Active Deformation of the Corinth Rift, Greece: Results from Repeated Global Positioning System Surveys Between 1990 and 1995 Pierre Briole, Alexis Rigo, H

Active deformation of the Corinth rift, Greece: Results from repeated Global Positioning System surveys between 1990 and 1995 Pierre Briole, Alexis Rigo, H. Lyon-Caen, Jean Claude Ruegg, K. Papazissi, C. Mitsakaki, A. Balodimou, G. Veis, D. Hatzfeld, Anne Deschamps

To cite this version:

Pierre Briole, Alexis Rigo, H. Lyon-Caen, Jean Claude Ruegg, K. Papazissi, et al.. Active deformation of the Corinth rift, Greece: Results from repeated Global Positioning System surveys between 1990 and 1995. Journal of Geophysical Research : Solid Earth, American Geophysical Union, 2000, 105 (B11), pp.25605 - 25625. 10.1029/2000JB900148. hal-01416924

HAL Id: hal-01416924 https://hal.archives-ouvertes.fr/hal-01416924 Submitted on 26 May 2020

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 105, NO. Bll, PAGES 25,605-25,625, NOVEMBER 10, 2000

Active deformation of the Corinth rift Greece' Results from repeated Global Positioning System surveys between 1990 and 1995 P. Briole,1 A. Rigo,e H. Lyon-Caen,3 J. C. Ruegg,1 K. Papazissi,4 C. Mitsakaki,4 A. Balodimou,4 (3. Veis,4 D. Hatzfeld,• and A. Deschamps•

Abstract. Between 1990 and 1995, we carried out seven Global Positioning System(GPS) c•mp•ignsin the Corinth rift •re• in order to constrainthe spatial •nd temporal crustal deformation of this •ctive zone. The network, 193 points over--10,000 km e, s•mplesmost of the •ctive f•ults. In order to estimatethe deformationover • longer period, 159 of those points •re •lso Greek triangulation pillars previouslymeasured between 1966 •nd 1972. Two e•rthqu•kes of m•gnitude 6.2 •nd 5.9 h•ve occurred in the network since it w•s inst•lled. The extension rate deducedfrom the analysisof the different GPS data sets is 14 4- 2 mm/yr orientedN9øE in the west, 13 4- 3 mm/yr orientedS-N in the center, and 10 4- 4 mm/yr orientedN19øW in the east of the gulf. The comparisonbetween GPS and triangulationgives higher rates and lessangular divergence (25 4- 7 mm/yr, N4øE; 22 4- 7 mm/yr, S-N; 20 4- 7 mm/yr, N15øW,respectively). Both setsof data indicate that the deformingzone is very narrow(10-15 kin) in the west,might be widerin the center(15-20 kin), and is morediffuse in the east. The analysisof the displacements observedafter the M•-6.2, June 15, 1995, and the M•=5.9, November 18, 1992, earthquakes,both located in the west of the gulf, together with seismologicaland tectonicobservations shows that thesetwo earthquakesoccurred on low-angle(_< 35ø) north dippingnormal faults locatedbetween 4.5 and 10 km depth in the inner part of the rift. Assumingthat the deformationis concentratedin relatively narrow deformingzones, we use a simple model of a dislocationin an elastic half-space to study the implication of the localization. Using the geometry of the known seismogenicfaults, our observationsimply continuousaseismic deformation in the uppermostcrust of the inner rift. This model predicts geodeticstrain rates closeto seismicstrain rates in oppositionto previousestimates. This is becauseour model takes into accountthe activity on low-angle normal faults in the inner rift and an effectiveseismogenic layer of 6-7 kin, about half that usually assumed.

1. Introduction the world. [e.g., Mercier et al., 1977; McKenzie, 1978; Le Pichon and Angelier, 1981; Jackson and McKen- The Aegeanis the most seismicallyactive part of Eu- zie, 1988; Le Pichon et al., 1995]. In northern Greece rope and one of the most rapidly extending provincesof the deformation is accommodated across a series of ex- tendinggrabens (North Aegeantrough, Evvia graben, •D6partementde Sismologie,UMR-CNRS 7580, Institut Corinthrift) delimitedby activenormal faults oriented de Physique du Globe, Paris, France. NW-SE to E-W. These grabens connect the western 2Groupede Rechercheen G6od•sieSpatiale, Observatoire part of the North Anatolian fault to the Hellenic Arc Midi Pyrenees, Toulouse, France. [Armijo et al., 1996]. The Corinth rift (Figure 1) is 3Laboratoirede G•ologie,Ecole Normale Sup•rieure,Paris, France. the most active of these grabens and the most acces- 4HigherGeodesy Laboratory, National TechnicalUniver- sible one to observations because only its central part sity, Athens, Greece. is presently under sea level. It thus provides a very SLaboratoirede G6ophysiqueInterne et Tectonophysique, good opportunity to study in detail crustal deforma- Observatoire de Grenoble, Grenoble, France. tion processesinvolved in suchactive rifting. The work 6Laboratoire de G•odynamique, CNRS Sophia-Antipolis, presentedin this paper is part of a European multidisci- Valbonne, France. plinary effort aimed at a better understandingof these processesin that area. Results from seven Global Po- Copyright 2000 by the American GeophysicalUnion. sitioningSystem (GPS) campaignmeasurements of a Paper number 2000JB900148. densenetwork first installedin 1990 and now covering 0148-0227/00/2000JB900148509.00 the entire gulf are given. Besidesproviding accurate in- 25,605 25,606 BRIOLE ET AL.- DEFORMATION OF THE CORINTH RIFT

& Livadia

1981t03/05 1981/03/04

1ø, . 24Alklom 'des • ..... 1981•3/0

1981t0•2

Corinth

22 00' 22 30' 23 00' Figure 1. Seismotectonicmap of the Corinth rift showing topography, active faults, and fault plane solutionsof earthquakes > Me = 5.5 for the period 1965 to present. Adapted fi'om Armtie nigo .t solution parameters are listed in Table 3. Ps, -He, and Xi refer to the Psatopirgos, Hclike, and Xilokastro faults, respectively.

stantaneous extension rates acrossthe rift, the density ßthe seismicity of the Corinth rift included six events of of observationsboth in spaceand time allows us to give magnitude ]1.I• _> 6 (Eratini, 1965, M•-6.4; Antikira, new insightson the localizationof the deformation, the 1970, M•=6.2; Corinth 1981, M•=6.7, ;lL=6.4, and effective thickness of the brittle crustal layer, and the M•=6.2; Aigion, 1995, M•=6.2). The focal mechanisms relation between the continuouspart of the deformation of all these earthquakes represent almost pure normal (loadingprocess) and the discontinuouspart associated faulting with a N-S to NNW-SSE extension direction with earthquakes. For this purpose, we use available [e.g., Baker et al., 1997] (Figure 1). Althoughthe two geodetic, tectonic, and seismologicalobservations out- main events of the 1981 sequence ruptured known 45- lined below. 500 north dipping faults outcropping on the southern The Corinth rift is an asymmetric Quaternary rift. coast of the gulf [Jacksonet al., 1982; Hubert et al., The most active normal faults are located on the south- 1996],the 1995Aigion earthquake ruptured a fairly low ern coastof the gulf whichis subjectto uplift [$Sbricr, angle(33 ø) north dippingoff shorenormal fault not out- 1977;Armijo et al., 1996].The typical length of the E- croppingon land [Bernard et al., 1997a]. This earth- W striking en •chelonfault segmentsalong the southern quake, together with the 1992 M,:5.9 Galaxidi earth- coastof the gulf (fromXilokastro to Psathopirgos)is 15 quake[Hatzfeld et al., 1996],demonstrated that inner- + 5 kin. Tectonic studies based on deformation model- rift normal faults could play an important role in the ing of marine terracesin the footwall of the Xilokastro deformation of the rift and that these faults may have fault, a major fault on the central part of the south coast somewhat lower dips than the on shore ones. In addi- (Figure1), indicatethat the overalllong-term morphol- tion, a 2-month microseismicstudy conducted in 1991 ogy of the southerncoast of the gulf can be explainedby in the westernpart of the gulf [Rigo et al., 1996], and the repetition of earthquakeson 400-600 north dipping covering the Aigion 1995 earthquake epicentral area, faults, slippingat 11 + 3 mm/yr [Armijo et al., 1996]. revealed a strong clustering of micro earthquakes. A This would correspond to an extension rate of 6 + 1 significantnumber of theseindicate north dipping, low- mm/yr. The level of historicaland instrumentalseis- angle, normal-faulting focal mechanismsat 7-11 kin micity is quite high [Papazachosand Papazachos,1989; depth beneath the northern coast of the gulf. This re- Ambraseys and Jackson, 1990; Rigo et al., 1996; Am- sult suggestedthat the steeply north dipping normal braseysand Jackson,1997]. During the last 40 years faults seen at the surface root at these depths on a low- BRIOLE ET AL.' DEFORMATION OF THE CORINTH RIFT 25,607

21 ø 22 ø 23 ø 24 ø 40300 . . 40 ø

0 o [] o

39 ø 39 ø

38 ø

37 ø 37 ø

SLRpoints Aegean seanetwork Central Greece network Corinth & Central networks 36 ø 36: 20 ø 21ø 22ø 23ø 24ø Figure 2. Geodeticnetworks in centralGreece. Triangles are the satellitelaser ranging (SLR) points.The SLR pad of Dionysosis usedas referencefor all GPS campaignsin centralGreece. Shadedsquares indicate the Ioniansea network [Kahle et al., 1995]. Circlesare the 6(3points of the centralGreece network [Den!Is et al., 1994]installed in 1989. The eight solid circles, displayedwith their code, are the central Greecepoints commonwith the Corinth rift network. Correspondingcentral Greece code can be foundin TableA1 (availableas electronic supporting material). The box indicatesthe locationof Figure3.

