Physics of the Earth and Planetary Interiors 277 (2018) 99–112 ContentslistsavailableatScienceDirect Physics of the Earth and Planetary Interiors journal homepage: www.elsevier.com/locate/pepi Seismic anisotropy in central north Anatolian Fault Zone and its implications T on crustal deformation A. Licciardia,⁎, T. Ekenb, T. Taymazb, N. Piana Agostinettic, S. Yolsal-Çevikbilenb aGeophysicsSection,SchoolofCosmicPhysics,DublinInstituteforAdvancedStudies,Dublin,Ireland bDepartmentofGeophysicalEngineering,FacultyofMines,IstanbulTechnicalUniversity,Istanbul,Turkey cDepartmentofGeodynamicsandSedimentology,FacultyofEarthSciences,GeographyandAstronomy,UniversityofVienna,Vienna,Austria A R T I C L E I N F O A B S T R A C T Keywords: WeinvestigatethecrustalseismicstructureandanisotropyaroundthecentralportionoftheNorthAnatolian Seismicanisotropy FaultZone,amajorplateboundary,usingreceiverfunctionanalysis.Thecharacterizationofcrustalseismic Crustaldeformation anisotropyplaysakeyroleinourunderstandingofpresentandpastdeformationprocessesatplateboundaries. Seismology Thedevelopmentofseismicanisotropyinthecrustarisesfromtheresponseoftherockstocomplicatedde- Receiverfunction formation regimes inducedbyplate interaction.Through theanalysis of azimuthally-varying signals oftele- seismicreceiverfunctions,wemaptheanisotropicpropertiesofthecrustasafunctionofdepth,byemploying theharmonicdecompositiontechnique.AlthoughtheMohoislocatedatadepthofabout40km,withnomajor offsetacrossthearea,ourresultsshowaclearasymmetricdistributionofcrustalpropertiesbetweenthenorthern andsouthernblocks,dividedbytheNorthAnatolianFaultZone.Heterogeneousandstronglyanisotropiccrustis presentinthesouthernblock,wherecomplexintra-crustalsignalsaretheresultsofstrongdeformation.Inthe north,asimplerandweaklyanisotropiccrustistypicallyobserved.Thestrongestanisotropicsignalislocatedin thefirst15kmofthecrustandiswidespreadinthesouthernblock.Stationslocatedontopofthemainactive faultsintheareaindicatethehighestamplitudes,togetherwithfault-parallelstrikesofthefastplaneofani- sotropy.Weinterprettheoriginofthissignalasduetostructure-inducedanisotropy,androughlydetermineits depthextentupto15–20kmforthesestations.Awayfromthefaults,wesuggestthecontributionofpreviously documentedupliftedbasementblockstoexplaintheobservedanisotropyatupperandmiddlecrustaldepths. Finally,weinterpretcoherentNE-SWorientationsbelowtheMohoasaresultoffrozen-inanisotropyinthe uppermantle,assuggestedbypreviousstudies. 1. Introduction The determination of the directional dependence of seismic wave speed, also known as seismic anisotropy, plays a fundamental role in Asanintercontinentaldextralstrike-slipfaultwithsignificantstrain the elucidation of the complicated deformation regimes induced by localization,the1600-km-longNorthAnatolianFaultZone(NAFZ)re- plateinteractionalongsuchplatemargin. presents a major plate boundary between the Eurasian plate in the Crustalseismicanisotropyisgenerallyattributedtothealignmentof northandtheAnatolianplateinthesouth.Althoughcollisionbetween joints or microcracks, to lattice preferred orientation (LPO) of aniso- theArabianandEurasianplates(intheeast)wasinitiallythoughttobe tropic minerals, or to highly foliated metamorphic rocks (e.g., themaindrivingforceforthewestwardmotionoftheAnatolianplate Sherringtonetal.2004).Intheuppercrust,possiblesourcesofseismic (e.g.DeweyandŞengör,1979),recentadvancesinhigh-resolutionGPS anisotropy can be either stress-induced or structure-induced (Boness datahaverevealedaclearroleofthesouthwest-trendingrollbackofthe andZoback,2006).Stress-inducedanisotropycanbegeneratedeither Hellenic subduction zone in the south Aegean Sea for the rapid de- bytheextensivedilatancyoffluid-filledmicrocracks(Crampin,1987) formationoftheAegean-Anatolianregion(e.g.McCluskyetal.,2000; or by the preferential closure of fractures by the in situ stress field Reilinger et al., 2006). In this respect, the deformation history of the (Boness and Zoback, 2006). In the latter case, the orientation of fast rocks atvariousdepth rangesremainsenigmatic withinthecrustand waves of vertically propagating shear waves aligns parallel to the mantle ofthis complextectonic setting. Adetailed sketchoftheAna- maximum horizontal stress (SH ). When structure-induced mechan- max toliantectonicsettingcanbefoundinFig.1. isms are dominant, seismic anisotropy may be associated to the ⁎Correspondingauthorat:Géosciences,UniversitédeRennes1,Rennes,France. E-mailaddress:[email protected](A.Licciardi). https://doi.org/10.1016/j.pepi.2018.01.012 Received4October2017;Receivedinrevisedform14December2017;Accepted26January2018 Available online 06 February 2018 0031-9201/ © 2018 Elsevier B.V. All rights reserved. A.Licciardietal. Physics of the Earth and Planetary Interiors 277 (2018) 99–112 Fig.1.Asketchmapofactivetectonicboundariesinthe studyareaandsurroundings.Forthesourceofcompiled data,pleaseseeTaymazetal.(1990,1991,2004,2007a,b), Yolsal-Çevikbilen and Taymaz (2012) and references therein. Abbreviations: AS: Apşeron Sill, ASM: Anaxi- mander Sea Mountains, BF: Bozova Fault, BGF: Beyşehir Gölü Fault, BMG: Büyük Menderes Graben, BuF: Burdur Fault, CTF: Cephalonia Transform Fault, DSF: Dead Sea Transform Fault, DF: Deliler Fault, EAF: East Anatolian Fault, EcF: Ecemiş Fault, EF: Elbistan Fault, EPF: Ezine Pazarı Fault, ErF: Erciyes Fault, ESM: Eratosthenes Sea Mountains,G:Gökova,Ge:GedizGraben,GF:GarniFault, HB:HerodotusBasin,IF:IğdırFault,KBF:KavakbaşıFault, KF: Kağızman Fault, KFZ: Karataş-Osmaniye Fault Zone, MF: Malatya Fault, MRF: Main Recent Fault, MT: Muş Thrust, NAF: North Anatolian Fault, NEAF: North East Anatolian Fault, OF: Ovacık Fault, PSF: Pampak-Savan Fault,PTF:PaphosTransformFault,RB:RhodesBasin,SaF: SalmasFault,Si:SimavGraben,SuF:SultandağFault,TeF: TebrizFault,TF:TatarlıFault,TGF:TuzGölüFault.Black