3D crustal structure along the North Anatolian Fault Zone in northcentral Anatolia revealed by local

31 Pages·2012·3.26 MB·English

Geophysical Journal International
Geophys.J.Int. (2012)188,819–849 doi:10.1111/j.1365-246X.2011.05313.x
3-D crustal structure along the North Anatolian Fault Zone in
north-central Anatolia revealed by local earthquake tomography
Seda Yolsal-C¸evikbilen,1 C. Berk Biryol,2 Susan Beck,2 George Zandt,2 Tuncay Taymaz,1
Hande E. Adıyaman2 and A. Arda O¨zacar3 s
c
ni
1DepartmentofGeophysicalEngineering,theFacultyofMines,IstanbulTechnicalUniversity,MaslakTR-34469,Istanbul,Turkey. o
E-mail:[email protected] ct
2DepartmentofGeosciences,UniversityofArizona,Tucson,AZ85721,USA te
3DepartmentofGeologicalEngineering,MiddleEastTechnicalUniversity,I˙no¨nu¨Bulvarı,06531Ankara,Turkey d
n
aD
o
sw
cn
Accepted2011November21.Received2011November21;inoriginalform2010September13 milo
a
ad
ne
SUMMARY odyd from
3-D P-wave velocity structure and V /V variations in the crust along the North Anatolian Ge h
p s ttp
FanaudltSZ-woanvee(NtraAvFeZlti)minesn,otrothg-acienntarableAttneratuonlidaewrsetarendininvgesotifgathteedsbeyismthoeloingvicearlsicohnaorafcltoecriasltiPcs- GJIs://a
c
a
of the region. The 3-D local earthquake tomography inversions included 5444 P- and 3200 d
e
m
S-wavereadingsobtainedfrom168well-locatedearthquakesbetween2006Januaryand2008
ic
May.Denseraycoverageyieldsgoodresolution,particularlyinthecentralpartofthestudy .o
u
p
area. The 3-D Vp and Vp/Vs tomographic images reveal clear correlations with both the .c
o
surfacegeologyandsignificanttectonicunitsintheregion.Weobservedthelowerlimitofthe m
/g
seismogeniczonefornorth-centralAnatoliaat15kmdepth.Finalearthquakelocationsdisplay ji/a
adistributedpatternthroughoutthestudyarea,withmostoftheearthquakesoccurringonthe rtic
majorsplaysoftheNAFZ,ratherthanitsmasterstrand.Weidentifythreemajorhigh-velocity le
-a
blocks in the mid-crust separated by the ˙Izmir-Ankara-Erzincan Suture and interpret these b
s
blockstobecontinentalbasementfragmentsthatwereaccretedontothemarginfollowingthe tra
c
closureofNeo-TethyanOcean.Thesebasementblocksmayhaveinpartinfluencedtherupture t/1
8
propagationsofthehistorical1939,1942and1943earthquakes.Inaddition,largevariations 8
/3
inthe V /V ratiointhe mid-crustwere observed and have been correlated withthe varying /8
p s 1
fluidcontentsoftheexistinglithologiesandrelatedtectonicstructures. 9/6
9
0
Key words: Seismicity and tectonics; Body waves; Seismic tomography; Continental tec- 2
1
6
tonics:strike-slipandtransform;Crustalstructure. b
y
g
u
e
s
t o
n
0
5
b; Figs 1b and c). The Intra-Pontide and ˙Izmir-Ankara-Erzincan A
1 INTRODUCTION p
Inthisstudyweinvestigatethe3-DcrustalstructurealongtheNorth Stiudteusr,eKZıros¸neehsir(MIAaEsSsiZf,)E,˙IzsitnaenbPuazlaZroı–nSeu,nSgaukralruyFaaZuoltnaen,dCtehnetrC¸alanPkoınr-ı ril 20
1
AnatolianFaultZone(NAFZ)innorthern-centralAnatolia.Itisa Basinareexamplesofthesetectonicstructures(Fig.1b).The2000- 9
seismicallyactivedextralstrike-slipfaultzoneextendingforabout km-long˙Izmir-Ankara-ErzincanSutureZone(Fig.2b),aremnant
1500kmfromKarlıovaineasternTurkeytotheGulfofSarosinthe oftheNeo-TethysOceanandamajorcompressionalpalaeotectonic
AegeanSea(Fig.1a).ActivetectonicsoftheNorthCentralAnato- structureinnorthernTurkey,trendsinanapproximatelyE–Wdi-
lianregionaregovernedbythecomplexcollisionoftheArabian, rectiontothewestofAnkara,thenturnsnearly90◦andhasaNNE
Africanand Eurasianplates and theresultingwestward extrusion trend towards C¸ankırı (see Koc¸yig˘it 1991; Okay & Tu¨ysu¨z 1999;
of Anatolian Plate between the right-lateral NAFZ and the left- Kaymakc¸ı2000;Taymazetal.2007a,b).TheC¸ankırıBasinislo-
lateralEastAnatolianFaultZone(EAFZ,McKenzie1972;S¸engo¨r catedbetweentheSakaryaContinenttothenorthandtheKırs¸ehir
1979;S¸engo¨retal.1985;Deweyetal.1986;Taymaz1996;Taymaz BlocktothesouthandboundedbytheNorthAnatolianOphiolitic
etal.2004,2007a).Ascommonlyemphasized,theNAFZandsur- Me´lange in the west, north and east (Fig. 1c). It has a basement
roundingregionareimportantcontributorstotheseismotectonics consisting of an upper Cretaceous ophiolitic me´lange and grani-
oftheAnatolianregion,becausetheyappeartorepresentageologi- toidsoftheKırs¸ehirBlock(Kaymakc¸ıetal.2009,2010;Robertson
calknotwheremanydifferenttectonicbeltsconverge(Ketin1966; etal.2009).Duetoitsmultideformationalgeologicalhistory,the
Go¨ru¨r et al. 1998; Okay & Tu¨ysu¨z 1999; Taymaz et al. 2007a, C¸ankırıBasinandsurroundingregionhavebeenthefocusofseveral
