The North Anatolian Fault on the Hersek Peninsula, Turkey: Its

Home , Armutlu Peninsula, North Anatolian Fault

Th e North Anatolian Fault on the Hersek Peninsula, Turkey: Its Geometry and Implications for the 1999 İzmit Earthquake Rupture Propagation

ÖZGÜR KOZACI1,2, ERHAN ALTUNEL3, SCOTT LINDVALL2, CHARLIE BRANKMAN2,4 & WILLIAM LETTIS2

1 İstanbul Technical University, Eurasian Earth Sciences Institute, Maslak, TR−34469 İstanbul, Turkey 2 now at Fugro William Lettis & Associates, Inc., Walnut Creek, 94596 California, USA (E-mail: [email protected]) 3 Eskişehir Osmangazi University, Engineering Faculty, Department of Geological Engineering, TR−26040 Eskişehir, Turkey 4 now at Department of Earth & Planetary Sciences, Harvard University, Cambridge, Massachusetts, 02138, USA

Received 02 November 2009; revised typescripts receipt 23 June 2010 & 04 August 2010; accepted 03 September 2010

‘We dedicate this study to Aykut Barka who devoted his life to understanding the earthquake phenomenon. He was a brilliant scientist, a true friend and a giving advisor besides his humble personality. He will be remembered as a source of inspiration and kindness.’

Abstract: Th e western termination of the 1999 İzmit earthquake still remains as an intriguing problem for researchers and the people residing around the Sea of Marmara. Th ere have been numerous off shore mapping and modelling studies performed in the Gulf of İzmit. However, the main debate about the western termination of the 1999 İzmit surface rupture is linked to the Hersek Peninsula and corresponding fault geometry. We focused our eff orts at resolving the fault geometry on the Hersek Peninsula by applying geological mapping, geomorphic analyses, palaeoseismic trenching and geophysical surveys. Our studies reveal that the North Anatolian Fault forms a restraining stepover and did not experience surface rupture during 1999 İzmit earthquake in the vicinity of Hersek Peninsula. We tested this fault geometry with a fi nite element model in half elastic space and correlated the results successfully with the existing topography. In addition, we ran a simple Coulomb model to explain the possible cause of surface rupture termination at this specifi c location. Our studies, combined with detailed off shore bathymetry data, suggest that the restraining step of the North Anatolian Fault on the Hersek Peninsula is capable of creating an effi cient earthquake rupture barrier.

Key Words: North Anatolian Fault, Hersek Peninsula, fault geometry, rupture termination, active tectonics

Kuzey Anadolu Fayı’nın Hersek Deltası’ndaki Geometrisi ve 1999 İzmit Depremi Kırığının İlerlemesine Etkileri

Özet: 1999 İzmit depreminin batıda sonlandığı yer araştırıcılar ve Marmara Denizi civarında yaşayanlar için önemli bir sorun oluşturmaya devam etmektedir. İzmit Körfezi’ni konu alan pek çok kıyı ötesi haritalama ve modelleme çalışmaları yapılmasına rağmen 1999 İzmit depremi yüzey kırığının sonlandığı yerle ilgili tartışmalar Hersek Deltası’na ve Kuzey Anadolu Fayı’nın buradaki geometrisine düğümlenmiştir. Bu sorunu anlamak üzere çalışmalarımız Hersek Deltası’ndaki fay geometrisini anlamamıza yardım edecek şekilde jeomorfolojik analizler, paleosismik hendek kazıları, ve jeofi zik araştırmalar üzerinde yoğunlaştırılmıştır. Çalışmalarımız Kuzey Anadolu Fayı’nın bu bölgede sıkışma oluşturan bir geometriye sahip olduğunu ve 1999 İzmit depremi sırasında yüzey kırığı meydana getirmediğini ortaya koymuştur. Yarı uzayda sonlu elemanlar yöntemiyle modellenen bu fay geometrisi çalışma alanının güncel topoğrafyası ile uyum göstermektedir. Ayrıca, basit bir Coulomb modellemesi ile yüzey kırığının neden burada sonuçlanmış olduğu açıklanmıştır. Deniz çalışmaları ile karada yaptığımız çalışmaların biraraya getirilmesi Kuzey Anadolu Fayı’nın Hersek Deltası’ında sıkışmalı bir sıçrama yaptığını ve bu fay geometrisinin etkin bir deprem kırığı engeli oluşturduğunu ortaya koymaktadır.

Anahtar Sözcükler: Kuzey Anadolu Fayı, Hersek Deltası, fay geometrisi, kırık sonlanması, aktif tektonik

359 NORTH ANATOLIAN FAULT ON THE HERSEK PENINSULA

Introduction that specifi c location are essential for estimating the On August 17th 1999, the M7.4 İzmit earthquake location and potentially the magnitude of future large struck the Marmara region of Turkey causing much earthquakes. Researchers (e.g., Barka & Kadinsky- devastation. Th e İzmit earthquake is the seventh Cade 1988; Stein et al. 1997; Wesnousky 2008) have surface rupturing, large-magnitude earthquake in convincingly demonstrated that fault geometry and a westward migrating earthquake sequence on the Coulomb stress loading can signifi cantly aff ect the North Anatolian fault (NAF) during the 20th century initiation point of the next large earthquake on a fault (e.g., Barka et al. 2000, Figure 1a). Th e section of system. Furthermore, it has been noted that rupture the NAF within the Sea of Marmara remains as a end points usually coincide with discontinuities on seismic gap between the 1912 Saros and 1999 İzmit faults, such as stepovers (e.g., Segal & Pollard 1980; earthquake ruptures and the probability of a surface Sibson 1985). Th us, gaining insights into the western rupturing earthquake event is heightened for this extent of the 1999 İzmit earthquake rupture is region (e.g., Parsons 2004). Th e ~1500-km-long essential to estimate the magnitude of the expected dextral transform NAF is one of the major tectonic Marmara earthquake. Th e Hersek Peninsula is central structures of Anatolia, accommodating ~90% of to the debate on the western termination of the 1999 the deformation between the Eurasian Plate and surface rupture because it is the westernmost locality Anatolian Block (McClusky et al. 2000; Reilinger et where the NAF can be observed directly before it al. 2006). During the İzmit earthquake, four segments enters the Sea of Marmara (Figure 1b). Th is paper (Karadere, Sakarya, Sapanca, and Gölcük) of the aims to describe the geometry of the NAF on the NAF experienced surface rupture with right-lateral Hersek Peninsula and discusses its implications on displacements of up to fi ve metres. Th e ~126-km- the fault rupture of the 1999 İzmit earthquake. long surface rupture terminated near Gölyaka in In this study we employed a comprehensive, the east (Figure 1b), but the western termination of multi-technique approach on the Hersek Peninsula. the İzmit earthquake is more uncertain since it lies Specifi cally, we performed geomorphic analyses, off shore in İzmit Bay. According to some geodetic geological mapping, palaeoseismic trenching, models (i.e. Wright et al. 2001; Reilinger et al. 2000; geophysical surveying, modelling of deformation in Bürgmann et al. 2002) and seismicity analysis (i.e. half-elastic space with fi nite elements and Coulomb Pınar et al. 2001) it was suggested that the 1999 stress change modelling. We also combined our on surface rupture extended 10–30 km west of the land results with the existing off shore data in order Hersek Peninsula. Off shore studies within the Gulf to present a complete fault model for the Hersek of İzmit demonstrated the presence of underwater Peninsula. We then present a detailed discussion of fault scarps (Polonia et al. 2004; Cormier et al. the implications of fault geometry at our study area. 2006; Uçarkuş et al. 2008), but these were somewhat inconclusive in addressing the location of the 1999 rupture termination. Study Site Understanding where earthquake ruptures A Historical Background terminate has fundamental implications for Th e Hersek Peninsula is a triangular fan-delta with Probabilistic Seismic Hazard Analysis (PSHA) and an area of ~25 km2 in the Gulf of İzmit (Figures 1b & earthquake physics. Structural complexities along 2). Th e tip of the Hersek Peninsula extends ~5.5 km faults (i.e. asperities, stepovers, bends, and structural northward into the Gulf of İzmit creating the shortest junctions) may arrest rupture propagation and cause distance (~2.7 km) between the northern and perturbation of the state of stress on adjacent fault southern shores. Th e location and physiography of segments. Th e fi rst and most vital step is documenting the Hersek Peninsula not only allows for a shortened the characteristics (i.e. hypocentre, extent, geometry, gulf crossing but also controls the entrance to the and slip distribution) of individual ruptures. gulf and the route to İzmit (Nicomedia) and İznik Documenting earthquake rupture endpoints and (Nicaea) while providing a suitable landfall area with understanding what caused a rupture to terminate at its beaches and delta plain. Consequently it has been

360 Ö. KOZACI ET AL. 0 40 45’ 0 0 31 00’ 31 00’ 999 İzmit 999 İzmit 0 Erzincan

40 0 10 km e segment e

Location map of the study the study of Location map Karader (b) Suşehri

1939 SEA BLACK Niksar 0

0 1942 0 30 30’ 36 30 30’ Adapazarı

Sakarya segment 1943 Ilgaz Lake Sapanca

0 1944 32 0 İzmit 0 30 00’ 30 00’ Bolu

Sapanca segment 2000). Dashed lines in the Gulf of İzmit are our interpretation of the fault geometry based on Kuşçu geometry Kuşçu based on the fault of interpretation our are İzmit of 2000). Dashed lines in the Gulf 1957 1999b et al.