arrowsexhibitrelativeplatemotionswithrespecttoEur- asia(McCluskyetal.,2003;Reilingeretal.,2006). alignment of macroscopic features due to shear-induced deformation et al., 2010; Park and Levin, 2016) has proven to be effective to nearactivefaults(ZhangandSchwartz,1994;ZinkeandZoback,2000; quantifyseismicanisotropyinvariousgeodynamical settingsoverthe Tadokoro et al., 2002), sedimentary bedding planes (Kern and Wenk, last decade and at different scales of investigation (Piana Agostinetti 1990),andpreferredmineralalignments(Sayers,1994). et al., 2011; Bianchi et al., 2015; Olugboji and Park (2016); Vinnik Local shear waves splitting analyses, performed over detected et al., 2016) including the shallow crust (Licciardi and Piana micro-seismic earthquakes along various segments of the NAFZ, have Agostinetti, 2017; Piana Agostinetti et al., 2017). In particular, RF indicatedaspatialcorrelationbetweenthefastpolarizationdirections harmonicshavebeenusedtomapthedepth-dependentdistributionof (FPDs)andstationdistancefromthefault(e.g.Tadokoroetal.,2002; seismicanisotropyinareasofintensecrustaldeformation,e.g.,around PengandBen-Zion,2004;HurdandBohnhoff,2012;Ekenetal.,2013). the San Andreas Fault (SAF) (Audet 2015), the Tibetan Plateau (Liu LateralvariationsoftheFPDsinferredfromlocalshearwavesimplied et al., 2015), the Cyclades (Cossette et al., 2016), the Canadian Cor- thepresenceofbothstress-andstructureinducedmechanismscausing dillera(Tarayounetal.,2017)andtheAppennines(Bianchietal.,2010; seismicanisotropyintheupper8–10kmofthecrust.Inaddition,these 2016). local splitting studies have highlighted the structural control of the In this work, we analyse RFharmonics using data from theNorth complexgeologicandtectonicenvironmentalongthewesternsegments Anatolian Fault passiveseismic experiment([dataset]BeckandZandt, oftheNAFZonthespatialvariationofFPDs. 2005;Biryoletal.,2010),inordertodelineatethefirst-orderseismic For the deeper part of the crust, i.e.>20–25km, several studies structure of the crust and to map crustal anisotropy as a function of have shown that aligned minerals are the most likely cause of aniso- depth.Inparticular,ourmainobjectivesaretoelucidatei)orientation tropy (Sherrington et al., 2004, and references therein) and that hex- andstrengthofdeformationinthecrustatvariousdepthrangesii)how agonal anisotropy with a unique slow symmetry axis can explain se- much the strain fields within crust and upper mantle are coupled iii) ventypercentoftheobservations(Brownleeetal.,2017).Inparticular, possiblelinkbetweenlateralvariationofcrustalanisotropyparameters thealignmentofmicasalongtheplaneoffoliationisoftentheprimary and existing lithology contrast across the NAF. These results yield in- causeofbulkanisotropyinthisdepthrange(Sherringtonetal.,2004; sight into the poorly known role of crustal seismic anisotropy in the Audet2015). area. Overthelastdecade,receiverfunction(RF)datahavebeenwidely usedforthecharacterizationofseismicanisotropy.RFsaretimeseries 2. GeologicalsettingofNorth-CentralAnatolia that represent the impulse response of the near receiver structure in terms of P-to-S conversions contained in the P-coda of teleseismic Thestudyregionislocatedinanimportantareaoforogenicamal- events (Vinnik, 1977; Langston, 1979). After deconvolution of the gamation of Anatolia, a transition zone between compressional-de- verticaltracefromthehorizontalones,P-to-SVandP-to-SHconverted formed eastern Anatolia and extensional western Anatolia. There are phasesareisolatedontheRadial(R)andTransverse(T)componentsof numerous key structures developed under the complex deformation, theRFs,respectively.Inparticular,P-to-SHconversionsaregenerated suchastheEzinePazarı–SungurluFault,theİzmir–Ankara–Erzincan fromtherotationoftheenergyoutofthesource-receiverplaneinduced (IAESZ) and Intra – Pontide Suture Zones, the İstanbul Zone, the byanisotropyand/ordippingvelocitycontrastsatdepth(Sherrington Sakarya Continent, the Central Pontides, the Kırşehir Massif and the etal.,2004;MaupinandPark,2007;PianaAgostinettiandChiarabba, ÇankırıBasin(OkayandTüysüz,1999;Fig.2).Itisreportedthatsome 2008; Schulte-Pelkum and Mahan, 2014a). The analysis of the azi- ofthesestructures(e.g.,theIstanbulZone;Şengör,1979)werepartsof muthally varying characteristic of the P-to-S conversions (amplitudes Eurasia, while other fragments were separated from the Arabian- anddelaytimes)canproviderobustinformationaboutthelocationof AfricanPlate.Görüretal.(1998)furtherinformthatthemajorbasinsin anisotropy at depth (e.g. Rümpker et al., 2014; Licciardi and Piana Central Anatolia were formed on continental units; i.e., the Sakarya Agostinetti, 2016). RFs provide complementary depth-dependent in- Continent andtheKırşehirMassifadjacent to thesuture zones. These formationaboutseismicanisotropythatisdifficulttoobtainwithother structures have significant importance on understanding the tectonic common seismological data (e.g. shear wave splitting and surface evolutionoftheregion.Forexample,theİzmir–Ankara–Erzincansuture waves dispersion), since RFs are strongly sensitive to the depth of zone(IAESZ)isaremnantoftheNeo-TethysOceanandhenceconsists contrastsinanisotropicproperties. ofophioliticunits(Rojay,2013).ItseparatesthePontidestothenorth Moreindetail,theRFharmonicdecompositiontechnique(Bianchi from the Anatolide–Tauride and the Kırşehir blocks to the south 100 A.Licciardietal. Physics of the Earth and Planetary Interiors 277 (2018) 99–112 Fig.2.Mapwiththelocationofthe38broadbandseismic stations used in this study. Stations belong to the North AnatolianFault(NAF)passiveseismicexperiment(Biryol etal.,2010),whichoperatedfromJanuary2006toMay 2008. Major faults are indicated in black. Suture zones modifiedafterOkayandTüysüz(1999)areindicatedwith redlines.TracesoftheprofilesshowninFig.7areindicated inblueandlabelled.StationsusedinFigs.4and5(ALIC and DOGL) are indicated by red triangles. CB, Çankırı Basin; CP, Central Pontides; ESF, Ezinepazari-Sungurlu Fault; IAESZ, Izmir-Ankara-Erzincan Suture Zone; IPS, Intra-PontideSuture;IZ,IstanbulZone;ITS,Intra-Tauride Suture;KM,KırşehirMassif;NAF,NorthAnatolianFault;.SZ SakaryaZone. (Koçyiğit, 1991; Okay and Tüysüz, 1999; Taymaz et al., 2007a,b). 