(cid:3)C 2012TheAuthors 819
GeophysicalJournalInternational(cid:3)C 2012RAS
820 S.Yolsal-C¸evikbilenetal.
D
o
w
n
lo
a
d
e
d
fro
m
h
ttp
s
://a
c
a
d
e
m
ic
.o
u
p
.c
o
m
/g
ji/a
rtic
Figure1. (a)MajortectonicunitsandbathymetryintheEasternMediterraneanSearegion(seefordetailsTaymazetal.2004,2007a,b;Yolsaletal.2007; le
Yolsal2008andYolsal&Taymaz2010).NAFZ,NorthAnatolianFaultZone;NEAF,NorthEastAnatolianFault;EAFZ,EastAnatolianFaultZone;DSF,Dead -a
b
TSeraanTsfroarnmsfoFramultF;aDulFt,;DAeSli,leArpFs¸aeuroltn;EScilFl,;EBcFe,mBios¸zFoavualtF;aEuFl,t;EBlbGisFt,aBneFyas¸uelht;irEGSFo¨Zlu¨,FEazuinlte;PBaMzaGrı,–BSuu¨ynug¨kurMlueFnaduelrteZsoGnera;bEernF;,BErucFi,yBesuFrdauurlt;FGau,lGt;o¨CkToFv,a;CGepeh,aGloedniiza strac
Graben;GF,GarniFault;IF,Ig˘dırFault;K,Karlıova;KBF,Kavakbas¸ıFault;KF,Kag˘ızmanFault;KFZ,Karatas¸-OsmaniyeFaultZone;MF,MalatyaFault; t/1
8
MRF,MainRecentFault;MT,Mus¸Thrust;OF,OvacıkFault;PSF,Pampak-SavanFault;PTF,PaphosTransformFault;SaF,SalmasFault;Si,SimavGraben; 8
/3
SuF,Sultandag˘Fault;TeF,TebrizFault;TF,TatarlıFault;TGF,TuzGo¨lu¨Fault;TuF,TutakFault.Largeblackarrowsshowrelativeplatemotionswithrespect /8
toEurasia(McCluskyetal.2000,2003).Bathymetriccontoursareshownat500,1000,1500and2000m,andareobtainedfromGEBCO-BODC(1997)and 19
Smith&Sandwell(1997a,b).(b)MajortectonicunitsandblocksinTurkey.AFZ,AlmusFaultZone;CAFZ,CentralAnatolianFaultZone;DFZ,DodurgaFault /69
Zone;ETGFZ,Eskis¸ehir-TuzGo¨lu¨FaultZone;EFZ,EldivanFaultZone;LFZ,Lac¸inFaultZone;ESFZ,EzinePazarı–SungurluFaultZone;KFZ,Kızılırmak 02
1
FaultZone;ODFZ,Orta–DevrezFaultZone;C¸KFZ,C¸erkes¸–Kurs¸unluFaultZone(revisedfromBozkurt&Koc¸yig˘it1996;Koc¸yig˘it&Beyhan1998;Okay& 6
Tu¨ysu¨z1999;Kaymakc¸ı2000;Koc¸yig˘itetal.2001;Kaymakc¸ıetal.2003;Taymazetal.2007b).(c)GeneralizedgeologicalmapoftheNorthAnatolianFault by
Zone(NAFZ)innorth-centralAnatoliashowingtheC¸ankırıBasin(Okay&Tu¨ysu¨z1999;Taymazetal.2007b). gu
e
s
t o
n
geological,seismologicalandgeophysicalstudies(e.g.Go¨ru¨retal. 2009) compared to the SAF, which has ∼315–730 km of offset 0
5
1984; S¸engo¨r et al. 1985; Piper et al. 1996; Ates¸ et al. 1999; (e.g.McKenzie&Morgan1969;Atwater1970;Dickinson&Wer- A
p
2K0a1y0m;aTkac¸yım20a0z0e;tAadl.ıy2a0m0a7na,ebt;aSl.e2y0it0o1g˘;luKaeytmala.k2c¸0ı0et9a).l.I2n0n0o3r,t2h0e0rn9-, nthicekNeA1F99c7o)n.stSitiumteilsarthiennloarrtgheerenarbtohuqnudaakreyaocftitvhietyAwnaittholitahnePSlaAteF., ril 20
1
centralAnatolia,thisstudyarea,therearealsoseveralsmall-and The scattered earthquake activity along the NAF and the abun- 9
large-scalesedimentarybasins(e.g.C¸erkes¸-Kurs¸unlu-Ilgaz,Tosya, danceofoff-faultseismicitymightbeanindicationofitsyoungage
Kargı, Vezirko¨pru¨, Havza–Ladik, Niksar, Sus¸ehri, Erzincan and and geological and structural complexity. Fig. 2(a) shows recent
Karlıova)associatedwithbendsandoffsetsofthestrike-slipfault seismicity in the Eastern Mediterranean and surrounding region
