2015-Turkoglu-NAF-EA..

14 Pages·2015·4.64 MB·English

by  ErÅŸan TürkoÄŸlu

PhysicsoftheEarthandPlanetaryInteriors241(2015)1–14
ContentslistsavailableatScienceDirect
Physics of the Earth and Planetary Interiors
journal homepage: www.elsevier.com/locate/pepi
Crustal structure of the North Anatolian and East Anatolian Fault
Systems from magnetotelluric data
Er(cid:2)san Türkog˘lua, Martyn Unswortha,⇑, Fatih Bulutb, I_lyas Çag˘larc
aDepartmentofPhysics,UniversityofAlberta,Edmonton,ABT6G2G7,Canada
bIstanbulAydınUniversityAFAMDisasterResearchCentre,InonuCad.38,34295Istanbul,Turkey
cDepartmentofGeophysicalEngineering,FacultyofMines,IstanbulTechnicalUniversity,Maslak,Istanbul34469,Turkey
a r t i c l e i n f o a b s t r a c t
Articlehistory: Magnetotelluric (MT) studies can map subsurface resistivity structure and have located zones of low
Received28March2014 resistivity(highconductivity)withinmajorstrike-slipfaultzonesworldwidewhichhavebeeninterpret-
Receivedinrevisedform20December2014 edasregionsofelevatedfluidcontent.ThisstudydescribesMTdatafromtheeasternpartoftheNorth
Accepted9January2015
AnatolianandtheEastAnatolianFaultSystems(NAFSandEAFS)andpresentstheresultsofthefirstMT
Availableonline30January2015
studiesofthesefaults.TheinversionoftheMTdataproduced2-Dresistivitymodelswhichshowedthat
bothfaultsystemsareunderlainbyabroadlowresistivityzonethatextendedintothelowercrust.How-
Keywords:
ever,theresistivitybeneaththeEastAnatolianFaultSystemwasmuchlowerthanbeneaththeeastern
Magnetotellurics
partoftheNorthAnatolianFaultSystem.Theseconductorsbeginatadepthof10km–notatthesurface
Resistivity
asonthecentralSanAndreasFault(SAFS).Thisdifferenceisinterpretedasbeingduetothefactthatthe
Faultzoneconductor
EasternAnatolia EAFSandNAFSareyoungfaultsystemscharacterizedintheuppercrustbymultiplefaulttraces–as
NorthAnatolianFault opposedtotheSAFSthathasevolvedintoasinglethroughgoingfault.Differentstagesoftheseismic
EastAnatolianFault cyclemayalsoinfluencetheresistivitystructure,althoughthisisdifficulttoconstrainwithoutknowl-
edgeoftimevariationsinresistivitystructureateachlocation.
(cid:2)2015ElsevierB.V.Allrightsreserved.
1.Introduction suggestthatdeformationintheductilepartofthecrustmayoccur
by creep processes that are enhanced by the presence of water
Transform faults represent one of the three classes of plate (Tullis et al., 1996). It has also been proposed that in the brittle
boundaries and pose a significant seismic hazard, as evidenced uppercrust,thebehaviorofseismogenicfaultsmaybecontrolled
by recent earthquakes on the North Anatolian, San Andreas, and by spatial and temporal variations in fluid content (Byerlee,
other faults (Barka and Kadinsky-Cade, 1988; Fuenzalida et al., 1993; Sleep and Blanpied, 1992). Over-pressured fluids in the
1997;Barkaetal.,2002;ToppozadaandBranum,2002).Strike-slip fault-zonemaytriggerearthquakesthroughreducingtheeffective
faults exhibit a wide variability in their seismogenic behavior. normal stress and thereby lowering the shear stress needed for
Somefaultsarecharacterizedbysegmentsthatexhibitcontinuous failure (Sleep and Blanpied, 1992). A related observation is that
creep,withadjacentsegmentslockedandrupturingduringmajor strike-slipfaultsareoftenobservedtobemechanicallyweak.For
earthquakes.Thephysicalcauseofthesevariationsinbehavioris example, the San Andreas Fault System (SAFS) appears to move
notwellunderstood.Thestructureofstrike-slipfaultsintheduc- withashearstressofjust10–20MPa,thatcorrespondstoacoeffi-
tilelowercrustisalsoanongoingresearchquestion(Beckenetal., cient of friction in the range 0.1–0.2 (Zoback et al., 1987; Mount
2008).Geologicalstudiesofexhumedfaultsandshearzonessug- and Suppe, 1987). Similar values (0.05–0.2) have been reported
gest that the zone of deformation broadens with depth (Sibson, on the NAFS based on geodetic data (Provost et al., 2003). These
1977; Hanmer, 1988). Geophysical studies of the lower crust frictional values are significantly lower than those obtained in
beneathshearzonesarelimitedinnumber,butindicatethatfluid laboratory studies of rock friction (Byerlee, 1978; Sibson, 1974)
composition is an important parameter that may influence the andareconsistentwiththelowvaluesoffrictionalheatgeneration
style of deformation (Bedrosian et al., 2004). Deformation in the reportedfromtheSanAndreasFaultbyWilliamsetal.(2004).
ductile lower crust is influenced by fluids and laboratory studies Akeyquestionthatarisesinthisdebateishowtheamountof
fluidandtypeofdeformationarerelatedinstrike-slipfaults.Itis
⇑ possible that either (a) an increase in the amount of fluid causes
Correspondingauthor.Tel.:[phone].
a fault to weaken and rupture, or (b) that the amount of fluid is
E-mailaddress:[email protected](M.Unsworth).
http://dx.doi.org/10.1016/j.pepi.2015.01.003
0031-9201/(cid:2)2015ElsevierB.V.Allrightsreserved.