angledipping (150 4- 10ø) detachmentzone [Rigo et al., rate observed. According to Ambraseys and Jackson 1996;Rietbrock et al., 1996]. [1997]the frequencyof large earthquakesin the eastern Using the comparisonbetween old triangulationob- gulf during the period 1690-1890has been higher than servationsand a GPS survey,Billiris et al. [1991]gave during this century, and they suggestthat the deficit of the first large-scaledetermination of geodeticdeforma- seismic strain release in the western Corinth rift could tion rate (10 mm/yr) and extensionorientation (N-S) be met by severalsuch earthquakes in the medium term. of the Corinth rift over a 100-yearperiod. Clarke et al. However,the former comparisonsof seismicand geode- [1997]give a more detailedpicture of the deformation tic balancesof energyrelease are stronglydependent on at the regionalscale around the gulf based both on a the effectivethickness of the seismogeniclayer. 100-year comparison and on G PS results between 1991 On the basis of our densegeodetic data and on the and 1996. In particular, they show that the velocities knowledge gained from the 1995 Aigion earthquake, increasesmoothly from east (6.4 4- 1.0 mm/yr) to west we will refine the extension rate across the rift and (12.7 4- 1.0 mm/yr). They foundthat the geodeticve- demonstratethat the deformationis strongly localized. locity in the easterngulf is in goodagreement with that We then study the consequencesof this localization by deducedby summing100 yearsof energyreleased by means of simple dislocation models that allow us to dis- earthquakes.This is not the casein the west, where the cussthe inner/outer rift repartitionof the deformation rate of seismicmoment releasepredicts extension rates as well as the effectivethickness of the seismogeniclayer 12to 3 times smallerthan the lowestgeodetic extension and the seismic/geodeticenergy release. 25,608 BRIOLE ET AL.: DEFORMATION OF THE CORINTH RIFT

2. Data tional criteria suchas accessibility,safety of the place, and local terrain stability have also been taken into ac- 2.1. Geode•ic Networks count. This latter criterion has been difficult to satisfy During the last 10 years,several groups have installed along the southerncoast of the gulf where most of the and measured GPS networks in continental Greece at area is coveredby youngsediments and alluvial deposits differentspatial scales and in differentareas (Figure 2). (e.g.,points D,T,Y,CI,CQ). In principle,the densityof These networks are well tied together and have been the second-ordernetwork (Figure 3) is high enoughto colocatedwith the SLR pad of Dionysosduring most of ensurea sufficientnumber of displacedpoints for mod- the GPS campaigns. In 1989 the "central Greece" net- erate earthquakes like the M•=5.9 1992 Galaxidi and work (66 points) coveringan area of about 150 x 150 M•=6.2 1995 Aigion ones. A network with sucha spa- km was installed and measured by the National Tech- tial density was particularly useful for measuringthe nical University of Athens; the Universities of Newcas- grounddeformation associated with the June 15, 1995, tle, Nottingham, Oxford, and Cambridge;and the ETH Aigionearthquake and constrainingthe fault plane that Zurich for the purpose of monitoring the deformation ruptured during this event, the rupture size, and the of Central Greeceat a large scale[Denys et al., 1994; amountof slip [Bernardet al., 1997a].All second-order Clarke et al., 1998]. Our networkwas progressively in- points are pillars of the Hellenic triangulationnetwork stalled around the Gulf of Corinth starting in 1990, with measured between 1966 and 1973. "Old" coordinates of the purpose of monitoring the deformation associated these points have been recalculated in the Greek Geode- with the seismiccycle around the main active faults. It tic ReferenceSystem (GGRS 87) [Vels et al., 1992],and includes nine central Greece points. In October 1995 we present here a comparisonof these coordinateswith the network included 193 points. Among these, 51 the GPS ones in addition to the comparisonof GPS are first-orderpoints (Figure 3) observedduring sev- coordinates between 1990 and 1995. eral long sessions(8 hourstypically) in eachcampaign and with a precision of localization of a few millimeters, 2.2. GPS Campaigns and 142 are second-orderpoints increasingthe density Overall, seven GPS campaigns took place in 1990, of the first-ordernetwork (Figure 3) but observeddur- 1991, 1992, 1993, 1994, June 1995, and October 1995. ing short sessions(1-2 hours). The first-ordernetwork Table 1 summarizes the characteristics of each cam- distribution has been established in order to ensure an paign. During each campaign, the first-order points accurate and pertinent determination of the deforma- were measured in three or four sessions of 8 to 24 hours. tion around the active faults where both spatial and The measurementswere carried out using varioustypes temporal gradients of deformation are expected. Addi- of Ashtechreceivers (LD-XII, MD-XII, P-XII, Z-XII3).

22 ø 23 ø

ß•ao• ß ßoo4 -co "a,; ] ß .

3• ø

22 ø 23 ø Figure 3. Corinth rift GPS network. The nine large solid points are commonwith the central Greece network. %Janglesare the additional first-order points. Small points indicate the location of second-orderpoints. All these points are pillars of the "old" Greek triangulation. Table A1 (electronic)displays coordinates and detailsabout the first-orderpoints, and Table A2 (electronic)displays coordinates and detailsabout the pillars. BRIOLE ET AL.' DEFORMATION OF THE CORINTH RIFT 25,609

Table 1. Overview of the SevenGPS Campaigns

1990 1991 1992 1993 1994 1995a 1995b

Starting date May 8 Aug. 30 Nov. 27 May 10 Sept. 20 June 16 Oct. 3 Ending date May 18 Sept. 10 Dec. 5 May 22 Oct. 2 June 24 Oct. 14 Number of codeless receivers 3 8 3 12 3 4 16 Number of code P-Z receivers 5 4 3 16 Number of permanent points I 3 2 3 3

Number of first-order stations 7 23 9 43 16 23 51 Duration of one session, hours 4 6 4 8 8-12 12-18 8-24 Number of sessionsper point 3 4 2-3 3-4 2-3 2-3 3-4 •Tf• •T •-• • iGS iGS iGS 1• Software used for analysis G G-B G G-B B B B Vertical sessionto sessionscatter, mm 13 17 18 16 17 18 13 Horizontal sessionto sessionscatter, mm 3.5 3 4 4 4.5 4 4 rms in 3-D network adjustment, mm 10 12 8 9 14 13 9

Number of second-order stations 9 34 24 22 84 Duration of one session,hours 2 2-3 1-3 1-6 1-12 Number of session per point 1 1-2 1-2 1-3 1-3 Ephemeris used for analysis b b b b b Software used for analysis G-A A A B-A B-A

G, GAMIT software; B, BERNESE software; A, Ashtech GPPS software; IGS, International GPS Service for Geody- namics;NGS, National GeodeticSurvey (U.S.); b, broadcastorbits.

A few of the 1995 measurements were carried out us- tions measured, this selection of a reference point at ing Leica and Trimble receivers.During the first cam- each campaign is purely technical, as errors as large paigns,especially 1991 and 1993, most of the measure- as 20 cm in the absolute coordinates would not affect ments were performed during eveningsand nights to significantlythe relative positioning. For the same rea- try to minimizeboth ionosphericand troposphericef- son, using the IGS stations as "fiducial" sites was not fects. Later, becauseof the increasingnumber of re- mandatory in our study. For the 1994 and 1995 cam- ceiversand decreasingnumber of operators, we often paigns we processedthe IGS data together with the left the receiversunattended for longer periods of time, others, especially to improve the accuracy of vertical thus with day and night observationsin the samedata positioning.The relative horizontal coordinateswithin files at the same point. Starting in 1992, at least one the network do not depend upon the introduction of IGS station was maintained in permanent recording during stations in the processing. For all campaigns we used the wholecampaign. In the 1994 and 1995 campaigns, the troposphericparameters measured in the field (tem- therewere four permanentrecording stations (Dionysos perature,pressure, humidity) to fit the a priori model and pointsT, CH, and CT in the network). (the sametropospheric model of Saastamoinenis used by the GPPS, GAMIT, and BERNESE software)and 2.3. GPS Data Processing let the software estimate an additional term of tropo- Processingof the GPS data was performedusing sphericdelay correctionin windowsof 6 hours. For the GAMIT and BERNESE software for the first-order second-order points the Ashtech GPPS software does pointsand AshtechGeodetic Post ProcessingSoftware not estimate an additional term of tropospheric delay, (GPPS) softwarefor the second-orderpoints. Avail- and the data were processedby introducing one unique abledata from the InternationalGPS Service(IGS) sta- set of tropospheric parameters for the whole session. tionsMatera, Graz, Wettzell, and Madrid wereincluded The data from the later campaignsare fairly good, es- in the processingfor the 1993, 1994, and 1995 cam- pecially those obtained from Z-code Ashtech receivers. paigns. The coordinatesof our network are expressed In contrast, some 1990-1991 data were difficult to clean within the International Terrestrial Reference Frame for cycleslips because of the lower signal/noiseratio of 1992 (ITRF92) [Bouchere! al., 1993]for epochApril the first generationsof receiversand the higher level of 1993. The referencepoint used for all campaignswas solar activity in early 1990s inducing high ionospheric the main GPS point of DionysosObservatory (DION) perturbations. Because of the moderate size of the net- exceptfor the first campaigns(1990 to 1992) in which work we were able to fix almost all phase ambiguities DION was not measured. For those campaignsthe ref- for all campaigns. The accuracyof the solutionsis sum- erencehas been arbitrarily fixed at point E. Owing to marized in Table 1. For each campaign we tested the the small size of the network and the large deforma- quality of our processingaccording to two criteria: 25,610 BRIOLE ET AL.' DEFORMATION OF THE CORINTH RIFT

a 1991 Baselinelength b 1993 Baselinelength •'2o i i i i i i fi 15 m 10 • 5 : 3 0 ' • f o 10 20 30 40 50 60

North component North component •' 20 •' 20 • 15 • 15 m 10 m 10 m 5 c• 5 3 0 B: 0' 10 20 30 40 5'0 6'0 0

East component East component •' 20 •' 20 • 15 • 15 m 10 m 10 m 5 3 0 3 0 0 10 20 30 40 5'0 6'0 1'0 2'0 3'0 40 5'0 6'0

Vertical Vertical • 1oo •' 100 • 80 • 80 m 60 • 60 • 4020 • 4o20 •: o • o 0 10 20 30 40 50 60 0 10 20 30 40 50 Length (km) Length (km) Figure 4. Sessionto sessionscatter versus baseline length for (a) the 1991and (b) the 1993 GPS surveys. For each day the data were processedwith the GAMIT softwareusing precise orbits(NGS orbitsin 1991and IGS orbitsin 19!13).Each GPS stationwas occupied during four sessionsof about 8 hours. Severalbaselines have been measuredtwo, three, or four times. The plotted circlesare the weightedrms deviationfrom the weightedmean of differentcomponents (length,north, east, and vertical)of thesebaselines. The averagerms for all the baselinesare reportedin Table 1. For the baselinelength the averagerms is 3 mm for the 1991campaign and 4 mm for the 1993 campaign.