2005;Biryoletal.,2010).Foreachstation,weselectedtheeventswith Yolsal-Çevikbilen et al. (2012) identified three major high-velocity high signal-to-noise ratio (SNR) waveforms using a visual inspection blocks in the mid-crust of north-central Anatolia, separated by the process. These waveforms were extracted from a list of about 1000 İzmir-Ankara-Erzincan SutureZone,andtheyinterpretedtheseblocks teleseismic events with magnitude≥5.5 that occurred at epicentral to be continental basement fragments that were accreted onto the distances(Δ)between30°and100°.Mostoftheeventscomefromthe margin, following the closure of Neo-Tethyan Ocean. They also pro- first quadrant for the selected Δ range, where the major seismogenic posedthatthesehigh-velocitybasementblocksmightcontroltherup- zonesarelocated.Nonetheless,agoodazimuthalcoverageisachieved turepropagationsofthehistorical1939,1942and1943earthquakes. foralmostallstations.AnexampleofsuchcoverageisshowninFig.3 Similarly, the Çankırı basin is a natural laboratory for studying the forstationDOGL.Theselectedthree-componentseismogramsarethen subduction and collision processes, and therefore it is accepted as an rotatedintotheZRTreferencesystem,whereRistheRadialdirection essentialunitshapingactivetectonicsofthenorth-centralAnatolia.It along the great circle path from the epicenter to the station, T is the locatesatthecontactregionbetweenthePontidesandtheAnatolide- TransversedirectionperpendiculartoRinthehorizontalplane,andZis Tauride Block, where the 1200km long North Anatolian Fault Zone the vertical component. In RF analysis, the Z component of the bifurcates into a number of small-scale splay faults deforming the Anatolian Block internally, with a complex rotational deformation (Seyitoğlu et al., 2004; Kaymakçı et al., 2009; Lucifora et al., 2013). Recently, Lucifora et al. (2013) have obtained asymmetric tectonic rotations along the opposite edges of the Çankırı Basin from palaeo- magnetic data. They have correlated that complex pattern of paleo- magneticrotationswithalocalblockrotationmechanismcausedbythe strike-slipfaultactivityalongthemarginsoftheBasin. Theotheressentialtectonicunitsinnorth-centralAnatoliaarethe Sakarya Continent andthe KırşehirMassif, also known astheCentral Anatolian Crystalline Complex (Akiman et al., 1993). The Kırşehir Massif has a triangular shape that spans over an area of ∼300km× ∼200km (Şengör and Yilmaz, 1981), and is bordered by the right- lateralTuzGölüFaultZoneonthewest(Çemenetal.,1999)andbythe left-lateralstrikeslipEcemişFaultZoneontheeast(seeFig.2).Itisalso surrounded by ophiolitic melanges associated with subduction-accre- tion complexes of the Izmir-Ankara-Erzincan and the Inner Tauride suture zones. This Massif is dominated by Late Cretaceous age high- grademetamorphism,intrudedbybatholiths(Ketin,1966;Göncüoğlu, 1977;WhitneyandHamilton,2004).Itconsists ofmetamorphiccrys- talline rocks, i.e., metasediments, ophiolitic sequences, and magmatic intrusions(Kaymakçıetal.,2009).Similarly,theSakaryaContinenthas a high-grade metamorphic basement which consists of gneisses, mar- bles, amphibolites, and metaperidotites, with granodiorite intrusions (Okay,1996;Görüretal.,1998). 3. Dataandmethods Fig.3.AzimuthalequidistantprojectioncenteredatstationDOGLwiththesourceloca- WeusedteleseismicdatarecordedbetweenJanuary2006andMay tionoftheteleseismiceventsusedtocomputetheRFdatasetforthisstation(blacktri- 2008at38broadbandseismicstationsbelongingtotheNorthAnatolian angle).Onlyeventsfrom30°to100°ofepicentraldistancehavebeenused.Sizeofcircles Fault (NAF) passive seismic experiment ([dataset]Beck and Zandt, isproportionaltoeventmagnitude(M)asperlegend. 101 A.Licciardietal. Physics of the Earth and Planetary Interiors 277 (2018) 99–112 observedseismogramsistakenasaproxyofthesourcewavelet,while (red) color is used for positive (negative) pulses associated with an horizontal components hold information about the near receiver S- increase (decrease) of velocity with depth. On the RRF sweeps, the wavevelocitystructure.Thelattercanberecoveredafterdeconvolution strongpositivepulseatabout5.5scanbeassociatedwiththeconver- oftheverticalcomponentfromthehorizontal ones.Inthiswork,RFs sionoriginatedattheMoho.InthetimewindowprecedingtheMoho are computed through the frequency domain deconvolution method arrival, crustal complexities beneath both stations are represented by developedbyDiBona(1998). high-amplitude pulses (both positive and negative). High amplitudes AGaussianfilterwithlow-passfrequencyof∼1Hz(Gaussianfilter are also observed on the TRF panels at both stations for delay times width a=2) is applied to each RF, in order to rule out the effect of between 0 and 6s. This observation indicates that strong P-SH con- high-frequency noise. This choice limits the vertical resolution of the versions take place at crustal depths. Moreover, clear patterns and RFs to about 2–4km (Piana Agostinetti and Malinverno, 2017). Both periodicity canbe observed inthewaveformson theTRF panels sug- radial receiver function (RRF) and transverse receiver function (TRF) gestingthepresenceof3-Dfeaturesatdepth.Insuchcomplextectonic arecomputedfromthethree-componentseismogramsofasingleevent. settings,usefulinformationaboutthesubsurfacecanbeextractedfrom AfterRFsareobtained,anothervisualinspectionisperformed,andonly bothRRFandTRF. RFwithacceptableSNR,small“ringing”andsmallamplitudesfordelay Inthenextsection,asimple,yeteffectivemethodispresentedfor times<0s areretained.Thisproduces atotalnumber of∼7100 RFs