GEOPHYSICAL RESEARCH LETTERS, VOL. 17, NO. 7, PAGES 1017-1020, JUNE 1990 RUPTURE PROCESSOF THE MACQUARIE RIDGE EARTHQUAKE OF MAY 23, 1989 Joehen Braunmiller and John N•b[lek Collegeof Oceanography, OregonStateUniversity, Corvallis, OR 97331 Australia plateboundary, in agreement withmotionpredicted IncorporatedResearchInstitutefor Seismology(IRIS) and GEOSCOPE are used to study the ruptureprocessof the Macquarie Ridgeearthquake. Firstwedetermine thecentroidal parameters: mechanism, depth,seismic momentandthesource time historyof therupture.Thenwe examineif thedatashow rupturespreading in a unilateralor bilateral fashionawayfrom its nucleationpoint and estimatethe fault lengthand stress byplatetectonic models. The totalseismicmomentis drop. Abstract.Broadband bodywavesrecordedat 15 digital seismic stationsworldwideare usedto studythe rupture pr•ess oftheMay23,1989Macquarie Ridgeearthquake. Thecentroidal solution(strike211o,dip 86%rake 180ø, and depth of10kmbelowtheseafloor) indicates shallow rupture •ithpureright-lateral strike-slip motionalongthePacific13.4x1020 Nm, 80% of whichwasreleasedin thefirst 24 s of fierupture process. Modeling favorsa bilaterally propagating rupture withslightly different dipandrakeforthenorthward 158øE 159' i 60 ø 161ø 162' 164"' 163ø andsouthward fault segmentsand similar momentrelease along bothdirections. Theestimated faultlengthis quiteshort, 50øS .: about 90 km, andthe derivedstressdropof 180 bar andaver- ... •-. age displacement of 17m areunusually high.Thebathymetry intheepicentral regionshows topographic segmentation ofthe ridge, possibly indicating faultsegmentation whichconfines ruptures to shortsegments. 51' Introduction ,::.:-. ...::.., TheMacquarie Ridgeearthquake of May 23, 1989with a .. o surfacewave magnitude(Ms) of 8.3 is the largestevent ,oddwidein more than a decade. The focal parameters according to the PreliminaryDeterminationof Epicenters ,PDE)catalogueare: latitude 52.3ø S, longitude160.6ø E, •,epth 10 km and origin time 10h54m46.3s.The-earthquake occurred in thecentralpart of the MacquarieRidge Complex , 5/B3/89 / t ßO,T'• i..:i •Figure1), which formsthe Pacific-Australiaplate boundary. andrunsapproximately north-south for 2000 km betweenthe South Island of New Zealand and the Pacific-Antarctic- )7 Australia triplejunction.Globalplatetectonicmodelsplacethe Pacific-Australia poleof rotationcloseto theMacquarieRidge '%..,.. .... •l• P•c,rlc {MinsterandJordan,1978; DeMets et al., 1990), thusrelative platemotionsalongthe plate boundarycan changerapidly, •hich is reflected in its evolution. The region originally formed in an ocean-spreading environment(Molnar et al., 1975) withtheyoungest radiometrically datedoceanicbasalts fromMacquarieIslandof about10 Ma age (Duncanand Va:ne,1988).Due to thesouthward migrationof therotation pole( Molnaret al., 1975; StockandMolnar, 1982), theridge subsequently servedasa strike-slipplateboundarywith the northernmost part(northof 50ø S) developing intoanoblique convergent boundary(Ruff et al., 1989)wheretheAustralian plate isbeingsubducted beneath thePacificplate. Bodywavesrecordedat 15 stationsfrom theGlobalDigital Seismic Network(GDSN)andthedigitalnetworksof the locations and source mechanisms of the 1989 earthquake (largestar)andtwothrustevents25 May 1966,M=6.5and23 May 1984,M=5.9 areshown.Thefirstweekof aftershocks is indicatedby solidcirclesandthoserecorded sincethenuntil December1989by opencircles(onlyaftershocks recorded by 20 or more stationsare depicted;PDE data).The regionin which 80% of the momentreleaseoccurredduring the 1989 eventis dashed.Bathymetryis shownby 500 m contours. Dark shadingindicatesareasdeeperthan 4500 m, light shading2000-4500m, andwhite areasare shallowerthan 2000 m. Topographicsegmentation of the ridge in the epicentralarea is apparent.The insertshowsa tectonic overview. The arrow indicates the relative plate motion Cop)a'ight 1990bytheAmerican Geophysical Union. betweenthe Pacific and Australianplatesaccordingto the NUVEL-1 model (DeMets et al., 1990). 