GEOPHYSICALRESEARCHLETTERS, VOL. 23, NO. 14, PAGES 1817-1820,JULY 1, 1996

Seismicslip on a low anglenormalfault in the Gulf of Corinth: Evidencefrom high-resolutionclusteranalysisof m icroear th quakes

Andreas Rietbrock •, Christel Tiberi 2,Frank Scherbaum • andH61•ne Lyon-Caen 2 Abstract.

The Gulf of Corinth in Western Greece is one nodal planes(Rigo, 1994; Rigo et al., 1996). This area is a of the most active extensional zones in the Aegean very active asymmetricgrabenwith quaternaryslip ratesof

region.It is still an openquestionwhetherextension can the orderof 1 cm/yr (Armijo et al., 1996) anda presentrate of (Briole et al., be activelyaccommodated on low anglefaultingor if openingof 1.5 cm/yr from GPS measurements those faults as seen in geologicalrecordshave been rotated.Whilst numerousfault planesolutionsobtained froma densetemporarynetworkdeployedin thewestern part of the gulf in July-Augustof 1991 showedone of the nodal planesas a subhorizomalplane, slip on the high-angleconjugateplaneis equallyprobablefrom the focal mechanismdata (Rigo et al. 1996). Sincepart of theactivityoccurredin spatialclusterswith similarfocal mechanisms, we useda high resolutionclusteranalysis to determinethe mostlikely activeplane.Exploitingthe waveformsimilarityof theseevents,relativeonsettimes of P and S waves could be determinedat subsample accuracy(less than 0.01 s). The clusteranalyzedhere contains12 evems, among these 8 have a well constrainednormalfaultingfault planesolutionwith a shal-

1993; Rigo, 1994). The main active faults located at the

southern edgeof the Gulf strikeE-W anddip northat 45-50ø (Fig. 1). No clear activity could be associatedwith the main active faults observed at the surface (Fig. 1). Rather, the evems are concentrated in a 2-5 km thick zone at 6-10 km

depth. On the basisof microearthquake locationsand fault planesolutionsRigo et al. (1996) suggested the existenceof an activenorthdippinglow angledetachmentfault at a depth of about10 lcm.This had alreadybeensuggested(King et al., 1985) on the basis of geometricalconstraintsin the eastern part of the Gulf. However,becauseof the ambiguitybetween fault plane and aux'fiiaryplane in the fault plane solution determinations,Rigo et al. (1996) could not rule out that slip occunexton a seriesof steeplysouthdippingplanesdefining an overall shallownorthdippingdetachmentzone. low (12-20ø) northdippingplaneanda steeplysouth In this study,we make use of the fact that a considerable dippingplane.A masterevemrelocationshowsthatthe part of the seismicactivity in the Gulf of Corinth occunextin relocated12 hypocentercemroidsare alignedalongthe spatialclustersto demonstratethat a subhorizontalplanewas low angleplane showingclear evidencefor active low seismicallyactive. We do this by determininghigh precision relative locations for cluster events which axe candidates for

anglenormalfaulting.

shallow dipping normal faults. While the conventionallocation accuracy(depthresolutionlessthan 1 kin) is insufficient to conclusivelydeterminewhethersome of the eventsoccur Introduction on the subhorizontalplane, the relative location accuracy The generationof shallowdippingnormal faultsin regions reached for spatially clusteredevents has been shown in of extensionalstressis a puzzlestill waitingto be solvedcon- manycasesto be sufficientto determinethe activefault plane clusively. Although there is some evidence suggestingthe (depth resolution better than 100 m) (e.g. Fremont and possibilityof activeseismicslip on subhorizontal planes(e.g. Malone, 1987; Deichmann and Gracia-Femandez, 1992; Got Abers, 1991; Doser, 1987), conclusive seismologicalevi- et al., 1994). The main prerequisitefor the achievementof denceis still lacking.Globalcompilationsof fault planesolu- this kind of accuracyis sufficientwaveformsimilarityfor the tionsfor normalfaultingshowthe vastmajorityof the nodal individualclusterevents.This allows the useof high resoluplanesin the rangebetween30ø - 60ø (Jackson, 1987;Jack- tion onsettime determinationtechniquessuchas cross-specson and White, 1989). Moreover, based on Coulomb's frac- tral or cross-correlation methods(Nakamura,1978; Poupinet ture criterion,normal faulting is mechanicallyimpossibleon et al., 1984; Ito, 1985; Scherbaumand Wendlet, 1986; Deich1 shallowdippingplanesundernormalconditions(o vertical) mann and Gmcia-Femandez, 1992; Maurer and Deichmann, (e.g. Jaegerand Cook, 1979; Buck, 1988). 1996). In general,waveformcoherencerequiresproximityof Recently, however, a microseismicstudy in the Guff of the sourcesand similarity of the focal mechanismsand the Corinth revealed a significant number of well constrained sourcetime functions.We usedthe completewaveformdatanormalfaultingmechanisms with shallow(10-20ø) dipping set from the 1991 seismologicalexperimem to searchfor earthqu•e clusterswith similar waveforms.