zones. reported by the USGS-NEIC (United States Geological Survey –
TheNorthAnatolianFault(NAF)isanalogoustotheSanAndreas National Earthquake Information Center; 1973–2010) and focal
Fault(SAF)inCaliforniainsomeways,withthetwocontinental mechanism solutions of major earthquakes (M > 6.0) along the
transformssharingsimilarsliprates,totalfaultlengthandstraight- NAFZ. Major destructive earthquakes occur on the NAFZ with
ness relative to their poles of rotation (Stein et al. 1997). How- a westward migrating pattern, from the Karlıova Triple Junction
ever, the age and offset values are significantly different for each to the Marmara region, in series, every ∼200–400 yr, causing
faultzone.TheNAFisyounger(∼5Ma)thantheSAF(∼17–30 serious risk for population centres (for details see Barka 1996;
Ma)andhasabout85kmofcumulativedisplacement(Westaway Stein et al. 1997; Ambraseys 2002). One of the largest histor-
1994;Armijoetal.1999;Barkaetal.2000;Hubert-Ferrarietal. ical earthquakes occurred in 1939, in Erzincan, eastern Turkey
(cid:3)C 2012TheAuthors,GJI,188,819–849
GeophysicalJournalInternational(cid:3)C 2012RAS
Structurerevealedbyearthquaketomography 821
D
o
w
n
lo
a
d
e
d
fro
m
h
ttp
s
://a
c
a
d
e
m
ic
.o
u
p
.c
o
m
/g
ji/a
rtic
le
-a
b
s
tra
c
t/1
8
8
/3
/8
1
9
/6
9
0
2
1
6
b
y
g
u
e
s
t o
n
0
5
A
p
ril 2
0
1
9
Figure1. (Continued.)
(Mw 7.9) and ruptured over 360 km of the NAFZ, with a maxi- and 7.2 ˙Izmit and Du¨zce earthquakes, respectively (Barka et al.
mumright-lateraloffsetof∼7.5m.Theseismicity(Fig.2a)pro- 2000).
gressedalongthefaultzonewithmanyMs>6earthquakes,reach- In this study we used local earthquake tomography (LET) to
ing Hersek Peninsula and beyond with the recent 1999, Mw 7.6 obtainabetterunderstandingofthe3-DP-wavecrustalstructure
(cid:3)C 2012TheAuthors,GJI,188,819–849
GeophysicalJournalInternational(cid:3)C 2012RAS
822 S.Yolsal-C¸evikbilenetal.
D
o
w
n
lo
a
d
e
d
fro
m
h
ttp
s
://a
c
a
d
e
m
ic
.o
u
p
.c
o
m
/g
ji/a
rtic
le
-a
b
s
tra
c
t/1
8
8
/3
/8
1
9
/6
9
0
2
1
6
b
y
g
u
e
s
t o
n
0
5
A
p
ril 2
0
1
9
Figure1. (Continued.)
and V /V ratios for the region. LET studies are able to image 1997, 2009; Husen et al. 2000; Paul et al. 2001; Rowlands et al.
p s
heterogeneitiesinthecrustanduppermantle,withspatialresolution 2005;Koulakovetal.,2007;Kaypak2008;Haberlandetal.2009).
dependingonthedensityofraysampling(e.g.Thurber1981,1992). A detailed knowledge of the velocity structure of the crust and
Manylocaltomographystudieshavebeenperformedduringthelast upper mantle not only helps to improve our understanding of the
decadeindifferenttectonicallyactiveregions,includingstrike-slip, tectonicsofaregion,butalsoplaysakeyroleinobtainingaccurate
subduction, collisional and/or volcanic zones (see Thurber et al. earthquakelocations.