2 E.Türkog˘luetal./PhysicsoftheEarthandPlanetaryInteriors241(2015)1–14
theresultofelevatedporositycausedbydeformation.Thisques- (2002).Thesestudieshavedetectedconductorsintwodistincttec-
tion can be addressed by mapping the subsurface distribution of tonic environments. To make a clear distinction between these,
fluids within major strike-slip faults using surface based geo- separateabbreviationsareusedinthispaper:
physical studies. Electromagnetic (EM) methods are effective in
thisregardbecausetheyimagesubsurfaceresistivity–arockprop- (1) Shallow conductors associated with the damage zone of a
erty that is largely controlled in the upper crust by the amount, fault, and caused by groundwater present in regions of
salinity and geometry of the pore fluids. In this context, magne- elevatedporosity,andperhapssupplementedwithclaymin-
totellurics(MT)isthemostusefulbecauseitcanimagetheentire eralization.Thesearetermeddamagezoneconductors(DZC).
crustusingnaturalEMsignals,withouttheneedforatransmitter. (2) Deeper conductors in the mid-crust, generally extending
Depth sounding in MT is achieved through the skin-depth phe- acrossthebrittle–ductiletransitionarecalledcrustalfault-
nomenon that gives a penetration depth that is inversely related zoneconductors(CFZC).
tothesignalfrequency(Vozoff,1991). Theobjectofthispaperistopresentnewimagesoftheelectri-
AnumberofMTsurveyshavebeenusedtostudythefluiddis- cal resistivity structure of the major strike-slip faults in Eastern
tribution within active strike-slip fault zones. Studies of the San Anatolia,withthegoalofbeingabletorelatetheelectricalresis-
AndreasFault(SAF)inCaliforniarevealedthatsomefaultsegments tivity structure to the style of deformation. The NAFS and EAFS
are characterized by a zone of low resistivity (elevated conduc- arerelativelyyoungstrike-slipfaultsthathaverelativelysmalloff-
tivity).AtParkfield,theSAFisintransitionfromcreepingtolocked setsandhavenotyetdevelopedintomaturestrike-slipfaultzones.
andtheconductorextendsfromthesurfacetoadepthof2–3km Studyingfaultsatanearlystageofdevelopmentallowstheoppor-
(Unsworth et al., 1997). The conductor extends to mid-crustal tunitytoaddressthequestionofhowthestructureoffaultzones
depths at Hollister where the fault creeps at 10–15mm/yr evolvesovertime.
(Burford and Harsh, 1980; Evans et al., 1981; Bedrosian et al.,
2002). A small fault-zone conductor was observed on the locked
2.Tectonicsettingandpreviousstudies
Carrizosegment(Unsworthetal.,1999;Mackieetal.,1997).Deep-
erfaultzonestructureoftheSAFwasinvestigatedwiththelonger
ThetectonicsofEasternAnatoliaisdominatedbytheongoing
profile and 3-D array of Becken et al. (2008) and showed a con-
collision of the Eurasian and Arabian Plates (Fig. 1). Convergence
tinuouszoneoflowresistivityextendingthroughtheentirecrust.
intheEocenewasinitiallyaccommodatedbyshorteningandthick-
InnorthwestTurkey,anumberofMTstudieshaveinvestigated
eningoftheArabiancontinentalmargin(Hempton,1985;S(cid:2)engör
the resistivity structure of the North Anatolian Fault close to the
etal.,1985).TheNAFSandEAFSsubsequentlydevelopedtoaccom-
rupturesoftheI_zmitandDüzceearthquakesandrevealeddeeper modatethewestwardmotionoftheAnatolianblocktowardsthe
zones of low resistivity in the mid-crust (Tank et al., 2005; Kaya Aegeanarc(BurkeandSengör,1986).Forcesgeneratedbytrench
etal.,2009). MorerecentstudieshaveusedseafloorMTtostudy rollbackandthesouthwardmigrationattheAegeanarccontribute
the fault strand of the NAF under the Sea of Marmara (Kaya tothewestwardmotionoftheAnatolianblockalongthelarge-s-
et al., 2013). In China, a number of MT studies have investigated calestrikeslipfaultsystems(Reilingeretal.,2006).Today,theAra-
theKunlunandAltynTaghFaultswhicharemajorstrike-slipfaults bian Plate moves northward at a velocity of 15mm/yr and only
associated with the northern margin of India–Asia collision 10%ofthisconvergenceisaccommodatedbylithosphericshorten-
(Unsworth et al., 2004; Le Pape et al., 2012; Zhang et al., 2015). ing with the remainder of the convergence accommodated by
Baietal.(2010)describedmajorconductorsinthemid-crustthat strike-slip motion on the NAFS and EAFS (Reilinger et al., 2006).
werecoincidentwithmajorshearzonesinsouthwestChinaover Recentstudieshaveshownthatthepresentdaydrivingforcefor
horizontal distances of hundreds of kilometers. The Alpine Fault the westward movement of the Anatolian plate is primarily
in New Zealand was studied with MT by Wannamaker et al. derivedfromtheAegeansubduction,withonlyasmallcomponent
Fig.1. SimplifiedtectonicmapofTurkeyandsurroundings(modifiedfromS(cid:2)engöretal.,1985;Barka,1992).BlackrectanglesindicatetheareasstudiedwithMTdatainthis
paper.Thegrayrectangle,eastoftheMarmaraSea,showsthesurveyareaofTanketal.(2005).StarsymbolsshowthelocationofpreviouslyreportedcreepontheNAFand
EAF.FilledcirclesshowMTsites(Türkog˘luetal.,2008)NAFS:NorthAnatolianFaultSystem.NEAFS:NorthEastAnatolianFaultSystem.EAFS:EastAnatolianFaultSystem.