1. We checkedthe day-to-day repeatability of the average scatter of the individual baselines before ad- baselines(for the baselinesmeasured at leasttwo times). justment and the rms of the global adjustment(Table In Figure 4 we show for example the root-mean-square 1). (rms) scattersin east, north, and up componentsand in distanceversus baseline length for the 1991 and 1993 3. Results campaigns. All the campaignswere analyzed in the 3.1. Basic Results and Their Implications same manner and the baseline repeatability of each of them is given in Table 1. The horizontal repeatability Figure 5 showsthe 1991-1993displacements in the is •3-4 mm, and the vertical one is between 1 and 2 cm. western Gulf of Corinth where measurements started Also, there is no correlationbetween the length of the in 1990 and 1991. It clearlyindicates that, overthe 3 baselineand the qualityof the repeatability(Figure 4), years sampled, most of the deformation occursoff shore the repeatability of a given baselinebeing more depen- in the central part of the rift. The extensionoccurs on dent on the difference in elevation between the points averagealong an axis orientedN9øE, consistentwith than on the length of the baselineitself. This is related the regional tectonics. to nonmodeled tropospheric effects. Figure 6 displaysthe time seriesof 1990-1995relative 2. For each campaignthe coordinateswere adjusted displacements(values in Table2) of lincA-G, projected to fit the baselinesusing a least squares adjustment alongthe N9øE axis. At the first-orderand according technique[Tarantola and Valette, 1982], in a code to the errorbars on the coordinatesat eachepoch of ob- (AG3D) developedat Institut de Physiquedu Globe servations, the extension of the rift is a linear function de Paris (IPGP) [Rueggand Bougault,1992]. The rms of the elapsedtime between 1990 and 1994. The southof the residualsafter adjustment gave us another esti- crnside of the gulfappears to behaveas an almostrigid mate of the global consistencyof the network. For all block.The northernside of the gulf,between points B the campaigns,there is a good agreementbetween the and A appearsalso to behaveas a rigid block,whereas BRIOLEET AL.' DEFORMATIONOF THE CORINTH RIFT 25,611

1991-1993displacements c, • 50 mm A/

38 30'

38 00' 22 00' 22 30' Figure 5. The 1991-1993displacements in the westernpart of the Corinth riff; G is fixed.

150 the extensionrate at point C is somewhatlower than at B and A. This trend changesbetween 1994 and 1995 owing to the M•=6.2, June 15, 1995, earthquake. Between 1993 and 1994 the data suggestthat the extensionrate •B of the northernside might havebeen lower than during lOO the previousyears, but the changeis subtle except at point C. We estimate that the hypothesisof a change of extensionrate before the 1995 earthquake is not suf- ficiently supported by our data to be discussedhere. In this paper, we use these raw results to present a more refined analysis. 50 1. From the resultsof Figure 6, we argue that at the first order, over short periods of time, the deformation of the rift is the sum of a linear deformation field and of the contribution of the earthquakes. 2. We use this argumentto separatein our geodetic data the contribution of the 1992 and 1995 earthquakes from the total observed deformation. 3. We use the relative rigidity of the southern and northern blocks to scale and rotate coordinates issued from classicalgeodetic observations within the GPS coordinates frame.

-50 1990 • 1991 • 1992 •_1993•_ 1994 L 1995 , Time (years) 3.2. Coseismic Deformation Figure 6. N9øEprojection of thedisplacements (mm) of points A,B,C,D,E,F,G as a function of time. Er- Two significantearthquakes occurred during the pe- ror bars for eachindividual campaignare drawn. The riod of our surveys,on November18, 1992(Ms=5.9, changein the slopeof the curvesbetween 1994 and 1995 Galaxidi),and on June 15, 1995 (M,=6.2, Aigion). Co- correspondsto the coseismicdisplacements associated seismicdisplacements were measured at six pointsfor with the June 15, 1995, Aigion earthquake. Points A the 1992earthquake (C, D, S, T, X, V) (Platela) and and G are not sensitiveto that earthquake. at 29 pointsfor the 1995one (Plate lb). We alsocal- 25,612 BRIOLE ET AL.' DEFORMATION OF THE CORINTH RIFT

Table 2. Horizontal Displacementsof Points A to G 1965event, which occurred before the first triangulation as a Function of Time measurement)has also been calculated and is discussed in section 3.4. Year Months A B C D E F G 3.3. Continuous Deformation on the 1991 16 24 21 22 5 4 0 0 First-Order Network 1993 36 57 49 43 7 4 0 0 1994 53 68 63 46 0 3 0 0 Figure7 displaysthe velocityvectors in the Corinth 1995 65 78 114 90 -38 -31 -19 -5 rift after removingthe coseismiccontributions of the 1992 Galaxidi and 1995 Aigion earthquakes.The val- Displacementsare in mm. The numberof monthsrefers ues are listed in Table A1 in the electronic supporting to the date of the first campaign(May 1990). material1. To bringthe coordinates issued from the processingof eachcampaign into the samereference frame, we have consideredthe Peloponnisosas a fixed refer- culatedERS syntheticaperture radar (SAR) interfero- ence. For the period 1990-1994, all Peloponnisospoints gramsfor time intervalscovering both earthquakes.We except points X and T that were affected by the 1992 used the velocity field deduced from Figure 5 to cor- earthquakehave been usedto minimize the residualdis- rect our measurementsfor the long-term motion and to placement vectorsbetween epochs. For the comparison properly extract the coseismiccontribution. The dis- of the 1995 campaignswith the previousones, points P, placementsassociated with the June 15, 1995, earth- Q, R, D, E, F, U, and T close to the 1995 earthquake quake are relatively large (up to 10 cm) and above have been discarded and the Peloponnisospoints used the level of noise at more than 20 points. These data for minimization were K, L, G, and all points located were usedtogether with SAR interferometry(InSAR) southeastof point X (seeFigure 3). No long-termve- and seismologicaldata to constrain a dislocationmodel locity has been calculated for points S and T that are [Bernardet al., 1997a]. Comparedto the modelpro- affected by both earthquakes. Figure 7 also displays posedby Bernard et al. [1997a],the model presented the velocities obtained over 100 years by comparing re- here (Plate lb and Table 3) takesinto accounta few ad- cent GPS coordinates with coordinates obtained from ditional GPS vectors, but there are no significantdiffer- triangulation fieldworkcarried out in 1892 [Davies ½t encesin the parameters of both models. The displace- al., 1997],the velocitiesobtained by Kahl½½t al. [1995] ments associated with the November 18, 1992, earth- over the Ionian sea network and the velocities obtained quakeare small (5 20 mm). Plate la displaysthe co- by Anzid½iet al. [1996]at CHLE point (westPelopon- seismicdata for the 1992 earthquake and a dislocation nisos). With respectto the deformationlocated in the model assuming a north dipping plane located at the inner Gulf of Corinth, little internal deformation occurs eastern end of the 1995 fault. A preliminary analysis in Peloponnisos,even along its northern coast. This (basedon displacementsobserved at pointsC, X, S, T, result is consistentwith the other geodeticstudies car- U, V) proposedthat the 600 south dippingplane was ried out at a broaderscale [Billiris et al., 1991;Davies the fault plane[Briole et al., 1993]but, further seimolog- et al., 1997; Clarke et al., 1998; Nahle et al., 1995; ical analysis[Hatzfeld et al., 1996]indicated that a 300 V½is½t al., 1992], and its originality lies in the higher north dipping plane, as for the 1995 Aigion earthquake number of points that allows a better constraint on the [Bernardel al., 1997a],is the lnost probablesolution. width of the deforming zone. Most of the deformation Also the InSAR data agree better with this interpreta- occurs offshore between the southern and the northern tion' the synthetic interferogramcalculated assuming a coastsof the gulf. North of the Gulf of Corinth, the north dippingplane (Plate lc) exhibitsno fringesin the vectors are almost parallel to each other in the cen- northern coast of the gulf, as it is the casein the data. tral and western areas but rotate in the northeastern In contrast.,the synthetic interferogram calculated as- region from N10øE to N30øW, so they remain almost sumingthe antithetic south dippingfault plane (Plate perpendicularto the tectonicstructures there (Kapar- ld) exhibitstwo fringesabove the northwesternedge of elli fault zone). In this area, deformationassociated the fault. In this case, to obtain a synthetic interfero- with the progressivechange of orientation of the ex- gram without fringes, it is necessaryto shift the fault tension is probably occurring. However, the sampling plane 3 km to the south, inducing a displacement at interval of only 2 yearsin the east (1993-1995)limits point S not compatible with the observationsany more. There is no way of fitting both the GPS data and the InSAR data if one assumesa south dipping plane for 1Supporting data Tables A1 and A2 are available on the 1992 earthquake. The parameters of both the 1992 diskette or via Anonymous FTP from kosmos.agu.org, di- rectory APEND (Username=anonymous,Password=guest). and 1995 dislocation models are given in Table 3 to- Diskette may be ordered from American Geophysical Union, gether with the parametersof the other earthquakesof 2000 Florida Avenue, N.W., Washington, DC 20009 or by magnitude Ms >_5.5 that occurredsince 1965. The co- phone at 800-966-2481; 15.00. Payment must accompany seismiccontribution of all the earthquakes(except the order. BRIOLE ET AL.: DEFORMATION OF THE CORINTH RIFT 25,613