(cid:3)C 2012TheAuthors,GJI,188,819–849
GeophysicalJournalInternational(cid:3)C 2012RAS
Structurerevealedbyearthquaketomography 823
D
o
w
n
lo
a
d
e
d
fro
m
h
ttp
s
://a
c
a
d
e
m
ic
.o
u
p
.c
o
m
/g
ji/a
rtic
le
-a
b
s
tra
c
t/1
8
8
/3
/8
1
9
/6
9
0
2
Figure2. (a)EarthquakeactivityreportedbyUSGS-NEICandfocalmechanismsolutionsofmajorearthquakes(M>6.0)alongtheNorthAnatolianFault 16
Zone(NAFZ)andsurroundingregionfortheperiodof1973–2010(TableB1;compiledfromMcKenzie1972;Taymazetal.1991,2007b;Taymaz1999; by
S¸engo¨retal.2005)and(b)thesettingoftheNAFPassiveSeismicExperimentinthenorth-centralAnatolia.Theseismicnetwork(whitesquares)consisted gu
ofNAFbroad-bandstationsoperatingfrom2006Januaryto2008May.TheblacksquaresdenoteKOERI(KandilliObservatory)stationsusedtoimprovethe es
azimuthalcoverageinthisstudy(seeTable1).Newearthquake(M≥3.0)locationsobtainedbyusingtheANTELOPEdbloc2commercialsoftwaresuitefrom t o
n
BRTT(1998)andPandSreadingsrecordedbyNAFandKOERIstationsarealsoshown.Blacklinesindicatethemajorfaults(S¸arog˘luetal.1992)andred 0
thesuturezones(Okay&Tu¨ysu¨z1999)intheregion,respectively.IAESZ,˙Izmir-Ankara-ErzincanSutureZone;ESFZ,EzinePazarı–SungurluFault;NAFZ, 5 A
NorthAnatolianFaultZone;ITS,Intra-TaurideSuture;IPS,Intra-PontideSuture. pril 2
0
1
9
Despitenumerousgeologicalandgeophysicalstudies,theshal- S¸engo¨retal.2005;Taymazetal.2007a,b;Toketal.2008;Gans
lowcomplexstructuresinnorth-centralAnatoliaremainpoorlyun- et al. 2009; Biryol et al. 2010). To image the crustal structure in
derstood.Afterthedevastating1999˙IzmitandDu¨zceearthquakes, north-central Anatolia, Gu¨rbu¨z et al. (2003) carried out seismic
theMarmaraSeaBasinandthewesternpartoftheNAFZwerein- refractionstudies,whichsuggestedthecrustisabout36kmthick
vestigatedinnumerousseismologicalstudies.Severaltomography at south of NAFZ, and decreases to 30 km north of the NAFZ,
studieswereperformedtounderstandtheshallowcrustalstructure withanaverageP velocityof7.8kms–1.Similarly,Kulelietal.
n
of the western continuation of the NAFZ (e.g. Nakamura et al. (2004) found that the crustal thickness varies between 37 and 42
2002;Karabulutetal.2003;Barıs¸ etal.2005;Salahetal.2007; kmincentralAnatolia,withaP velocityofabout8kms–1,also
n
Beceletal.2010;Koulakovetal.2010).Incontrast,thereisstill basedonseismicrefractiondata.Theyobservedlaterallychanging
nodetailedinformationaboutthe3-Dcrustalvelocitystructureand crustalthicknessandanabruptchangeacrossthemajorfaultzones,
deformationstylesinthecentralpartoftheNAFZ,eventhoughit suchastheNAFZ,suggestingthatthecrustalthicknessisthinner
hasbeenextensivelystudiedbyvariousmethods(e.g.Tataretal. innorthwestAnatoliaandslightlythickerineasternAnatolia.Gans
1996;Gu¨rbu¨zetal.2003;Tokso¨zetal.2003;Kulelietal.2004; et al. (2009) performed P tomography and reported the lack of
n
(cid:3)C 2012TheAuthors,GJI,188,819–849
GeophysicalJournalInternational(cid:3)C 2012RAS
824 S.Yolsal-C¸evikbilenetal.
a P velocity contrast across the NAFZ. They also found that P
n n
velocitiesintheEasternandWesternPontidesareslowerthaninthe
CentralPontides.Inaddition,thereareanumberofcrustalmodels
fromreceiverfunctionstudiesbySaundersetal.(1998),Toketal.
(2008) and C¸akır & Erduran (2011) for the Anatolian region to
nameafewamongothers.
We investigate the shallow crustal structure and seismicity in
thecentralpartoftheNorthAnatolianFaultusingLET.Weused
high-qualityearthquakedatarecordedby47broad-bandseismome-
terstodeterminea3-DP-wavevelocitymodelandtheV /V dis-
p s
tribution. We compare our tomographic images with well-known
geologicalstructuressuchastheEzinePazarı–SungurluFaultZone,
C¸ankırıBasinand˙Izmir-Ankara-ErzincanSutureZone.
