GEOPHYSICALKESEARCH LETTERS, VOL. 19,NO. 3, PAGES297-300, FEBRUARY 7, 1992 THELARGE NORMAL-FAULTING MAPdANA EARTHQUAKE OF APRIL 5, 1990 IN UNCOUPLED SUBDUCTiON ZONE YasuhiroYoshida 1,Kenji Satake 2 and Katsuyuki Abe 1 Abstract.A large,Ms = 7.5, shallowearthquake occurred The epicentraldata suggest that the Marianaearthquake is beneaththe Mariana trench on April 5, 1990. From the an outerrise earthquakewhich is definedas an earthquake relocatedaftershock distribution, the fault areais estimatedto originatedwithin the subductingplate in the vicinity of the be70 x 40km 2. A tsunamiobserved on the Japanese islands trenchaxis. In theMariana region, the Pacific plate is aseismi- verifiesthat the depth of the main shockis shallow. For cally subductingbeneath the PhilippineSea plate without waveformanalysis, we use long-period surfacewaves and causinglarge underthrusting earthquakes at theplate interface. bodywaves recorded at globalnetworks of GDSN, IRIS, Thisresults from the weak coupling between the downgoing GEOSCOPEand ERIOS. The centroidmoment tensor (CMT) andoverlying plates. In suchuncoupled subduction zones, the solutionfrom surfacewaves indicates normal faulting on a outerrise earthquakesgenerally possess normal-fault type faultwhose strike is parallel to the local axisof the Mariana mechanismswith their tensional stress axes nearly perpendic- trench,with the tensionaxis perpendicular to it. The seismic ular to the trench strike [Christensenand Ruff, 1988]. It is momentis 1.4 x !020Nm (x 1027dyn.cm) which gives Mw = veryintriguing whether the Mariana earthquake is oneof such 7.3. Far-field P and SH waves from 13 stations are used to tensionalouter-rise events in an uncoupledsubducfion zone. determinethe source time function. Since the sea around the epicentralregion is about5 km deep,body waveforms are Main Shock and Aftershocks contaminated with water reverberations. The inversion results in a sourcetime functionwith a predominantlysingle event The hypocenter parameters of the 1990 Mariana witha durationof 10 sec, a seismicmoment of 2.1 x 1020 earthquakegiven by N'EIC are: origin time 21:12:35.5 on Nm,and a focalmechanism given by strike= 198ø, dip = 48ø, April 5 (UT), epicenter15.125øN, 147.596øE, depth 11 km, slip= 90ø. The shortduration indicates a small area of the and Ms 7.5. The location of the main shock is shown in rupture.The locationof the main shockwith respectto the Figure 1 with the 3-day aftershockdistribution. The main aftershockarea suggeststhat the nodalplane dipping to the shockis locatedjust beneaththe trenchaxis. An accurateesti- westis preferredfor the fault plane.The local stressdrop of mateof focaldepth from teleseismic data is generallydifficult thesingle subevent is estimatedto be 150 MPa (1.5 Kbars). andthe focal depthof the main shockwas constrained to be TheMariana earthquake is consideredto haveoccurred in an 11 km by NEIC. The followinganalysis of surfaceand body uncoupledregion, in responseto the gravitationalpull caused wavessupports this shalloworigin. The T-phasewas clearly bythe downgoing Pacific plate. recordedat ChichijimaIsland, 1400 km from the epicenter, andthis observation is consistentwith the shalloworigin. Introduction The NE!C locationsof aftershocksextend as long as 70 km alongthe trench and the depthdown to 46 km thatis de- A largeshallow earthquake occurred on April 5, 1990just terminedfrom depth phases. For morereliable locations, we beneaththe Mariana trench where the Pacific plate is subduct- applieda masterevent relocation method for arrivaltime dam ingbeneath the Philippine Sea plate. This is thelargest shal- from commonstations. The largestaftershock is selectedas a lowevent ever recorded in theregion; the last event with a similarmagnitude was on September22, 1902with Ms = 7.5 [Abeand Noguchi, 1983]. Since that event is veryold for studyingthe focal mechanism,the 1990Mariana earthquake providesus with a goodopportunity to studythe seismotec- toniccharacteristics of this region. In thispaper, we analyze waveformdata to estimatethe source process. This earthquake alsocaused a tsunami that was observed on the Japanese is- landsas well as the Pacific islands.The unusualbehavior of thetsunami and the implication for hazardassessments of tsunamipotential is describedelsewhere [Satake et al., 1992]. 147OE 148oE 1E•akeResearch Institute, University ofTokyo 2DepartmentofGeological Sciences, University ofMichigan 140'E 145'E 150'E ,-- Fig. 1. Mapof thesource area. Star symbol denotes the main shockand circles denote the 3-dayaftershocks. Data arefrom Copyright1992 by theAmerican Geophysical Union. NE!C. Dotted line shows the trench axis. Bathymetric Papernumber 92GL000!65 contoursare also shown. Inset shows relocated epicenters and 0094-8534/92/92GL-00165503.00 triangledenotes the master event used for relocation. 