DSFS:DeadSeaFaultSystem.KTJ:KarlıovaTripleJunction.MTJ:Mara(cid:2)sTripleJunction.D:Düzce.E:Erzincan.
E.Türkog˘luetal./PhysicsoftheEarthandPlanetaryInteriors241(2015)1–14 3
coming from the collision in Eastern Anatolia (Reilinger et al., 2.2.EastAnatolianFaultSystem(EAFS)
2006;Bulutetal.,2012).
ThetotaloffsetontheEAFSisaround22km(ArpatandS(cid:2)arog˘lu,
1972; S(cid:2)arog˘lu et al., 1992), which is less than the 75–85km
2.1.NorthAnatolianFaultSystem(NAFS) observedontheNAFS(LePichonetal.,2001).Thissignificantdif-
ferencecouldbeaconsequenceoftherollbackofthesubducting
RecentseismicactivityontheNAFShasbeencharacterizedbya African lithosphere beneath the Aegean arc causing a counter-
sequence of westward propagating earthquakes that began with clockwise rotation of the Anatolian block (Reilinger et al., 2006).
the M =8.2 Erzincan earthquake in 1939 (Parsons et al., 2000). IncontrasttothehighlevelsofseismicactivityontheNAFS,the
s
The most recent earthquakes in this sequence occurred on the EAFS has been characterized by low seismicity levels during the
westernpartoftheNAFSin1999whentheAugust171999I_zmit last century (Jackson and McKenzie, 1988; Ambraseys and
(M =7.4) and November 12 1999 Düzce (M =7.2) earthquakes Jackson,1998).However,historicalrecordsclearlyshowthatmajor
w w
ruptured 110–140km and 40km of the NAFS, respectively earthquakes have occurred on this fault in the last 900years
(Armijoetal., 1999;Barka, 1999). Stress modelingby Steinetal. (Ambraseys, 1971; Ambraseys and Jackson, 1998; Nalbant et al.,
(1997) suggested that these earthquakes have increased the 2002; Taymaz et al., 2007). There is also evidence that seismic
chance of a major earthquake on the strand of the NAFS in the activity jumps every 800–900years from the NAFS to the EAFS
SeaofMarmara(Kingetal.,2001;Parsonsetal.,2000;LePichon andviceversa(Ambraseys,1971).Continuousbroad-bandseismo-
et al., 2003). Two geodetic studies have identified segments of graphrecordingsbetweenOctober1999andAugust2001showed
theNAFSthatappeartocreepaseismically: that the EAFS was more active than previously thought and the
mostactiveseismogeniczoneisatadepth10–30km,incontrast
(a) ThewesternNAFSatI_smetpa(cid:2)sahasbeenmonitoredforcreep totheNAFSwhereitoccursinthedepthrange0–10km(Turkelli
etal.,2003).ThemorerecentstudybyBulutetal.(2012)revealed
since1957aftermajorearthquakesin1944and1951(Çakır
that the EAFS comprises a 20km wide band of seismicity with
etal.,2005;KutogluandAkcin,2006).Thecreepingmotion
complexsegmentationwithbothRiedelandanti-Riedelstructures.
wasreportedtobelinearlydecreasingfrom(cid:2)2cm/yrinthe
Preciseearthquakelocationsdonotoutlineasinglethrough-going
period 1957–1969 to (cid:2)0.8cm/yr in the period 1992–2002
fault.Thefocaldepthspredominantlyrangebetween5and20km.
(KutogluandAkcin,2006).However,itisnotclearwhether
thelargeearthquakestriggeredthecreepmotionat I_smet- The kinematics are dominated by SW–NE oriented left-lateral
strike slip mechanisms reflecting the overall characteristic of the
pa(cid:2)sa orcausedachangeintheexistingcreeprateasthere
EAFS.ThefaultzonealsocontainsEast–Westorientedthrustfaults
werenomeasurementsbeforetheseearthquakes.
andNorth–Southorientednormalfaults.Thesefeaturesmayrep-
(b) Karabacak et al. (2011) used a LIDAR system to study the
resentanearlystageofdevelopmentofashearzone,ratherthan
NAFSintheI_smetpa(cid:2)saandDestekregions(Fig.1).Between
acollisional/orextensionalregime(Bulutetal.,2012).
2007 and 2009, creep rates at I_smetpa(cid:2)sa were found to be Combinedinterpretationofmagnetic,seismologicalandgeolog-
0.9–1.0cm/yrwhichisclosetothe0.8cm/yrreportedprevi- icdatabyDolmazetal.(2008)showedthatalowseismicactivity
ously.TheaseismicsectionofNAFSaroundDestekwasalso zoneontheEAFScoincideswithashallowCuriedepth((cid:2)14km)
reportedtocreepatarateof0.6–0.7cm/yr(Karabacaketal., and strong magnetic anomalies (LSZ in Fig. 2). They interpreted
2011).TherearenoMTstudiesatI_smetpa(cid:2)saorDestek,soit thisareatohaverelativelythincrust(36–38km)andhighercrus-
cannot be determined if there is a correlation of electric taltemperaturescomparedtoadjacentregions.Therefore,thispart
resistivitystructureandseismicbehavior,asmaybethecase oftheEAFSmaybebehavinginaductilewayandallowingcreepin
fortheSanAndreasFaultinCentralCalifornia. theuppercrust.
Fig.2. MagnetotelluricstationsusedforstudiesoftheEastAnatolianFaultSystemaroundHazarGölü.SmallgraydotsrepresentearthquakeepicentersfromBulutetal.
(2012).Blackarrowsshowrealinductionvectorsaveragedovertheperiodrange2–40sandplottedintheParkinsonconvention.Rosediagramshowgeoelectricstrike
directionsderivedfromthephasesensitiveskew.BSZ:BitlisSutureZone;LSZ:LowSeismicityZone(Dolmazetal.,2008).