D
o
2 DATA AND METHOD w
n
lo
Weused39broad-bandseismicstationsinstalledinthestudyarea a
d
starting in 2006 January and operating until 2008 May, as a part Figure 3. Wadati diagram using P- andS -wave traveltimes recorded by ed
ofthe‘NorthAnatolianFault(NAF)passiveseismicexperiment’, theNAFExperimentandKOERIstationsusedinthisstudy.Tp,P-wave fro
anicjoailnUt pnriovejercsittyb,etMwiedednlethEeaUstnTiveecrhsnitiycaolfUAnriivzeornsait,yIsatnadnbBuolg˘Taezcihc¸-i tthraevnelotrimthe-c;eTnst,raSl-Awnaavteoltiraavreelgtiiomne..TheapproximateVp/Vsratiois1.76for m http
s
University Kandilli Observatory and Earthquake Research Insti- ://a
tute (KOERI; http://www.koeri.boun.edu.tr/, last accessed 2010). c
a
d
Wealsocomplementedourstationswithadditionaldatafromeight e
m
permanentbroad-bandKOERInetworkstationslocatedinthesame ic
region(Table1,Fig.2b). .o
u
p
ThedatasetincludesarrivaltimesofP-wavesandS–Ptimesob- .c
tainedfrom∼300localearthquakes(M≥3.0)thatoccurredwithin om
north-centralAnatoliafortheperiodof2006January–2008May. /g
Aanrdrihvoalritziomnetsalocfo9m8p9o0nPenatnsdus7in9g17thSeAwaNvTesELwOerPeEpcicokmedmoerncivaelrstiocfat-l ji/article
waresuitefromBoulderReal-TimeTechnologies(BRTT1998).The -a
b
Wadatidiagramandtraveltimeversusdistancegraphswereplotted stra
c
t/1
Table1. Coordinatesofbroad-bandseismicstationswhichbelongto 88
the‘NAFPassiveSeismicExperiment’andKOERIusedinthisstudy /3
(seeFig.2b). Figure 4. Traveltime Vs. distance plot indicating average direct and re- /81
Station Lat.(◦N) Lon.(◦E) Station Lat.(◦N) Lon.(◦E) fractedcrustalP-andS-wavevelocities. 9/69
0
2
ALIC 40.978 33.487 INSU 39.842 35.366 1
6
ALIN 41.061 32.879 ISKE 40.764 37.067 toestimatetheaverageV /V ratio(1.76),aswellasthedirectand b
p s y
ALOR 41.301 32.870 KARA 40.688 35.245 refractedcrustalP-andS-wavevelocities(Figs3and4).Wewere gu
ARSL 40.955 35.887 KARG 40.291 33.552 e
abletolocateabout200earthquakes,whichhadatleastsixPand s
BBBAEEDKGIIB 444011...213721815 333634...452106063 KKKAGIYVAIAC 444000...219834011 333254...833712836 SLOrePaEdidnbglso,cw2iathlgaonriathzmimpuathckalaggeapanodftlheessvtehlaonci2ty00m◦o,dueslinogfAGuN¨rTbuE¨z- t on 05
etal.(2003)andTokso¨zetal.(2003).Inthefinalinversions,only A
BOKE 40.552 36.211 KIZI 40.048 36.536 p
BCZAKKM 4410..906105 3347..030637 KKUUYZAL 4410..549401 3346..323428 1ti6o8necaorvtehrqaugaek(e≤s1w8e0r◦e),sethleecnteudmbbaesreodfoPn-tahnedirSa-zwimavuethraeladgianpgisn(>staa-t ril 201
CALT 41.328 35.125 KUZO 40.904 32.861 least10P-wavearrivals)andsmallrmsvalues(Fig.2b). 9
CANT 40.606 33.619 KVT 41.008 36.046
CAYA 40.373 34.269 LOD 39.889 32.764
CORM 40.178 34.630 OGUR 41.109 35.165 2.1 1-Dvelocityinversion
CRLU 40.064 34.357 PANC 40.647 34.301
CUKU 40.604 33.441 PELI 41.113 34.299 Toobtainanappropriate1-Dstartingmodel,wecomputedthebest
DERE 41.477 35.064 RSDY 40.397 37.327 minimum 1-D P-wave velocity model using the VELEST code
DIKM 41.649 35.025 SEYH 40.856 32.900 (Kissling1988;Kisslingetal.1994,1995).Thisalgorithmcalcu-
DOGL 40.391 35.284 SVSK 39.917 36.992 latestheraypathsfromsourcetoreceiver,includingdirect,refracted
DUMA 40.918 35.140 SYUN 40.838 33.529 andreflectedrayspassingthroughthegiven1-Dvelocitymodel.Be-
EKIN 41.147 35.787 TEPE 41.369 35.743
causetheproblemisnon-linear,thesolutionisobtainediteratively.