llnstitutfar Allgemeine undAngewandte Geophysik der Ludwig-Maximilians Universititt Mtinchen, Mtinchen, FRG

Cluster

2Laboratoire deSismologie LIRACNRS195,Institut dePhysiqueduGlobe,Paris,France

Copyright1996by the AmericanGeophysical Union. Papernumber96GL01257 0094-8534/96/96GL-01257505.00

detection

For the detection of earthquake clusters we used the method proposedby Maurer and Deichmann (1996). It is basedon the evaluationof waveformcoherencefor all possible combinationsof events in a dataset.As a quantitative measureof waveform similarity, the cross-correlation coefficientsare computedfor time windowscontainingthe P andS

1817

1818

R.II•TBROCK ET AL.: SEISMIC SLIP ON A LOW ANGLE NORMAL FAULT IN THE GULF OF CORINTH

wave portionof the seismograms. The lengthof the time window usedwas 1 s starting 10% before the wave onsettime. These values are taken as the elements of a correlation matrix

ß

ß

with the row and column indicescorresponding to event numbers.This matrix is subsequently analyzedto identify rows with similar row patternswhich indicatethat the correspondingeventsshowa similar amountof waveformsimilarity for the samestationcombinations. In contrastto Maurer and Deichmann (1996) we did not

restrictthe analysisto singlecomponent recordsbut usedall componentsavailable.The highestcorrelationcoefficiemon eithercomponemwas takenas a measureof waveformsimilarity. Only thosewaveformswhichresultedin a correlation coefficientlarger than 0.75 where treated as similar. We found that the high correlationcoefficientsfor the S waves couldonly be foundon the horizontalcomponents. In total, we determined about 20 clusters based on waveform similar-

•.

ß•//

-0.6 -0.4 -0.2 0.0 0.2

0.4•.',

•/

ity. In this work we focusour interestontoonecluster,which correspondsto normal faulting eventswith one of the nodal Figure 2. Locationsof the clustereventsbefore(circles)and planesshowingdip anglesbelow25ø (Fig. 1 andFig. 2). To after (triangles)relocation.The starindicatesthe locationof showthe degreeof waveformsimilarityobtained,the vertical the masterevent.Magnitudesrangefrom 2.0 and2.8. Lower componentrecordsfor all clustereventsfor stations"kamb" hemisphere focal sphereswith observedpolaritiesare shown and "anoz" are displayedin Fig. 3. for all events.Well constrained nodalplanesareplotted. Relocation

(lessthan 0.01 s). This is doneby usingthe phaseof the

Since in the context of the current problem we are not interested in absolute locations, we can use a master event

cross-spectrum of the waveforms(Ito, 1985; Scherbaumand

Wendlet,1986). In the first step of the analysis,the two

in a small time winrelocationtechnique,for which the relative locationerror is phasesare alignedby cross-correlation dow (~ 0.2 s) centered around the predetermined onsettime. considerablyreduced.We exploitthe waveformsimilarityto Subsequently, the cross-spectrum is computed with the determine the relative onset times at subsampleaccuracy aligned traces and the relative onset times are determined

from the slopeof the phaseof the crossspectrumby least

a)

21'154' _

22'• 03' thio

I

22'112' •

k M

I

squaresfit. We addedtwo constraintsto the determinationof

the slope.Firstly,the regression line was only calculatedfor thosefrequencies for whichthecoherence of the phaseswas greaterthan0.8 andthenormalizedcrossspectrum amplitude was greaterthan 0.1. Secondly,we requiredthe regression line to passthroughthe origin as expectedfor noise-free coherentsignals.The errorof the onsettime wascomputed fromtheerrorof theslopedetermination at thehighestusable frequency, thusprovidingan upperlimit estimate.In cases wherethe slopeof thephasecorresponds to time differences

largerthan1 sample(afterpre-alignment by thecrosscorrelation) we assumed that the onset time determination o •