4 E.Türkog˘luetal./PhysicsoftheEarthandPlanetaryInteriors241(2015)1–14
2.3.PreviousmagnetotelluricstudiesoftheNAFandEAF wires,culturalnoiseorinstrumentfailure,andwereexcludedfrom
theprocessing.Theremainingdatasegmentswerepre-whitened,
PreviousstudieshaveshownthatMThasthepotentialtomap Fourier transformed, and estimates of the magnetotelluric impe-
fluidswithinmajorstrike-slipfaults.AnumberofMTstudiesofthe dance were computed using the statistically robust algorithm of
North and East Anatolian Fault Systems have been published in Egbert (1997). Smooth MT transfer functions were obtained in
recent years. Tank et al. (2005) described the first detailed MT the period ranges 0.01–2000s and 10–10,000s for broad-band
study of the western NAFS (Fig. 1) close to the rupture of the and long-period MT data, respectively. Fig. 4 shows sample MT
1999 I_zmit earthquake (Barka, 1999). This resistivity structure curves from both the NAF and EAF profiles. A complete set of
determinedbythisMTstudyshowedazoneoflowresistivitythat observedMTdatasoundingsandthecomputed2Dinversionmod-
was located betweenthe main fault strands, and extending from elresponsesareshownintheAppendix.
10km depth through the entire crust. Hypocenters of the after-
shocksoftheI_zmitearthquakewerelocatedonthesouthernedge
3.1.DimensionalityanddirectionalityoftheMTdata
of this low resistivity zone, that was interpreted by Tank et al.
(2005)asbeingcausedbyazonewithelevatedfluidcontent.This
ThedimensionalityofMTdatacanbeinvestigatedinanumber
fluid-richlayerjustbelowtheseismogeniczonewasinterpretedas
of ways. The simple definition of skew proposed by Swift (1967)
triggering earthquakes and being responsible for post-seismic
canbe an unreliablewayof distinguishingbetween2-Dand 3-D
creep. Kaya et al. (2009) reported two MT profiles crossing the
dataiftheMTdataareaffectedbygalvanicdistortion.Thephase
NAF near Düzce that were collected to the west and east of the
sensitive skew proposed by Bahr (1991), is relatively insensitive
1999Düzceearthquakeepicenter.TheseMTprofilesshowedhigh-
togalvanicdistortioneffectssuchasstaticshifts,andgivesabetter
erresistivitiesontheeasternMTprofile,thatwereinterpretedas
estimateofthedimensionalitythantheSwiftskew.Askewvalue
an explanation for the observed high rupture velocity and after-
greaterthan0.3canbeanindicationof3Deffectsinthedata,while
shockactivityontheeasternsideoftheDüzcefault,comparedto
a smaller skew value (<0.3) is a necessary, but not sufficient,
thewest.Türkog˘luetal.(2008)reportedaregionalscaleMTstudy
requirement for the impedance data to be considered 2D (Bahr,
oftheArabia–EurasiacollisioninEasternAnatolia(Fig.1).Awide-
1991).Thephasesensitiveskewthresholdcitedaboveisnotstatis-
spreadlowresistivitylayer(10–30Xm)beneaththeAnatolianPla-
tically defined in terms of noise in the MT impedances, so Marti
teauwasdiscovered,andinterpretedasbeingduetothepresence
et al. (2005) introduced the Bahr-Q method. When the MT data
ofafluid-richlowercrustthatwasmechanicallyweak.Thislayer
are of good quality, Bahr skew and Bahr-Q methods give similar
may well control deformation by (a) permitting crustal flow or
results.TheMTdatapresentedinthispaperaregenerallyofgood
(b) locally decoupling of the upper and lower parts of the litho-
quality,sotheBahrskewwasusedtoinvestigatethedimension-
sphere (Türkog˘lu et al., 2008). This study also showed that the
ality,asshowninFig.5.Overall,theNAFandEAFMTdataarechar-
resistivity in the lithospheric upper mantle beneath the Eastern
acterized by relatively low Bahr skew values (<0.25) over the
AnatolianPlateauwas10–30Xm,whichisanorderofmagnitude
wholeperiodrange.IsolatedBahrskewvalues(>0.5)areobserved
less than expected and could be an indication of a shallow
atshortperiods(T<10s)forthelong-periodMTstationsandare
asthenosphere.
duetonoisyorbiaseddata.Thesedatawereexcludedfromsubse-
Av(cid:2)sar et al. (2012) reported two MT profiles crossing the NAF
quentdataanalysis.
withintheErzincanBasin(Fig.1).ThewesternMTprofileimaged
The geoelectric strike direction is needed for subsequent MT
aDZCspatiallycorrelatedwiththeNAFintheErzincanbasin.The
data analysis and can be determined using a range of methods
absence of a deeper CFZC under the eastern MT profile was
attributedtopermeabilityandfluidcontentoftheNAFSintheErz-
incanBasinatshallowdepths.