GOCE 39.743 34.348 YESI 40.405 37.229
The inversion process consists of checking the space parameters,
HASA 41.469 33.565 YIKI 40.748 35.954
takingintoaccountdifferent1-Dinitialmodelsthatincludevarying
INCE 40.581 32.906
the numbers of initial velocities and thicknesses of the layers. It
(cid:3)C 2012TheAuthors,GJI,188,819–849
GeophysicalJournalInternational(cid:3)C 2012RAS
Structurerevealedbyearthquaketomography 825
Table2. Thefinal1-DP-wavevelocitymodel Next,wedeterminednewearthquakehypocentresusingthebest
obtainedforthestudyarea. minimum1-DP-wavevelocitymodel,P-wavestationcorrections
andtraveltimesoflocalPandSwavesinthesingleeventmodeof
Layerdepth Layerthickness
(km) (km) Vp(kms–1) theVELESTalgorithm.Wealsoplottedapproximateraycoverage
usinglinestoconnectindividualstationsandevents(Fig.7).These
–2.0 –2.0 2.15
rays demonstrate the capability of resolving power for each grid
0.0 5.0 3.86
5.0 2.0 5.66 nodeinthe3-DLETinversion.
7.0 2.0 5.99
9.0 2.0 6.09
11.0 4.0 6.10 2.2 3-DLET
15.0 5.0 6.18
20.0 10.0 6.20 We used the damped least-squares iterative inverse code
30.0 10.0 6.62 SIMULPS14, which includes a standard approximate pseudo-
40.0∗ 10.0 7.67 bending ray tracer (ART) and a full 3-D shooting ray tracer al-
50.0∗ – 8.10 gorithm(RKP,Eberhart-Phillips1990,1993;Virieux&Farra1991; D
∗Denotesanundeterminedlayervelocityby Evansetal.1994;Husenetal.2000;Haslinger&Kissling2001), ow
n
the1-Dinversion. toinvertforthe3-Dvelocitymodels.Thefinaldatasetconsistsof lo
a
168 well-located earthquakes recorded by 47 broad-band seismic d
e
alsocalculatesstationcorrectionstocompensateforheterogeneous slotactiitoynms,owdiethl i5s4s4p4ePci-fiweadveonanads3e2t0o0fS3--wDavsepaatriraivlaglrtiidmpeosin.tTsh,ewviteh- d from
v1e9l8o8c;itSyeasrtrcuyc2tu0r0e3)n.eFaurritnhderivdideutaaillssotaftitohnespr(oDcoeussgilnagss1t9ep6s7;foPrutjhoel athleinperaerviionutesrpLoElaTtiostnudaideospcteitdedbeetawrleieenr.tWheepcooinnststr,uccotendsisthteentinwitiitahl https
VELESTcodecanbefoundinKissling(1988)andKisslingetal. P-wave velocity model using the best minimum 1-D horizontally ://a
c
(1994,1995). layeredmodeldeterminedfromtheVELESTcode,andareference ad
WeappliedtheVELESTcodeinasimultaneousmode,inverting Vp/Vsvalueof1.75.Weinitiallytriedseveraldifferentgridintervals, em
fsotarrbtiontgh1ve-Dlocvietyloacnitdyhmypoodceelnitnrealcoccoartdioanncs.eWweitchhoresferalacytieorninsgeoisfmthice vfoarrythineginbveetrwseioenn.1A5fiannadl3g0rikdmsp,atocionbgtaoifn2t5hekmop×tim2u5mkmgriindtsheettihnogrs- ic.oup
studiescarriedoutinthisregion(Gu¨rbu¨zetal.2003;Tokso¨zetal. izontaldirectionswaschosenforthecentralpartofthestudyarea, .co
duetothedenseraypathcoverageandstationdistribution(Fig.7). m
2003). Because the VELEST code does not alter layer geometry /g
andinvolvesatrial-and-errorprocess,weperformedtrialrunsfor Gridnodesintheverticaldirectionwerepositionedat–2,0,5,7,9, ji/a
several models with different layering and velocities to select the 11,15,20,30,40and50km,consistentwithourbest1-DP-wave rtic
bbeesstt-mfititninimgulamye1r-Dthimckondeesls.Tahnedvgeeloomcietytrmy,oadnedlgthivuisntgothesetmabilnisimhuthme v5ealoncdity11mkomdelyi(eTladbeldel2o)w.Trhmesthvianluveesrtficoarltnhoedieniltaiyaelr1in-Dgbsetatwrteinegn le-ab
s
rmsvalue(0.51)waschosenasthefinal1-DP-wavevelocitymodel modelandinitialearthquakelocations.Thethinlayersallowbet- tra
(Fig.5andTable2).Thelayersat40and50kmdeptharepoorly termodellingofverticalgradientswhichimprovesraytracingfor ct/1
raysbottominginthisdepthrangeandprovidemodelflexibilityto 8
constrained due to the absence of deep earthquakes in the study 8
moreaccuratelyreflecttopographyofthebasin–basementinterface. /3
area.Theinversionresultsweretestedbyusingmanyinitialmodels /8
Modelresolutionstudiesshownlaterindicatethatthereissomede- 1
withaveragevelocitiessignificantlyhigherandlowerthanthoseof 9