1800 m depth contours aredashed, depthsgreaterthan4800m areshaded. Paper number 90GL01162 0094-8276/ 90 / 90GL-0116250 3. O0 1017 1018 BraunmillerandN•ib•lek: The 1989MacquarieRidgeEarthquake Data and Methods TABLE 1. SourceModelsDerivedby BodyWaveInversion The analysisfollowsthe procedures describedby N•ib•lek (!984). P andSH bodywavesrecordedby GDSN, IRIS and GEeSCOPE digitalstationsin the epicentraldistancerange Moment Depth Strike Dip Rake Duration cr 1020Nm km 13.4 deg deg deg s 10 211 52 86 180 between 30 ø and 100 ø for P waves and 30 ø and 90 ø for SH PS wavesareused.In thisdistancerange,the waveformsarenot contaminated by stronguppermanfieor corephases. The large magnitudeof the earthquakecausedsomeseismograms from stationsin the abovedistancerangeto behavenon-linearly; thesedatawereexcludedfrom the analysis.We deconvolved the instrumentresponseto obtainbroadbanddisplacement seismograms;dependingon the data quality and the seismo•aph type, the data are broadbandfrom the Nyquist periodof 0.6 s to a periodof 100 s or 200 s.For stations BJI andMDJ, whereonlylong-perioddatawereusedto obtainthe broadband seismograms, thesheri-period cutoffis 4 s. Dueto a highnoiselevelat stationZOBO thelong-period cutoffis 50 PSs 9.7 10' 211' 86* 180' 26 0.30 USS 9.3 10' 211' 86* 180' 26 0.34 USN 10.2 10' 211' 86* 180' 26 0.28 4.8 5.2 10' 10' 211' 21!* 86* 86* 180' 180' 16 28 4.7 5.5 10' 10' 210 212 95 82 174 187 16 30 S. The crustalstructurein the sourceregionis assumed to be a half-spaceoverlain by a 3.8 km thick water layer. At the receiverswe chosea crustalhalf-space.The P- and S-wave velocitiesof the half spaceare: 6.0 km/s,3.46 kin/s,with a BS1 0.26 south north BS2 0.24 south north The formalunceaaimies (2c0 are:depth0.5 km, orientation 2ø,andmerrett 20%. PS - pointsource,PSs- pointsourcewith durationrestricted to •mah moment release, USS - unilateral rupture towards southwest,USN unilateralrupturetowardsnortheast,BS1 - bilateralrupturewith soutin mechanismfixed, BS2 - bilateralrupturewith sourcemechanismfree. normalized variance of residuals. "*" - fixed. rockdensity of 2.8 g/cm 3. Theseismograms aresimultaneouslyinvertedin a least-squares sensefor the sourcemodel parameters.A point source is used to obtain a average (centroidal) description of thesource. To examinecomplexities aliasingstructuraleffectsinto the estimatedsourcetimehistory. Becauseabout80% of the seismicmomentis released in dueto sourcefiniteness andrupturepropagation, we useda propagatingline sourceparameterization as describedby N•ib•lek(1984)andBoydandN•ib•lek(1988).To investigate suchdetailsof theruptureprocess, it is necessary to precisely determine thearrivaltimes.The onsetof manyP wavesfor the MacquarieRidgeearthquake is quiteemergent; we determined the f•rstmotionusingbothbroadband andhigh-pass filtered waveformsandby checkingfor consistency with stationsat on thisdominantpartof theruptureprocess. For an eventof thissize,thepointsourcemodelprovides a surprisinglygoodfit to the data,indicatinga shortrupture length,although,it doesnot adequately matchtheamplitudes for stationsfrom azimuthsalongthe strikeof thefault (Figure 4). Moreover,it failsto matchthefirstmotionpolarities