297 298 Yoshidaet al.'The Mariana Earthquake of 1990 masterevent. It occurredat 14:57:20.1(LIT) on April 6 (rob= TABLE 1. Centroid moment tensor solution 5.9) near the southernend of the aftershockarea. The results ' Momenttensor elements (10'19 Nm) were almost same as the NEIC locations(Figure 1). The lengthof the aftershockarea is a little shorter,about 65 km. .... ...... ñ 0"10 The aftershocks are distributed in the N-S direction with a Moo 1.38 ñ 0.08 sparseregion in thecentral part. The main shock is locatedin M• 11.45 ñ 0.12 the southerncluster. Almost the same distribution is obtained Mro -2.73 ñ 0.59 for a masterevent near the northernend. This patternof the Mr• -6.86 -+ 0.56 aftershockactivity suggests an inhomogeneous distribution of MO• 0.09 stressdrop or strengthon thefault plane. Bestdouble couple ThepP phases were reported on NEIC for 14aftershocks. (!0z9Nm). '14A ThepP-P times range from !0 to 15 sec.Since the sea depth Plane a Plane b in theepicentral area is 5 km or so,it is probablethat pwP '(•(strike) ' "16.0ø .............i'86"iø phaseswere misread to pP phases.We consultedseismo- 15(dip) 60.4 ø 29.9 ø gramsrecorded at Dodaira Micro-earthquake Observatory and •, (slip) ,, -85.1 ø -98.6 ø its satellitestations at Kantoarea in Japan.The epicentraldis- tancesare about25 ø. Clearphases are observed 10 to 15 sec after the P wave arrival. For example,the event of the will get a steeperdip from the followinganalysis of body 22:52:59.9(UT) on April 5 showsclear phases 4 and 13 sec waves. after the P wave arrival. If we assume that the thickness of the waterlayer as 5 km andthe sourcedepth as 25 km, thephases BodyWave Analysis arriving4 and 13 secafter P wavecorrespond to pP andpwP phase,respectively. This suggeststhat the pP phasesreported For body wave analysis, we used P and SH waves by NEIC arepwP phases. This misreadingresults in overes- recordedat 13 stationsat epicentraldistances ranging from 300 timationof the focal depth.We calculatedthe sourcedepth to 100ø. The arrivaltimes were read directly on the broadband usingthe pP-P times of NEIC by assumingthat the phases are channel.All therecords were deconvolved to displacements. pwP and got the depthrange of 12 to 34 km for 12 after- We applieda multipledeconvolution method developed by shockswith an averageof 23 km anda standarddeviation of 6 Kikuchi andKanamori [ 1991]. By minimizingthe difference km. The fault width is estimated to be 40 km when the fault betweenthe observedand syntheticwaveforms, we deter- dip is 42ø which is derivedfrom following body wave minedthe mechanismsof subevents.For the computationof analysis. Green's functions, we assumeda structure consistingof a waterlayer 5 km thick and an oceaniccrust 10 km thicknear SurfaceWave Analysis the sourceregion. We computedGreen's functionsby assumingsource We performed a Centroid Moment Tensor (CMT) depths of 6, 10, 14, 16, 20, 25 and 30 km, and derived inversion[Dziewonski eta!., 198!; Kawakatsu,1989] using sourcemechanism for eachdepth with a sourcetime function long-periodsurface waveforms (mainly R1 andG 1) recorded of triangularshape. The residualerror was minimized for the at 19 stationsof the IRIS, GDSN GEOSCOPE [Romanowicz caseof 16 km. Fixingthe depthat 16 km andusing a point eta!., !984] and ERIOS [Takano et al., !990] networks.The source with the same mechanism as that obtained from the stations used were COL, HRV, BCAO, CAY, COR, SCZ, aboveanalysis, we frrstmade only oneiteration. The seismic PAS, KIP, PPT, CTAO, SLR, CMB, KEV, KMI, SSB, moment is obtainedto be 2.1 x 1020 Nm. The sourcetime INU, MAJO, KONO and TSK. The data were deconvolvedto functionis shownin Figure2 (a) andthe synthetic waveforms displacementsand bandpassfiltered betweenthe frequency areshown in Figure3. They showwater reverberation similar range from 3 to 7 mHz. Combining the data from these to the observation,although the detailsare different. The dif- stations,we havea goodazimuthal coverage. ferencemay be due to our simplifiedwater layer (uniform The final solutionfor the momenttensor elements is given depth)or dueto realcomplexity of the sourceprocess. If the in Table 1. The centroid location is 15.288øN and 147.397øE lattercase is true,more iterations should reveal the complex for a depthof 10 km. The inversionswere performed for the sourceprocess. We made five iterations.The seismicmoment selectivedepths of 10, 20 and33 km. The smallestvariance of is obtainedto be 3.0 x 1020Nm. Figure2 (b) showsthe thesolution was obtained for the depthof 10-20kin. The non- double-couplecomponent that is expressedas the ratio of the a) b) minimumeigenvalue to themaximum is only 8 % andnegli- gibly small.The momentof the bestdouble couple is 1.4 x (x10• Nm/s) (x10•'