1967 picenter Mw 7.4 17 August 1999 e 1999a Gölcük segment İstanbul 0 Karamürsel Gulf of İzmit 28 ed map of the North Anatolian fault and westward migrating earthquakes since 1939 (from Barka 1999). since 1939 (from earthquakes migrating westward and fault Anatolian the North of ed map 0 0 (2002) bathymetry. (2002) bathymetry. 29 30’ Simplifi 29 30’ Hersek

E et al. area (dashed square) on LANDSAT image. Star symbol shows the epicentre of the 17 August 1999 earthquake. White bold lines are 1 bold lines are White 1999 earthquake. the 17 August of the epicentre shows symbol Star image. LANDSAT on (dashed square) area Lettis (from surfaceearthquake rupture

S 1912

40 W W 0 b a Figure 1. Figure (a) 0 40 45’

361 NORTH ANATOLIAN FAULT ON THE HERSEK PENINSULA

occupied for centuries as a strategic location in the era), baths, a cistern, and the Hersekzade Ahmed Gulf of İzmit (Supplementary fi gure 1). Paşa Mosque can still be readily observed in the Th e settlement on the Hersek Peninsula has vicinity of Hersek Village. Th e Hersekzade Ahmed been known as various names by diff erent cultures Paşa Mosque experienced extensive damage only throughout history. It was known as Drepanon until one year aft er its construction during the great 1509 318 A.D. when Byzantine emperor Constantine earthquake. It experienced less extensive damage in renamed Drepanon as Helenopolis aft er his mother other large earthquakes aff ecting the region including who was born there. By 1087, the name Cibotos the 1999 İzmit earthquake. and/or Civetot were used by Europeans. However, with the eff ects of repetitive earthquakes and Geology/Geomorphology of the Study Site battles Helenopolis was, sometimes, called ‘Eleinou Polis’ meaning ‘the wretched town’ (Th e Catholic Th e Hersek Peninsula has four main geologic/ Encyclopaedia 1910). Later in the 16th century geomorphic units; (1) delta plain deposits, (2) marine during the Ottoman Empire it was called Hersek terrace deposits, (3) beach ridge deposits, and (4) aft er Hersekzade Ahmed Paşa. Today, it is still called lagoon deposits (Figure 3). Hersek Village. Th e oldest deltaic unit is the Upper Pleistocene Th e settlement on the Hersek Peninsula has Altınova formation (Chaput 1957; Akartuna 1968; undergone three major construction phases during Sakınç & Bargu 1989), which includes sand with history. Th e fi rst major construction took place aft er widespread Ostrea shells, clayey sand, silty sand, Constantine renamed Drepanon as Helenopolis. marl and sandy marl, and uncomformably overlies Constantine stayed in Helenopolis on the way the Yalakdere and Taşköprü sandstone. Dedeler Hill, back to İstanbul (Constantinople) from the Yalova 28 m a.s.l. (above sea level), is the most prominent thermal baths, especially during his last years. Aft er geomorphic feature on the peninsula. Uplift ed Constantine, especially during Justinian’s time, marine terraces on its fl anks indicate it is an area Helenopolis gained more importance when the of active uplift . Dedeler hill is a NE–SW-trending gulf crossing traffi c was shift ed between here and ridge, bounded by a steep scarp on its south-eastern Dakibyza (Gebze). Justinian rebuilt Helenopolis by fl ank (Figure 2) and more gentle slopes on its north- adding an aqueduct, a second public bath (a rare western fl ank. situation for the time), churches, a palace and other Th e delta is ~2–3 m a.s.l. and constitutes most of buildings (Supplementary fi gure 2). He also cleared the Hersek Peninsula (Kozacı 2002) (Figure 2). It is the entrance of the Drakon River (currently known formed by the north-fl owing Yalakdere River. Th e as Yalakdere), built bridges and widened the road headwaters of Yalakdere in the Samanlı Mountains to Nicaea (İznik). During this period the Drakon are ~480 m a.s.l. and ~17 km south of the Hersek River valley was used as the route connecting Peninsula. Recent deposition occurs in the northwest Constantinople (İstanbul), Helenopolis (Hersek) and portion of the delta (Figure 2). th Nicaea (İznik). Later, in the 16 century, Hersekzade Th e youngest marine terraces are composed of Ahmed Paşa built a small harbour, 700 houses, a marine sand with loose fabric and coarse Gastropod mosque with two minarets named aft er him, two packages, which in some places uncomfortably inns, and a care house for the poor and a school of overlie the Altınova formation. Th ey are exposed Islamic theology. approximately 400–500 m inland near Hersek Many great earthquakes (Supplementary table 1) Village at an average elevation of about 1–2 m a.s.l. as well as battles throughout history aff ected the study (Figures 2 & 3). Th e middle and youngest marine site. During the palaeoseismic excavations by Witter terrace deposits overlie the oldest marine terrace et al. (2000) following the 1999 İzmit earthquake deposits with angular unconformity. Although all two destruction horizons were identifi ed within the marine terrace deposits have a similar lithology they trenches. In addition, many graves and bones were can be easily diff erentiated on aerial photographs by recovered. Th e remnants of an aqueduct (Justinian their elevation diff erence. Th e oldest marine terrace

362 Ö. KOZACI ET AL.

Figure 2. Map showing the vicinity of the study area. Coloured contours are extracted from the 20X exaggerated digital elevation model (DEM) and overlaid on the aerial photo. Colour-coded contour intervals represent 5-m elevation changes. Note that Dedeler Hill has a NE–SW-trending elongated shape located at the north of the peninsula with an elevation of 28 m (a.s.l.). Th e delta morphology with its active and passive lobes became easily recognized as a result of using 1/1000 scale survey data. Trench locations are shown as yellow lines (T4, T5, T6…). Seismic refl ection profi le location is shown as white bold line (SRP). Very Low Frequency-Electromagnetic profi le locations are shown as a white box (VLF). Previous palaeoseismic study site by Witter et al. (2000) is shown as a yellow box (1999). Dashed white box shows the area of Figure 3 and yellow box north of Hersek Lagoon shows the location of Figure 4. Th e DEM was created using 1/1000 scale topographic survey of T.C. İller Bankası.