3.MagnetotelluricdatacollectedontheNAFSandEAFS
ThesurveybyTürkog˘luetal.(2008)definedtheregionalscale
resistivitystructureoftheArabia–EurasiacollisionzoneinEastern
Anatolia for the first time. This survey also used more detailed
transectstoinvestigatethegeoelectricstructureoftheNorthand
East Anatolian FaultSystems. Ineach of thetwo studyareas, MT
data were collected on profiles that were approximately perpen-
dicularto thesurface traceof thefaults (denotedas theEAF and
NAF profiles). The EAF-profile crosses the EAFS and Bitlis-Zagros
SutureZone(BSZ)closetoHazarGölü(Fig.2).Thisprofileconsists
of 10 long-period and 11 broad-band MT stations (1 broad-band
MTstationwasexcludedfromtheinversions).TheNAFprofileis
located just east of Refahiye (Fig. 3). This profile consists of 9
long-period and 4 broad-band MT stations. The broad-band MT
stationstypicallyrecordeddatafor1–3dayswhilethelong-period
MTstationsrecordeddatafor20–30days.Typically,3broadband
MT stations and 15 long-period stations were running syn-
chronouslyandwereusedformutualremotereferencetoremove
uncorrelatednoise.Aquietstationusedforthereferencethatwas Fig.3. MagnetotelluricstationsusedforthestudyoftheNorthAnatolianFault
typicallymorethan10kmdistantfromthemeasurementsite.The System(NAFS)aroundRefahiye.Smallgraydotsrepresentearthquakeepicenters
from1960to2004KOERI,Turkellietal.(2003)andKaypakandEyidog˘an(2005).
MT time series data were visually inspected and compared to
Blackarrowsshowrealinductionvectorsaveragedovertheperiodrange2–40s
neighboring stations in order to identify noisy data segments.
andplottedintheParkinsonconvention.Rosediagramsshowthegeoelectricstrike
Theseweregenerallycausedbyelectrodeproblems,brokentelluric directionderivedfromthephasesensitivestrike.
E.Türkog˘luetal./PhysicsoftheEarthandPlanetaryInteriors241(2015)1–14 5
Fig.4. ApparentresistivityandphasecurvesfortypicalMTstationsfromtheNAF(lowerrow)andEAF-profiles(upperrow).Dataareshownafterrotationtothestrike
direction.SeeFigs.2and3forstationlocations.
(Simpson and Bahr, 2005). The whole period range was grouped directions of N105(cid:3)E for the NAF-profile and N75(cid:3)E for the EAF-
intofourperiodbandstoinvestigatevariationsofthegeoelectric profile were used. This choice was made because the target of
strikedirectionwithincreasingperiod.Notethatincreasingsignal the survey was the upper crustal structure of the fault, and not
periodcorrespondstoanincreasingdepthofpenetrationintothe thestructureatdepthsampledbythelongerperiodMTdata.
Earth. Phase sensitive strike directions determined from Bahr’s TheinductionvectorsarealsoplottedonthemapsinFigs.2and
(Bahr,1991)methodareshowninFig.6.Approximatedepthesti- 3. The real part of the induction vectors were averaged over the
matesforeachofthefourperiodbandsweremadeusingtheMT period range of 2–40s and plotted in the Parkinson convention
skindepthforarangeofexpectedresistivityvaluesbasedonthe with vectors pointing towards conductors. The induction vectors
apparentresistivitycurvesandarelistedinFig.6. showevidenceforareversalaboveboththeNAFSandEAFSwhich
indicatesthefaultzonesaremoreconductivethantheirsurround-
ings.However,thesurfacetraceoftheNAFSandthelocationofthe
3.1.1.NAF-profile
induction vector reversal are not exactly coincident (Fig. 3). On
The short period data (T<2s) show no well-defined strike
both profiles, the reversals are complicated by other factors,
directionas thenear surfaceresistivity structureisessentially1-
includingasignificantfaultparallelcomponentontheEAFSprofile.
D and/or has a variable strike direction. The average geoelectric
strike direction of the NAF-profile is relatively well defined at
N105(cid:3)Efromthefirsttwoperiodbands(2–40s).Thisgeoelectric
3.2.GeneralaspectsoftheMTdata
strike direction is similar to the strike direction of the NAF as
defined by thefocal mechanismsofthe large earthquakes (Barka
The observed MT data were rotated to the strike coordinate
and Eyidogan, 1993; Grosser et al., 1998; Aktar et al., 2004) and
framesdeterminedaboveandsamplesoundingcurvesareshown
thestrikeofthefaulttrace(Ketin,1948;SengörandKidd,1979).
inFig.4.TypicalMTstationsontheArabianPlateshowrelatively
MT signals with periods greater than 40s show a well-defined
low resistivity values ((cid:2)10Xm) at short periods (T<10s) and
strike direction that is east–west independent of period, and
increasingapparentresistivityastheperiodincreases.Thisbehav-
roughlyparalleltotheBitlisSutureZone.
ior could be attributed to a layer of low resistivity sedimentary
rocks in the Arabian Foreland overlyingthe crystalline(resistive)
3.1.2.EAF-profile basementrocksofthecrustanduppermantle.TheMTcurvesfrom
Asimilarchangeinthegeoelectricstrikedirectionwithperiod theAnatolianBlockshowmorespatialvariationthanthosecollect-
occurs on the EAF-profile (Fig. 6) where a N75(cid:3)E strike direction ed on the Arabian Plate. Relatively high apparent resistivities
changestoeast–westatlongerperiods(T>40s).TheN75(cid:3)Egeo- ((cid:2)100Xm) are observed at short periods (T<10s). Decreasing
electricstrikedirectionisclosetotheN65(cid:3)–70(cid:3)Estrikeofthesur- resistivities are observed at intermediate periods
facetraceoftheEAF.Thusfortwo-dimensionalinversions,strike (10s<T<1000s) and a slight increase of apparent resistivity is
6 E.Türkog˘luetal./PhysicsoftheEarthandPlanetaryInteriors241(2015)1–14
Fig.5. Bahr’sphasesensitiveskew(Bahr,1991)forEAF-profile(top)andNAF-profile(bottom).Whitespaceindicatesthatthedataarenotavailableatthoseparticular
periods.