theanalysisofboththe“isotropic“and“anisotropic”componentsofthe withanaveragevalueof∼180RFsperstation. RFdataset,i.e.theenergycontainedintheRFdatasetgeneratedfrom Inaflat-layeredandisotropicmedium,onlyP-to-SVconversionsare bulk S-wave velocity discontinuities, also referred as “isotropic struc- generated and no energy should be observed on the TRF (Savage, ture”, and from anisotropic materials at depth, also referred as “ani- 1998). However, the presence of anisotropy and/or dipping velocity sotropicstructure”. contrasts at depth (called 3-D features hereinafter) rotates the energy outofthesource-receiverplaneandproducesP-to-SHconversionsthat arecommonlyobservedontheTRF(ParkandLevin,2016).Inthiscase, 3.1. Harmonicdecomposition two-lobed (360°) or four-lobed (180°) periodicity in amplitudes and delaytimesasafunctionofback-azimuth(ф)areexpectedonbothTRF Classical RF studies assume isotropic and flat-layered media for andRRF(LevinandPark,1998;Audet,2015;Bianchietal.,2015). which the response of the subsurface is invariant with the back-azi- InordertoassessthequalityofthefinalRFdatasetandtohavea muth,resultinginalltheP-to-Sconvertedenergytobeconfinedinthe first glimpse atthecontribution of3-D features beneath each station, RRF.Anisotropyand/ordippingvelocitycontrastsintroduceamplitude weplotboththeRRFandTRFasafunctionofф,afterstackinginфand and delay time variations as a function of back-azimuth on both RRF Δwithbinsof20°and40°widthrespectively,and50%overlaptoin- and TRF with distinct periodicity (either π or 2π). These effects are creasetheSNR.InFig.4weshowtwoexamplesforstationsALICand coupledontheRRFandTRF,drasticallyincreasingthecomplexityofa DOGL,whicharelocatedontopofthesurfacetraceoftwomajoractive given RF dataset. This makes even a qualitative analysis of raw RF faultsofthearea,i.e.theNAFandtheESF,respectively.Althoughthe datasets (as those shown in Fig. 4) difficult, leading to strongly non- finalback-azimuthalsweepsaredifferentbetweenthesestations,both unique interpretations. To overcome this difficulty, RRF and TRF ofthemshowazimuthalvariationsinamplitudeanddelaytimesofthe should be analysed jointly. The harmonic decomposition technique conversions along the RRF and TRF components. In this work, blue providesasimplewayforthesimultaneousanalysisoftheRRFandTRF (Bianchietal.,2010;Audet,2015).Thistechniqueaimsatseparating Fig.4.BinnedRadialandTransverseRFsforstations(a)ALICand(b)DOGL,plottedasafunctionofback-azimuth.Graynumbersindicatethecentralepicentraldistanceusedinthe binningprocedure.SmallnumbersindicatethetotalnumberofRFsmakingupeachindividualbin.Blue(red)pulsescorrespondtoaninterfacewithapositive(negative)downward velocityjump.(Forinterpretationofthereferencestocolourinthisfigurelegend,thereaderisreferredtothewebversionofthisarticle.) 102 A.Licciardietal. Physics of the Earth and Planetary Interiors 277 (2018) 99–112 the isotropic response from the response induced by 3-D features at (derivativepulses)onbothk=1componentsinthefirst4s,although depth exploiting the azimuth-dependent signatures produced by the station DOGL is dominated by the derivative pulse on the cos(ф) latter. Thisrepresents a clearadvantage over classical RFanalysis, as componentwithpeaksat1.6and3s.Smallamplitudesareobservedon the RF dataset is partitioned into azimuth-invariant and azimuth-de- the k=2 harmonics compared to k=1. In addition, for the k=2 pendentcomponents,thusfacilitatingsubsequentinterpretations. harmonics, comparable amplitudes are observed between the “mod- Moreindetail,themethodisbasedontheassumptionthatanyRF eled”and“unmodeled”terms.Theseobservationsholdforthemajority dataset can be approximated as a series of cos[kф(t)] and sin[kф(t)], ofthestations,thus,inthefollowing,weexcludethek=2harmonics withk=0,1,2,…Inthisway,theazimuth-dependentfeaturescanbe and restrict the analysis only to the first two order harmonics. We isolatedfromtheazimuth-invariantcomponents.Thek=0component displaythecomponentsofcos(ф)(orientedN-S)andsin(ф)(orientedE- istheconstantazimuthtermanddescribestheisotropicstructure(i.e. W) separately for k=1, giving a total of three harmonic coefficients contrastsinVS).AtthecoreofthemethodisthattheSNRofthetwo- (i.e.threetimeseries). lobed or four-lobed periodicity is enhanced on the k=1 and k=2 harmonics,respectively,bystackingtheRRFandTRFwithapositive 3.2. Energyandmap-orientationof3-Dfeatures shiftinback-azimuth(+π/2k),asshownbyShiomiandPark(2008). The two-lobed periodicity (k=1) is associated with the presence of The main aim of this work is to map the position of anisotropic dipping velocity contrasts or hexagonal anisotropy with plunging bodies and/or dipping velocity contrasts at crustal depths and to de- symmetryaxisatdepth.Ontheotherhand,thefour-lobedperiodicity terminetheirorientations. Basedontheharmonicsdatasetpreviously (k=2)isuniquelyproducedbyanisotropywithhorizontalsymmetry described,wefirstperformatime-to-depthconversionofthek=0and axis(GirardinandFarra,1998). k=1harmonicsateachstation.ThestandardIASP91velocitymodel FollowingBianchietal.(2010)andLiuetal.