atKIP' andAFI. Matchingthe first motionpolaritiesfor all stations simultaneously wasnotpossible.Slightchanges in themechanismas the rupturepropagates,as well as deviationsfrom greatcirclepathsof the rays,can affectthe first motionsfor similar azimuths. Inversion Results The resultsof our analysisaresummarized in Table 1. The centroidalsolutionindicatesa purestrike-slipmechanism: strike211o, dip 86ø andrake180ø.Thevariationof themismatchbetweenthe observedandsyntheticseismograms asa functionof depth(Figure2) indicates thatthecentreidis less than20 km belowthe seafloor.Themechanism in thisdepth rangeremainsvery stable,whiletheseismicmomentvariesby about10% (Figure2). The leastmismatchis obtainedfor a pointsourceat 10 km belowthe seafloorwith a seismicmo- thefirst24 s,we concentrate in thesubsequent analysis only stations close to the P-wave nodes. To determinetheprimarydirectionof therupturepropagationwe introduce kinematicmodelswithruptures propagating unilaterallyandbilaterallyalongthestrikeof thefault.A rapture velocity of 2.0 km/s is usedand the depthand source mechanism arefixedto thepointsource estimate; in eachca we invert for the momentreleasealongthe fault. Sincewe havefour stations alongthestrikeof thefault (KIP andAFim 0.60 ............. mentof !3.4x1020Nm, whichcorresponds to a moment magnitude(Mw) of 8.02. Followinganemergentbeginning, about80% of the momentis releasedin a 22-second-long pulsefollowedby 3 smallerpulses(Figure3a). The fit to the 17 0.55 = 0.50 ;• 0.45 ._• 0.40 E 0.35 dataatrepresentative azimuthally distributed stations for•the pointsourcemodelis shownin Figure4. The P waveforms,especiallythosenear the nodesof the radiationpattern(KIP, AFI, RPN, CRZ andPAF),appearto bestrongly affected by waterreverberations. Thebathymetry playsanimportant rolein thecharacter of thesewaves(Wiens, 19.89)andourhorizontally layeredmodelis probably inadequatein explainingthe P-waveseismograms nearthe nodes. The later partof the sourcetimefunctionshowsperiodic pulseswith theperiodof thewaterreverberations, probably 0.30 0.25 0 . '• 10 ß , "' ß 20 3; ß , 40 ß 11 50 depth (km below seafloor) Fig. 2. Residualvariance(filled squares) andseismicmo_•:t (opensquares) asa functionof centmiddepth. BraunmillerandNfib•lek: The 1989MacquarieRidgeEarthquake 1.2 the waveformsand amplitudesbetterin thefirst 30-40 s than the othermodels.Laterpartsof the P waveforms, probably contaminated by waterreverberations (e.g.AFI P wave),are not modelledadequatelyby any of our models.Our final 1.0 0.8 0.6 model(Figure5 andTable1) hasthefollowing parameters: the 0.4 northern segment (Mo= 5.5x1020 Nm)hasa strikeof 212ø, 0.2 dip 82ø and rake 187ø, and the rupturelastsabout30 s, 0.0 whereas thesouthward segment (Mo= 4.7x1020 Nm)hasa 0 10 20 30 40 60 50 time (s) 0.3 0.2 0.! 0.0 1, s{ 1019 , i , -20 0 •uthwest ß ß i 0 ...... i 20 distance (kin) ß ß 40 ß 60 strikeof 210ø, dip95ø, rake174%andrupturelastingabout16 s. The kinematicbilateralmodel indicatesonly very small momentreleasein thevicinityof theepicenter. Thepossibility of additionalslip at the epicenterduringlater stagesof the ruptureis investigated by allowingtheslipat theepicenterto occurat arbitrarytime in a 25 s time frameandwe find that additionalslipmighthaveoccurredthere.Theestimated spatial distribution of theseismic momentfor a bilateral rupturemodel with and withoutreslip in the epicentralarea is shownin Figure 3b. northeast Fig.3. a) Seismic moment release rateasa functionof time (sometimefunction)for bestpointsourcePS (Table!). b) Seismicmoment release along the fault length for the bilaterally propagating