the starting 1-D model. However, we calculated high rms values cline in individual parameter resolution but we did not interpret /69
for both trial cases. We used damping parameters of 0.01 for the individual node anomalies in the resulting tomography. Another 02
1
hypocentral,0.1forthestationand1.0forthevelocityparameters, important input parameter is the minimum number of rays sam- 6 b
assuggestedbyKissling(1988)andKisslingetal.(1994). pling in each grid node; the resolution of the inversion improves y g
TocalculateP-wavestationcorrections,weselectedstationPELI as the number of rays increases. Using a trial-and-error method, ue
asthereferencestationbecauseofitscentrallocationwithinthenet- we chose a minimum of five rays for both the synthetic and the st o
workandtheabundanceofrecordedearthquakes(Fig.2b).Station realmodels.Basedonpreviousstudies,thecrustalthicknesswas n 0
corrections represent the averaged characteristics of local surface assumed to be about 40 km for the entire model and to account 5 A
goueorlcoagsye,astmeaalclhstasittieon(Kdeislasylisnwge1r9e8o8b;taHinuesdeninetthaelc.e2n0t0ra0l,p2a0r0to3f).thIne fcoornsatrrraivinaelsdobfyPandpeheapseers,ltahyeeruplopceart-emdanattle50pakrmto.fTthhreoumgohdoeultwthaes pril 2
0
seismicnetwork,whereaslargeP-wavestationcorrections(>0.7s) tomography inversion, we applied weighting to each observation 19
basedonbothresidualsizeandsource–receiverdistance.Wealso
were found towards the outer regions of our study area (Fig. 6).
testedvariabilitywithlargerandequallayergridnodespacingand
Negativestationdelayscorrespondtohavingtruevelocitiesfaster
obtainedverysimilarresultsforthewell-resolvedcentralportion
than the model, and positive delays to true velocities slower than
ofourmodelledregion(seeAppendicesAandB).
the model. Absolute station delay values tend to be larger at the
Oneofthekeyparametersintheinversionisthedampingvalue,
outsideofthenetworkduetothecombinedeffectsoflocalgeology
whichdependsontheearthquakesource-stationdistributionandthe
and the limited azimuthal coverage and distances of observations
selectedgridintervals.Thedampingparameterinfluencesboththe
at these stations. Thus the resultant 1-D velocity model is a bet-
inversionresultsandresolutionestimates(e.g.Husenetal.2000;
terrepresentationofthestructureinthewell-sampledcentralparts
Kisslingetal.2001).Inthisstudy,wefollowedtheEberhart-Phillips
anddeviatesmorefromthestructureintheperipheralregionsand
(1986)proceduretoselecttheappropriatedampingvaluebyeval-
structurebeneathcorrespondingstations.Thisdeviationmightbe
uatingtrade-offcurvesbetweenthedataandthesolution(model)
mappingintostationcorrections(andcompensatedbytheseterms)
variances. These trade-off curves are built with a single iteration
oftheperipheralstations,makingthemlarger.
(cid:3)C 2012TheAuthors,GJI,188,819–849
GeophysicalJournalInternational(cid:3)C 2012RAS
826 S.Yolsal-C¸evikbilenetal.
D
o
w
n
lo
a
d
e
d
fro
m
h
ttp
s
://a
c
a
d
e
m
ic
.o
u
p
.c
o
m
/g
ji/a
rtic
le
-a
b
s
tra
c
t/1
8
8
/3
/8
1
9
/6
9
0
2
1
6
b
y
g
u
e
s
t o
Figure5 (a)Thefinal1-DP-wavevelocitymodel(blacksolidline)afterseveralinversionsteps(formodeldetailsseeTable2).Thedashedlineshowsthe n
starting(initial)1-Dvelocitymodelusedintheinversion.(b)Testsonthestabilityoftheminimum1-DP-wavevelocitymodel.Thedashedlinesindicatethe 05
inputmodelsforthetestswithhighandwithlowinitialvelocities,andthesolidlinesrepresenttheresultingmodelsaftertheinversion. A
p
ril 2
0
1
inversion for the coupled hypocentre-model parameters, using a node(KHIT),thederivativeweightedsum(DWS,geometricaver- 9
widerangeofdampingparameters.Becausesuitabledampingval- age)andthediagonalelementoftheresolutionmatrix,resolution
ues show a significant decrease in data variance without a strong diagonalelement(RDE,seeHaslingeretal.1999;Husenetal.2000;
increaseinmodelvariance,wechoseadampingvalueof50forboth Diasetal.2007;Kaypak2008;Arroyoetal.2009;Thurberetal.