1992 earthquake 1995 earthquake

Data (20 mm) o l Data (50 mm) c -- •A Model (20 ram) Model(50mm) •

38030 '

b ' 10km 38000 ' 22'00' 22030 ' 22000, 22030'

Model A Model B

Data (20 mm) c Data (20 mm) c -• Model(20 mm) Model(20 mm) •

38030 '

•v •v

10 km 38000 ' 22000, 22d30' 22000' 22030ß

Plate 1. (a) Observedand modeledcoseismic displacements associated with the November18, 1992, Galaxidi earthquake. Projection of the fault at the surface is shown with the slip vector. Parameters of the model are given in Table 3. Also plotted is the interferogram obtained using the pair of imagesERS1 5162 (July 11, 1992) and ERS1 10172 (June 26, 1993) with altitude of ambiguityof 603 m. Note that no fringesare visible. (b) Observedand modeledcoseismic displacementsassociated with the June 15, 1995, Aigion earthquake. Projection of the fault at the surface is shown with the slip vector. Parameters of the model are given in Table 3. Also plotted is the interferogramobtained using the pair of imagesERS1 6164 (September19, 1992) and ERS1 22039 (October 2, 1995) with altitude of ambiguityof 972 m. Note that this interferogramis almost identical to that shownby Bernard et al. [1997a]that was calculated over a period of 4 monthsonly. (c) Modelingof November18, 1992, Galaxidi earthquakeand predicted interferogramfor a 300 north dipping plane of length 14 km and width 9 km. No fringes are predictedby this model (model A). (d) Model B correspondsto a 600 south dipping fault plane of same area as model A and predicts two fringes on the northern coast. 25,614 BRIOLE ET AL.: DEFORMATION OF THE CORINTH RIFT

Table 3. Parameters of the Recent Earthquakes in the Gulf of Corinth

Event M0, Longitude, Latitude, Strike, Dip, Rake, Length, Width, Centroid, Slip, 1018N in øE øN deg deg deg km km km in

July 6, 1965 a 1.7 22.40 38.37 281 34 -72 13. 8. 10. 0.5 April 8, 1970 a 0.9 22.56 38.34 265 23 -81 12. 7. 9. 0.32 Feb. 24, 1981b 8.8 22.97 38.23 264 42 -80 18. 14. 12. 1.07 Feb. 25, 1981b 4. 23.12 38.17 241 44 -85 16. 14. 8. 0.54 March4, 1981• 2.7 23.26 38.24 50 45 -90 12. 14. 7. 0.49 Nov. 18, 1992 c 0.9 22.45 38.30 270 30 -81 14. 9. 7.4 0.21 June 15, 1995 a 3.6 22.20 38.36 277 35 -81 14. 9. 7.1 0.87

The coordinates refer to the projection at the surface of the center of the upper edge of the fault. • Baker et al. [1997]. b Taymazet al. [1991]. CRevisedfrom Hatzfeldet al. [1996]and Briole et al. [1993].Both studiesindicate that the surfaceof the rupturedplane is ratherlarge for an earthquakeof this magnitude.The seismicmoment found by the waveformmodeling is 0.4 x 10ls N m, the Harwardcentroid moment tensor (CMT) is 0.5 x 10ls N in andthe seismicmoment deduced from the inversion of the geodeticdata is 0.9 x 10ls N m. dBernard et al. [1997a]modified to take into accounta few additionalGPS data. The seismicmoment for this earthquake is deducedfrom the modeling of the GPS and SAR data. It is 5% higher than the seismicmoment deducedfrom the waveformmodeling (3.4 x 10ls N m) and30% lower than the revised Harward CMT (5.1 x 10ls N m).

21 ø 22 ø 23 ø

39 ø f/---•-%'•• 39ø N ' * . •'•

38 ø

DOXa < I

30km "• . . 21 ø 22 ø 23 ø Figure 7. Relative velocitiesof GPS points acrossthe Corinth rift. The 1990-199,5velocities of 41 first-orderpoints of the Corinthrift networkare shown,with Peloponnisosfixed (obtained by minimizationof the residualof the velocityvectors of all pointslocated south of the gulf). Pointswith star havebeen measured during three or more campaigns.The velocitiesobtained in other studiesare alsoplotted: 1989-1993velocities on the Ionian seanetwork [Kahle et al., 1995], of particular interest at points KRPN, Q, SNDO, and DOXA that completeour network to the west; the thicker arrowsare 1892-1992velocity vectors [Davies et al., 1997]at four locationsto the north of the gulf (one betweenpoints A and B, one betweenpoints CV and CN, and two north of CQ), onelocation in the east (eastof CL), and three locationsto the southof the gulf (onesouth of CK, oneat CG, and onebetween K and L). There is, in general,a goodagreement betweenthe determinationsof the other groupsand our velocityvectors, except at point Q, where the result of Kahle et al. [1995]is not consistentwith our determinationand with the general observation that the Peloponnisoshas not internally deformed over the observed interval. At point CHLE the three vectorsshow three different determinationsof the velocity of the points, with referenceto Peloponnisosfixed, two determinationsbeing from the above references,the last one (thin arrow) givenby Anzideiet al. [1996].Profiles A-G and CS-CD usedin the further discussion are drawn. BRIOLE ET AL.' DEFORMATION OF THE CORINTH RIFT 25,615

A-G profile CS-CD profile 1990 to 1995 GPS measurements 1991-1995 GPS measurements 20

Gulf Gulf

0 20 40 60 80 0 20 40 60 80 100 Kilometers Kilometers

Figure 8. The 1990-1995GPS velocityalong profiles A-G and CS-CD (all pointslocated closest than 12 km to the profilehave beenincluded) and error bars. The extensionrate is 14 mm/yr alongline A-G in a N9øE azimuth and 13 mm/yr alongline CS-CD in a north azimuth.

the accuracy.Figure 8 showsthe projection of the ve- this area (Livadia-Arakhova)the orientationof the velocity vectorsplotted in Figure 7 alongtwo lines (A-G locity vector changesin a relatively narrow band from and CS-CD) crossingthe rift at two differentlocations, a N30øW to a N-N30øE direction. Figure 7 also indi- and samplingtwo of the main structuresdiscussed in cates that the orientation of the velocity vector at CS is section5, the Helikefault (line A-G) and the Derveni- consistent with the orientation of the two velocity vec- Xilokastrofault (line CS-CD). These sectionsindicate tors computedfor the time interval of 100 years on the extensionrates and azimuthsof 14 4- 2 mm/yr along two points located south of the Evvia Gulf. This result line A-G in a N9øE directionand 13 + 3 mm/yr along has to be confirmed by additional measurementsover a line CS-CD oriented N-S. These values of extension are longer period. in relatively good agreementwith the resultsof Clarke 3.4. Continuous Deformation of the et al. [1997]but are somewhatlower, in particularin Second-Order Network the western part of the gulf, where we do not measure extensionshigher than 15 mm/yr while Clarke et al. The second-orderGPS network consistsof pillars pre- find valuesup to 20 mm/yr. The only areaswhere gra- viously measuredby classicalmeans during campaigns dients of deformation are observed on land are around carried out between 1966 and 1973. Because we did point C (northernside) and to the westof point C as not try to occupy remote points located on the highest mentionedby Bernard et al. [1997b]and Clarke et al. mountains, most of these pillars are third- or fourth- [998] ound pointsJ N (southernside). The order triangulation. The 1966-1973coordinates of those maximum width of the deformingarea has thus an up- points were computedas a singleepoch network by forc- per boundequal to the width of the gulf, that is 10 km ing it to conform with the first- and second-ordertri- on the H-L line, 15 km on the A-G line, and 30 km on angulation carried out in 1974. The average standard the two eastern lines. Its lower bound seems to be of error (lrr) for the positionsof thesethird- and fourth- the order of 10 km as suggestedby the observedoff- order points were calculated to be of the order of 3 shoregradients of deformation. Figure 7 suggeststhat cm [l/•is et al., 1992]. For the entirecountry of Greece internal deformationis occurringin the easternpart of the 25,000 coordinatescalculated for this epoch of mea- the northern block, in particular, in the area located surement(early 1970on average)constitute the "Greek betweenCR, V (Antikira area), CQ, and CS. Inside GeodeticReference System 1987" (GGRS-87) [Veis et 25,616 BRIOLE ET AL.: DEFORMATION OF THE CORINTH RIFT

Table 4. Rotation and Scale Correction of the Trian- Aigion, is arbitrarily kept fixed for the comparisonof gulation Data the two sets of data, although any other point could have been kept fixed. The general trend of the raw dis- Area Points Rotation, Scale, rms S, rms N, placementvectors in the northeast and in the southeast ppm ppm mm mm does not agree with the results of the GPS-GPS com-