source BS2(solidline) (Table1) anda bilateral rapturemodelwith additionalslip in the epicentral region duringlaterstagesof the rupture(dashedline).Both models fit thedataequallywell. thenorthandCRZ andPAF to the south),rupturepropagation effectsshouldbe detectable.A southwestwardpropagating rupture fits the datatheleast,whereasa rupturepropagating towards northeast slightlyimproves themisfitcompared tothe pointsourcesolution.The bestfit is obtainedby a bilateral rapture, in whichslightlymoremomentis releasedalongthe northward than the southwardsegment(Table 1). Figure 4 shows thematches to thedataatrepresentative stations. $1•2:•5o pm .... SH Becausethe P waves near the nodes continue to show sub- stantial mismatch, weinvestigate theimprovement thatcanbe achieved by allowingvariationin the mechanisms of the northward andsouthward segments. A smalldifferencein dip. aMrakeappears to considerably improvethemismatch. The bilateral modelwitha smallvariationin dipandrakematches AFI.P 41' PPT,P 68ø PAF,P 235' BJI,P 327" AF!.$H 4I ø CRZ,$H 227' Fig. 5. Observed(solid) and theoretical(dashed)P and SH wavesfor the modelBS2 in Table 1. Faultplanesolutionfor the sum of the two subevents is shown at the center of the figure.Seismograrn amplitudes arenormalizedto an epicentral distance of 40 ø. SummaryandDiscussion The centroidalsolutionindicatesalmostpureright-lateral strike-slipwith a strikeof 31o, whichis in goodagreement with thelocaltrendof theridgein theepicentral regionwhere it hasa strikeof 33ø. Globalplatetectonic modelspredicta very similardirectionof platemotion.The RM2 modelof MinsterandJordan(!978) predictsanazimuthof 53ø anda rate of 2.7 cm/a, and the NUVEL-1 model(DeMetset al., anazimuthof 41ø andarateof 3.0cm/afor the Fig,4. Comparison of P andSH waveforms for different 199.0)assigns ...models listedin Table 1. Number next to the stationcodeis the •ufft fromthe source. platemotionin theepicentral region. Our total seismic moment estimate is close to that of the 1020 Braunmiller andN•ib•lek:The1989Macquarie RidgeEarthquake long-period CMTsolution (14x1020 Nm).Ourresults show DavidReinerr drafted thebaihymetric map.Thisproject was that about 80% of this moment was releasedin the first 24 s supported by the NationalScienceFoundation•ant EAR- of theruptureprocess. Observable directivitydueto rupture 8896187. propagation is surprisingly small,partlydueto shortfault lengthandpartlyto bilaterally propagating rupture. A model withrupturepropagating unilaterally towards northeast fits betterthana modelwith a purelysouthwestward rupture.A bilaterally propagating rupturewitha smallvariation in the sourcemechanismsbetweenthe southernand northernseg- ments prov/des thebestfit tothedata.Thelengthof thenortheastsegment is about55 kmandtheseismic moment released alongit is about5.5x1020 Nm. Thesouthwest segment is shorter(35 km) and the seismicmoment slightly lower (4.7x1020Nm). Initially,onlya smallamount of moment wasreleased in the epicentral region,butthereis a possibility thattheregionreslippedduringlaterstages of theruptureprocess. Theinitial lowmomentrelease in theepicentral regionandthesmalldifferencein dip of thetwo faultsegments suggest a possibility thattheMacquarieRidgeearthquake nucleated neara geometric barrierformedby a smallchangein the dip of the fault. King and N•b•lek (1985) and King (1986) showedthat earthquakes oftennucleate nearfaultbendsor othergeometricalcomplexities of thefaultzone. References Banghat,A. R., and L. R. Sykes,Focal mechanisms of earthquakes in theIndianOceanandadjacent regions, J. Geophys.Res., 74, 632-649, 1969. Boyd, T. M., and