•o

•_

•

-38' 15'

o I•

b)

s 0

I

•

•-•5

0

sis wasperformedfor the P and S wave onsets.

n I

=-1oo•• -20

obtainedfrom the cross-correlation couldnot be improved andthetimingerrorwasassumed to be 1 sample.Thisanaly-

1•

20

I

•

O

3'0 40

km

Figure 1. a) Microse•micity •o• d•g 2 monks • 1991 by 55 •git• smtio• (one stafi• eve• 4-5 • •ord•g at 1• or 2•Hz) •igo, 1994). O•y the statio• us• • • pa•r for •e rel•ation m shorn. Active noml fauEs • •e

m•k•. C1 •dicates •e NS •h• •e •cates

l•ation of •e studi• cluster. •e l•ati• of •e cross s•-

tion.b) vefficalcross•fion •ough •e sm•

clnsmr,

00

0'2

0'4 '

0'6 '

0'8 '

Time (s)

02

04

06

08

Time (s)

Figure 3. Waveformsof the verticalcomponents recordedat stations"kamb" and "anoz" for the 12 eventsforming the cluster.The mastereventis shownat the top.

RIETBROCK

ET AL.: SEISMIC SLIP ON A LOW ANGLE NORMAL FAULT IN THE GULF OF CORINTH

1819

N

Relative relocation of the cluster events was performed

basedonhomogeneity of velocityin the clusterregion,with the raysfor all the eventsto a particularstationleavingthe clusterregionunderthe sameslowness vector(e.g. Console and Di Giovambattista, 1987). This line•s

the location

problemwhichwassolvedby singularvaluedecomposition (Presset al. 1992). For the relocationwe used6-11 phase readings for eachevent(atleast4 P and2 S wavephasereadings).The distribution of the relocatedeventsis shownin Fig. 4. In orderto determine thelocationaccuracy, we useda Monte Carlo techniqueto perturbthe onsettimes and the velocitymodel,assuming thatbotharesubjectto error.It was assumedthat the onset time errors follow a Gaussiandistribu-

tion with the errorof the phaseslopeas standarddeviation. Furthermore,it was also assumedthat the velocity of the clusterregionwas6.0 + 0.1 km (Rigoet al, 1996).1000synthetic data setswere computedin which the relative onset timesandthe velocitymodelaccordingto thesedistributions wereperturbed,and a subsequent relocationwasperformed. S For eachof the spatialcoordina_tes the rangecontaining95% Figure 5. Pole distributionof best fitting planescomputed of the relocationswas determined(95% being taken as the from 1000 randomlyselectedevent combinationsaccording uncertaintymeasure).The correspondingerror bars are to the determinedlocation errors (gray circles). Superimshownin Fig. 4. Accordingto thisnumericalexperimentthe posedare with a black triangle the pole of bestfittingplane relocated hypocenters areknownwithina fewtensof meters. for the final location,with a black square the maximumof the pole distributiondeterminedby the three point method (Fehleret al., 1992), and with a blackcircle the pole of the shallow dipping plane determined from a composite fault To determine whether the relocatedevents are aligned plane solution.

Determination of the active plane

alongany of the two nodalplanesobtainedfrom the focal mechanisms,two different methods were used.

Firstly the bestfitting planefor the final eventlocations using a least squarefit was determined.To investigatethe accuracyof the predetermined best fitting plane a Monte Carlo techniquewas employed.Randomly1000 eventcombinationswere selectedaccordingto the determinedlocation errors.For each of thesecombinationsthe best fitting plane wascomputedandthe pole of the planewasplottedin a stereographicprojection.The obtainedpoledistributionis shown in Fig. 5 (graycircles).Superimposed with a blacktriangleis the pole of the bestfittingplanefor the final eventlocations. As onecan seethe plungeof the planeis determinedwith an

As an independentapproach,the three-pointmethodproposedby Fehler et al. (1992) was also used to determine internal alignmentsin spatially distributedpoint sets. The philosophybehindthisapproachis thateverycombinationof threepointsdefinesa planewhichcan be assigneda pole. If thepolesof all theseplanesareplottedin a stereographic projection, preferred planes should show up by clusteringof poles. The maximum of the pole distributioncorresponds to the preferredplanein the clusterdistributionandis indicated in Fig. 5 with a black square.Superimposedin Fig. 5 is also accuracyof 10ø. The shortestaveragedistancebetweenall the pole of the shallow dipping nodal plane obtainedfrom a used event combinationsto the best fittings plane is about compositefault plane solutionfor all selectedevents(black circle). All determinedpoles have a very small plunge and 43 m. concentrate in an area of 10ø. S