(2015),foreachtime (KennettandEngdahl,1991)isusedforthispurpose.Weusethemean step we can recover five coefficients, A(t), B(t), C(t), D(t), E(t), by slowness value at each station to perform the depth conversion. In solvingthefollowinglinearsystemofequationsforeachtime-steptin principle,thischoicecouldproducedefocusedresultsintheharmonics, theRFdataset: if the range of slowness considered is broad. In order to estimate the ⎡⎢R1…(t)⎤⎥ ⎡⎢…1 cos(ϕ1) sin(ϕ1) cos(2ϕ1) sin(2ϕ1) ⎤⎥⎡⎢AB((tt))⎤⎥ ⎢⎢RTN((tt))⎥⎥=⎢⎢10 cosc(ϕos(+ϕNπ)/2) sin(sϕin(+ϕNπ)/2) cos(c2o(ϕs(2+ϕNπ)/4)) sin(2si(nϕ2(+ϕNπ)/4))⎥⎥⎢⎢C(t)⎥⎥, ⎢ 1… ⎥ ⎢… 1 1 1 1 ⎥⎢D(t)⎥ ⎣⎢TN(t)⎦⎥ ⎣⎢0 cos(ϕN +π/2) sin(ϕN +π/2) cos(2(ϕN +π/4)) sin(2(ϕN +π/4))⎦⎥⎣⎢E(t)⎦⎥ (1) whereNisthetotalnumberofeventsconsideredateachstation.A(t) error associated with this choice, we compare our results with an al- corresponds to the k=0 harmonic, B(t) and D(t) are the cos[kф(t)] ternativeapproachinwhichtheharmonicdecompositionisperformed components and C(t) and E(t) are the sin[kф(t)] components of the in depth domain after time-to-depth conversion of the RF dataset k=1andk=2harmonics.Withthisconfiguration,cos[kф(t)]andsin (Audet 2015; Cossette et al., 2016; Tarayoun et al., 2017), thus pre- [kф(t)] for k=1 are orthogonal and oriented N-S and E-W, respec- serving the information about slowness. The comparison points out tively. The same linear system can be solved using a negative phase shift(−π/2k),inwhichtheeffectofdippingstructuresandanisotropy issuppressed(Liuetal.,2015).Inthiscase,theretrievedcoefficients quantifytheeffectofcomplex3-Dstructures,scattering,orincomplete geometrical coverage (“unmodeled” components hereinafter). The “unmodeled”coefficientscanbeseenasanestimateoftheerrorasso- ciated with the harmonic decomposition. For further details on the method,thereaderisreferredtoParkandLevin(2016). The harmonic decomposition is applied to the RF dataset at each station, after stacking in ф and Δ with bin-width of 10° and 20° re- spectively, and with no overlap. This choice prevents the use of the same RF twice in the analysis, thus reducing any possible bias in- troduced by smoothing. As an example, we show the harmonic de- compositionoftheRFdatasetsatstationsALICandDOGLinFig.5.The values ofф andΔofeachbinusedtocomputetheharmonicsarein- dicatedwithredstarsforeachstation. The two stations show different isotropic responses on the k=0 plot.TheP-to-SconversionoriginatedattheMohocanbeidentifiedas the positive pulses at about 5.5 and6.0s on ALIC andDOGL, respec- tively.Atearlierdelaytimes,ALICshowspositivepulsesatabout1and 2.4swhileabroadnegativepulsecanbeobservedat4s.Onthecon- trary, the pulse with highest amplitude at station DOGL is found at about0.8s,indicatingthepresenceofastrongandshallowimpedance contrast,followedbyanegativepulseat2.3s. The k=1 harmonics (cos(ф) and sin(ф) traces) show higher am- plitudes on the “modeled” part than on the “unmodeled”, thus in- Fig.5.HarmonicdecompositionofRFsforstations(a)ALICand(b)DOGL.k=0trace dicatingasignificantandcoherentcontributionfromtheazimuthally- correspondstothe“constant”,zero-orderharmonic.cos[kф(t)]andsin[kф(t)]harmonics fork=1,2areshowninthebottompanelsformodeled(+π/2kshiftinthestack)and varying structure beneath the two stations. Higher amplitudes are unmodeled(−π/2kshiftinthestack)parts.Seetextfordetails.Toprightpanelshowsthe foundinthefirst6sforbothstations,suggestingacrustaloriginofthis geometricaldistribution(inback-azimuth,фandepicentraldistance,Δ)ofRFsusedfor signal.Bothstationsdisplayaseriesofpulseswithoppositepolarities theharmonicdecomposition.Thesameamplitudescaleisusedinallplots. 103 A.Licciardietal. Physics of the Earth and Planetary Interiors 277 (2018) 99–112 that, given the distribution of events for the NAF experiment, small interpretations.Itcould,forexample,representthetrendoftheplun- errorsareproducedbyusingthemeanrayparameterandthiseffectis gingsymmetryaxisofhexagonalanisotropy.Inthiscase,theaxiswould negligibleinthecrust(seeAppendixA). plungetowardseitherNWorSE.Someambiguitiesariseinthiscase,as Consideringthattheinter-stationdistanceisgenerallygreaterthan noinformationaboutthenatureofthesymmetryaxis(fastorslow)is 30km, we build depth-profiles by juxtaposing single-station data and available (see Sherrington et al., 2004 for details). A second possible focusourattentiononverticalandlateralvariationsofcrustalstructure interpretation provides the dipping direction of a dipping velocity (k=0harmonics)andpositionof3-Dfeatures(fromk=1harmonics) contraststrikingNE-SW.Thisambiguityisstrongerintheuppercrust, alongtheprofiles. butitdiminisheswithincreasingdepth,wherestronglydippingvelocity Asaproxyforthepositionof3-Dfeaturesatdepth,wecomputethe jumps(morethan30°)arelesslikelytooccur,andhenceanisotropyis sumofthesquaresofthe(depthconverted)“modeled”cos(ф)andsin assumedasthemaincauseoftheobservedsignals. (ф)termsforthek=1harmonics(PianaAgostinettiandMiller,2014) In the next section, to facilitate our interpretation, we follow and subtract the contribution of the “unmodeled” counterpart ac- Schulte-Pelkum and Mahan (2014b) and plot the direction normal to cordingto: the retrieved trend direction (i.e. the strike). Therefore, for a dipping velocitycontrastourmap-projectedstrikecanbedirectlyinterpretedas E(z)=[B(z)2M +C(z)2M]−[B(z)U2 +C(z)U2] (2) the strike of the dipping velocity contrast. On the other hand, for hexagonalanisotropy,thisgivesusthemap-projectedorientationofthe wherethesubscriptsMandUreferstothe“modeled”and“unmodeled” planeperpendiculartotheplungingsymmetryaxis.Incaseofnegative harmoniccoefficientsandzisthedepthaxis.TheresultingquantityE(z) anisotropy (unique slow symmetry axis), this is the strike of the fast (referredas“energy”hereinafter)highlightsthepositionof3-Dfeatures plane,whichcanbeinterpretedasthestrikeoffoliationorthestrikeof atdepthinanintuitiveway,thusmakingtheinterpretationoftheresults theplaneofpreferredcracksormineralalignment.Byplottingthemap- simpler. High energy at a certain depth may suggests the presence of projected strike together with the energy in a given depth range, we either a contrast in anisotropic properties (top or bottom of an aniso- buildenergymapswithorientationofthe3-DfeaturesasseenbytheRF tropicbodywithplungingsymmetryaxis)oradippingvelocitycontrast. data. Thisprovides uswith spatialinformation abouttheazimuthally In addition, the map-projected orientation of 3-D features can be varyingstructureandtheirfirst-orderdepthdependence. retrievedfromthedepth-convertedk=1harmonics.Infact,fromthe particlemotionofthecos(ф)(N-Scomponent)andsin(ф)terms(E-W 4. Resultsanddiscussion component)itispossibletoinferthetrendoftheuniquesymmetryaxis