363 NORTH ANATOLIAN FAULT ON THE HERSEK PENINSULA

deposits, about 5–6 metres thick and 10–15 m a.s.l, across a south-facing scarp forming the southern fl ank represent the shore facies with sand lenses and local of Dedeler Hill and the shore of the lagoon (Figures Ostrea rich zones. 2 & 4). Trench T-10 was excavated as a series of short Beach ridges of well-rounded pebbly sands trenches down the southern fl ank of Dedeler Hill are well exposed west of Hersek Village. Modern (Figure 4). It exposed a marine terrace that abruptly rounded pebbly beach sand is well exposed on both thickened and a drop of the abrasion platform, most probably indicative of fault deformation. Strands of the east and west shores of the Hersek Peninsula. the North Anatolian fault and related deformation Modern basin deposits and tidal marsh is composed were exposed in Trenches T-12, T-14, and T-16. of sandy silts and can be observed around Hersek Trench T-15 was excavated perpendicular to T-16 Lagoon (Figure 3). and parallel to the NAF (Figure 4), and exposed secondary strands of the NAF at this locality. Th ere Palaeoseismic Trenching was no compelling evidence of deformation within trench T-17. Following the 17 August 1999 İzmit earthquake, Witter et al. (2000) excavated several palaeoseismic trenches ~250 m northwest of the Hersek Lagoon Trench T-12 (Figure 2) in an eff ort to document the rupture Trench T-12 was excavated across a N70°E-trending history of the North Anatolian Fault on the Hersek dilatational crack that was formed during the 17 Peninsula. However, this trench site unearthed August 1999 İzmit earthquake in south of Dedeler remnants of an ancient settlement (Witter et al. Hill (Figures 4 & 5a, b). Trench T-12 is 16 metres long, 2000). Walls, foundations, clay water pipes, graves, 1.5 metres wide and 2.5 metres deep, and exposes bone fragments, and evidence of destruction were the North Anatolian fault at station eight. Th e fault documented during these excavations and the site strikes N70°E with a near vertical-dip and extends was abandoned. to the surface (Figures 5b, c & 6). South-dipping During the summer of 2000, we performed (30°), shell-rich units south of the fault and massive additional palaeoseismic trenching in two diff erent clay with sand and gravel are juxtaposed along the locations on the Hersek Peninsula (Figure 2). Th e main fault. Secondary deformation is expressed as a fi rst set of trenches was located across the tonal N65°W-trending near-vertical fi ssure at station two. and vegetation lineaments that were mapped on Th e tilting of the units south of the fault indicates the delta plain as a result of our aerial photography north-side-up deformation. interpretations. We excavated six, approximately north–south-oriented slot trenches (T-4, T-5, T-6, Trench T-14 T-7, T-8, and T-9) on the delta plain west of the Witter et al. (2000) site (Figure 2). Th e total length Trench T-14, 22 metres long, 1.5 metres wide, and approximately 2 metres deep, was excavated east of of these 1.5-m-wide trenches is ~604 m, with depths trench T-12 (Figure 4). Th e fault zone is exposed ranging between 1 to 2.2 metres, depending on between stations six and seven with an orientation ground water conditions and trench wall stability. Th e of N70°E. Marine terrace deposits and fl uvial trenches located on the delta plain exposed laterally units are juxtaposed along the fault zone (Figure continuous and undeformed strata consisting of 7a). Units south of the fault zone dip gently to the predominantly marine sand overlying silty sand, south consistent with T-12 stratigraphy (Figure 7b). sand, and clay of deltaic and lagoonal origin, but no Radiocarbon samples T14-6, T14-9, and T14-14 faults were exposed. Nevertheless, these trenches yielded calibrated (2-sigma) ages of 2215 (+133,-65) provide a spatial constraint for the fault locations on ybp (years before present), 1562 (+129,-39) ybp, and the delta plain. 3785 (+174,-93) ybp, respectively. Th ese ages indicate Th e second set of palaeoseismic trenches (T-10, faults in the trench have experienced recurrent late T-12, T-14, T-15, T-16 and T-17) were excavated Holocene ruptures.

364 Ö. KOZACI ET AL.

N Qhb W E

Qhpk S

Qhb

Qmt1 Gulf of İ zmit

Qmt 3 Qhpk

Qmt1

Qpu Qhp Qmt2 T10 Qhpk T10 Qmt2

Qmt1

Hersek Lagoon t Faul Qhb Anatolian North Qha

0 1 km

Explanations Qmt middle marine terrace Qhpk modern beach sand 2 oldest marine terrace Qhb modern basin and tidal marsh Qmt3 Qhp Holocene beach ridge deposits Qpu Pleistocene Altıf nova ormation Qha Holocene alluvium terrace riser young marine terrace drainage system Qmt1 Figure 3. Geomorphic and geologic map of the Hersek Peninsula.

Trench T-16 N

W E T-14 Trench T-16, 27 m long, 1.5 m wide, with its deepest S T-17 section reaching 2.2 metres in depth, is located T-16 between trenches T-12 and T-14 (Figure 4). Th e 80E fault zone was observed between stations zero N6 and eight (Figure 8a, b). A vertical fault juxtaposes T-12a horizontal units in the south against north dipping T-15 units in the north at station 0.5. Th e main fault zone, Hersek Lagoon however, is oriented ~N65°E and exposed between stations fi ve and eight. Th is north-dipping reverse fractures T-12b fault is accompanied with almost vertical antithetic

0 30 m deformation around station seven. Furthermore, the stratigraphic units north of the main fault zone are Figure 4. Detailed map showing trench locations (T12, T14, folded and uplift ed as a result of transpression in this T15, T16, and T17) and mapped fault traces on the area. Radiocarbon samples T16-1, T16-2, and T16- south-facing scarp of Dedeler Hill. 11 yielded calibrated (2-sigma) ages of 6662 (+117,-

365 NORTH ANATOLIAN FAULT ON THE HERSEK PENINSULA

Fault: N70 °E

e 5c

Figur

c b

Figure 5. (a) A fracture formed during the August 1999 earthquake. (b) Two diff erent units are juxtaposed on both sides of the North Anatolian fault in Trench T-12. (c) Close-up view of the fault on the western wall of the trench.

164) ybp, 6131 (+146,-139) ybp, and 5922 (+68,- Th ese results suggest that units north of the fault zone 152) ybp, respectively. Th e ~6.6 ka-old T16-1 was did not override the units south of the fault as a result recovered from marine terrace deposits (Unit K). Th e of reverse faulting until ~5.9 ka years before present. ~6.1 ka-old T16-2A, however, was recovered from a Th e trench exposures provided direct large anthropologic excavation (Unit L) cutting and confi rmation of the location of fault strands of the postdating units F, I, K, and J. Th ese dates suggest NAF and demonstrated that the style of deformation that the marine terrace deposits emerged above sea (right-lateral with a considerable north-side-up level some time between ~6.6 and 6.1 ka years before reverse component) is consistent with the long-term present. Th e units north of the fault zone are older style of deformation produced from repeated surface than the buried soil horizon (Unit D) south of the rupturing earthquakes refl ected in the uplift and tilt fault zone, where Sample T16-11 was recovered. of Dedeler Hill.

366 Ö. KOZACI ET AL.

T-12a S N 0 m

A F F A root carbonate lining B contact ? F ? F F C carbonate D lining 2 D contact F E º fissure N65 W, sub-vertical FAULT ZONE N70º E, sub-vertical 0 2 4 6 8 10 12 14 16 m

A very coarse shell hash; minor sand minor amount of recrystalized fibrous material, weakly cemented; localized alterations

B medium coarse shell hash to shell rich zone; upper 20 cm of unit contains fewer shell fragments and greater clay content C medium coarse shell to shell rich zone; upper 30-40 cm of zone contains very few shells, fine sand to silty sand, weakly cemented D medium coarse shell hash and some sand, fining upwards, fine sand to silty sand matrix supported E thin lens of fine sand; no shell fragments F clay (bedrock) interbedded with sand, gravel, cobbles; clay is massive, mottled; sands range from fine to well sorted and well

Figure 6. Log of trench T-12 (western wall).

Geophysical Surveys plots and demonstrate a structural anomaly between Seismic refl ection and Very Low Frequency – Electro metres 50 and 70, in agreement with the observed Magnetometer (VLF-EM) surveys were performed deformation on the seismic refl ection profi le (Figure on the delta plain in order to locate the westward 10). continuation of the North Anatolian Fault (Figure 2). Th e north–south-oriented, 650-m-long seismic Model refl ection profi le is located ~600 m west of the lagoon Combination of our palaeoseismic investigations on (see Figure 2). A sledge hammer was used as the energy Hersek Peninsula and off shore geophysical surveys source. A 12-channel recording system was used with (Kuşçu et al. 2002 and Cormier et al. 2006) revealed fi ve-metre geophone spacing. Interpretation of the a left -stepping geometry for the North Anatolian low-fold stacked profi le indicates the presence of a Fault (Figure 11). As a further test, we utilized fi nite discontinuity 200 metres north of the southern end element modelling in half elastic space for comparing of the seismic profi le (Figure 9). the resultant deformation of this fault geometry VLF-EM surveys, which have been successfully with the present day geomorphology (Figure 12). In used for non-mineralized shallow fault zone addition, a simple Coulomb model (Figure 13) was investigations (e.g., Jeng et al. 2004), were focused employed to provide a plausible physical explanation on the area of deformation in seismic refl ection data on how this restraining stepover might have aff ected (Figure 2). Four parallel, 90-metre-long profi les were the 1999 rupture propagation. performed fi ve metres apart in order to confi rm this deformation both laterally and vertically. Data were collected using an ENVI Scintrex VLF instrument Finite Element Modelling in Half Elastic Space with 2 metre intervals. Th e in-phase (IP), out- We tested the fault geometry documented during of-phase (OP), and TILT values were measured our fi eld studies by using fi nite element modelling in simultaneously in three diff erent frequencies between half-elastic space (Figure 12). Coulomb 2.0 (King et 15 kHz to 30 kHz (16.0, 23.4, and 26.8 kHz). All al. 1994 and Toda et al. 1998) was used to correlate measurements were stacked into three dimensional the modelled deformation patterns of various fault