South 69 4.25 2.3 125 164 parisonspresented in section3.3 (Figure 7) and indi- North 61 6.1 2.55 154 146 cates that a correction for orientation and scale should Average 5.2 2.4 133 150 be applied for the old coordinates. On the basis of the precisionof the old data, we limited the correction for the old coordinatesto a two-parametercorrection (scale Rotation and scaleadjustment of the triangulation coordinates to the GPS coordinates have been calculated assum- andorientation), an approximationfurther validated by ing no internal deformation of both southern and northern the results obtained on the corrected coordinatesof pil- block. The rms scatter between the two sets of coordinates is lars. To find the best values for scale and orientation displayed after correction is applied. Two estimations have correction, we minimized the displacement vectors of beendone: one with the northernside rigid (61 points),and the 69 points located in Peloponnisos,according to the onewith the southernside rigid (69 points). Underboth as- observation from the repeated GPS surveys that this sumptions, the rms have been calculated separately for both sides of the gulf. Rigid north side assumptionleads to rms block is not significantly deforming with respect to the values slightly higher. Scale correction is almost the same inner gulf zone. The best fit was found with a coun- for both estimations: the differenceof 0.25 ppm corresponds terclockwiserotation of 4.25 ppm and dilatation of 2.3 to an error of 25 mm at a distance of 100 km, which is well ppm. Those values of scale and rotation are somewhat belowthe bestresidual rms (125 mm). The rangeof possible higher than were predicted at the regional scale from rotation parameters(1.85 ppm) leadsto maximumchanges of 185 mm at 100 km that is comparable with the residual the adjustmentof the triangulationdata [Veis et al., error level. Assuming that the errors in the secondepoch 1992]. This suggeststhat there could be locally some of observations(GPS) are negligibleand providedthat the bias higher than the global average,due to nonhomoge- uncorrected tectonic motions have no reason to be corre- nousnetworks (offshore areas) and problemsof connec- lated with the measurement errors, 125 mm is a measure tions between networks measuredduring different field of the maximum statistic error on the triangulation coordinates(including both measurementand processingerrors). campaigns.Assuming the northern block of the Gulf of Indeed, a detailed analysis of the residual vectors between Corinthto be rigid (61 points),the minimizationleads close points in the southern block shows that the errors in to valuesof +6.1 ppm for the rotation and +2.5 ppm for the triangulation coordinatesare more likely to be of the the scale. The most important fact is the similarity of order of 50 mm, that is, roughly consistentwith the value the values found for the scale correction, the difference of 20-30 mm givenby Veiset al. [1992]. between the two determinations being below the value of the errors on the determination itself. In Table 4 we summarize the values of the rms found on both sides al., 1992]. They are givento an accuracyof a few cen- of the gulf for the above two sets of rotation and scale timetersplus one part in 106, while the scaleand ori- factors and for an intermediate value. As the scatter entationare on averagewithin a few parts of l0 -? of between both solutions remains below the errors on the the internationally accepted reference system ITRF89 pillar coordinatesdetermination, in the following,we [Boucherand Altamimi, 1989; Boucheret al., 1993]. adopt the valuesof +4.25 ppm and +2.3 ppm for the We measureda total of 159 triangulationpillars (Ta- orientation and scale correction; these are the values ble A2, electronic),17 belongingalso to the first-order that give the best fit for the 69 points located in Pelo- G PS network. We were able to compare the coordi- ponnisos.The amplitude of coseismicdisplacements es- natesfor 146 of them (in somecases, the comparison timated over 24 years from the parameters in Table 3 indicated that the pillars had been displacedand in few (Figure 10) is muchless than the extensionmeasured cases,the old coordinateswere not available).We ne- geodeticallyover the sameperiod, except possiblyin the glectedthe durationof eachseries of campaigns(7 years Alkyonidesarea where a relatively large part of the defor the first one that occurred between 1966 and 1973 formation could be explained by the 1981 earthquakes and 4 yearsfor the G PS campaigns)with respectto [Rigo,1994]. Figure 11 showsthe projectionof the vec- the mean time interval of 24 years betweenthe two se- tors plotted on Figure 9, along the lines A-G and CD- ries of campaigns. This approximationis justified by CS (seeFigure 7) usedfor projectingthe GPS velocities the a priori size of the errors associated with the tri- and along an additional line crossingthe Thivai area at angulationmeasurements and further validatedby the the eastern end of the gulf. The high rates found from standarderror deducedfrom our study (Table4). We th• comparisonGPS triangulation are not compatible divided the pillars into four categoriesbased on their with the GPS-GPS estimation of velocities in the Gulf location:69 in the southof the gulf, 61 in the north, 14 of Corinth. On linesA-G and CS-CD, the averageover- in the east, and 2 in the west. The direct comparison estimation of the velocitiesis 80% with respectto the of the 1991-1995 GPS coordinates and the 1966-1973 GPS-GPS ones. This overestimation of the velocities coordinatesis representedin Figure 9a. Point E, near across the Corinth rift could be due to errors in con- BRIOLE ET AL.' DEFORMATION OF THE CORINTH RIFT 25,617 necting classical geodetic data from one to the other networks were measuredduring severalfield campaigns, side of the gulf. Indeed, although the old coordinates implying possibleproblems of connection between each are correctly scaled at the global scale, as indicated by survey during the •djustment phase. Veiset al. [1992],they couldpresent distortions at the intermediate scale due to the nonhomogeneity of the 3.5. Summary of the Observations network and to the fact that the third- and fourth-order The main results of the data analysis are as follows: 1. The extension rate measured over 5 years across

22 23 •9 39 the Corinth rift rangesbetween 15 and 10 mm/yr from 30 'kin 250n•n west to east. The extension rate measured over 24 years by comparing GPS and triangulation coordinates over- estimatesthe above valuesby 50 to 80%, probably due to distortion in the triangulation data at the regional scale due to both spatial inhomogeneity of the data (presenceof the sea)and temporalinhomogeneity of the data (severalfield campaignsbetween 1966 and 1973). 2. Both data sets indicate that the southern side of the gulf behaves like a rigid block. On the northern side, there is a regional east-west compressionand local gradients of deformation at points located near the 38 i•• 38northern coast, west of point C. 3. In consequence,almost all the deformation is localized offshore, within the inner part of the Corinth a • rift. The deformationis particularly localized (,-•10 km width) in the western part of the rift (Aigion-

39 39 Psathopirgos)oThe localizationis lesswell definedin Rotation= 4.25 ppml the centerof the rift (Derveni, Xilokastro), where the Scale=2.3 ppm gulf is wider. The deformation seems to be more dif- Southfixed• It fuse in the easternpart of the rift (Corinth, Thivai). •)"-"1•" 2$0mmIL The largest openingrate (14 mm/yr) is measuredat the longitude of Aigion, where the width of the deforming zone is apparently the lowest. The implications of such a localization are very important and are discussed in section 4. 4. No movement is observed across the active faults bordering the southern coast of the gulf like the Helike and the Xilokastro faults. These faults thus appear to behave in a locked-seismic mode. 5. The extensiondirection (Peloponnisosassumed fixed) rot.at, o,• from about N9øE in the west to about O• 250nun b o• 2$(hmn N20øW in the east of the rift, in relatively good agree- ment with the focal mechanismsof earthquakes and the orientation of the active faults observed in the field.

Figure 9. (a) Comparisonof the 1991-1995GPS coordinatesof 146 pillars of the Hellenictriangulation network with their 1966-1972 coordinates obtained from terrestrial measurements. First-order point E is used as commonreference for comparisonof the two epochs. (b) Displacementof the 146 pillarsafter counterclock- wiserotation of +4.25 ppm and dilatation of +2.3 ppm that minimizes the residualson the 69 pillars located in the southern part of the gulf. The barycenter of the 69 pillars is used as commonreference for comparisonof the two epochs(the sum of the displacementvectors of the 69 pointsequals 0). (c) Samedisplacement field as in Figure 9b but with the barycenterof the 61 pillars located north of the gulf kept fixed. 25,618 BRIOLE ET AL.' DEFORMATION OF THE CORINTH RIFT

38 30'

38 00' -

37 30'

250mm22 bO' •,.• 22 •0' • 23 00' 23 •0' Figure 10. Modeled coseismicdisplacements associated with all earthquakesof magnitude >_5.5 that occurred within the network between 1966 and 1995. The parameters of the modeled faults are those of Table 3. The area where displacementscould be observedgiven the accuracy of the old measurements(50 to 125 ram) is limited. The amplitudeof the predicteddisplacements is much lower than that of the observed motions.

4. Deformation Modeling ure 1) arealso good candidates for inner-riftactivity on a shallownorth dippingplane, althoughno constraints In order to interpret our observations we construct are available on the rupture plane for these events. In a model based on th• simple assumption of deforma- addition, the observeddeformation presented here oc- tion in an elastic half-space. We make the hypothesis curs mostly offshore. As we will see below, this ob- that the deformation is localized along active faults or servationcannot be explainedwith faults locatedonly in relatively narrow deformingzones for which a planar alongthe southerncoast of the gulf. We thus impose fault model is a good approximation to predict deformation in the near and far field. This hypothesis has the geometry shown in Figure 12 where the two sys- been discussedby Gilbert et al. [1994]in the caseof tems of faults (F5 for outer and F1 and F2 for inner the San Andreas fault and is also supported by numer- rift) are assumedto join at depth at the brittle-ductile ical modelingof extendingareas [Hassaniand Chary, transition zone representedas a gently north dipping 1996]. The geometryof the faults that enter our model decollementsurface (F3 and F4) as proposedby Rigo is controlled by available tectonic and seismologicalin- et al. [1996]and alsosupported by the multipletanal- formation. The study of uplifted marine terraces in ysis of Rietbrocket al. [1996]. We assumethat the the Xilokastro as well as the Derveni and Aigion areas extensionis drivingcreep on thisdecollement (F3) that [Rigo, 1994;Armijo et al., 1996]gives estimates of slip we model as continuousslip. F2 and F5 are assumedto rates on the main faults bordering the southern coast be seismicfaults, lockedduring the interseismicperiod of the gulf. Armijo et al. [1996]estimated a long-term and slipping during earthquakes. Characteristic earth- slip rate on the Xilokastro fault of 11 + 3 mm/yr on quakesfor F2 and F5 are the M•:6.2 Aigion and the a 500 dipping plane. On the other hand, the analysis M•=6.6 Helike earthquakes,respectively. Either contin- of the June 15, 1995, Aigion earthquake[Bernard et uous or locked-seismicslipping modes can be assumed al., 1997a]pointed out the importanceof the inner-rift for F1 and F4. At eachfault junction (F1-F2 and F4- active low-dipping faults in the processof extension of F5) the amplitudeof horizontalmotion is assumedto be the Corinth rift. The inner-rift activity is further cor- continuous. At the F2-F3-F4 triple junction the kine- roborated by the smaller M•=5.9 Galaxidi earthquake matic compatibility for horizontal motion is assumed. [Hatzfeldet al., 1996]that occurredon a •030ø north These conditionsallow us to calculate slip rates on each dipping fault. The 1965 and 1970 earthquakesthat oc- fault, given that v3, the slip rate on F3, is deduceddi- curred closeto the northern coast of the gulf with simi- rectly from the data and assuminga given partition- lar mechanismsas the Aigionand Galaxidievents (Fig- ing of the opening betweenthe i_nnerand the southern BRIOLE ET AL.- DEFORMATION OF THE CORINTH RIFT 25,619