ofhexagonalanisotropy(orthedippingdirectionofadippingvelocity Weapplytheharmonicdecompositiontechniquetoallthestations contrast)responsiblefortheobservedsignal(Bianchietal.,2010).Ina inthearea,andplottheresultssidebysidealongsevenprofiles(Fig.7). simplelinearregressionoftheparticlemotionateachstation,theor- IneachpanelofFig.7,weplotthetopographyofeachprofilewiththe ientation of 3-D features is recovered from the slope parameter. In locationofthemainactivefaults(NAFandESF)togetherwiththere- principle, we could estimate a value of the slope parameter at each sultsoftheharmonicanalysis.Foreachstation,thewaveformsofthe depthpoint;however,asinAudet(2015),wefoundmorestableresults resulting harmonics are plotted individually. Blue lines are used to ifarangeofdepthsisconsidered.Anexampleoftheparticlemotionfor highlightourobservationsontheconstantcomponent.Inthelastplot the0–15kmdepthrangeatstationALICisshowninFig.6.Weestimate of each panel, the energy associated with the k=1 harmonics is dis- the trend direction (green arrow) from the slope parameter, together played in grayscale. Here, green and orange boxes indicate areas of withitsuncertainties(grayarrows),foreachstation,findingerrorsof strongenergyatdepthssmallerandgreaterthan30km.Locationofthe usuallylessthan15°(95%confidenceinterval). profiles is shown in Fig. 2. Five profiles (from AA′ to EE′) follow S-N The retrieved orientation (green arrow) can have multiple direction and cross themain active faults inthe area (NAF and ESF). The last two profiles, FF' and GG', run W-E and are designed to in- vestigatethecrustalsettingofthenorthernandsouthernblocksdivided bytheNAFZ.Lateralvariationsinbothk=0andk=1harmonicsare evidentinthemajorityoftheprofiles. 4.1. CrustalstructureandMohotopography Itisbeyondthescopeofthispapertodiscussindetailtheisotropic structureofthearea,asthefocusismainlyon3-Dfeatures.Giventhat theinter-stationdistancedoesnotallowforoverlappingzonesofsen- sitivityinthecrust,ourobservationsarethereforegenerallylimitedtoa single-station approach. With a vertical resolution of 2–4km, k=0 harmonicsatthestationsanalysedinthisstudyindicatethattheMoho topography is generally a continuous feature at about 40km (Fig. 7). Thisresultisingoodagreementwithearliercrustalthicknessestimates fromRFsinthesamearea,whichreportedfairlyconsistentvalueswith thepresentstudy,i.e.∼35–40km(Özacaretal.,2010;Vanacoreetal., 2013).Nevertheless,ourresultsshowafewlocalexampleswherethe identificationoftheMohoisdifficultorambiguous.AlongprofileCC′ (Fig.7c)forexample,apositivepulseislocatedatabout40kmdepth foralmostallstations.However,thecontinuityoftheMohoPsphaseis brokenattwoneighbouringstations(CAYAandPANC)between75and Fig.6.Particlemotionforthe0–15kmdepthrangeatstationALIC.Particlemotion(red points)oftheN-SandE-Wcomponentofthefirstorderharmonic.Trendofthesymmetry 105kmalongtheprofile.StationCAYAshowsalmostnosignalbetween axis(greenarrow)isestimatedthroughlinearregressionoftheparticlemotion.Grey 25 and 50km depth. The following station to the north, PANC, pos- arrows indicate three standard deviations of the estimated slope parameter from the sessesapositivepulseatabout30km,whichisquiteshallowerthanthe linearfit.(Forinterpretationofthereferencestocolourinthisfigurelegend,thereaderis general Moho depth along this profile. These two contiguous stations referredtothewebversionofthisarticle.) arelocatedbetweentheESFtothesouthandtheIAESZtothenorth.In 104 A.Licciardietal. Physics of the Earth and Planetary Interiors 277 (2018) 99–112 Fig.7.Harmonicdecompositionresults.Time-to-depthconvertedharmonicsareshownalongsevenprofiles.ThesurfacetraceofeachprofileisindicatedinFig.2.Foreachprofile,the toppanelindicatesthesurfacetopographyforreference.Thefollowingthreepanelsdisplaythek=0harmonic,andthecos[kф(t)]andsin[kф(t)]componentsofthek=1harmonic. Wigglesareplottedontopofaninterpolatedimageforclarity.Thebottompanelshowstheenergycalculatedfromthek=1harmonics,asexplainedinthemaintext.Redarrows indicatethepositionofthesurfacetracesoftheNorthAnatolianFault(NAF)andtheEzinepazari-SungurluFault(ESF).Thesamescaleisusedforharmonicsandenergyamplitudes betweendifferentpanels. 105 A.Licciardietal. Physics of the Earth and Planetary Interiors 277 (2018) 99–112 Fig.7. (continued) addition,thek=0harmonicatstationPANCshowsthehighestenergy Çankırı Basin and its sedimentary infill is likely responsible for pro- atabout5kmdepth,andverybroaddirectP-wavepulsesareobserved ducing reverberations in the sedimentary pile that can interfere with inthefirst110kmoftheprofile.StationsCAYAandPANCarelocated theprimaryPsconversionfromtheMoho,thusproducingtheobserved within the Kırşehir Massif tectonic unit. Here, the presence of the broaddirectP-wavepulse. 106 A.Licciardietal. Physics of the Earth and Planetary Interiors 277 (2018) 99–112 Themainstructuraldifferencesbetweenthecrustoftheblockslo- catedsouthandnorthoftheNAFZareevidentwhencomparingthetwo longitudinalprofilesFF′andGG′showninFig.7gandf,respectively.In FF′, the Moho Ps pulse is continuous along the length of the profile, althoughittendstogetshallowerandbroadertowardstheeast.InGG′, the pulse related to the Moho is located at a depth of about 40km, exceptforstationCAYA,whichdoesnotshowaclearMohosignal,and forstationDOGL,whichhasadeeperMoho(50kmdepth).Atthesame time,alongGG′,nocontinuityofintra-crustalsignalscanbeobserved, indicating a more heterogeneous setting in the crust. The intense de- formation history that affected the southern block, together with its tectonic activity, result in a more heterogeneous crust, in accordance with3-Dlocalearthquaketomography(Yolsal-Çevikbilenetal.,2012). This dramatically increases the complexity of the RF waveforms with respecttothemoregeologicallystablenorthernblock. Finally, there is no clear evidence of Moho offset related to the NAFZ,whichconfirmspreviouslower frequencyobservations from S- RF(Kindetal.,2015).However,theNAFZclearlyaffectsintra-crustal structures in the first 20km (see Fig. 7. profiles AA′, EE’). A smaller inter-station distance would be required in order to study the lateral continuityofintra-crustalPsconversionsacrossthismajorfaultsystem. 