367 NORTH ANATOLIAN FAULT ON THE HERSEK PENINSULA

Trench T-14 S 14-T14/9 N C 14-T14/14 C14-T14/6 C A1

0 m A2 A2 E F F B H H C F F D G D 2 a 0 2 4 6 8 10 12 14 16 18 20 22 m

A1 Ahorizon, similar toA2 but includes some silt b A2 cl,olay with sand and grave rganic rich

B sl,mandy clay to clayey sand with grave any gravel clasts with diameters up to 3 cm;f ew tile fragment s, weakly disseminated carbonate concentrated along roots and pores

C similar to unit B but no carbonate concentration; fewer and smaller tile fragments;a mount of sand increases towards the bottom of the unit

D terrace deposits composed of gravelly sand to locally clayey sand;u pper contact is formed by well-rounded cobbles with diameters up to 8 cm;c ommon shell fragments

E palaeo-soil;s andy gravel with rounded clasts up to 1 cm in diameter,c ommon shell fragments ,w eakly disseminated carbonate along the roots and pores F marine terrace deposits composed of interbedded medium to very coarse sand and fine gravel lenses;c ommon to many shell fragments concentrated along beds

G clay, gravely clay, gravely sand and clay interbedding

H palaeo-trenches by human activity

charcoal samples C14-T14/6 CC14-T14/9 14-T14/14

Figure 7. (a) Photo showing two diff erent units (white and black arrow heads) juxtaposing both sides of the fault (red arrow). (b) Log of trench T-14 (western wall). geometries with the current morphology of the same fault parameters were applied: 1 m of dextral Hersek Peninsula. We ran diff erent models with and 0.3 m vertical slip for the segments to the east of various plausible fault geometries (such as right the peninsula and on the peninsula. Th ese values are stepping, left stepping, overlapping, no overlap) both compatible with the InSAR inversions (please see determined by onshore and off shore studies (see discussions for details) and a potential segmentation online data repository for results). In all models the boundary-type deformation. Assuming that the fault

368

Ö. KOZACI ET AL.

2 0 m 0 N 26 m 14-T16/1 tile fragments C A 24 K 22 L 14-T16/2 C J 1 20 A K K F 2 I shell fragments 18 1 I 16 A 2 J 14-T16/11 C 14 F 14-T16/2 C 12 Trench T-16 14-T16/1 C A 0E 10 H v,f ell sorted fine sand riable ell bedded ery coarse, very poorly sorted shell hash ockets of sand (fine-coarse) ostly oyster shells and some gastropods ilty fine sand ell sorted andy silt-silty sand ontains few shell fragments and more commonly charcoal fragments ense/stiff clay with few pockets of sand and sandy clay assive ine to medium sand with fine rounded gravels oorly sorted common to variable few to common shell fragments Ahorizon; modern soil silty sand with gravel sized shell fragments s,w s,c E- ery coarse, very poorly1- sorted very shell friable, debris same organics, >90 shell fragments 2- ragment more size indurated, varies matrix from supported, medium-coarse more sand dissemenated size carbonate than3- E more dissemenated carbonate than E v,p,m poorly sorted fine to coarse sand with few gravels and common shell fragments andy clay with few-manyouth shells end derived from ommon shells tile fragments in F ense and rades stiff diffusely upward into ighly modern variable A in horizon clast and shell arbonate content nodules near base charcoal samples F w,f,w f,p silty fine sand with few small shell fragments d,m interbedded silty fine sand with large shell fragments, moderately indurated with loose friable shell fragments Fault: N65 8 s,c,d,h s,g,c, G 2 1 3 1 2 I J L I K F E B H D E E G A C A photo showing trenches T-15 and T-16. White dashed bold line shows the trend of the North Anatolian fault. Anatolian the North of the trend dashed bold line shows White T-16. and T-15 trenches showing A photo A b) 6 2 E Fault Zone 3 E 1 E 14-T16/11 4 C T-15 A D 2 B Log of trench T-16 (western wall). ( (western T-16 trench Log of 0 B S C a b Figure 8. Figure (a)

369 NORTH ANATOLIAN FAULT ON THE HERSEK PENINSULA

Figure 9. Seismic refl ection profi le and interpretation (see Figure 2 for location).

steps or overlaps, both segments are considered to with the steeper slope to the south adjacent to be dipping north at 84°. Th e depth of seismogenic the fault. South of the fault, however, basin fi ll zone is taken to be 15 km. Of these models one stratigraphy can be observed on that subsided side fault geometry provided a good match with fault (Figure 12b). topography at Dedeler Hill and a selected off shore Th e best fi t fault geometry model, with left - profi le (see Figure 12a, b, respectively). stepping faults that do not overlap, has the highest Cross-section locations on the deformation correlation between both the onshore topography models are targeted for optimal correlation with and the off shore seismic profi le (Figure 12, see real topography (Figure 12). Th e NW–SE-oriented supplementary data for other fault models). In this western profi le on the Hersek Peninsula is normal to model the eastern segment (off shore) does not the NE–SW orientation of Dedeler Hill. Th e north- extend to the peninsula. Th is model does not favour western fl ank of the Dedeler Hill is a gentle slope an overlap or right-stepping geometry between these whereas its southeast slope is steep and abruptly two fault segments. Th e deformation obtained in this abuts the lagoon along the fault scarp (Figure 12a). model successfully correlates with the pressure ridge However, the approximately N–S-oriented eastern located north of the delta and the depression area of profi le corresponds to an off shore seismic profi le the Hersek Lagoon (Figure12a, c, d). Correlation of location (from Kuşçu et al. 2002). In this profi le these geomorphic features is not only limited to their bedding north of the fault is folded asymmetrically locations, but the amount of vertical deformation

370 Ö. KOZACI ET AL. 3.0 2.0 1.0 0.0 -1.0 -2.0 -3.0 -4.0 -5.0 4.5 3.5 2.5 1.5 0.5 -0.5 -1.5 -2.5 5.5 4.5 3.5 2.5 1.5 0.5 -0.5 -1.5 -2.5 le line, le. Black arrows arrows Black le. approx. depth 0-5 m approx. depth 5-10 m approx. depth 10- 15 m ection profi TILT 3-D (16.0 kHz - 23.4 kHz - 26.8 kHz)

cant change in the electrical conductivity can be can electrical in the conductivity change cant

/%Hp Hs % %Hs/%Hp %Hs/%Hp the profi is along the 3-D diagram e x-axis of 1.0 0.5 0.0 -0.5 -1.0 -1.5 -2.0 -2.5 -3.0 2.0 1.5 1.0 0.5 0.0 -0.5 -1.0 -1.5 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 -0.5 -1.0 le results are combined and evaluated as a block diagram. diagram. as a block evaluated and combined are le results to right) for 16 kHz, 23.4 kHz and 26.8 kHz frequencies representing representing frequencies 26.8 kHz and 23.4 kHz 16 kHz, for right) to approx. depth 0-5 m profi e four approx. depth 5-10 m approx. depth 10- 15 m les. Note that in these diagrams a signifi in these diagrams that les. Note %OP3-D(16.0kHz-23.4kHz-26.8kHz)

les, which are on the same trend as the anomaly on the seismic refl on as the anomaly trend the same on les, which are

%Hs/%Hp %Hs/%Hp Hp %Hs/% 5.0 3.0 1.0 -1.0 -3.0 -5.0 -7.0 -9.0 8.0 6.0 4.0 2.0 0.0 -2.0 -4.0 10.0 8.0 6.0 4.0 2.0 0.0 -2.0 -4.0 approx. depth 0-5 m approx. depth 5-10 m approx. depth 10- 15 m 3-D %IP (In Phase), %OP (Out of Phase) and TILT values are plotted (left plotted are values TILT Phase) and of (Out Phase), %OP 3-D %IP (In Th respectively. bottom) to intervals (top depth 10–15 metre 0–5, 5–10 and approximately VLF-EM measurement results are shown (see Figure 2 for location). Th location). 2 for (see Figure shown are results measurement VLF-EM where the y-axis shows the width of the four parallel profi parallel the four of the width the y-axis shows where the profi in the middle of observed approximately indicate fault location. fault indicate

% IP 3-D (16.0 kHz - 23.4 kHz - 26.8 kHz)