Fig.6. Bahr’sphasesensitivestrikedirectionforfourperiodbands(Bahr,1991).TheMTskindepthequationwasusedtoestimatetheapproximatedepthtowhichtheMT
signalspenetrateineachperiodband.Typicalapparentresistivitiesofthisdataset(10Xmand100Xm)wereusedtoestimatetherangeofskindepths.Rosediagrams
representhistogramplotsofthestrikedirectionsforallthestations.Blackandwhitehistogramsrepresentthe90(cid:3)ambiguityinthestrikedirectioninherenttoMTdata.
observedatlongperiods.Qualitativeinterpretationoftheapparent the fault zones. The 2-D NLCG algorithm of Rodi and Mackie
resistivity curves from Anatolian Block typically indicates a low (2001) was used for inversion of both apparent resistivity and
resistivityzoneatcrustalorsub-crustaldepths. phase for the transverse electric (TE) and transverse magnetic
(TM) modes. The vertical magnetic field transfer functions were
3.3.Two-dimensionalMTinversion alsoinverted.Staticshiftcoefficientswereestimatedbytheinver-
sionalgorithm.Errorfloorsof20%,5%and0.02%wereassignedfor
TheMTstationswereprojectedontoprofilesperpendicularto apparentresistivity,phaseandtipperdata,respectively.Notethat
the geoelectric strike directions determined above. Where avail- 5% in phase is equivalent to 1.43(cid:3). The starting model was a
able, the short period MT data (0.01–10s) were also included in homogenoushalf-spacewitharesistivityof100Xmwith171cells
theinversionstoobtainmoredetailedresistivitymodelscloseto inthehorizontaldirection((cid:2)1kmwidth)and111inthevertical
E.Türkog˘luetal./PhysicsoftheEarthandPlanetaryInteriors241(2015)1–14 7
direction (first cell thickness (cid:2)13m). Initial r.m.s. misfits for the
NAFandEAFprofileswere7.87and6.36.Identicalinversionpara-
meterswereusedforboththeNAFandEAF-profiles.
The final inversion models, static shift coefficients, and r.m.s.
misfitsobtainedafter200iterationsareshowninFig.7.Itiswell
known that the MT inverse problem is non-unique (Menke,
1989; Whittall and Oldenburg, 1992) and the solution of the
inverseproblemrequiresthatadditionalconstraintstobeapplied
totheresistivitymodel.Thisisgenerallyachievedwithregulariza-
tion that seeks a resistivity model that minimizes the misfit
between the observed and predicted MT data and also keeps the
resistivity model as spatially smooth as possible (Tikhonov and
Arsenin,1977).IntheNLCGinversionalgorithm,theregularization
parameter (tau) controls the overall smoothing of the model. As
thevalueoftauincreases,theresistivitymodelbecomessmoother
and the measured MT data are not fit as well. Fig. 8 shows the
trade-off between the MT data fit and resistivity model smooth-
ness.ThecornerofthesocalledL-curvedefinestheoptimumvalue
oftaufor agivenMTdata setanda compromisebetweenfitting
theMTdataandgeneratingasmoothresistivitymodel.Thecorner Fig.8. Resistivitymodelnormalizedroughnessversusr.m.s.datamisfitforarange
oftauvaluesfortheNAF(dashedline)andEAF(solidline)profiles.
of the L-curve indicates the preferred tau values for this study
ranges from tau=2–10. Relatively large station spacing caused
unwantedroughstructuresbetweenthestationswhenlowertau isobtained,withallperiodsfitequallywell.Ifaparticularrangeof
valueswereused.Therefore,arelativelyhightauvalueof10was periodsarenotwellfitthenthefinalinversionmodelmaybemis-
used. leadingandsmallererrorfloorvaluesmaybeneededtofocusthe
The final inversion models for the NAF and EAF-profiles have inversiononaparticularrangeofperiodscorrespondingtoresis-
r.m.s. misfit values of 1.40 and 1.48, respectively. These values tivity structures of interest. Fig. 9 shows that a satisfactory fit to
are statistically acceptable and as expected since the dimension- themeasureddatawasobtainedfortheentireperiodrangeused
ality analysis suggests the impedance tensor is relatively 2-D at for inversion (0.01–10,000s). Sharp lateral resistivity changes in
mostoftheMTstations.Staticshiftcoefficientsestimatedbythe the pseudo sections occur because computed static shifts have
inversionwerelessthanhalfadecadeinmagnitudeatmostsites, been applied to the apparent resistivities of the model response.
butthereweresomesiteswithhigherestimatedstaticshiftvalues Infact,theresponsedataarelaterallysmootherthantheyappear
(Fig.7).Staticshiftcoefficientestimationusedadampingfactorof in Fig. 9. The data fit is also shown as sounding curves in
10,000intheinitialstepsoftheinversionprocesstoobtainanopti- Figs.SP1andSP2forallstationsonbothprofiles.
mumfitwiththemodelfeatures,andonlyallowedstaticshiftcoef- TheinversionmodelfortheEAF-profileshowninFig.7revealsa
ficients in the later stages of the inversion. Fig. 9 shows the number of significant features. The EAF represents the main
observed and calculated inversion responses as pseudosections boundary between the resistive block to the south and the low
forallthreedatacomponents.Pseudosectionsshowthatcomputed resistivity lower crust (C1) to the north at a distance of around
responsesfromthefinalinversionmodelwereabletofitallaspects 42km. This lateral resistivity discontinuity may represent the
oftheobservedMTdata. boundarybetweentheArabianPlateandtheAnatolianBlockand
ThefitoftheinversionresponsetotheobservedMTdatashould appearstodipsoutheast.Anothersignificantfeatureofthismodel
beexaminedtoseeifauniformfitisobtained.Ideallya‘‘white’’fit istheverticalconductor(C2)locatedbeneaththeEAFatadistance
Fig.7. Upperpanelsshowther.m.smisfitfor2-Dinversionswiththefollowingparameters.NAF-profilemodelused105(cid:3)NEstrikeangleandEAF-profilemodelused75(cid:3)NE;
20%,5%,0.02%errorfloorsforapparentresistivity,phaseandtipper,respectively;tau=10.Dashedlineshowsthepreferredr.m.s.misfitvalueofunity.Middlepanelsshow
shiftcoefficientsestimatedbytheinversionalgorithm.Thedashedlineshowsashiftofonewhichrepresentsnostaticshift.LowerpanelsshowinversionmodelsfortheEAF
(left)andNAF(right)profiles.