1966-1972 to 1991-1995 displacements faults. We use these slip rates to compute the defor-

lOOO mation induced by the continuousslipping segmentsas AG pr6file well as the predicted seismicenergy release and recurrence time of earthquakes on the seismicfaults. We also tested models where slip is continuous from F1 to F2 and from F4 to F5. This type of model did not yield 500 significantly different results, and here we discussonly models based on the continuity of horizontal motion, which seems more reMistic. We first ran a series of tests to demonstrate the im- possibility for fault F1 to behave in a locked-seismic mode. For this test we assume continuous slip on F3- F4 only, for three variabledepths of thesefaults (correspondingto 8, 6, and 4 km for the ES-F4 junction). The predicted rates of opening are comparedto the ob- 500 0 20 40 60 80 100 servations for profile A-G in Figure 13. None of the

lOOO three models fits the highly localized observedopening zone, and the only way to localize this deformation is to CS-CD profile assume some continuous deformation at shallow level in the uppercrust beneath the gulf. We chooseto confine this continuousdeformation to the upper 2.5 krn and to 500 model it as continuous slip on Fl. This does not mean we believe F1 really slips continuously. Rather, we think that the uppermostfew kilometers of the crust, which include thick sedimentsin the gulf, are deforming by creep. This has been suggestedfor the westernUnited States[King et al., 1994],partly on the basisof the ab- senceof seismicityat shallow depths which we find here also[Rigo et al., 1996;Bernard et al., 1997a].Whether this creep is localized on one fault as assumedhere, or 500 0 50 100 150 on a few faults, or is distributed in volume, cannot be Kfimnetres resolved from our data and thus is not relevant for the Figure 11. Displacementprofiles obtained from the modelingpresented here. triangulationGPS comparison along profiles A-G and CS-CD overan averageperiod of 24 years. On A-G 4.1. Aigion Profile (Line A-G) the averageopening between the two sides of thegulf is 600mm, which corresponds to a rateof 25 mm/yr. On The opening rate deduced from the G PS data is 14 CS-CDthe averageopening is 534mm or 22 mm/yr. mm/yr. Unlike the Xilokastrofault farther east, no re-

l0 (km) 20 30 Figure 12. Five-fault model used in the modeling. Location of the five fault segmentsis dis- cussedin the text and is based on the observedactive structures along the A-G profile. Each fault Fi has a slip rate vi. 25,620 BRIOLE ET AL.' DEFORMATION OF THE CORINTH RIFT

Horizontalvelocities associated with a decollementplane at depth (functionof the depthof the upperedge of the decollementplane)

Gulf

-40 -30 -20 - 10 0 10 20 30 40

N9øEA-G profile Figure 13. Horizontal velocitiespredicted by a 10ø north dipping decollementlocated at three differentdepths (4, 6, and 8 kin, with the upperedge of the planelocated at x:0 km). For each depth, the slip velocity is adjustedin order to obtain 14 mm/yr betweenG and A. Note that even a very shallow decollement is unable to explain the observations. liable estimate of the long-term slip rate exists for the ent models correspondingto two different hypothesis Helike fault. However, on the basis of terrace uplift for partitioning(Figure 14): model la (Table 5) has analysis,Rigo [1994]estimated that the slip on the He- 10 and 4 mm/yr of openingon the inner and southern like fault could be at least half that on the Xilokastro faults, respectively,and model lb (Table 6) has equal fault, that is 3-4 mm/yr. We thus presenttwo differ- openingrates on the inner and southernfaults. A good

16 Model lb + fault F4 slipping ......

14 Model

l0 c

• 6

Gulf

i i i i i ! !

-40 -30 -20 - 10 0 10 20 30

N9øE crosssection of the Gulf of Corinth (km) . Figure 14. Comparisonof observedand modeledinterseismic extension rates alongline A-G for the two modelsdiscussed in the text. Model la (Table5) correspondsto an average4 mm/yr of slip on the Helikefault and modellb (Table6) correspondsto 7 mm/yr. BRIOLE ET AL.' DEFORMATION OF THE CORINTH RIFT 25,621

Gulf

CH X -40 -30 -20 - 10 0 10 20 30 40 NS crosssection of the Gulf of Corinth (km) Figure 15. Comparisonof observedand modeledextension rates a,long line CS-CD. The model assumesan average7 mm/yr of slip on the Xilokastrofault (Table 7, model2). fit of the GPS observationsand, in particular, the nar- tioningat 10 and 4 mm/yr. This is thusin poor agree- row deforming zone is obtained with model la if F4 ment with the seismicrates deducedfrom the seismicity is assumedto slip in a continuous mode. We checked as discussed in section 5. that the solutiondoes not changesignificantly when the angle of faults F3-F4 varies between 0ø and 20ø. This 4.2. Xilokastro Profile (Line CS-CD) model predicts recurrencetimes of 242+60 years for the Existingconstraints for this profileare the opening Helike fault and of 71+7 years for the offshorefaults. rate deducedfrom the GPS data (13 mm/yr) and the Model lb assumesopening rates of 7 mm/yr on both horizontalopening rate across the Xilokastro fault (74-2 fault segments.Figure 14 showsthat the fit to the data mm/yr) measuredover geological timescale [Armijo et is poor when continuousslip is assumedon the upper al., 1996].Assuming the samegeometry for the faultsas decollement(fault F4). To fit the data, this fault needs that •sed for the A-G line and the same characteristic to behave in a brittle-elastic manner instead of in a earthquakes,the predictedopening rates across the rift creepingmode. This solution, however, implies a rate are plotted in Figure 15. The parameters of the model of seismicmoment about twice that required for parti- are detailed in Table 7. The fit with the extrapolated

Table5. Modella (AigionArea) With 4 mm/yrExtension Rate on Southern Faults and 10 mm/yr on Inner-Rift Faults

Fault Rate, Slipper Fault Recurrence M, Mo, Mo/100years/15 km, earthquake, lenght, time, geodetic, mm/yr mm km years 10•8 N m 10•8 N m

F1 15.6 F2 12.2 870 15 71 6.35 3.9 5.5 F3 14.2 F4 4.1 F5 6.2 1500 25 242 6.7 8.7 2.2

Theslip rates are deduced from the assumed geometry and the adopted slip balance of 10and 4 mm/yrfor inner gulf faultsand southerngulf faults. The Aigionearthquake is assumedto be characteristicfor F2. The recurrencetime on the innerrift fault planeis deducedfrom the comparisonof the modeland the 1990-1994data. We use the relation log(Mo)=12.24+ M,. Thetypical M8 on the southern fault F5 is assumedto be 6.7. This corresponds to a seismic slip of 1.5m ona 25-km-longsegment, which roughly corresponds to the observed offsets and fault length of the1861 Helike earthquake.The recurrence times on the faults are deduced from the slip rate and the assumed coseismic offsets (870 mm and 1500mm, respectively). 25,622 BRIOLE ET AL.: DEFORMATION OF THE CORINTH RIFT

Table 6. Model lb (AigionArea) With 7 mm/yr ExtensionRate on SouthernFaults and 7 mm/yr on Inner-Rift Faults

Fault Rate, Slip per Fault Recurrence Ms Mo, Mo/100 years/15km, earthquake, lenght, time, geodetic, mm/yr mm km years 10•s N m 10•s N m

F1 10.9 F2 8.5 870 15 102 6.35 3.9 3.9 F3 14.2 F4 7.1 975 25 138 F5 10.9 1500 25 138 7.05 19.3 13.6

It is not possibleto fit the data when continuousslip is assumedon the upper decollementF4. The total seismicmoment related to this model is 2 times larger than that correspondingto model la. The recurrencetime on the southernfaults is not compatiblewith the record of large earthquakesin the gulf, and the predictedmagnitude is too large with respectto the typical magnitude of earthquakesin the gulf.