%Hs/%Hp %Hs/%Hp %Hs/%Hp Figure 10. Figure

371 NORTH ANATOLIAN FAULT ON THE HERSEK PENINSULA

29º30É 29º 4 0É 29º 5 0É N s-trike slip fault normal fault W E HEREKE

S İZMİT DARICA 40º45´N DERİNCE 40°45´N Gulf of İzmit

GÖLCÜK

DEĞİRMENDERE 40°42´N

2 HERSEK PENINSULA

40°4 ´N 0 2 4 km KARAMÜRSEL

Figure 11. Interpretation of fault geometries (from Kozacı 2002) based on detailed bathymetry data by Kuşçu et al. (2002). Th ere is a 2-km-wide stepover west of Gölcük. Note that there are two ~1-km-wide, rhomboidal releasing stepovers between Değirmendere and the Hersek Peninsula. Th e North Anatolian fault makes a restraining stepover just east of the Hersek Peninsula and continues westwards with a trend of N70°E.

ratios also shows similarities. Furthermore, the Hersek onshore segment (Figure 13). Th e modelled observed folding geometry on the seismic profi les, stress shadow caused by a rupture on the NAF east run in the east of the Hersek peninsula by Kuşçu et of the Hersek Peninsula reaches –0.2 bars and creates al. (2002), can be correlated very accurately with the an unfavourable Coulomb failure condition for the modelled deformation pattern (Figures 12b, f). receiver fault on the Hersek Peninsula.

Coulomb Model Results and Discussion Stein et al. (1997) demonstrated that calculation of As presented above, the Hersek Delta is a fl at plain static stress changes resulting from large earthquakes ~2–3 m a.s.l. and the NE–SW-oriented, oval-shaped can help identify how earthquakes on adjacent fault Dedeler Hill is a prominent topographic high (28 segments interact. Th eir study of the 20th century m a.s.l.) at the tip of the peninsula (Figure 2). Th us, earthquakes along the NAF showed that nine out it is necessary to discuss its presence because the of ten epicentres struck within the area of increased following observations suggest that Dedeler Hill was stress (2–4 bars) caused by the preceding earthquake. raised tectonically. (1) Th e NE–SW-trending Dedeler Most segments within decreased stress changes (–0.1 Hill is an asymmetric high; its NW fl ank is gentle to –0.6 bars), however, did not experience rupture. while its SE fl ank is steep (Figure 2); (2) Detailed Coulomb stress change calculations depend on geomorphologic mapping revealed marine terraces fault geometry, slip, and the coeffi cient of friction of a nestled around the hill at diff erent elevations above source fault, and on the fault orientation of a receiver sea level (Figure 3); (3) Trenching on the south-facing fault. We used the best fi t fault model parameters scarp exposed tilted (Figure 6) and folded units along (location, geometry, and physical fault parameters) a major fault (Figure 8), and (4) the presence of the from a fi nite element model to calculate a simple Hersek Lagoon is evidence of subsidence south of Coulomb static stress change. Th e coeffi cient of Dedeler Hill (Figure 2). Based on these observations, friction value was set at 0.4, as suggested by Stein et it can be concluded that the Dedeler Hill has been al. (1997). A model rupture with a 1 m dextral and actively uplift ing in this part of the Hersek Delta and 0.3 m vertical slip on a source fault corresponding to its existence provides key evidence for the style of the East–West-oriented off shore segment east of the tectonic deformation in the study area. peninsula generates a stress shadow area in the Hersek Off shore bathymetry and seismic data showed that Peninsula area for faults oriented ~N70°E, like the the NAF trends nearly east–west east of the Hersek

372 Ö. KOZACI ET AL.

Figure 12. Modelling of deformation in the vicinity of the Hersek Peninsula using fi nite elements in elastic half-space. Th e two dextral faults (left -stepping faults with no overlap) used for this model are located according to the fi ndings of onshore (Receiver Fault - RF) and off shore (Source Fault – SF) studies. (a) Topography of Dedeler Hill on the peninsula is correlated with the modelled deformation. Plan view of the grid model showing the fault geometry and the NW–SE (A–A’) cross-section is compared to the morphology of the Hersek Peninsula. Th e NW–SE-oriented cross-section (A–A’) across Dedeler Hill presents a NW facing gentle slope and a SE facing scarp. (b) An off shore seismic profi le (Kuşçu et al. 2002) east of the peninsula is correlated with the modelled deformation. Th e N–S seismic profi le (B–B’) displays folded sediments with a sudden dip towards the fault (south). Plan view of the grid model showing the fault geometry and the N–S (B–B’) cross- section is correlated with the off shore seismic profi le east of the peninsula. Note that many geometries were run but only the displayed geometry provides an acceptable fi t to the observed morphology.

373 NORTH ANATOLIAN FAULT ON THE HERSEK PENINSULA

Th e Western Extent of the 1999 İzmit Surface Rupture Geodetic measurements and models suggest that the rupture of the 1999 İzmit earthquake propagated 10–30 km west of the Hersek Peninsula, with displacements of up to 2 metres at depth. Feigl (2002) correlated the geodetic and seismic moment SF estimates and gave a detailed list of potential causes RF for the discrepancies between fi eld and geophysical observations. A lengthy discussion of these causes was provided by Feigl et al. (2002) specifi cally for the İzmit earthquake. We would like to emphasize that the discrepancy between the geodetic and geologic techniques may bars very well be within the uncertainties of geodetic measurement techniques and modelling parameters. -0.2 -0.1 -0.02 0.02 0.1 0.2 Th e discrepancy between the geodetic studies and our observations may be the product of a few reasons: (1) methodological uncertainties, (2) some artefact of Figure 13. Coulomb stress change model for the faults in the geodynamic models used to model deformation with vicinity of the Hersek Peninsula. A source fault (SF), geodetic data, (3) comparison of observations from corresponding to the off shore fault in the east of the Hersek Peninsula with an east–west trend, and a diff erent depths, and (4) the length of observation receiver fault (RF– the fault on which the Coulomb period. Th e fi rst reason for the apparent discrepancy failure is calculated) in the west corresponding to between the fi eld observations and the geodetic the N70°E-trending onshore segment, were placed models is demonstrated by Feigl et al. (2002) who according to the fi eld observations and best-fi t report that the uncertainties of inferred slip from deformation model. Th e source fault was divided into patches and the slip on this fault was assigned geodetic models for the 1999 İzmit rupture could be to be 2 m in the middle and decreasing towards the as high as 1 metre. Also, using smooth versus stepping ends to 1 m. In addition, a 0.3 m vertical slip was put fault geometry and, more importantly, using the most in as a fault parameter. Note that the receiver fault accurate fault geometry, are some of the important in the west that represents the fault on the Hersek parameters used in geodynamic models with a direct Peninsula falls in the stress shadow area. eff ect on the modelling results (Feigl et al. 2002). In addition, Hearn & Bürgmann (2005) demonstrated Peninsula (Figures 11 & 12). Our palaeoseismic that using depth-dependent versus uniform trenches on the south-eastern fl ank of Dedeler Hill geodynamic models may aff ect the seismic moment provide evidence for major strike-slip faulting with calculations up to a factor of three. Th irdly, our fi eld a north-side up reverse component (Figures 5–8). observations are limited to only very shallow depths Although other trenches on the delta plain did (< 40 m), in contrast to the geodetic measurements. not provide any evidence for faulting, geophysical And lastly, the extent of an earthquake rupture may continue to evolve aft er days, weeks or even years surveys confi rm the westward extension of this fault following the main shock. Feigl et al. (2002) pointed at depth (Figures 9 & 10). Palaeoseismic trenches, in out that the GPS data and interferograms record combination with the geophysical survey, indicate 75 and 30 days of deformation and may contribute that the NAF trends N70°E in the study area (Figures to the uncertainties up to 10%. Hence, some of the 3 & 4). Both fi eld observations and existing data deformation inferred from geodetic measurements suggest that Dedeler Hill has been rising as a result of could be the result of post-seismic deformation a restraining bend on this part of the North Anatolian beyond rupture termination that was included within Fault (Figures 11 & 12). the measurement over an extended interval.

374 Ö. KOZACI ET AL.

