seismic tectonics of the Persian Karig, D. E., Origin of marginal basins in the plateau, Eastern Turkey, Caucasusand Hindu- western Pacific, J. Geophys. Res., 76, 2542-2561, Kush regions, Bull. Seismol. Soc. Amer., 61, 1971a. 317-342, 1971. Karig, D. E., Structural history of the Mariana Okada, A., Active faulting of the median tectonic island arc system, Bull. Geol. Soc. Amer., 82, line (in Japanese), Kagaku, 41, 666-•9, 1971. 323-344, 1971b. Richter, C. F., Elementary Seismology, p. 768, Karig, D. E., and J. Mammerickx, Tectonic frame- W. H. Freeman, San Francisco, Calif., 1958. work of the New Hebrides island arc, Mar. Rodolfo, K. S., Bathymetry and marine geology Geol., 12, 187-206, 1972. of the Andaman basin and tectonic implications Katili, J. A., Large transcurrent faults in south- for southeast Asia, Geol. Soc. Amer. Bull., 80, east Asia with spherical reference to Indonesia, 1203-1230, 1969. Geol. Rundsch., 59, 581-600, 1970. Rothe, J. 0., The Seismicity o] the Earth 1953- Katsumata, M., and L. R. Sykes, Seismicity and 1965, p. 336, Unesco, Tournai, Belgium, 1967. tectonics of the western Pacific: Izu-Mariana Scholz, C. H., M. Barazangi, and M. L. Sbar, Caroline and Ryukyu-Taiwan regions, J. Geo- Late Cenozo.ic evolution of the great basin, phys. Res., 74, 5923-5948, 1969. western United States as an ensialic inter-arc Krause, D.C., Tectonics, marine geology and basin, Geol. So,c..Amer. Bull., 82, 2979-2990',1971. bathymetry of the Celebes Sea-Sula Sea region, Smith, W. D., Earthquakes at shallow and inter- Geol. Soc. Amer. Bull., 77, 813-832, 1966. mediate depths in Fiordland, New Zealand, J. Le Pichon, X., Sea floor spreading and conti- Geophys. Res., 76, 4901-4907, 1971. nental drift, J. Geophys. Res., 73, 3661-3705, Sykes, L. R., Seismicity in the Indian Ocean and 1968. a possible nascent island arc between Ceylon Maldonada-Koerdell, M., Geological studies in the and Australia, J. Geophys. Res., 75, 5041-5055, Gulf of Fonseca-Nicaraguan depression area, 1970.. Central America, Pap. 66-14, pp. 220-238, Geol. Vening Meinesz, F. A., Developments in Solid Surv. of Can., Ottawa, Ont., 1966. Earth Geophysics: The Earth's Crust and Upper Matsuda, T., K. Nakamura, and A. Sugimura, Late Mantle, p. 124, Elsevier, New York, 1964. Cenozic orogeny in Japan, Tectonophysics,4, Visser, W. A., and J. J. Hermes, Geological results 349-366, 1967. of the exploration for oil in Netherlands, New McKenzie, D. P., and W. J. Morgan, Evolution of Guinea, Verb. Ned. Geol. Mijnbouwk. Genoot. trible junctions, N•ture, 224, 12•-133, 1969. Geol. Ser., 20, 1-265, 1962. McKenzie, D. P., and R. L. Parker, The north Wageman, J. M., T. W. C. Hilde, and K. O. Pacific: An example of tectonics on a sphere, Emery, Structural framework of east China. Sea Nature, 216, 1276-1280, 1967. and Yellow Sea, Amer. Ass. Petrol. Geol. Bull., Miyashiro, A., Orogeny, regional metamorphism 54, 1611-1643, 1970. and magmatism in the Japanese islands, Medd. Watanabe, T., D. Epp, S. Uyeda, M. Langseth, Da•n. Geol. Foren., 17, 390-446, 1967. and M. Yasui, Heat flow on the Philippine Sea, Molnar, P., and L. R. Sykes, Tectonics of the Tectonophysics,10, 205-224, 1970. Caribbean and middle America regions from Weeks, A. L., R. M. Harbison, and G. Peter, focal mechanisms and seismicity, Geol. Soc. Island arc system in the Andaman Sea, Amer. Amer. Bull., 80, 1639-1684, 1969. Ass. Petrol. Geol. Bull., 51, 1803-1815, 1967. Moores, E., Ultramafics and orogeny, with models Wu, F. T., Focal mechanism and tectonics in the of the U.S. cordillera and the tethys, Nature, vicinity of Taiwan, Bull. Seismol. Soc. Amer., 228, 837-842, 1970. 60, 2045-2056, 1970. Morgan, W. J., Plate motion and deep mantle Wu, F. T., The Philippine sea plate: A 'sinking convection, in Studies in Earth and Space towel?,' Geol. Soc. Amer. Bull., 82, 2671-2672, Science, Mem. 132, edited by R. Shagarn, Geo- 1971. logical Society of America, Boulder, Colo., in press, 1972. Neuman Van Padang, M., Catalogue o• Active (Received November 23, 1971; Volcanoes of the World, vol. 2, Philippine revised April 12, 1972.)