1993-1995rate is good. The typical recurrencetime for et al. [1997]found values of 854-15x 10•6 N m/yr. In the earthquakesin this area is longer for the offshore our modeling, the effective seismogeniclayer is 10-12 faults(1194-20 years) and shorter for the southernfaults km outsidethe gulf and only about 7-8 km beneath the (1384-20years) than that estimatedfor the Aigionarea. gulf. It is thus logicalthat our seismicmoment rate be somewhat smaller than the value estimated by Clarke 5. Discussion and Implications for et al. [1997]. Two setsof observationsargue in favor the Deformation of the Corinth Rift of a thin seismogeniclayer beneath the gulf. A detailed microearthquakestudy carried out in the western part of the Corinth rift [Rigoet al., 1996]located no earth- 5.1. Energy Release in the Corinth Rif• quake deeper than 12 km in the area and no earthquake A direct output of the models presentedis the esti- deeperthan 10 km beneaththe gulf itself. On the other mate of rates of seismicmoment releasealong the pro- hand, the lack of seismicityin the upper 3-4 km of the files A-G and CS-CD. In order to calculate the predicted crust is also evident in Rigo et al.'s study. Although it seismicmoment releaserate for the entire length of the is difficult to reject the fact that this may be an artifact rift we first calculated the predicted seismicmoment re- due to the hypocenterselection criteria, Bernard et al. lease rate for 15-km-longsegments for both the Aigion [1997a]came up with the sameconclusion for the Aigion and Xilokastroprofiles (Table 8). Assuminga mean earthquake that on the basis of deformation and after- rate of 124-2mm/yr of extensionfor the entire rift and shock data, this event did not rupture the upper 4 km taking into accountthe fact that eight 15-km-longseg- beneaththe gulf. Cattin et al. [1999]have shown that ments are needed for the entire rift, the total seismic for a Young's modulus contrast of 10 between super- momentrate estimateis 534-9x 10•6 N m/yr. This ficial layers and crustal material below, the superficial is to be comparedto the 42 x 10•6 N m/yr obtained layers simply follow the elastic deformation of the crust by Ambraseysand Jackson[1990] from seismicmoment below, meaning that the superficiallayers do not con- tensor summation of the last century. Assuminga seis- tribute to the effectiveelastic thickness in this case[see mogeniclayer thicknessbetween 10 and 15 km, Clarke alsoSavage, 1998]. We think that the thin seismogenic

Table 7. Model2 (XilokastroArea) With 6 mm/yr ExtensionRate on SouthernFaults and 7 mm/yr onInner-Rift Faults

Fault Rate, Slip per Fault Recurrence Ms Mo, Mo/100 years/15km, earthquake, lenght, time, geodetic, mm/yr mm km years 10is N m 10•s N m

F1 9.3 F2 7.3 870 15 119 6.35 3.9 3.3 F3 13.2 F4 7.1 F5 10.9 15,00 25 138 6.7 8.7 3.8

We assumethat typical inner rift earthquakesare similar to the June 15, 1995, Aigion earthquake, with averageslip on fault 870 mm. The recurrencetime is longer than in the Aigion area due to the lower openingrate acrossthe gulf here. BRIOLE ET AL.' DEFORMATION OF THE CORINTH RIFT 25,623

Table 8. Estimated Values of Seismic Moment Release in the Corinth Rift

Segment Rate, mm/yr Inner Gulf SouthernFaults Total Clarke et al. Seismic

Aigion 14 5.5 4- 0.5 2.2 4- 0.5 7.74-1 Recurrence time, years 71 4- 7 242 4- 60 Xilokastro 13 3.3 4- 0.5 3.8 4- 0.5 7.14-1 Recurrence time, years 119 4- 20 138 4- 20 All (eight) 12 4- 2a 52 4- 10 85 4- 15 42 4- 18b

Valueof t• usedis 3.3 x 10•ø N m-2. The seismicinoments in Aigionand Xilokastroareas are computedfor a 15-km lengthzone. Our estimateis 1.5 to 2 timeslower than that of Clarkeet al. [1997]and closeto the valuefound by Arnbraseys and Jackson[1990]. Error estimateson the recurrencetimes are internalerrors of the modelingitself, assumingno errors on the extensionrate. All momentsare in unitsof 10•8 N m/100 years. aAverage value for the whole Corinth rift. bSeismicmoment release estimated by Arnbraseysand Jackson[1990]. layer in the Corinth rift is due both to the fairly shal- more extensionfor a given energy release. In our model low depth of the brittle-ductiletransition (10-12 km) the predicted recurrencetimes for magnitude 6.7 earth- and to the presenceof relativelythick (2-4 km) uncon- quakeson the main faults bordering the southern Gulf solidatedsediments within the gulf below 0.5 to 0.8 km of Corinth are typically 138-242 years. For the entire of water. The other factor that contributes to the re- Gulf of Corinth, Ambraseysand Jackson[1997] listed duction of our estimate of seismic moment rate is the eight events of magnitude )_ 6.5 in the last 300 years fact that we introduced seismicfaults with low dips in that could correspondto this type of event. This sug- our modeling.For a givenseismic moment, the amount geststhat the modeledrecurrence time of 2424-60years of extensionwill be ~25% largerif slip occurson a 300 (Aigion)is possiblebut that 1384-20years (Xilokastro) dippingfault rather than on a 500 dippingfault. Am- is probably too small. Alternatively, the energy may be braseysand Jackson[1997] evaluated the openingve- partly releasedthere by smaller-magnitudeearthquakes locity deducedfrom earthquakesof magnitudeslarger or the assumedslip rate on the Xilokastro fault is too than 6 over the last 300 years and concludedthat it large. may be 5-10%larger than the velocityestimates based on only 100years of seismicity[Ambraseys and Jackson, 6. Conclusion 1990].This indicatesthat the seismicmoment rate re- leasecalculated over 300 yearsis not significantlydiffer- The GPS data analyzed in this study show that the ent from that calculatedover 100 years. Our estimate Corinth rift is extending at rates ranging between 14 basedon deformationmodeling is quite closeto Am- mm/yr in the westand 10 mm/yr in the east. The den- braseysand Jackson's[1990] determination. This indi- sity of our observationsallows us to show that the de- catesthat if our assumptionof creep at shallowdepth formation is localizedin a very narrow deformingzone, is valid, there is no need to invoke aseismicstrain at in particular, in the westernpart of the rift where strain midcrustal depths. ratesreach values of 4.5 x 10-x4 s-X. Assumingthat this high rate over 5 years representsthe steady state 5.2. Partitioning of the Deformation and deformation acrossthe rift, we proposea simple model Recurrence Time of Earthquakes of fault partitioning that explainsthe GPS observations Our modelingpredicts that 71% of the energyrelease and is consistentwith available seismologicaland tec- acrossthe Corinth rift at the longitudeof Aigionis asso- tonic observations. In our model a large part of the ciatedwith earthquakesin the inner part of the Corinth seismicdeformation is localized on relatively low-angle rift. On average,for the entirerift we find that •_ 50% faults located off shoreat depth rangingbetween 4 and of the energyis releasedin the inner Gulf of Corinth. 10 km, and the upper crust located above this depth In terms of contributionto the long-term extensionof deformsmostly in a creepingmode. This thicknessof 4 the Corinth rift, the contributionof the extensionlo- km could represent the total thicknessof the layer con- cated off shore to the total extension is also • 50%. stitutedby the water (0.5-0.8km), the youngsediments As we discussedabove, we suggestthat the contribu- of the gulf (2-4 km dependingon areas),and the upper- tion of the earthquakeslocated in the inner Corinth most fractured rockslocated below the youngsediments rift to its extension is increasedby two factors' first, (1 km). Below 10 km, the deformationis modeledas the fact that the seismogenicthickness is reducedthere creep along a decollementsuggested by seismological (presenceof waterand unconsolidated sediments at the studies[Rigo et al., 1996;Rietbrock et al., 1996]. The top of the uppercrust there); and second,the fact that important implication is that only about one half of faults havelower dip anglethere and thus accommodate the observeddeformation, and even less in the western 25,624 BRIOLE ET AL.: DEFORMATION OF THE CORINTH RIFT partof thegulf, is stored as elastic energy in theseismo- in the Gulf of Corinth(Greece), J. EarthquakeEng., 1(3), geniclayer. If thisis true, the particularly high observed 433-474, 1997. strain rate in the central and westernpart of the gulf Anzidei, M., P. Baldi, G. Casula, M. Crespi, and F. Riguzzi, Repeated GPS surveysacross the Ionian Sea: evidence doesnot imply a largedeficit of seismicenergy release of crustal deformations, Geophys. J. Int., 127, 257-267, asproposed by Clarkeet al. [1997].In the easternmost 1996. part of the gulf,where the strain rate is smaller,no low Armijo, R., B. Meyer, G. King, A. Rigo, and D. Papanas- dip normalfaults are known. The percentageof the de- tassiou, Quaternary evolution of the Corinth rift and its formationstored as elasticenergy could be larger than implicationsfor the late Cenozoicevolution of the Aegean, Geophys. J. Int., 126, 11-53, 1996. in the central and westernpart of the gulf and closeto Baker, C., D. Hatzfeld, H. Lyon-Caen, E. Papadimitriou, 100%,as proposedby Clarkeet al. [1997].This is con- and A. Rigo, Earthquake mechanismsof the Adriatic firmedby the coseismiccalculations shown on Figure sea and western Greece: Implications for the oceanic 10. One difficultywith the modelproposed here, where subduction-continental collision transition, Geophys. J. Int., 131, 559-594, 1997. a largepart of the deformationoccurs within the gulf Bernard, P., et al., The M•=6.2, June 15, 1995Aigion earth- on low-dippingfaults, concerns the predictionof verti- quake(Greece): Evidence for low anglenormal faulting cal movements.A detailedinvestigation of this problem in the Corinth rift, J. $eismol., 1, 131-150, 1997a. is beyondthe scopeof thispaper, but it is unlikelythat Bernard, P., M. Chouliaras, A. Tzanis, P. Briole, M.P. the modelwe propose can explain the topographyof the Bouin, J. Tellez, G. Stavrakakis, and K. Makropoulos, Seismic and electrical anisotropy in the Mornos delta, rift flanksin the Aigionarea. One possibilityis that Gulf of Corinth, Greece, and its relationship with GPS the faultingconditions changed fairly recentlyrelated strain measurements, Geophys. Res. Lett., 2J, 2227-2230, in particularto an increasein strainrate. Faultingmay 1997b. havestarted and may havecontinued for sometime at Billiris, H., et al., Geodetic determination of tectonic defor- highangle. As proposedby Melosh[1990], low-angle mations in Greece from 1900 to 1988, Nature, 350, 124- 129, 1991. faultingmay evolvefrom high-angle faults due to stress Boucher, C., and Z. Altamimi, ITRF89 and other realiza- rotation associatedwith a low-viscositylower crust. Ev- tions of the IERS Terrestrial Reference System for 1989, idence for such a low-viscositylower crust at 10 km IER$ Tech. Rep., 6, 1989. below the Gulf of Corinth comesfrom magnetotelluric Boucher, C., Z. Altamimi, and L. Duhem, ITRF92 and its observationsindicating an extremelyconductive layer associated velocity field, edited by S.O. Dickey and M. Feissel, IERS Tech. Note, 15, Obs. de Paris, 1993. below10 km suggestinga ductilerheology with possi- Briole, P., A. Deschamps, H. Lyon-Caen, K. Papazissi, and bly fluid-enrichedrocks [Pham et al., 1996]. Further J. Martinod, The Itea (M•=5.9) earthquakeof Novem- investigationof this problemshould involve numerical ber 18, 1992: Characteristics of the main shock inferred modelingwith a realisticdepth dependent theology as from body wave and ground displacement analysis, pa- proposedby Hassaniand Chary [1996]. Recently, Hager per presented at 2nd Hellenic Geophysical Union, Florina, Greece, May 1993. et al. [1999]have shown that steepgradients of defor- Cattin, R., P. Briole, H. Lyon-Caen, P. Bernard, and P. mation(somewhat smaller than those observed here) Pinettes, Effects of superficial layers on coseismicdisplace- couldbe explainedin part by a viscoelasticmodel with ments for a dip-slip fault and geophysical implications, lateral variations of elastic modulus, which favors the Geophys. J. Int., 137, 149-158, 1999. localization of elastic strain in the more compliant re- Clarke, P.J., et al., Geodetic estimation of seismic hazard in the Gulf of Corinth, Geophys. Res. Lett., 2J, 1303-1306, gion.It wouldbe interestingto discusssuch a possible 1997. mechanism for the Corinth rift to estimate how it could Clarke, P.J., et al., Crustal strain in central Greece from localize the deformation and in which conditions this repeated GPS measurements in the interval 1989-1997, could be an alternative model to creeping at shallow Geophys. J. Int., 135, 195-214, 1998. depth as proposedhere. Davies, R.R., P.C. England, B. Parsons, H. Billiris, D. Par- adissis, and G. Veis, Geodetic strain of Greece in the interval 1892-1992, J. Geophys. Res., 102, 24,571-24,588, Acknowledgments. More than 60 peopleparticipated 1997. in one or more GPS campaigns.We thank all of them. R. Denys, P., et al., GPS networks for determining the accu- Armijo participatedin the initiationof the projectand the mulation of current crustal strain in central Greece, paper networkdesign. This work benefitedfrom severaldiscus- presented at 1st International Symposium on Deforma- sions with P. Bernard and J. Chary. We thank Barry Par- tion in Turkey, Istambul Technical University, Istambul, sons,John Beavan,and an anonymousreviewer for their Turkey, 1994. commentsand suggestions.This work was supportedby Gilbert, L.E., C.H. Scholz, and J. Beavan, Strain local- CE EPOC CT 91-0043program, CNRS/INSU DBT, IDYL, ization along the San Andreas fault: Consequencesfor PNRN programs, and Greek-FrenchPLATON program. loading mechanisms,J. Geophys. Res., 99, 23,975-23,984, IPGP contribution 1697. 1994. Hager, B., G. Lyzenga, A. Donnellan, and D. Dory, Rec- References onciling rapid strain accumulation with deep seismogenic fault planes in the Ventura Basin, California, J. Geophys. Ambraseys, N.N., and J.A. Jackson, Seismicity and asso- Res., 10•, 25,207-25,219, 1999. ciated strain of central Greece between 1890 and 1988, Hassani, R., and J. Chary, Anelasticity explains topography Geophys. J. Int., 101,663-708, 1990. associatedwith Basin and Range normal faulting, Geol- Ambraseys, N.N., and J.A. Jackson, Seismicity and strain ogy, 2J, 1095-1098, 1996. BRIOLE ET AL.: DEFORMATION OF THE CORINTH RIFT 25,625