Our fi eld observations agree with Barka et What Stopped the 1999 Surface Rupture at Hersek al. (2000) that the Hersek Peninsula did not Peninsula? experience any surface rupture during the 1999 Our fi eld observations (the geomorphology of İzmit earthquake. It is possible that the saturated and Dedeler Hill, uplift ed marine terraces, depression unconsolidated delta sediments may have masked of Hersek Lagoon, multiple exposures of the NAF minor slip hence making surface rupture recognition within trenches on the south facing scarp of Dedeler diffi cult. Alternatively, the surface rupture may have Hill, and presence of a deformation zone within ‘skipped’ the Hersek Peninsula. However, we fi nd geophysical profi les) and fi eld evidence-based simple these alternatives highly unlikely, based on our fi eld models demonstrate that the stepover of the NAF on observations. In palaeoseismic trench exposures the Hersek Peninsula creates a restraining bend. Our north of the lagoon, it is evident that the NAF has investigations suggest that this was effi cient enough to ruptured to the surface on the Hersek Peninsula terminate the 1999 rupture propagation. Contrary to during the late to middle Holocene. Th erefore, there the geodetic models, the absence of a surface rupture is no mechanical reason why the rupture should by- beyond this stepover makes a compelling argument pass only the on-land location between the off -shore against extended rupture. Th e eff ects of geometrical basins that are suggested to have evidence for 1999 fault discontinuities are discussed in various studies surface rupture. Although submarine investigations as a candidate mechanism for stopping earthquake documented surface deformation west of the Hersek rupture propagation (e.g., Sibson 1985; Kadinsky- Peninsula (Uçarkuş et al. 2008), there are no piercing Cade & Barka 1989; Harris & Day 1999; Harris et al. features that would suggest consistent dextral 2002; Lettis et al. 2002). Many factors may aff ect this displacement. We suggest that the observed ‘fresh process: (1) the type of step (releasing or restraining); looking’ surface deformation on the Yalova fault (2) the width of this step; (3) the presence of transfer segment to the west of Hersek Peninsula, most likely faults within the basin (Harris & Day 1993; Oglesby represents secondary sympathetic deformation. 2005); (4) the amount of remaining energy for the During our fi eld observations following the 1999 continuation of the rupture propagation; (5) the İzmit earthquake we observed similar secondary amount of accumulated slip on the neighbouring segment in the rupture direction, and (6) the deformation on the Taşköprü Delta to the west. direction of the source directivity. Although this deformation was parallel to the general strike of the North Anatolian fault in this Kozacı (2002) re-interpreted the fault geometry area, the associated lateral displacements at the between Gölcük and the Hersek Peninsula, based on surface were in the order of a few centimetres only. the very detailed bathymetry study results published Th is kind of secondary soft sediment deformation by Kuşçu et al. (2002). Unlike previous studies that can be explained by strong ground shaking induced had suggested a ~5-km-wide releasing Karamürsel slope failure or lateral spreading, depending on the stepover (Barka 1999; Lettis et al. 2000) or a single morphologic location. east–west-trending Karamürsel segment (Harris et al. 2002) based on previous bathymetry data, we As a result, our studies on the Hersek Peninsula propose that there are two relatively narrow (~1 suggest that the surface rupture of the 1999 İzmit km wide) releasing stepovers between the Hersek earthquake did not extend west of the Hersek Peninsula and Değirmendere (Figure 11). As a result, Peninsula. It is likely that the ‘required’ slip at depth a signifi cant amount of energy should have been to the west of the peninsula is the result of triggered dissipated within this stretch of the fault. Th e location aft ershocks similar to the Düzce earthquake of these geometrical discontinuities coincides with (Reilinger et al. 2000; Langridge et al. 2002; Cormier et the sudden decrease in the slip amount anticipated al. 2006) and may potentially indicate the nucleation by the models of Feigl et al. (2002) and Çakır et al. point of the next earthquake or simply post-seismic (2003). However, perhaps these releasing stepover deformation. Th is section may also re-rupture during basins are not wide enough to completely arrest the the expected Marmara earthquake similar to the rupture. Moreover, although the restraining step on M7.2 1999 Düzce earthquake. the Hersek Peninsula is less than 1 km wide, the fault

375 NORTH ANATOLIAN FAULT ON THE HERSEK PENINSULA

segment on the peninsula remains within the stress Conclusions shadow area of the adjacent fault rupture to the east Our study demonstrates that the pressure ridge (Figure 13). Our two-dimensional coulomb model (Dedeler Hill) located in the middle of the İzmit may be used to explain why restraining stepovers, are Bay is a product of a restraining stepover at this equally, if not more, likely to arrest rupture than a location. In addition, it implies that the 1999 surface wider releasing stepover (Figure 13). rupture did not extend west of the Hersek Peninsula. Another cause for the rupture propagation to We suggest that this conclusion corresponds to the run out of energy might be the 1894 earthquake, location where the North Anatolian Fault begins to which occurred between the Hersek Peninsula and bifurcate into the Yalova and Çınarcık segments to the Çınarcık Basin (Eginitis 1895; Ambraseys 2001; the southwest and the Princes Islands segment to the Harris et al. 2002). Alternately, this geographical northwest. limitation itself, as suggested by Cormier et al. (2006), Although, the geodetic models suggest 2 m of suggests a structural discontinuity during a historic slip beneath the Hersek Peninsula at depth (i.e. 10– earthquake at the same location. 20 km) diminishing within 10–30 km to the west Lastly, the North Anatolian fault begins to (Reilinger et al. 2000; Wright et al. 2001; Bürgmann bifurcate into two branches (Princes Islands segment et al. 2002; Feigl et al. 2002; Çakır et al. 2003) the in the north and Çınarcık segment in the south) lack of surface rupture implies that the 1999 rupture at the Hersek Peninsula. Th is structural junction did not extend beyond Hersek Peninsula to the west. As a consequence we speculate that the next large on its own could be considered as a signifi cant earthquake in this area with a surface rupture will segmentation location capable of rupture arrest probably break the segment crossing the Hersek (Kame et al. 2003). Th e Yalova segment exposed Peninsula, as well as the faults to the west with within our palaeoseismic trenches along the scarp displacements similar to the 1999 earthquakes (3–5 north of the lagoon strikes ~N70°E and possibly m) at or near the surface. We therefore propose that is the continuation of the Çınarcık segment. Th e the restraining step over at the Hersek Peninsula structural connection between the Princes Islands presents an effi cient structural barrier for earthquake segment and Karamürsel segment, however, is not rupture propagation at shallow crustal levels. well developed. Cormier et al. (2006) bathymetry data showed a ~250-m-wide, 5-km-long en-échelon style deformation zone between the Hersek Peninsula Acknowledgments and where the Princes Islands segment becomes well We would like to thank PG&E and Lloyd Cluff for defi ned further west. In addition, two potential buried their support that made this study possible. We also faults were observed on the northern extent of our thank AK Kağıt A.S. for their generous hospitality seismic refl ection profi le west of Dedeler Hill. Th ese at their facilities in Yalova. Emre Evren and Uğur buried fault splays are potentially the continuation of Meray provided much appreciated help in the fi eld the en-échelon fault structure observed by Cormier and trenches. Çağlar Yalçıner performed the VLF et al. (2006) using bathymetry west of the Hersek measurements and processing. Ziyadin Çakır and Peninsula. Based on the orientation and the degree of Matt McMackin provided much appreciated input geomorphic expressions we suggest that a westward- during discussions. We would like to further thank propagating rupture on the Karamürsel segment reviewers Helen Cormier, Aurelia Hubert-Ferrari, would preferentially propagate on to the southern and two anonymous reviewers for their constructive (Yalova and then to Çınarcık) segments. reviews.

References

Akartuna, M. 1968. Armutlu Yarımadasının Jeolojisi [Geology Ambraseys, N.N. 2001. Th e earthquake of 10 July 1894 in the Gulf of Armutlu Peninsula]. İ.Ü.F.F. Monografi leri (Tabii İlimler İzmit (Turkey) and its relation to the earthquake of 19 August Kısmı), 20. 1999. Journal of Seismology 5, 117–128.

376 Ö. KOZACI ET AL.