J. L. Ntib•lek, Ruptureprocessof the AndreanofIslandsEarthquake of May 7, 1986,Bull. Seisra. $oc. Am., 78, 1653-1673, 1988. DeMets, C., R. G. Gordon, D. F. Argus, and S. Stein, Currentplatemotions,Geophys. J. Int., in press. Duncan,R. A., andR. Vame, The age anddistributionof the igneousrocks of Macquarie Island, Pap. Proc. R. $oc. Tasm., I22(1), 45-50, 1988. Kanamori, H., and D. L. Anderson,Theoretical basisof sou• empiricalrelationsin seisinology,Bull. Seisin.Soc.Am. 65, 1073-1095, 1975. King, G. C. P., Speculations on the geometryof theinitiatim andterminationprocesses of earthquake ruptureanditsrelation to morphologyandgeologicalstructure,Pageoph,124, 567-585, 1986. The rupturelengthof 90 km associated with the main momentreleasebasedon the bilateral rupturemodel is unusually shortfor aneventof thismagnitude. Thefightclus- King, G. C. P., andJ. L. N/ib•lek,Theroleof bendsin faults in the initiation and termination of earthquakerupture, ter of the aftershocks (Figure1) alsoindicatesthattherupture Minster, J. B., and T. H. Jordan,Present-dayplate motions, J. Geophys.Res., 83, 5331-5354, 1978. Molnar, P., T. Atwater, J. Mamrnerickx, and S. M. Smith, Magneticanomalies,bathymetry,andthe tectonicevolu•on of the southPacific sincelate Cretaceous,Geophys.J. R. probablydid notextendmorethan100 km. The stressdrop associated withthedominant90 km of therupture,takingthe faultwidthof 20 km (twicethecentroidal depth)anda seismic momentof 10.2x1020Nm, is 180barandthe averagedisplacement is 17 m. For a 100-kin-long ruptureandthe total seismicmomentof 13.4x1020Nm, thesevaluesare2!0 bar and20 m. HencetheMacquarieRidgeearthquake showsone of thelargeststressdropsfor interplatestrike-slipeventsin thismagnitude range(seeKanamoriandAnderson,1975). The relativelyshortfault lengthandlargestressdropof this eventmay be relatedto the segmentation of the Macquarie Ridgein theepicentral region.Bathymetric chartsof theepicentralregion(Figure1) showthattheridgeis brokenup into smallersegments, on the orderof 50-100 km long. One of thesetopographic segments coincideswith what we determinedto be theregionof theprimarymomentreleaseduring the !989 event.Two thrustevents(BangharandSykes,1969; HarvardCMT solution)thatoccurredin 1966 and 1984 seem to brackettheregionthatbrokein thestrike-slip fashionduring the 1989 event,indicatingfault segmentationon the same orderas seenin thetopography. We submitthereforethatthe observedridge segmentation relatesto fault segmentation whichconfinesrupturesto shortsegments. Acknowledgements. We thankDavidChavezfor providing IDA data and Larry Ruff for providingGEOSCOPE data, whichultimatelywerecontributed by BarbaraRomanowicz. Science, 228, 984-987, 1985. Astr. $oc., 40, 383-420, 1975. N/ib•lek, J. L., Determination of earthquake sourceparameters frominversionof bodywaves,Ph.D. Dissertation, 360pp, Massachusetts Instituteof Technolog3r, Cambridge, 1984. Ruff, L. J., J. W. Given, C. O. Sanders,and C. M. Sperber, Large earthquakesin the Macquarie Ridge Complex: Transitional tectonics andsubducfion initiation,Pageop,q, 129, 71-129, 1989. Stock,J., andP. Molnar,Uncertainties in the relativepositionsof the Australian,Antarctica,Lord Howe, andPac•.fic platessincethe !ate Cretaceous, J. Geophys. Res.,87, 4697-4714, 1982. Wiens,D.A., Bathymetric effectson bodywaveforms from shallowsubduction zoneearthquakes andapplication m seismicprocesses in theKurileTrench,J. Geophys. Res., 94, 2955-2972, 1989. J. BraunmillerandJ. N•b•lek, Collegeof Oceanography, OregonStateUniversity,Corvallis,OR 97331 (ReceivedMarch 19, 1990; revisedMay 21, 1990; accepted May 21, 1990)