4.2. Seismicanisotropy IntheprofilesshowninFig.7,highvaluesoftheenergyassociated withthek=1harmonicsindicateawidespreadcontributionfrom3-D featuresthroughoutthecrustofthestudyarea.Twomainobservations canbedrawnfromtheseresults.First,asubstantialenergydropfrom southtonorthoftheNAFZisobservedalongprofilesBB′,DD′andEE′. Inaddition,highestenergywithcomplexpatternsislocatedinbetween activefaultswherecrustalstructureisalsocomplex(profilesAA′,CC′ andDD′ofFig.7)and,atleastlocally,theNAFcontrolsthetransitionof theenergyfromsouthtonorth(profileBB′inFig.7).Second,themain sourceoftheenergysignalislocatedwithintheupperandmiddlecrust, wherethehighestamplitudesareobserved.Deeper(greaterthan30km depth)sourcesoftheobservedenergyshowlowestamplitudesandless continuity.Finally,strongdifferencesintermsofenergyareobserved between profiles FF′ and GG′, both oriented W-E, and located respec- tively in the northern and southern blocks divided by the NAF. GG' showsmuchstrongerenergythanFF′throughoutthecrust. We complement the depth-dependent information about 3-D fea- turesshowninFig.7withaspatialmappinganalysisonthreediscrete depthranges:0–15,15–30and30–60km,roughlycorrespondingtothe uppercrust,middlecrustandlowercrustplusuppermantle.Foreachof these depth ranges, we build interpolated energy maps for the k=1 harmonicsbysummingthecontributionofeachpointinagivendepth Fig.8.Mapsofthetotalenergyatthreedifferentdepthranges.(a)0–15km,(b)15–30km range.Atthesametime,wecalculatethedirectionofthe3-Dfeatures, and(c)30–60kmdepth.Thecolorscaleindicatesthesumoftheenergyonthek=1har- monicsinthecorrespondingdepthrange.Strikeoftheplanenormaltotheaxisofsymmetry as explained in the previous section, from the slope parameter of the isindicatedwithagraybarateachstationandscaledbythetotalenergyinthesamedepth linear regression of the particle motion in the same depth range. The range.Themajortectonicfeaturesintheareaaredrawnwithblacklines(seealsoFig.2). resultsareshowninFig.8andsummarizedinTableS1ofthesupple- mentarymaterial. energyvalues,thusgivingrisetoahigherspatialfrequencyvariation In the following, we discuss our results regarding the energy dis- when compared to north of the NAF. A second order observation in- tribution,intermsofdepth-dependentandspatialfeatures. dicatesthatalmostallstations(13outof15stations)withhighenergy (greater than 0.5) are located on top or in close proximity of active 4.2.1. Uppercrust faults(NAFandESF)orancientsuturezones. The strongest azimuthally varying signal is generated in the first Thisisperhapsthemoststrikingfeatureofourresults:thespatial 15kmofthecrust.Overall,theenergyobservedinthisdepthrangeis distribution oftheenergy associated with the k=1 harmonics inthe higherthanthatoriginatedinthedeeperportionofthecrustandupper uppercrustisclearlyinfluencedbythestructuralsettingofthearea.In mantle(comparethetotalenergyinFig.8atdifferentdepths).Between particular, we identify the well defined partitioning between low-en- 0 and 15km depth (Fig. 8a), two spatial zones with distinctive char- ergystationsnorthoftheNAFandhigh-energystationsontopofthe acteristics are separated by the NAFZ, confirming what was already NAFandsouthwards(Figs.7and8). observedalongtheprofilesofFig.7.StationslocatednorthoftheNAF InourRFanalysis,bothdippingvelocityjumpsandseismicaniso- show consistently low energy in the k=1 harmonics. Conversely, on tropycontributetotheobservedenergy,especiallyintheuppercrust, topoftheNAFandsouthwards,higherenergyisfoundformorethan wherethecontributionofdippinglayerscanbestrong.In oursimple halfofthestations(15outof28stations).Inthisregion,inaddition, approach, the discrimination between the two effects is not easy and the pattern is more heterogeneous, with alternating low and high evenaninversionapproachwouldproducenon-uniqueresults.Adirect 107 A.Licciardietal. Physics of the Earth and Planetary Interiors 277 (2018) 99–112 interpretationofourresultsintheuppercrustisfurthercomplicatedby High-energystationslocatedontoporincloseproximityofancient thepossiblepresenceofreverberationsinducedbyshallowlow-velocity suture zones (IPS, IAESZ, and ITS) are more difficult to interpret, as layers,asthosepresentinsedimentarybasinswithcomplexgeometries they show both parallel (ALIN and CUKU) and oblique (KIZIK and (seeAppendixB). CAKM)strikes. Despite the inherent non-uniqueness of our results, shear wave Away from active faults, anisotropy is generally lower, but some splitting analyses performed on local seismicity across the different local high-energy anomalies are observed at different depth ranges, portionsoftheNAFZyieldedasignalwithamaximum∼0.2–0.3sof south of the NAF. We suggest that some of these spots of high aniso- time delay, which has been attributed to the presence of anisotropy tropyarecorrelatedwiththepresenceoftheupliftedblocksofaccreted within the first 8–10km of the upper crust (e.g., Peng and Ben-Zion, andstronglydeformedbasementrecognizedbyYolsal-Çevikbilenetal. 