Ambraseys, N.N. & Finkel, C.F. 1991. Long-term seismicity of Feigl, K., Sarti, F., Vadon, H., McClusky, S., Ergİntav, S., İstanbul and of the Marmara Sea region. Terra Nova 3, 527– Durand, P., Bürgmann, R., Rigo, A., Massonnet, D. & 539. Reilinger, R. 2002. Estimating slip distribution for the Ambraseys, N.N. & Fınkel, C.F. 1995. Seismicity of Turkey and İzmit mainshock from coseismic GPS, ERS-1, RADARSAT, Adjacent Areas, A historical Review,1500–1800. Eren Yayıncılık and SPOT measurements. Bulletin of Seismological Society of ve Kitapçılık Ltd. America 92, 138–160. Anna Comnena (Komnene). Th e Alexiad. Edited and translated by Harris, R.A. & Day, S.M. 1993. Dynamics of fault interaction: Elizabeth A. Dawes. London: Routledge, Kegan, Paul, 1928. parallel strike-slip faults. Journal of Geophysical Research 98, 4461–4472. Barka, A.A. 1999. Th e 17 August 1999 İzmit earthquake. Science 285, 1858–1859. Harris, R.A. & Day, S.M. 1999. Dynamic 3D simulations of earthquakes on en echelon faults. Geophysical Research Letters Barka, A.A., Akyüz, S., Altunel, E., Sunal, G., Çakır, Z., Dİkbaş, 26, 2089–2092. A., Yerlİ, B., Rockwell, T., Dolan, J., Hartleb, R., Dawson, T., Fumal, T., Langridge, R., Stenner, H., Christofferson, Harris, R.A., Dolan, J.F., Hartleb, R.D. & Day, S.M. 2002. Th e S., Tucker, A., Armijo, R., Meyer, B., Chabalier, J., B., 1999 İzmit, Turkey, earthquake: A 3D dynamic stress transfer Lettis, W., Page, W. & Bachhuber, J. 2000. Th e August 17, model for intra-earthquake triggering. Bulletin of Seismological 1999 İzmit earthquake, M= 7.4, Eastern Marmara region, Society of America 92, 245–255. Turkey: study of surface rupture and slip distribution. In: Hearn, E.H. & Bürgmann, R. 2005. Th e eff ect of elastic layering on Barka, A.A., Kozacı, Ö., Akyüz, S. & Altunel, E. (eds), inversions of GPS data for coseismic slip and resulting stress Th e 1999 İzmit and Düzce Earthquakes: Preliminary Results. changes: strike-slip earthquakes. Bulletin of the Seismological İstanbul Technical University, İstanbul, 15–30. Society of America 95, 1637–1653. Barka, A.A. & Kadinsky–Cade, K. 1988. Strike-slip fault geometry Jeng, Y., Lin, M.J. & Chen, C.S. 2004. A very low frequency-electro- in Turkey and its infl uence on earthquake activity. Tectonics 7, magnetic study of the geo-environmental hazardous areas in 663–684. Taiwan. Environmental Geology 46, 784–795. Burgmann, R., Ayhan, M.E., Fielding, E.J., Wright, T.J., Kadinsky-Cade, K. & Barka A.A. 1989. Eff ects of restraining McClusky, S., Aktuğ, B., Demİr, C., Lenk, O. & Türkezer, bends on rupture of strike-slip earthquakes. In: Scwhwartz, A. 2002. Deformation during the 12 November 1999 Düzce, D. & Sibson, R. (eds), Proceedings of Conference XLV on Turkey, earthquake, from GPS and InSAR data. Bulletin of Fault Segmentation and Controls of Rupture Initiation and Seismological Society of America 92, 161–171. Termination ,U.S. Geological Survev Open-File Report 89-315, Chaput, E. 1957. Etudes sur les terrasses marines du littoral de la mer 181–192. de Marmara, I. les terrasses de Yalova. Travaux du laboratoire Kame, N., Rice, J.R. & Dmowska, R. 2003. Eff ects of prestress state de Géologie de la Faculté des Sciences de Dijon 18, 129 –136. and rupture velocity on dynamic fault branching. Journal of Cormier, M., Seeber, L., McHugh, C.M.G., Polonia, A., Çağatay, Geophysical Research 108 (B5), 2265. N., Emre, Ö., Gasperini, L., Görür, N., Bortoluzzi, G., King, G.C.P., Stein, R. S. & Lin, J. 1994. Static stress changes and the Bonatti, E., Ryan, W.B.F. & Newman, K.R. 2006. North triggering of earthquakes. Bulletin of Seismological Society of Anatolian fault in the Gulf of İzmit (Turkey): rapid vertical America 84, 935–953. motion in response to minor bends of a nonvertical continental transform. Journal of Geophysical Research 111, B04102, 2006. Kozacı, Ö. 2002. Hersek Deltası’nda Kuzey Anadolu Fayı’nın Yalova Segmenti Üzerinde Palaeosismolojik Çalışmalar [Palaeoseismic Çakır, Z., de Chabalier, J.-P., Armijo, R., Meyer, B., Barka, Studies of the North Anatolian Fault in Hersek Peninsula]. MSc A.A. & Peltzer, G. 2003. Coseismic and early post-seismic Th esis, İstanbul Technical University, Eurasian Institute of slip associated with the 1999 İzmit earthquake (Turkey), from Earth Sciences, İstanbul, Turkey [unpublished, in Turkish with SAR Interferometry and tectonic fi eld observations. Geophysics English abstract]. Journal International 155, 93–110. Kuşçu, İ., Okamura, M., Matsuoka, H. & Awata, Y. 2002. Active Eginitis, M.D. 1895. Le tremblement de terre de Constantople du 10 faults in the Gulf of İzmit on the North Anatolian Fault NW Juillet 1894. Annales de Géographie 4, 151–165. Turkey: a high-resolution shallow seismic study. Marine Ergİn, K., Güçlü, U. & Uz, Z. 1967. Earthquake Catalog of Turkey Geology 190, 421–443. and its Region Between 11 A.D. and 1964. İTÜ Maden Fakültesi Langridge, R.M., Stenner, H.D., Fumal, T.E., Christofferson, Arz Fiziği Enstitüsü Yayınları, 24 [in Turkish]. S.A., Rockwell, T.K., Hartleb, R.D., Bachhuber, J. & Feigl, K. 2002. Measurement of coseismic deformation by satellite Barka, A.A. 2002. Geometry, slip distribution, and kinematics geodesy. In: Lee, W.H.K, Kanamori, H. & Jennings, P.C of surface rupture on the Sakarya Fault segment during the (eds), International Handbook of Earthquake Engineering and 17 August 1999 İzmit, Turkey, earthquake. Bulletin of the Seismo logy, 81A. Academic Press, 607–620. Seismological Society of America 92, 107–125.

377 NORTH ANATOLIAN FAULT ON THE HERSEK PENINSULA

Lettis, W.R., Bachhuber, J., Barka, A.A., Witter, R. & Brankman, Sakınç, M. & Bargu, S. 1989. İzmit körfezi güneyindeki Geç C. 2000. Surface Fault Rupture and Segmentation during the Pleistosen (Tireniyen) çökel stratigrafi si ve bölgenin Kocaeli Earthquake. In: Barka, A.A., Kozacı, Ö., Akyüz, S. neotektonik özellikleri [Upper Pleistocene stratigraphy in & Altunel, E. (eds), Th e 1999 İzmit and Düzce Earthquakes: south of İzmit Gulf and neotectonic characteristics of the Preliminary Results. İstanbul Technical University, İstanbul, region]. Geological Bulletin of Turkey 32, 51–64 [in Turkish 31–54. with English abstract]. Lettis, W.R., Bachhuber, J., Witter, R.C., Brankman, C., Segall, P. & Pollard, D.D. 1980. Mechanics of discontinuous Randolph, E., Barka, A.A., Page, W.D. & Kaya, A. 2002. faults. Journal of Geophysical Research 85, 4337–4350. Infl uence of releasing step-overs on surface fault rupture and fault segmentation: examples from the 17 August 1999 İzmit Sibson, R.H. 1985. Stopping of earthquake ruptures at dilatational earthquake on the North Anatolian fault, Turkey. Bulletin of fault jogs. Nature 316, 248–251. Seismological Society of America 92, 19–42. Stein, R.S., Barka, A.A. & Dietrich, J.H. 1997. Progressive failure