V andV /V inversionsbasedonthecalculatedtrade-offcurvesand 2009).Forexample,theDWSvaluedefinestherelativeraydensity
p p s
Occam’sRazorapproach(Fig.8).Adampingparameterof10also inthevicinityofamodelparameter(Thurber&Eberhart-Phillips
producesthesamedatavariance,butitscalculatedmodelvariance 1999)andisabettermeasureofinformationdensityprovidedby
ishigher.AccordingtotheOccam’sRazorapproach,iftwomodels theraycoveragethancountingthenumberofrays.TheRDEisthe
givethesamemisfit,thetendencyistochoosethesimplestmodel bestandquickestestimationofthequalityofthesolution,however,
(e.g. smallest number of parameter, smallest size and smoothest; asitpresentsthedegreeofindependenceofthemodelparameters
seeTrampert&Spetzler2006). inthesolution.Allresolutionparametersareaffectedbythechosen
Additionalcriteriondescribetheresolutionofthetomographyre- model parametrization of the tomography inversion (Husen et al.
sults,includingthenumberofrayhitscontributingtooneparticular 2000,2004;Kisslingetal.2001).
(cid:3)C 2012TheAuthors,GJI,188,819–849
GeophysicalJournalInternational(cid:3)C 2012RAS
Structurerevealedbyearthquaketomography 827
D
o
w
n
lo
a
d
e
d
fro
m
h
ttp
s
Figure6. P-wavestationcorrectionsobtainedfromthebest1-DP-wavevelocitymodelinversion(relativetostationPELI). ://a
c
a
d
e
m
ic
.o
u
p
.c
o
m
/g
ji/a
rtic
le
-a
b
s
tra
c
t/1
8
8
/3
/8
1
9
/6
9
0
2
1
6
b
y
g
u
e
s
t o
n
0
5
A
Figure7. Hypocentrelocationsof168selectedearthquakesobtainedfromtheVELESTsingleeventlocationandraypathcoverage(thingreylines)for3-D p
P-wavelocalearthquaketomography.Plussignsandwhitesquaresrepresent25km×25kmhorizontalgridnodespacing(forinteriorparts)andseismic ril 2
0
broad-bandstations,respectively. 1
9
3 RESOLUTION AND RESULTS 3.1 Resolution
The3-DP-wavevelocityandV /V ratiomodelswereobtainedby Ascommonlyemphasized,complex3-Dvelocitystructuresderived
p s
performingtwoseparatebutsequentialinversionsteps,eachhav- fromLETcanonlybeinterpretedmeaningfullyifthemodel’squal-
ingsixiterations,usingtheinversioncontrolparametersdescribed ityisknown(e.g.Thurber1993;Husenetal.2000;Arroyoetal.
earlier.Theoutputofthefirstinversionwasintroducedastheinput 2009).Wethereforeplottedtheresolutioncontrolparameterssuch
to the second inversion. After the final inversion, we reduced the asDWSandRDE(Fig.9).Further,weusedcheckerboardsensitiv-
rmstraveltimeresidualfrom0.51s(rmsfor1-DP-wavevelocity itytests,whichareusefulinassessingtheabilityofthetomographic
inversion)to0.077s(rmsafter3-DP-waveinversion)and0.078s inversiontoresolvestructuraldetailsintheEarth,togiveageneral
(rmsafter3-DV /V inversion). picture of the resolving power of the inversion. For this purpose,
p s
(cid:3)C 2012TheAuthors,GJI,188,819–849
GeophysicalJournalInternational(cid:3)C 2012RAS
828 S.Yolsal-C¸evikbilenetal.
D
o
w
n
lo
a
d
e
d
fro
m
h
ttp
s
://a
c
a
d
e
m
ic
.o
u
p
.c
o
m
/g
ji/a
rtic
le
-a
b
s
tra
c
t/1
8
8
/3
/8
1
9
/6
9
0
2
1
6
b
y
g
u
e
s
t o
n
0
5
A
p
ril 2
Figure8. Trade-offcurvesplottedfordifferentdampingvaluesforP-waveandVp/Vsratioone-iterationinversions.Wechoseadampingvalueof50forboth 01
inversions. 9
we set up a checkerboard velocity model by adding ±5 per cent Wealsoappliedjackknifeteststoobservetheinfluenceofvarious
velocityvariationtoourinitialmodel,andgeneratingsyntheticP- groupsofearthquakesontheinversionresults.Foreachcase,we
wavetraveltimesfromeachearthquakehypocentretotherecording removed10differentearthquakesselectedrandomlyfromtheinput
stations.Thevelocityperturbationsareapercentageoftheactual data, and then applied the 3-D inversion using the same model
velocityvalue.Beforetheinversion,weaddedGaussiandistributed parametrization. The mean velocities and standard deviations of
noise with a standard deviation of 0.1 s, with zero mean, to the velocitieswerecalculatedforeachgridpointineachmodellayer,
synthetictraveltimes.Finally,thesyntheticdatawereinvertedusing usingthejackknifetestresults(seeAppendixA,FigsA1–A4).The
thesameparametrization,constraintsandcontrolparametersused jackknifetestandresolutionparametersindicatethatwehavegood
intherealdatainversion(Fig.10). resolutioninthecentralpartofourstudyarea.
(cid:3)C 2012TheAuthors,GJI,188,819–849
GeophysicalJournalInternational(cid:3)C 2012RAS