8 E.Türkog˘luetal./PhysicsoftheEarthandPlanetaryInteriors241(2015)1–14
Fig.9. Pseudosectionsoftheobserveddata(apparentresistivity,phaseandtipperdata)andthecorrespondingresponsesfortheinversionmodelsshowninFig.8fortheEAF
andNAF-profiles,respectively.
of45kmwiththetopatadepthof10km,hereafterreferredtoas 3.4.SensitivityoftheMTdatatothegeometryoftheCFZCbeneaththe
thecrustalfaultzoneconductor(CFZC). EAF
TheinversionmodelfortheNAF-profileshowsaresistiveblock
locatedsouthofthesurfacetraceoftheNAFandanarrowpartof TheinversionofMTdataisanon-uniqueprocessandmustbe
this block reaches the surface. The crustal structure north of the understoodtodeterminethedegreetowhichfeaturesinthemodel
NAFSishighlyresistive(>1000Xm).Incontrasttotheresistivity arerequiredbytheMTdata.Itcanbeshownthattheconductance
structureoftheEAFS,theNAFSisrepresentedbyawiderzoneof ofalayercanberobustlydefinedbyMTdata(conductanceisthe
low resistivity but the resistivity beneath the fault zone is one product of conductivity and thickness). However, when the layer
orderofmagnitudehigherthanbeneaththeEAFS. isthincomparedtoitsdepth,thethicknessandconductivitycan-
E.Türkog˘luetal./PhysicsoftheEarthandPlanetaryInteriors241(2015)1–14 9
notbeindividuallyresolved.Thisnon-uniquenesscanalsobepar- were undertaken to determine if static shifts could have caused
tiallyovercomeifaconductivelayerisatthesurface.Whenacon- theinversiontoplaceaCFZCbeneaththeEAFS.Thiswasmotivated
ductorhasafinitehorizontalextent,suchastheCFZCbeneaththe by the observation that three MT stations just north of the EAFS
EAF, this non-uniqueness can be partially overcome, and more had the largest estimated static shifts on the EAF-profile (Fig. 7)
informationaboutthedepthvariationofconductivityderivedfrom andcouldhaveinfluencedthefinalresistivitymodel,andalsothat
the data. The spatial variation in the vertical magnetic fields are thenumberofthebroadbandMTstationsarelimitedaroundthe
especiallyusefulinthisrespect,asdemonstratedforanearvertical fault zone. The second panel from top in Fig. 11 shows that the
conductorimagedbyBertrandetal.(2009). presence of a CFZC does not depend on the data at stations
Toinvestigatethemodelresolution,agenericfaultzonemodel dan160, dan159 and dan158 which have high static shift coeffi-
wasdeveloped,asshowninFig.10andsyntheticMTdatawerecal- cients. The third panel from top in Fig. 11 shows that using only
culatedand5%Gaussiannoiseadded.Inversionsoftheindividual the long period data (T>10s) is sufficient to recover the CFZC
MT data components (TE, TM and Hz) do not reveal all features and the other major model features discussed in this paper. The
of the original resistivity model. However the joint inversion of final inversion model without statics found an even more pro-
TE,TMandHzrevealsallmodelfeatures,includingtheCFZC.Note nouncedCFZCandthisisexpectedastheinversioncannotaccount
thattheCFZCisbroaderintheinversionmodelbecauseofthespa- for model features with static shift coefficients (bottom panel in
tial smoothing that is used to regularize the MT inversion and Fig.11).
address the non-uniqueness inherent in inversion. Similar model
features are observed in both the synthetic inversion models
(Fig. 10), and those derived from the field data collected on the 4.Interpretationofresistivitymodels
EAF profile. This gives support for the presence of a strong CFZC
beneaththeEAF.Theimportanceofthetipper(Hz)dataisnosur- 4.1.Lowercrustalconductor
prisebecauseitiswellknownthatverticalmagneticfielddataare
sensitivetolateralresistivityvariationsandwidelyusedinmineral Alowercrustallowresistivitylayer(10–30Xm)waspreviously
explorationforlocatingdiscreteconductors.Asecondsetoftests reported at a depth of 20km throughout Eastern Anatolia by
Fig.10. LeftcolumnshowsindividualandjointinversionsoftheMTdataalongtheEAF-profile.Rightcolumnshowsasyntheticexperimenttoexplaintheabsenceofthe
verticalconductorforindividualinversions.Thethicknessoftheverticalconductorwas1kminthesyntheticmodel.TE+TMinversionsstarttorecoverthevertical
conductorwhenthethicknessoftheconductoris2kmorthicker(notshown).
10 E.Türkog˘luetal./PhysicsoftheEarthandPlanetaryInteriors241(2015)1–14
andBlanpied,1992)andthesefluidslowerthebulkresistivityof
arockandcanbeimagedwithmagnetotellurics(Unsworthetal.,
1997;UnsworthandBedrosian,2004).
4.2.1.EastAnatolianFault
TheverticallyintegratedtotalconductanceoftheEAFSatHazar
Gölüinthedepthrange0–50kmwasaround4000S.Theregular-
izationprocessresultsintheconductancevaluebeingatthelower
endoftherangeconsistentwiththeMTdata.ThetopoftheCFZC
(feature C2) is at 10km depth and well constrained by MT data.