McClusky, S., Balassanian, S., Barka, A.A., Demİr, C., Ergİntav, on the North Anatolian fault since 1939 by earthquake stress S., Georgiev, I., Gürkan, O., Hamburger, M., Hurst, K., triggering. Geophysics Journal International 128, 594–604. Kahle, H., Kastens, K., Kekelidze, G., King, R., Kotzev, The Catholic Encyclopedia, 1910. Volume VII. Published 1910. V., Lenk, O., Mahmoud, S., Mishin, A., Nadariya, M., New York: Robert Appleton Company. Nihil Obstat, June 1, Ouzounis, A., Paradissis, D., Peter, Y., Prilepin, M., 1910. Remy Lafort, S.T.D., Censor. Imprimatur: John Cardinal Reilinger, R., Şanli, I., Seeger, H., Tealeb, A., Toksöz, N. Farley, Archbishop of New York. & Veis, G. 2000. Global Positioning System constraints on the Toda, S., Stein, R.S., Reasenberg, P.A. & Dieterich, J.H. 1998. plate kinematics and dynamics in the eastern Mediterranean Stress transferred by the Mw= 6.5 Kobe, Japan, shock: eff ect and Caucasus. Journal of Geophysical Research 105, 5695–5719. on aft ershocks and future earthquake probabilities. Journal of Oglesby, D.D. 2005. Th e dynamics of strike-slip step-overs with Geophysical Research 103, 24543–24565. linking dip-slip faults. Bulletin of Seismological Society of Uçarkuş, G., Armijo, R., Çakır, Z., Schmidt, S. & Meyer, B. 2008. America 95, 1604–1622. Recent Earthquake Breaks at the Sea of Marmara Pull-apart Parsons, T. 2004. Recalculated probability of M= 7 earthquakes (North Anatolian Fault). American Geophysical Union Fall beneath the Sea of Marmara, Turkey. Journal of Geophysical Meeting, California, Abstracts, p. T24A–07. Research 109, B05304. Wesnousky, S.G. 2008. Displacement and geometrical characteristics Pınar, A., Honkura, Y. & Kuge, K. 2001. Seismic activity of earthquake surface ruptures: issues and implications for triggered by the 1999 İzmit earthquake and its implications seismic-hazard analysis and the process of earthquake rupture. for the assessment of future seismic risk. Geophysics Journal Bulletin of the Seismological Society of America 98, 1609–1632. International 146, F1–F7. Witter, R., C., Lettis, W., Bachhuber, J., Barka, A.A., Evren, Polonia, A., Gasperini, L., Amorosi, A., Bonatti, E., Bortoluzzi, E., Çakır, Z., Page, W., Hengesh, J. & Seitz, G. 2000. G., Çağatay, N., Capotondi, L., Cormier, M.-H., Görür, Palaeoseismic trenching study across the Yalova Segment N., McHugh, C. & Seeber, L. 2004. Holocene slip rate of the of the North Anatolian Fault, Hersek Peninsula, Turkey. Th e North Anatolian Fault beneath the Sea of Marmara. Earth and August 17, 1999 İzmit earthquake, M= 7.4, Eastern Marmara Planetary Science Letters 227, 411–426. region, Turkey: study of surface rupture and slip distribution. Reilinger, R.E., Ergintav, S., Bürgmann, R., McClusky, S., In: Barka, A.A., Kozacı, Ö., Akyüz, S. & Altunel, E. (eds), Lenk, O., Barka, A.A., Gürkan, O., Hearn, L., Feigl, K.L., Th e 1999 İzmit and Düzce Earthquakes: Preliminary Results. Çakmak, R., Aktuğ, B., Özener, H. & Toksöz, M.N. 2000. İstanbul Technical University, İstanbul, 329–339. Coseismic and postseismic fault slip for the 17 August 1999, M Wright T.J., Fielding, E.J. & Parsons, B. 2001. Triggered slip: = 7.5, İzmit Turkey earthquake. Science 289, 1519–1524. observations of the 17 August 1999 İzmit (Turkey) earthquake Reilinger, R., McClusky, S., Vernant, P., Lawrence, S., using radar interferometry. Geophysical Research Letters 28, Ergİntav, S., Çakmak, R., Özener, H., Kadırow, F., Gulıev, 1079–1082. I., Stepanyan, R., Nadarıya, M., Hahubıa, G., Mahmoud, Yüksel, F.A. 1995. Seismic Activity of the Gulf of İzmit Region. In: S., Sakr, K., Arrajehı, A., Paradıssıs, D., Al-Aydrus, A., Meriç, E. (ed), Quaternary Sequence in the Gulf of İzmit, ISBN Prılepın, M., Guseva, T., Evren, E., Dmıtrotsa, A., Fılıkov, 975-96123-0-5. S., Gomez, F., Al-Ghazzı, R. & Karam, G. 2006. GPS constraints on continental deformation in the Africa-Arabia- Eurasia continental collision zone and implications for the dynamics of plate interactions. Journal of Geophysical Research 111, B05411.

378 Ö. KOZACI ET AL.

SUPPLEMENTARY DATA

A- Archeoseismology

Figure 1. Mapshowing historical towns, cities and locations. Red dashed box indicates the study site of the Hersek Peninsula.

See next page for Table

I NORTH ANATOLIAN FAULT ON THE HERSEK PENINSULA

Table 1. Table showing the historical earthquakes aff ecting Hersek Peninsula and its vicinity. DATE INTENSITY LOCATION RESOURCE M.Ö. 19 VIII İznik, İzmit 1, 2 24.11.29 IX İznik, İzmit 1 33 VIII İznik, Kocaeli, Bursa and surroundings 1 02.01.69 VII İznik, İzmit 1, 2 120 VIII İznik, İzmit 1 121 İzmit 2 128 İzmit 2 170 VIII İzmit and surroundings 1 268 VIII İzmit and surroundings 1 269 İzmit-Gebze 2 ?.10.350 VIII İzmit, İznik 1 24.08.358 IX, VI Kocaeli, İznik, İstanbul 1 ?.11.359 VIII İzmit 1, 3 02.12.362 VIII, VI İznik, İzmit, İstanbul 1, 2, 3 26.01.446 VIII, VI İzmit Körfezi, İstanbul, İzmit 1, 2, 3 08.12.447 IX, VIII İzmit Körfezi, İstanbul, İznik 1, 3 448 VIII İzmit, Karamürsel 1 467 VI İzmit 1, 3 25.9.478 Karamürsel (Helenopolis), İzmit 2 488 VII İzmit-Yalova 3 500 VIII İzmit 1 551/554 VIII İzmit and surroundings 3 15.08.553 X İzmit, Kocaeli 1 16.8.554 İzmit 2 26.10.740 VIII İstanbul, İzmit, İznik 1, 2 25.10.989 Doğu Marmara 2 23.09.1064 VIII İstanbul-İzmit 2, 3 14.09.1509 IX İstanbul, Edirne, İzmit, Bolu, Bursa 3 01.10.1567 Sapanca 2 25.05.1672 VII İzmit 3 25.05.1672 VIII İzmit, İstanbul 1, 2 25.05.1719 IX İstanbul, İzmit, Karamürsel 1, 2 02.09.1754 IX, VII İzmit Körfezi, İstanbul, İzmit 1, 2, 3 13.01.1871 VI İzmit, Erdek 3 23.11.1875 VI İstanbul 3 19.04.1878 VIII İzmit, İstanbul, Bursa, Sapanca 1, 2, 3 10.05.1878 VIII, IX İzmit, İstanbul, Bursa 1, 2, 3 10.07.1894 IX İstanbul, İznik, Karamürsel, Tekirdağ, Lapseki 3 20.06.1943 M= 6.4 Adapazarı, Hendek, Akyazı, Arifi ye 2 26.05.1957 M= 7.0 Abant 2 18.09.1963 M= 6.4, 6.3 Yalova, Princes Islands 2, 3 22.07.1967 M= 7.1 Mudurnu 2 1. Yüksel 1995; 2. Ambraseys & Finkel 1991; 3. Ergin et al. 1967

II Ö. KOZACI ET AL.

Figure 2. Some of the historical remnant locations shown on a 1/10000 scale aerial photo (a), and their photos (b–e) (photos by Kozacı, 2000). B– Roman aqueduct tower, C– Cistern, D– Public bath, E– Hersekzade Ahmet Paşa Mosque.

III NORTH ANATOLIAN FAULT ON THE HERSEK PENINSULA

SUPPLEMENTARY DATA B- MODELS

N A A

A’

b a A’ A A’

B B’

N B

B’

d e B B’

Figure 3. Left -stepping model with gap or potential restraining bend. Best-fi t model.

IV Ö. KOZACI ET AL.

N A A

A’

a A’ b A A’

B B’

N B

B’

d e

B B’

Figure 4. Right-stepping model. Note that this fault geometry does not create a model that is comparable to the real geomorphology.

V NORTH ANATOLIAN FAULT ON THE HERSEK PENINSULA

N A

A’

b a A’ A A’

B B’

N B

B’

d e

B B’

Figure 5. Left -stepping model east of peninsula with overlap on the peninsula. Although this fault geometry creates a somewhat comparable onshore morphology the correlation between the model and off shore seismic profi le is poor.

VI Ö. KOZACI ET AL.

N A A

A’

b a A’ A A’

B B’

N B

B’ d e B B’

Figure 6. Left -stepping model with no overlap or gap. Th is fault geometry also creates a comparable model to the recent geomorphology of the study area. However, the off shore seismic profi le does not exhibit a back-facing scarp as suggested by this model.

VII