N

Discussion

The analysispresentedabovedemonstrates that the hypo-

centersof the 12 eventsstudiedlie on a planedipping10ø north.Uncertaintiesassociated with the dip of the northdip-

• -9.75 •1 • "' :

ping nodalplanein the fault planesolutionare due to uncertaintiesin the assumedvelocitystructure.Althoughthe mean velocitystructurewaswell constrained by Rigo (1994), inci-

::::::":::::::::::::::::::: -

denceanglescanvaryby about10ø whensmallperturbations in the velocity model are introduced.We thus estimatethe

-10.00

uncertainty on the nodalplanedip to be 15ø. The maximum

• 0.0

0.5

Relative distance(km)

1.0

possibledip for the north dipping plane would be less than

30ø.Consequently, it connotcompletely beruledoutthatslip occurson a seriesof either250-30ø northdippingplanesor on600-80ø southdippingplaneswhilethecentroids of the 12

Figure4. N-S crosssectionof therelocated events.The95% errorbarscomputedby the Monte Carlotechniquearegiven eventswould be alignedon a subhorizontalplane.This howfor eachevent.The thick gray lines indicatethe two nodal ever is not likely to have happened:the ruptureareaof each eventis of the orderof 100 m by 100 m andthereis no reason planesfroma composite faultplanesolutionfor all events.

1820

RIETBROCK

ET AL.: SEISMIC SLIP ON A LOW ANGLE NORMAL FAULT IN THE GULF OF CORINTH

that all centroidswould lie on the sameplane. The coincidenceof the internalalignmentplaneof the hypocenters with

oneof the nodalplanesdetermined for theseeventsis therefore interpreted as a clearexampleof seismicslip on a subhorizontal normal fault.

Mechanismscausingseismicslip on a low anglenormal fault requireeither very low frictioncoefficients(g of the

orderof 0.1),ornonvertical c•1.Thelattercouldbedueto

Got, J.-L., J. Frechet,andF. W. Klein, Deepfaultplanegeometry from multipictrelativerelocationbeneaththe southflank of Kilauea,J. Geophys.R., 99, 15375-15386,1994. Ito, A., High resolutionrelative hypocentersof similar earthquakesby crossspectralanalysismethod,J. Phys.Earth, 33, 279-294, 1985.

lacksonI., Active normalfaultingandcrustalextension,in ContinentalExtensionalTectonics,ed by M. Coward,I. Dewey andP. Hancock,GeologicalSocietySpecialPublication,28,

stressrefractionat interfaceswith strongviscositycontrasts 3-17,1987 (Bradshawand Zoback, 1988), or to stressperturbationnear JacksonJ., andN. White,Normalfaultingin theuppercontinenzonesof stressconcentration, or high porepressures closeto tal crust: Observationsfrom regions of active extension,J.

c•3 andconfined to thefaultzone(Sibson, 1985;Rice,1992).

We do not have enough informationat presentto favor or reject one or the other of these mechanisms. However we

postulatethatthe microseismicity recordedis restricted to a zone of creepat this depth,wherethe frictioncoefficientis likely to be low. Alternatively,fluid pressures may play an importantrole and independentinformationis urgently neoded in order to choose between the different mechanisms.

Struct. Geol.,11,15-36, 1989

laegerI. andN. Cook,Fundamentals of rockmechanics, ChapmanandHall, p.593,London,1979 King, G.C.P.,Z. Ouyang,P. Papadimitriou,A. Deschamps, I. Gagnepain,G. Houseman,I. Jackson,C. SouflerisandI. Virieux, The evolutionof the Gulf of Corinth (Greece):an aftershock study of the 1981 earthquakes,Geophys.J. R. astr. Soc.,80,677-683, 1985 Maurer, H. and N. Deichmann,Microearthquakeclusterdetec-

tionbasedon waveformsimilarities,with an applicationto the Acknowledgements. The Institut des Sciencesde l'Univers westernSwissAlps, Geophys. J. Int., in press,1996. (INSU) andthe DirectionauxRisquesMajeurs(DRM) provided Nakamura,Y., A 1 moonquakes: Sourcedistribution andmechasupportfor collectingthe data.We alsobenefited from theEC nism,Proc.HumanPlanet.Sci.Conf.,9th, 3589-3607,1978. EPOC-CT91-0043 contractand from a french-germanPRO- Poupinet,G., W. L. Ellsworth,andJ. Frechet,MonitoringVelocCOPE cooperation program.This work would not have been ity Variationsin the Crust Using EarthquakeDoublets:An possiblewithoutthe data analysisperformedby Alexis Rigo. Applicationto the CalaverasFault,California,J. Geophys.R., Hansruedi Maurersupported uswiththesourcecodeof hisclus89, 5719-5731, 1984. ter detectionalgorithm.