Thisisclearlyabovethedepthofthebrittle–ductiletransitionthat
wasdefinedtobearound20kmbyBektasetal.(2007).Aconduc-
tanceof4000Scanbeexplainedwithmanycombinationsofcon-
ductivityandthickness.Togiveanideaoftherangeofparameters
consistentwiththeMTdata,considerthelimitwherethisfeature
is thick and extends to the Moho at a depth of 40km (Zor et al.,
2003). In this case the resistivity of C2 should be 7.6Xm. In the
otherlimit,C2isthinandonlyextendstothebrittle–ductiletran-
sitionzoneatadepthof17–18km(Bektasetal.,2007)andwould
havearesistivityof4.5Xm.WhatfluidcontentforC2isimpliedby
aresistivityintherange7.6–4.5Xm?Thiscanbeestimatedusing
the empirical Archie’s Law (Archie, 1942) that relates the bulk
resistivity of the fluid-saturated rock (q) to the pore fluid resis-
tivity (q) and the porosity (/) as q¼q Phi(cid:3)m where m is the
f f
cementationfactorthatdescribesthespatialdistributionoftheflu-
idwithintherock,andcontainsinformationaboutpermeability.A
value of m=1 implies good interconnection with crack shaped
pores while m=2 implies poor interconnection with spherical
shapedpores.UnsworthandRondenay(2013)showedthatunder
ambient pressure and temperature, crustal saline fluids typically
Fig. 11. Inversion experiment to investigate the effect of static shifts on the
havearesistivityofintherange0.01–0.3Xmwheretherangeis
inversionmodelfortheEAF-profile.Fromtoptobottom;(1)preferredinversion
modelincludingallsitesandallavailableperiods.Staticsshiftcoefficientswere primarily controlled by the salinity. With m=1, a layer with
estimatedautomaticallybytheinversion,(2)threeMTsiteswithhighestimated 7.6ohm-mlayerrequiresaporosityintherange0.13–4.0%.Ifthis
staticshiftsexcluded,(3)onlylong-perioddata(T>10s)usedintheinversion,all layerisforcedtobethin,witharesistivityof4.5Xm,thenapor-
sites included, (4) all sites included but the statics are not estimated by the osityintherange0.22–6.7%isneeded.Ifm=2thentheseranges
inversionbutmodeled.
become 4.7–26% and 3.6–20% which are unrealistically high. The
m=1porosityvaluesarerealisticandreflectthemostlikelyfault
Türkog˘lu et al. (2008) who showed that it was bounded on the zone pore geometry. Reducing the range of porosity values
North and South by the NAF and EAF, respectively. The spatial requires additional knowledge of the fluid resistivity, which
extentofthelayeralsocoincideswiththeEasternAnatolianAccre- dependsprimarilyonthesalinity.
tionary Complex, a low Bouguer gravity anomaly, strong atten- Asurfaceconductor(C4)isimagedatthesoutheastendofthe
uation of S waves and low P velocities (Ates et al., 1999; EAFSprofileandisassociatedwiththesedimentarybasindeposit-
n n
Barazangi et al., 2006; Angus et al., 2006; Gök et al., 2007). The ed on the Arabian Platform. Another conductor (C3) is also
low resistivity likely indicates a region of elevated fluid content, observeddippingsoutheastawayfromthesurfacetrace.Detailed
mostlikelypartialmeltandbyinferenceisaregionofweakrhe- geologicalmappingisnotavailablefromthisarea,sothisfeature
ologywithinthelowercrust.Whilethislayerisimagedinthepre- cannotbereliablyinterpretedbutitislikelyassociatedwithstruc-
sentstudy(C1),afullerdiscussionofitssignificanceisdescribed tures in the sedimentary basin, perhaps formed by fault normal
by Türkog˘lu et al. (2008). Of relevance to this study, the layer deformation.
appearstobeconnectedtotheNAFandEAFwithalowresistivity
verticalfeature(Fig.12). 4.2.2.NorthAnatolianFault
The conductance of the NAFS around Refahiye over the depth
4.2.Crustalfaultzoneconductors range 0–50km is 1000S. Using the same approach to calculate
the porosity gives a value of 0.1–3% (m=1) for the region of
The MT profiles described in this paper lack the close station 10Xm material extending from the surface to approximately
spacingneededtodetectashallowDZCassociatedwiththeupper 5kmdepth(C5,Fig.12).Inthiscontextitshouldalsobenotedthat
crustaldamagezone,suchastheconductordiscoveredontheCen- claymineralsarealsocapableofloweringtheelectricalresistivity
tralSanAndreasFault.However,theprofilegeometryiscapableof ofafluidbearingrockthroughthepresenceofanelectricaldouble
resolvingthedeeperCFZCthathasbeenobservedinthemid-crust layer that forms at the fluid–clay interface which allows ions to
beneathanumberofstrike-slipfaults.Aconductorofthistypeis move more easily than in the fluid phase. However, it is difficult
the most prominent feature in the EAF-profile inversion model todistinguishbetweenclayandaqueousfluidsonthebasisofjust
(C2),andislocatedbeneaththesurfacetraceoftheEAF.Aweaker the resistivity values imaged with MT data. If Archie’s Law is
featureisassociatedwiththeNAF(C5).Thegeometryandphysical applied,andclayispresent,thentheporosityderivedfromArchie’s
propertiesoftheseconductorsarediscussedinthefollowingsec- Law will be an upper limit (Worthington, 1993). The clay is also
tions. Fault-zone fluids have been proposed as a mechanism for significantbecauseitcanlowerthefrictionandmodifytheseismo-
explaining the apparent low strength of major fault zones (Sleep tectonicbehaviorofthefault(Hironoetal.,2008).