Hatzfeld, D., et al., The Galaxidi earthquake of November Rigo, A., Etude sismotectoniqueet geodesiquedu Golfe de 18, 1992: A possible asperity within the normal fault sys- Corinthe (Grece), Ph.D. thesis,Univ. Paris VII, 1994. tem of the Gulf of Corinth (Greece),Bull. Seisrnol.Soc. Rigo, A., H. Lyon-Caen, R. Armijo, A. Deschamps, D. Am., 86, 1987-1991, 1996. Hatzfeld, K. Makropoulos, P. Papadimitriou, and I. Kas- Hubert, A., G.C.P. King, R. Armijo, and B. Meyer, Fault saras, A microseismic study in the western part of the reactivation, stress interaction and rupture propagation Gulf of Corinth (Greece):Implications for large-scalenor- in the 1981 Corinth earthquake sequence, Earth Planet. mal faulting mechanisms, Geophys. J. Int., 126, 663-688, Sci. Lett., 1•2, 573-586, 1996. 1996. Jackson, J.J., and D. McKenzie, The relationship between Ruegg, J.C., and C. Bougault, AG3D, Notice d'utilisation, plate motions and seismic moment tensors and the rate of IPGP report, 32 pp., Inst. de Phys. du Globe de Paris, active deformation in the Mediterranean and the Middle 1992. East, Geophys. J., 93, 45-73, 1988. Savage, J.C., Displacement field for an edge dislocation in Jackson, J.J., J. Gagnepain, G. Houseman, G. King, P. Pa- a layered half-space, J. Geophys. Res., 103, 2439-2446, padimitriou, P. Soufleris, and J. Virieux, Seismicity, nor- 1998. mal faulting and the geomorphologicaldevelopment of the Sabrier, M., Tectonique racente d'une transversale it l'Arc Gulf of Corinth (Greece): The Corinth earthquakesof Egaen: Le golfe de Corinthe et sesragions pariphariques, February and March 1981, Earth Planet. Sci. Lett., 57, thase de 3e cycle, Univ. Paris-Sud, Orsay, France, 1977. 377-397, 1982. Tarantola, A., and B. Valette, Generalized nonlinear in- Kahle, H.G., M.V. Muller, A. Geiger, G. Danuser, S. Mueller, verseproblem solvedusing the least squarescriterion, Rev. G. Veis, H. Billiris, and D. Paradissis, The strain field in Geophys., 20, 219-232, 1982. NW Greece and the Ionian Islands: Results inferred from Taymaz, T., J.A. Jackson,and D. McKenzie, Active tecton- GPS measurements, Tectonophysics,2.9, 677-680, 1995. ics of the north and centralAegean Sea, Geophys.J. Int., King, G.C.P., D. Oppenheimer, and F. Amelung, Block ver- 106, 433-490, 1991. sus continuum deformation in the western United States, Veis, G., H. Billiris, B. Nakos, and D. Paradissis,Tec- Earth Planet. Sci. Lett., 128, 55-64, 1994. tonic strain in Greecefrom geodeticmeasurements, Proc. Le Pichon, X., and J. Angelier, The Aegean sea, Philos. Athens Acad., 67, 129-166, 1992. Trans. R. Soc. London, Set. A, $00, 357-372, 1981. Le Pichon, X., N. Chamot-Rooke, S. Lallemant, R. Noomen, and G. Veis, Geodetic determination of the kinematics of A. Balodimou,C. Mitsakaki,K. Papazissi,and G. Veis, central Greece with respect to Europe: Implications for HigherGeodesy Laboratory, National Technical University, eastern Mediterranean tectonics, J. Geophys. Res., 100, Athens,9 HeroonPolytechniou str., GR-15780Zographos, 12,675-12,690, 1995. Greece. ([email protected];[email protected]; McKenzie, D.P., Active tectonics of the Alpine-Himalayan [email protected];[email protected]) belt: The Aegean Sea and surrounding regions, Geophys. P. Briole, and J.C. Ruegg, Dapartement de Sismolo- J. R. Astron. Soc., 55, 217-254, 1978. gie, UMR-CNRS 7580, Institut de Physiquedu Globe de Melosh, H.J., Mechanical basis for low-angle normal faulting Paris, 4 place Jussieu,F-75252 Paris Cedex 05, France. in the Basin and Range province, Nature, $J$, 331-335, ([email protected];[email protected]) 1990. A. Deschamps,Laborat&re de Gaodynamique,UMR Mercier, J.L., E. Carey, H. Philip, and D. Sorel, La nao- GeosciencesAzur, CNRS Sophia-Antipolis1, 250 av- tectonique plio-quaternaire de l'arc Egaen externe et de enueAlbert Einstein,F-06560 Valbonne, France. (de- la mer Egae et ses relations avec la sismicita, Bull. Soc. [email protected]) Geol. Ft., 18, 159-176, 1977. D. Hatzfeld, Laboratoire de GaophysiqueInterne et Papazachos, B.C., and C.B. Papazachos, The Earthquakes Tectonophysique(LGIT), UMR-CNRS 5559, Observatoire of Greece(in Greek), 356 pp., Ziti Thessaloniki,Greece, de Grenoble,BP 53, F-38041 GrenobleCedex 09, France. 1989. ([email protected]) Pham, V.N., D. Boyer, G. Chouliaras, and P. Bernard, Con- H. Lyon-Caen, Laborat&re de Gaologie,UMR-CNRS ductivita dectrique et structure de la crofite dans la ra- 8535, EcoleNormale Suparieure, 24 rue Lhomond,F-75231 giondu Golfede Corinthe(Grace) d'apr•s les rasultatsde ParisCedex 05, France.([email protected]) sondagemagnatotellurique (SMT), C. R. Acad. Sci., Set. A. Rigo, Observatoire Midi-Pyranaes, Groupe de II, 32.9, 651-656, 1996. Rechercheen GaodesieSpatiale (GRGS), UMR-CNRS Rietbrock, A., C. Tiberi, F. Scherbaum, and H. Lyon-Caen, 5562, 14 avenueEdouard Belin, F-31400 Toulouse,France. Seismic slip on a low angle normal fault in the Gulf of ([email protected]) Corinth: Evidence from high-resolution cluster analysis of microearthquakes, Geophys. Res. Lett., 2S, 1817-1820, (ReceivedAugust 5, 1999; revisedApril 7, 2000; 1996. acceptedApril 18, 2000.)