Press,W. H., Flannery,B. P.,Teukolsky,S. A. andVetteding,W. T, Numericalrecipesin C - TheArt of ScientificComputing, 735 pp., CambridgeUniversityPress,Cambridge,1992. References Rice J., Fault stressstates,pore pressuredistributionsand the weaknessof the San Andreasfault, in Fault mechanicsand Abers,G., Possibleseismogenic shallow-dipping normalfaults transportpropertiesof rocks,ed. By B. EvansandTF. Wong, in theWoodlark-D'Entrecasteaux extensionalprovince,Papua New Guinea,Geology,19, 1205-1208,1991. Academicpress,London,475-53, 1992. et g6odtsiquedu Golfe de CorArmijo R., B. Meyer,G. King, A. Rigo andD. Papanastassiou,Rigo, A., Etudesismotectonique inthe (Grace), Th•se de Doctorat, Universit6 Pads 7, France, Quaternaryevolutionof the Corinthrift and its implications 1994. for the late Cenozoicevolutionof the Aegean,Geophys.J. Rigo A., H. Lyon-Caen,R. Armijo, A. Deschamps, D. Hatzfeld, Int., in press. BradshawG. and M. Zoback, Listric normal faulting, stress K. Makropoulos,andP. Papadimitriou,A microseismic study refraction, and the state of stressin the Gulf Coast Basin, in the westernpart of the Gulf of Corinth(Greece):Implicationsfor largescalenormalfaultingmechanisms, Geophys. J. Geology,16, 271-274, 1988. Int, in press. Briole P., I.C. Ruegg,H. Lyon-Caen,A. Rigo,K. Papazissi, D. F. andWendler,I., Crossspectralanalysisof SwaHatzfeldandA. Deschamps; Activedeformation of thegulf of Scherbaum, bian lura (SW Germany)three-component microearthquake Corinth, Greece:Resultsof repeatedGPS surveysbetween recordings, J. Geophys.,60, 157-166,1986. 1990 and1993,Anna!esGeophysicae, 12, C65, 1994. Buck W.R., Flexural rotation of normal faults, Tectonics,7,959973, 1988.

Console,R. andR. Di Giovambattista,Local earthquakerelative locationby digital records,Phys.Earth Planet. Int., 43-49, 1987.

Deichmann,N. and Gracia-Fernandez, M., RuptureGeometry from High-Precision Relative Hypocenter Locations of Microearthquake Clusters,Geophys. J., 110,501-517,1992. DoserD., The Ancash,Peru,earthquakeof 1946 November10: evidencefor low angienormalfaultingin the high Andesof northernPeru,Geophys.J. R. Astr.Soc.,91, 57-71, 1987. Fehler,M., House,L. andKaieda,H., Determiningplanesalong which earthquakesoccur: Method and applicationto earthquakesaccompanying hydraulicfracturing,J. Geophys.R., 92, 9407-9419, 1987.

Sibson R., A note on fault reactivation, J. Struct. Geoi.,7,751754, 1985

H. Lyon-Caen,Institutde Physiquedu Globe,4 Placelussieu,

75252 Paris, Cedex 05, France,(e-mail: [email protected])

A. Rietbrock,Institutf'tirAllgemeineund Angewandte Geophysik, Theresienstr.41/IV, 80333 Mtinchen, FRG, (e-mall: andreas•ftaucher. geophysik.uni-muenchen.de) F. Scherbaum,Institutfar AllgemeineundAngewandte Geophysik, Theresienstr.41/IV, 80333 Mtinchen, FRG, (e-mall: [email protected]) C. Tiberi, Institut de Physiquedu Globe, 4 Place Iussieu,

75252Paris,Cedex05, France,(e-mail:[email protected])

Fremont, M.-I., and S. D. Malone, High PrecisionRelative Locationsof Earthquakesat Mount St. Helens,Washington, J. (ReceivedDecember2, 1995;revisedMarch 15, 1996; Geophys.R., 92, 10223-10236,1987. acceptedMarch 21, 1996)