2004;HurdandBohnhoff,2012;Ekenetal.,2013).Inaddition,are- (2012),above30kmdepthinthecentralandeasternpartofthestudy cent3-Dradiallyanisotropictomography(Çubuk-Sabuncuetal.,2017) area (south of the NAF). Alternatively, the localization of intense de- hasshownthepresenceofstrong(20%)negativeanisotropythroughout formationwhichhasbeentakingplacesouthoftheNAFmaybecon- theAnatoliancrustataregionalscale.Forthesereasons,wewilldiscuss sideredresponsiblefortheobservedseismicanisotropyatmid-crustal thepossibleimplicationsofourresultsintermsofanisotropyonly. depths, as heavily fractured and fluid-saturated rocks, alignments of Ifweassumeslowanisotropyintheuppercrust,ourresults(Fig.8a) anisotropicmineralsandcrustalflowcanproducedirectionalvelocity indicatethestrikeofthefastplaneperpendiculartoit,correspondingto changes at different crustal depths (Fouch and Rondenay, 2006; thestrike ofbedding, foliation oralignment ofcracks ormacroscopic Mainpriceetal.,2007). fractures.Underthisperspective,ourretrievedstrikesforhigh-energy stationslocatedontopofthemainactivefaultsofthearea(NAFand 4.3. Lowercrustanduppermantle ESF) match theorientations ofthefaults withintheerrors. Fault par- allelstrikeforthesestationsisingoodaccordancewithpreviousshear The interpretation of deep anisotropy (between 30 and 60km, wavesplittingstudiesindifferentportionsoftheNAFZ(Lietal.,2014) Fig.8c)isdifficult.Thenumberofhigh-energy(energy>0.5)stations andwithRFsstudiesalongtheSanAndreasFault(Porteretal.,2011; issmall(six)andtheirspatialdistributionisscattered.Nevertheless,we Audet,2015).ThedirectionofSH inthestudyareawasestimated noticethatthesestationsarelocatedclosetothecentral-westernpartof max fromlocalseismicity(Karasözenetal.,2014).ItshowsanaverageNW- theNAFZ,alongtheESF,andthatfiveofthemdisplaystrikesoriented SE orientation, thus being very different from our retrieved strikes of NE-SW. This agrees with the 3-D full-waveform tomography study of thefastplaneassumingslowanisotropy.Therefore,ourresultssupport Çubuk-Sabuncuetal.(2017),whichrevealeddeepnegativeradialan- a structure-induced origin of the observed anisotropy rather than a isotropy (V > V ) in the close vicinity of known weakness zones, SV SH stress-driven mechanism. This agrees with the findings of Li et al. including the NAFZ, as well as with the orientation of the FDPs from (2014).Inanycase,theinter-stationdistanceinthepresentstudydoes SKSsplittinginthesamearea(Biryoletal.,2010).Tentatively,wecan not allow for an estimate of the width of the damage zone, which also compare the results from this layer with the results obtained by however affects only the stations located on top of active faults Vinniketal.(2016)fromtheanalysisofthek=2componentsofP-RF, (Fig.8a).ThisisinaccordancewithLietal.(2014),wherethewidthof forthesamearrayofstationsusedinthisstudy.Usinganarray-aver- thedamagezonearoundtheKaradere–DüzcebranchoftheNAFZwas aged RF dataset for the harmonic decomposition, these authors re- estimated to be around 3km, as well as with Audet (2015), who re- ported intensity of horizontal-axis anisotropy between 3 and 5% (the portedsimilarestimatesaroundtheSanAndreasFault. highestvalueintheirstudy)andfastdirectionbetween30°and60°,ina Althoughourapproachdoesnotallowtoretrievethefullorienta- depthrangebetween35and60km.Ourresultsforhigh-energystations tion of the symmetry axis in 3-D, our results imply a dominant con- atthesamedepthlevelareingoodagreement,atleastforthedirection tribution from anisotropy with dipping symmetry axis, revealed from of anisotropy. Vinnik et al. (2016) argued that this sub-Moho aniso- theanalysisofthek=1harmonics.Thisobservationsuggestsastrong tropyislikelyfrozenintheuppermantle.Ourresultssupportthisidea. similarity between the deformation regime around the NAFZ and the SanAndreasFault,wheretheanisotropicinversionofRFdatashoweda 5. Conclusions substantialdip(about45°)oftheslowsymmetryaxisofanisotropyin thecloseproximityofthemainfault(Audet,2015). In this paper we use teleseismic receiver functions to study the Both the k=0 and k=1 harmonics in Fig. 7 and the spatial dis- crustalstructureandmapcrustalanisotropyaroundthecentralportion tributionofenergyintheuppercrustinFig.8suggestanasymmetric oftheNorthAnatolianFaultZone.Weshowthatasimpleyeteffective distribution ofelastic propertiesacross theNAFZ.Thisconfirmswhat harmonic decomposition of RFs can be used to reveal first-order fea- discussedindetailbyŞengöretal.(2005)andLePichonetal.(2005). turesofthecrust,aswellastolocatethepresenceofanisotropy asa Inparticular,thepresenceofabimaterialinterfacedevelopedasare- functionofdepth. sultofthecumulativeproductionofrockdamageassociatedwiththe Ourfindingsindicatestrongasymmetryofcrustalpropertiesaround faulting process (Ben-Zion and Sammis, 2003 and references therein) the North Anatolian Fault. The southern block possesses a complex was reported in several studies (e.g. Bulut et al. 2012; Ozakin et al., crustalstructureandishighlyanisotropicwhencomparedtothesim- 2012;Najdahmadietal.,2016),alongdifferentsegmentsoftheNAFZ. plerandmoreweaklyanisotropiccrustalblockinthenorth. Ourresultscorroboratethishypothesisinthestudyarea. The Moho is generally a continuous feature at around 40km showingnomajoroffsetsacrossthearea.Ontheotherhand,theintra- 4.2.2. Middlecrust crustal structure is strongly affected by intense deformation in the Between15and30kmdepth(Fig.8b),theintensityofanisotropy southernblock. diminishesforsomeofthenear-faultstations,suggestingaprogressive Theanisotropicsignalisstrongestinthefirst15kmofthecrustand closureoffracturesduetoincreasingconfiningstress,asobservedalso ispredominantlylocatedontopofthemajoractivefaultzones(NAFZ onamoreeasternportionoftheNAFfromshearwavesplittingmea- and ESF). Here, the strike of the fast plane of anisotropy is fault-par- surements(Lietal.,2014).However,afewnear-faultstationslocated allel,implyingastructure-inducedmechanismasthedominantsource on the westernmost part of the study area (in correspondence of the for observed crustal anisotropy. Progressively increasing confining NAF)andinthecentralpartoftheESFshowaconsiderableamountof stress around the faults contributes to the general observed trend of energy.Thismaysuggestadifferentialclosureoffracturesaroundthe decreasinganisotropywithdepth. faults, implying a depth extent ofthe damage zone below 15km (see Atmiddlecrustaldepths(15–30km),spotsofhighanisotropyinthe depthprofilesAA′,BB′andDD′)forsomestations. southernblockmayhavemultipleinterpretations.Forstationscloseto 108
Description: