A source study of the 6 July 2003 (M 5.7) earthquake ...users.auth.gr/~zroum/Publications/Papers/Saros.pdf · the Ganos fault closer to the northern margin of the trough. The aim

www.elsevier.com/locate/tecto

Tectonophysics 412 (

A source study of the 6 July 2003 (Mw 5.7) earthquake sequence in

the Gulf of Saros (Northern Aegean Sea): Seismological evidence for

the western continuation of the Ganos fault

Hayrullah Karabulut a, Zafeiria Roumelioti b, Christoforos Benetatos b,

Ahu Komec Mutlu a, Serdar Ozalaybey c, Mustafa Aktar a, Anastasia Kiratzi b,*

a Bogazici University, Kandilli Observatory and Earthquake Research Institute, Istanbul, Turkeyb Department of Geophysics, Aristotle University of Thessaloniki, 54124 Thessaloniki, Greece

c TUBITAK, Marmara Research Center, Earth and Marine Sciences Research Institute, Kocaeli, Turkey

Received 31 January 2005; received in revised form 12 September 2005; accepted 28 September 2005

Available online 21 November 2005

Abstract

The July 2003 sequence in the Gulf of Saros (Northeastern Aegean Sea) is investigated, in terms of accurate event locations and

source properties of the largest events. The distribution of epicenters shows the activation of a 25-km long zone, which extends in

depth between 9 and 20 km. The major slip patch of the 6 July 2003 Mw 5.7 mainshock is confined in a small area (~45 km2),

which coincides with the deeper (12–20 km) part of the activated zone. The epicenters of the sequence follow the northern margin

of the Saros depression. This observation supports recent studies, according to which the continuation of the Ganos fault in the Gulf

of Saros does not coincide with the fault along the northern coast of the Gelibolu peninsula, but it is located at the northern

boundary of the Saros depression. This is further supported by the fact that the focal mechanisms of the mainshock and of the

largest aftershocks of the 2003 sequence imply almost pure dextral strike-slip faulting, whereas the fault bounding the Gulf of Saros

to the south appears as a normal fault on seismic sections. Thus, we infer that the principle deformation zone consists of a major

strike-slip fault, which lies close to the northern margin of the Saros depression and this fault could be regarded as the continuation

of the northern branch of the North Anatolian Fault into the Saros Gulf and North Aegean Trough as suggested by regional tectonic

models. The northeastern extent of the 2003 sequence marks the western termination (at ~26.38 E) of a long-term seismic

quiescence observed in the period following the 1912 Ganos earthquake, which may be associated with the extend of the rupture of

the particular earthquake.

D 2005 Elsevier B.V. All rights reserved.

Keywords: Saros Gulf; Ganos earthquake; Focal mechanisms; Slip distribution; Aegean Sea

1. Introduction

In July 2003, a sequence of small to moderate

magnitude earthquakes occurred on the Gulf of Saros,

0040-1951/$ - see front matter D 2005 Elsevier B.V. All rights reserved.

doi:10.1016/j.tecto.2005.09.009

* Corresponding author. Tel.: +30 2310 998486.

E-mail address: [email protected] (A. Kiratzi).

close to the Greek–Turkish borders in Northeastern

Aegean Sea. An isolated event of Mw 4.0 occurred on

June 10, while the sequence began more in earnest on

July 5 and continued for a period of about one month.

The sequence included at least eleven events of mag-

nitude Mw 4.0 and larger, along with several smaller

ones. Two moderate earthquakes, Mw 5.7 at 19:10:28.0

UTC on July 6 and Mw 5.3 at 20:10:15.6 on the same

2006) 195–216

H. Karabulut et al. / Tectonophysics 412 (2006) 195–216196

day, were the strongest, in terms of magnitude, events

of the sequence. Although no damage was reported,

these earthquakes were strongly felt in wide parts of the

Greek and Turkish territories and caused panic to the

populations around the epicentral area.

The Saros Gulf is an E–W trending neotectonic

basinal structure located at the northeastern part of the

Aegean Sea, where the forces controlling the tectonic

escape of the Anatolian plate toward SW interact with

the back-arc, N–S extension of the Aegean Sea crust.

The wedge-shaped Saros Gulf extends parallel to the

coasts of the Thrace Shelf to the north and the Gelibolu

peninsula to the south (Fig. 1) and widens and deepens

toward WSW, where it becomes the easternmost part of

the North Aegean Trough.

A prevailing seismotectonic structure in this region

is the Ganos Fault (Fig. 1), which is known to have

ruptured during the 1912 catastrophic (Mw 7.4) Mur-

efte–Sarkoy earthquake (Ambraseys and Finkel, 1987,

1991; Ambraseys, 1990; Papazachos and Papazachou,

1997; Armijo et al., 2005), herein referred to as the

Ganos earthquake. Although the recent work of Armijo

et al. (2005) sheds light into the exact geometry and

continuation of the Ganos fault, the corresponding fea-

tures of its western termination, into the Gulf of Saros,

are poorly known. So far, most authors consider the

Fig. 1. Important features of the general geodynamic setting of the North Ae

correspond to NOA and squares to KOERI stations) whose waveforms were

the on land Ganos fault segment of the NAF are also shown.

southern margin of the Saros depression as the most

active one and consequently they identify it with the

main westward continuation of the Ganos fault in the

Gulf of Saros (Tuysuz et al., 1998; Okay et al., 1999;

SaatcVlar et al., 1999; Kurt et al., 2000), although such a

consideration is not adequately justified. Actually, the

most recent studies in the area (YaltVrak et al., 1998,

2000a,b; YaltVrak and Alpar, 2002; SakVnc et al., 1999)

discredit upon this theory and place the continuation of

the Ganos fault closer to the northern margin of the

trough.

The aim of this paper is to extract accurate informa-

tion from the high-quality seismological data of the

2003 Saros earthquake sequence to shed light into the

debating issue of the continuation of the Ganos fault in

the Gulf of Saros. To achieve this goal, we locate the

aftershock sequence by combining waveform and para-

metric data from Greek and Turkish seismological net-

works, determine the source parameters of the large

events of the sequence by applying a time-domain

moment tensor inversion technique (Pasyanos et al.,

1996; Dreger, 2002) and finally investigate the rupture

process of the mainshock (July 6, 2003, Mw 5.7) in

terms of its slip distribution through a source time

functions (STF) inversion procedure. We show that

the aftershock distribution of the 6 July 2003 sequence

gean Sea area and locations of the broadband stations (black triangles

used here. The location of the largest event of the 2003 sequence and

Fig. 2. One-dimensional velocity model used to locate the earthquakes

of the 2003 sequence.

H. Karabulut et al. / Tectonophysics 412 (2006) 195–216 197

and its focal mechanisms clearly indicate that the north-

ern margin of the Saros depression is presently the most

active one.

2. Earthquake location

Broadband data for selected events were available

from the Geodynamic Institute of the National Ob-

servatory of Athens (NOA) in Greece and the Mar-

mara Research Center (MRC) of the Scientific and

Technical Research Council of Turkey (TUBITAK)

and the Kandilli Observatory and Earthquake Re-

search Institute (KOERI) in Turkey (Table 1). Addi-

tional phase data were obtained from short-period

stations of the Department of Geophysics of the

Aristotle University of Thessaloniki (GLUT) in

Greece and KOERI.

We selected more than 130 events that were

recorded by at least three stations during the two-

month period following the mainshock. Arrival times

of both P and S phases were obtained for most of the

recording stations and were passed to the VELEST

inversion code (Kissling et al., 1994) to simultaneous-

ly invert for a one-dimensional velocity model and

station delays. The resulting velocity model (Fig. 2)

and station delays were then used in the HYPO71

code (Lee and Lahr, 1975) to accurately locate the

earthquakes of the sequence. In total, 101 events

(Appendix A), with average root mean square (rms)

travel-time residuals b0.12 s, were selected for the

final analysis (i.e. only well located events were

taken into account in subsequent steps of the analy-

Table 1

Information on the broadband stations, whose records were used in the analysis

Code Lat (8N) Lon (8E) Elev (m) Location Sensor Operating institute

CEV 40.3692 26.5831 118 Cevizli Gelibolu Peninsula L4-1Hz TUBITAK

EDR 41.5082 26.4462 209 Edirne CMG40T KOERI

ISK 41.0394 29.0355 132 Istanbul, Kandilli CMG3T KOERI

LIA 39.9000 25.1800 60 Limnos Isl. CMG40T/30 NOA

MFT 40.7900 27.3000 800 Murefte CMG40T KOERI

MRM 40.3619 27.3574 741 Marmara Island CMG40T KOERI

NEO 39.3100 23.2200 500 Neochori, Volos LE3D/20 NOA

NVR 41.3500 23.8600 595 Nevrokopi CMG40T/30 NOA

PLG 40.3740 23.4460 580 Poligiros, Chalkidiki Le3D/20 NOA

PRK 39.2460 26.2720 100 Ag. Paraskevi, Lesvos Isl. Le3D/20 NOA

RDO 41.1460 25.5380 100 Gratini, Rhodopi Le3D/20 NOA

SGT 40.7670 27.1081 310 Saglamtas, Tekirdag CMG40T TUBITAK

TRN 40.5061 27.7775 50 Erdek, BalVkesir CMG40T TUBITAK

YLV 40.3400 29.2234 829 Yalova CMG40T KOERI

TUBITAK = Scientific and Technical Research Council (Turkey).

KOERI = Kandilli Observatory and Earthquake Research Institute (Turkey).

NOA = National Observatory of Athens, Geodynamic Institute (Greece).

sis). The average horizontal and vertical uncertainties

of the selected events are less than 2.0 and 3.0 km,

respectively. Relocation of these events, using the

double-differences algorithm of Waldhauser and Ells-

worth (2000), did not result in significant changes of

the original locations.

From the distribution of the epicenters (Fig. 3) we

observe that the epicenters are aligned mainly along the

deep trough that forms the axis of the Saros Gulf. The

along strike dimension of the activated zone (Fig. 3a) is

large (~25 km) compared to what is expected from the


magnitude of the largest event of the sequence (6 July

2003, 20:10 UTM; Mw 5.7). We are tempted to observe

two distinct earthquake clusters in Fig. 3, which oper-

ated simultaneously. The eastern cluster is inherently

related to the largest event of the sequence whereas in

the western cluster the largest recorded event has Mw

Fig. 4. Focal mechanisms of the July 2003 Saros sequence as determined from moment tensor inversion, using regional broadband waveforms.

Event numbers are the same as in Table 2. Note the occurrence of the two major foreshocks (events 1 and 2) at the two ends of the rupture area. The

focal mechanism of a late aftershock (event 12; 15 July 2004, Mw 5.1), as well as the focal mechanism (Taymaz et al., 1991) of the 27 March 1975

(Mw 6.1) earthquake are also depicted. The parameters of the 12 moment tensor solutions are given in Appendix B.


4.7. In this cluster a late aftershock Mw 5.1 occurred

one year after, in 2004. In any case, the uncertainties of

the locations and the small spatial separation of the two

clusters do not permit any positive conclusions. No

seismic activity is observed on the northern shelf of

the Saros Gulf during the July 2003 earthquake se-

quence. Relatively diffused seismicity is observed on

the southern part of the depression, as well. The north–

south structural asymmetry of the Saros depression,

characterized by a stable northern margin and a distrib-

uted deformation zone toward the south, could be

compared to the asymmetric basin structures observed

in the Sea of Marmara (Le Pichon et al., 2001). Most

epicenters are concentrated close to the northern margin

of the depression, indicating that this branch of the

structure is the one that was activated during this recent

sequence.

Fig. 3. Distribution of epicenters (101 accurately located events of Appe

distribution; light shade color corresponds to the two largest events of the

occurrence of the largest event and dark shade color to the remaining e

perpendicular to the inferred fault strike, c) WSW–ENE cross-section (alon

distribution of regional seismicity (1970–2003, NEIC Catalog) shown for com

with the dashed ellipse). The 2003 earthquake sequence is also plotted on t

Vertical cross-sections (Fig. 3b, c) show that most

of the foci are concentrated at depths between 9 and

20 km. The computation of depths was based on an

extensive use of 3-component stations, which allowed

accurate S-wave arrivals selection. Similarities of

waveforms were taken into account to have consistent

picks. Although in some cases the available 3-com-

ponent stations were limited in number and provided

relatively poor azimuthal coverage of the activated

area, the consideration of low values for the residual

and the small error ellipses, assures the accuracy of

the depth solutions of the selected events. The ma-

jority of the hypocenters are concentrated in a narrow

zone, delineating the activated fault zone, which

appears to be pure vertical. Fig. 3d shows the distri-

bution of the regional seismicity in the studied area

for comparison purposes. A striking feature is the

ndix A) of the 2003 Saros Gulf earthquake sequence: a) horizontal

sequence, medium shade color to the events of the first 3 h after the

vents, b) three-dimensional view of the hypocenters in a direction

g line AB in Fig. 3a) of the accurately located earthquake sources, d)

parison (note the ~100 km seismic gap along the Ganos fault, marked

op of regional seismicity and in lighter shade color.


quiescence along the Ganos fault, which is later

discussed.

3. Focal mechanisms

3.1. Method applied

We determined earthquake focal mechanisms of the

Saros sequence from regional broadband seismograms,

using the moment tensor inversion technique of Dreger

and Helmberger (1990, 1991, 1993). This approach

uses synthetic Green’s functions in the inversion of

observed three-component full-waveform broadband

data and is capable of revealing the seismic moment

tensor of earthquakes of magnitudes as low as Mw 3.5

(Pasyanos et al., 1996). In the present work, Green’s

functions were calculated using the frequency-wave

number integration code (FKRPROG) developed by

Saikia (1994). For a detailed description of the method

and the analysis procedure, the reader is referred to the

work of Pasyanos et al. (1996) and Dreger (2002).

3.2. Distribution of focal mechanisms

Digital broadband waveforms obtained from NOA

and KOERI stations were used for the inversion. NOA

stations are equipped with Lennartz (LE-3D/20s) or

Guralp (CMG40T/30) sensors, whereas Kandilli sta-

tions are equipped with Guralp (CMG40T and

CMG3T) seismometers. Prior to the inversion, raw

data were cut into segments of at least 5-min duration,

band-pass filtered between 0.05–0.08 Hz and re-sam-

pled at 1 s. The velocity model used for the calculation

of the Green’s functions is the one proposed by

Novotny et al. (2001). Although this model is not as

Table 2

Parameters of the focal mechanisms obtained from moment tensor inversio

No Year Month Day Time Latitude

8NLongitude

8EDepth

km

Magnitude

Mw

1 2003 06 10 01:01:51.8 40.240 25.640 14 4.0

2 2003 07 05 21:58:30.0 40.426 26.079 16 4.3

3 2003 07 06 19:10:28.0 40.427 26.103 18 5.7

4 2003 07 06 19:39:50.7 40.411 25.996 19 4.2

5 2003 07 06 20:10:15.6 40.439 26.108 16 5.3

6 2003 07 06 20:48:53.3 40.406 26.006 20 4.7

7 2003 07 09 22:01:57.5 40.385 25.913 13 3.8

8 2003 07 09 22:08:49.5 40.386 25.902 12 4.1

9 2003 07 09 22:31:40.8 40.388 25.912 16 4.7

10 2003 07 13 06:32:08.1 40.389 25.923 14 4.0

11 2003 07 18 05:44:07.4 40.394 25.962 14 3.8

12 2004 07 15 12:02:38.5 40.373 25.901 13 5.1

The parameters of the strongest event of the sequence are marked in bold.

detailed as the one computed by the VELEST code in

the frame of the present work (see Section 2) it has been

previously used for various paths in the Aegean Sea

area (Benetatos et al., 2002; Zahradnik, 2002) and has

been proven adequate to explain the low frequency

content (0.05–0.08 Hz) of the recorded broadband

waveforms. We performed the inversion for depths

ranging from 6 to 25 km, with a 2-km increment.

The quality (good signal-to-noise ratio) and amount

of available data allowed the computation of 12 focal

mechanisms corresponding to two foreshocks, the

mainshock and 9 aftershocks. The focal spheres are

shown in Fig. 4, while the parameters of the moment

tensors are listed in Table 2. The results of the moment

tensor inversion for the 12 earthquakes are also pre-

sented in detail in Appendix 2.

These focal mechanisms show pure right-lateral

strike-slip faulting that is strongly associated with the

strike-slip motions characterizing the stress regimes of

the westernmost part of the NAF zone (Kiratzi, 2002).

The orientation of the fault planes of the studied earth-

quakes is ENE–WSW, which is in alignment with the

long axis of the basin bathymetry.

4. Slip distribution

4.1. Method used

The slip distribution of the Mw 5.7, July 6, 2003

earthquake in the Gulf of Saros was investigated by

combining the empirical Green’s function method

(Hartzell, 1978) with the earthquake source inversion

method introduced by Mori and Hartzell (1990) and

later extended by Dreger (1994) to be applicable at

regional distances. The combined methods are briefly

n using regional broadband waveforms

Strike1

(8)Dip1

(8)Rake1

(8)Strike2

(8)Dip2

(8)Rake2

(8)P az

(8)P dip

(8)T az

(8)T dip

(8)

51 81 152 146 62 10 101 13 5 26

78 73 171 171 81 17 304 6 35 19

257 89 �179 167 89 �1 122 1 212 0

89 53 173 183 84 37 310 21 53 30

253 89 175 343 85 1 298 3 208 4

252 85 �178 162 88 �5 117 5 207 2

74 89 �173 344 83 �1 299 6 209 4

75 87 �174 345 84 �3 300 6 210 2

71 78 �178 341 88 �12 295 10 27 7

69 83 165 161 75 7 116 6 24 16

244 87 �176 154 86 �3 109 5 19 1

74 82 178 164 88 8 299 4 29 7


described in the following paragraphs; readers are re-

ferred to the aforementioned papers (and references

therein) for a more detailed description.

In the frequency domain and in the case of a point

source, the far-field wave displacement, U(x)

recorded at distance r and azimuth / can be repre-

sented by:

U xð Þ ¼ MM xð ÞdG x; r;/ð ÞdRdI xð Þ ð1Þ

where M is the moment rate function (STF) of the

event, G is the Green’s function response of the

medium along the wave path, which includes the

effects of attenuation and geometrical spreading, R

is the radiation pattern factor and I is the response

of the recording instrument. In order to retrieve the

STFs (required for retrieving the slip distribution

pattern of the examined earthquake) from the dis-

placement records one has to remove the effects of

the propagation path and the recording instrument. To

accomplish this we deconvolved the waveforms of a

nearby smaller event with similar focal mechanism,

herein referred to as an empirical Green’s function

(eGf). In this approach the small event is assumed to

be a point source both in time and space and to the

extent that this assumption does not hold the STF of

the target event will be a coarse estimate of the true

one. Therefore, the eGf should be small enough to be

treated as a point source, but also large enough to

ensure a satisfactory signal to noise level at the

examined distances.

In the following stage and prior to the inversion,

the estimated STFs are normalized to unit area. This is

done to ensure that all of the STFs used integrate to

the appropriate scalar seismic moment and also tends

to equalize the weighting of individual STFs in the

inversion.

The next step involves the inversion of the STFs,

which are estimated as described previously. The

employed inversion technique (Mori and Hartzell,

1990; Dreger, 1994) is based on the assumption that

the variations in STF shape can be mapped onto the

spatial and temporal slip history of the event. The

source is parameterized through a radially propagating

rupture front, which expands with a constant rupture

velocity. Slip is confined to one of the nodal planes

indicated by the focal mechanism of the event. Then the

inversion method fits the STF of the examined event by

summing contributions from different subfaults, taking

into account the time delay due to wave propagation

and to the propagation of the rupture front, with respect

to the hypocenter. Distances to the different subfaults

are estimated using a half space ray trajectory approx-

imation. The contribution from each subfault can take

the form of several synthetic time functions. Here the

subfaults STFs have the form of boxcar functions.

The subfault STFs (B) are related to the observed

STF’s (D) through a system of equations of the form:

Di tð Þ ¼ STFi tð Þ ¼Xmj

Bj t � sij� �

wj ð2Þ

where s is the time delay due to wave and rupture

propagation, i is a station index, j is a subfault index

and w is a weight proportional to fault slip. In the

above system of equations positivity constraint is

also imposed to require all subfaults to have the

same slip direction. A spatial derivative minimization

constraint is applied to smooth the resulting slip

model. As a result, Eq. (2) can finally be written

in matrix form as:

B

kS

�dw ¼ D

0

��ð3Þ

where S is the matrix of first spatial derivatives and

k is a constant controlling the weight of the smooth-

ing equation.

The slip weight vector is obtained by standard least

squares. Slip amplitudes at each subfault, uj, are finally

obtained based on an independent seismic moment

estimate, M0, through the relation:

uj ¼M0dwj

Adlð4Þ

where A is the subfault area and l is the shear modulus,

usually taken equal to 3.5�1010 Pa.

4.2. Application — results

The selection of the appropriate eGf to be used in the

STF inversion method was based on a comparison of

the moment tensor parameters of the largest events of

the sequence to the corresponding parameters of the

mainshock, as well as on direct visual comparison of

the waveforms of the different events. Although many

small events present similar focal mechanisms with the

mainshock (Table 2), only one event (no 2 in Table 2; 5

July 2003, 21:58 UTM, Mw 4.3) provided low-noise

STFs. In Fig. 5 we indicatively compare horizontal

displacement seismograms at selected stations for the

mainshock and the eGf.

The STFs of the 6 July 2003, Mw 5.7, earthquake

were estimated using the broadband waveforms at

twelve stations (as depicted in Fig. 1 except from

Fig. 5. Comparison of the displacement waveforms of the 6 July 2003 Mw 5.7 earthquake and the corresponding waveforms of the employed eGf at

four selected stations located at various azimuths around the epicenter. All waveforms are band-passed filtered (0.05–1 Hz).


NEO and MFT, which were omitted due to the poor

quality of the eGf records), which provided satisfac-

tory azimuthal coverage of the epicentral area. The

largest azimuthal gap between the used stations is

observed to the southeast of the epicentral area (be-

tween stations CEV and PRK) and is of the order of

758. The original data from the Turkish stations were

down-sampled from 0.01 to 0.02 s, to match the

sampling interval of the Greek data. All waveforms

were then integrated to displacement and band-pass

filtered in the frequency range 0.05 to 1.0 Hz. The

deconvolution process was performed in the frequency

domain, by dividing the displacement spectra of the

two events. To avoid instability problems due to spec-

tral sags in the denominator we used a 1% water-level

correction (Clayton and Wiggins, 1976). Furthermore,

Fig. 6. Source time functions of the 6 July 2003, Mw 5.7 earthquake at the twelve examined stations. STFs are presented prior to normalization, with their absolute amplitudes as derived from the

eGf deconvolution (the x–y scale and units are shown on the lower left of the figure). Note the simple triangular shapes of the STFs at the eastern stations and the two-lobe shape of the STFs at the

western along strike stations (see text for further discussion).

H.Karabulutet

al./Tecto

nophysics

412(2006)195–216

203


to reduce the noise in the computed source time

functions we treated each one of the three-recorded

components separately and then stacked the results.

The final step in the extraction of the STFs included

their normalization to unit area.

The computed STFs at the 12 examined stations,

prior to normalization, are depicted in Fig. 6. Despite

the moderate magnitude of the event, the STFs ap-

pear complicated, with at least two lobes at the

western stations-located along the strike of the

fault- and simple triangular shapes at the eastern

stations — again to those along the strike of the

fault. An interesting observation is that the second

major lobe of the STFs that correspond to stations

LIA and PLG is similar, both in absolute amplitude

and duration, with the triangular STFs derived at the

eastern stations. If absolute amplitudes are correct,

then the STFs in Fig. 6 imply larger seismic moment

release toward west. It seems as if, prior to the main

event, there is energy onset from a smaller event,

which has only been recorded at the western stations.

A closer examination of the recorded waveforms

supports this hypothesis. Computation of correlation

Fig. 7. Comparison between mainshock and eGf velocity (left panels) and di

LIA. Top panels include waveforms aligned at the P-onset, while at low p

achieve maximum correlation. Amplitudes are normalized to unit and all w

functions between the mainshock and the eGf wave-

forms at the examined stations revealed that although

at the eastern stations the maximum correlation is

achieved when the waveforms are aligned at the P-

wave onset, at western stations the corresponding

correlation maximum is observed a few seconds

after the P-wave onset in the mainshock records. A

characteristic example of this shift is presented in

Fig. 7.

Assuming that a small event occurred close in

space and time to the main event, the STF inversion

will fail to correctly produce the slip distribution

pattern, unless the effect of the two events in the

STFs is separated. In the following, we present the

inversion results a) using the entire length of the STFs

in all stations and b) using only the second triangular

lobe in the eastern stations (the STFs of the western

stations remain the same).

In the inversion of the STF shapes we assumed a

planar fault model with dimensions 20�20 km, dis-

cretized into 1 km2 square subfaults. The dimensions

of the fault model were chosen to be larger than those

expected for an earthquake of Mw 5.7, to avoid an

splacement (right panels) waveforms (east–west component) at station

anels eGf waveforms are shifted to the right by approximately 2 s to

aveforms are band-passed filtered between 0.05 and 1 Hz.


imposed confinement of the slip. The orientation of

the fault model was taken equal to the orientation of

the almost E–W nodal plane of the focal mechanism

(Table 2) of the examined event, following the delin-

eation of the 2003 sequence epicenters (Fig. 3a) and

the rupture initiation point was assumed to coincide

with the hypocenter of the event. The rise time, H ,was given a constant value of 0.3 sec, [using the

empirical relation s =2.03d 10�9dM01 / 3 (Somerville et

al., 1999)], where M0 was given the value of

3.77d 1024 dynd cm, as computed from the time-do-

main moment tensor inversion. The rupture velocity,

Fig. 8. Sensitivity tests and slip distribution models for the 6 July 2003, M

presented in Fig. 6 (after normalization), b) inverting the STF at station LIA, o

MRM, TRN, SGT) and d) inverting the STFs at the eastern stations (as pres

stations (RDO and NVR were omitted due to difficulty in splitting the S

location. Our preferred model is the one shown in 8d. The major slip patc

released in an area 10 �6 km. Note that most of the slip (average slip 24

deepest 12–20 km of the vertical fault plane.

Vr, was also considered constant (2.7 km/s) and equal

to 80% of the shear-wave velocity at the source

region. It must be noted that usually when applying

the STFs inversion method a grid search for the

optimum values of the rise time and the rupture

velocity is possible. However, in our case the exam-

ined earthquake is relatively small in magnitude and

the correspondingly small dimensions of the slipped

area do not permit such a search. This was verified by

multiple inversion tests with different values of these

two parameters, which resulted in practically the same

variance reduction and slip distributions.

w 5.7 Saros Gulf earthquake a) inverting the entire set of STFs as

nly c) inverting the STFs of only the eastern stations (YLV, ISK, CEV,

ented in Fig. 6) and only the second large lobe of the STFs of western

TFs). The white star symbol corresponds to the adopted hypocenter

h is confined in an area ~45 km2, whereas most of the moment was

cm, with a maximum of 104 cm near the epicenter) is located at the


In Fig. 8 we present the calculated slip distribution

models for the 6 July 2003, Mw 5.7 Saros earthquake.

The results of the inversion are evaluated through the

Fig. 9. Comparison between the observed (continuous lines) and the synth

distribution models of (a) and (d) of Fig. 8. Both slip models produce synt

variance reduction, which corresponds to the goodness

of fit between the computed and synthetic STFs. In Fig.

8a, we present the distribution that was derived using

etic (dashed lines) source times functions corresponding to the slip

hetics with satisfactory fit.


the STFs (as shown in Fig. 6) after they were normal-

ized to unit area. The variance reduction (VR), in this

case, is of the order of 91% and the synthetic STFs

present satisfactory agreement with the corresponding

observed ones (Fig. 9a). In accordance with the STFs

shapes, the slip appears to localize in two patches

(marked as 1 and 2 in Fig. 8a), with the largest one

almost centered at the hypocenter area and the second,

smaller one, at shallower depths and to the west of the

hypocenter.

This model (Fig. 8a) provides high variance reduc-

tion and reflects the major slip patches satisfactorily,

but it does not adequately explain the difference in the

STFs durations between the western and the eastern

stations. Based on these durations, one should expect

source directivity effects toward east. However, emer-

gence of such effects is not supported by the model of

Fig. 8a. Tests performed using reduced sets of stations

in the inversions (Fig 8b, c) confirmed the inconsisten-

cy between the results from the eastern and western

stations. In Fig. 8b we indicatively present the distri-

bution derived by the inversion of a single STF com-

puted at station LIA. Similar distributions were

obtained when western stations PLG and NVR were

incorporated in the inversion. This model is quite dif-

ferent from the one of Fig. 8a. The slip appears dis-

persed in a much wider area and consequently the peak

slip amplitudes decrease (umax=65 cm in Fig. 8a com-

pared to umax=37 cm in Fig. 8b). Furthermore, a sig-

nificant amount of blateQ slip appears in an arched

patch, which seems rather unnatural. In Fig. 8c we

present the resulting model from the use of the eastern

stations solely (YLV, ISK, CEV, MRM, TRN, SGT;

VR=97%; umax=70 cm).

In Fig. 8d we present the model finally adopted

using the largest set of stations (all stations shown in

Fig. 6 except RDO and NVR; VR=92%; umax=106

cm). In this last case we have inverted only the second

large-amplitude lobe of the western STFs. Stations

RDO and NVR were excluded from the last trial be-

cause it was difficult to split the two lobes of their

STFs. In Fig. 9 we compare the observed STFs and

the synthetic STFs as computed using the models (a)

and (d) in Fig. 8.

The comparative examination of the four slip pat-

terns presented in Fig. 8, leads to the conclusion that

only the major slip concentration around the hypocenter

is well constrained. The second slip patch, which

appears clearly in Fig. 8a and vaguely in Fig. 8c,d

could be an artifact if the first lobe of the western

STFs really corresponds to another event. In this case,

the code simultaneously inverts the STFs of the first

event at the western stations and the STFs of the second

event at the eastern stations. As a result, the energy of

the second lobe of STFs at western stations, e.g., LIA

and PLG, is shifted away from the hypocenter and

appears as a second slip patch. Confirmation of this

hypothesis requires the application of a more sophisti-

cated slip inversion method that could handle a double

source.

Although the presence of a second slip patch in the

slip distribution model cannot be unambiguously con-

cluded by the applied method, the dimensions of the

ruptured area appear to be well resolved, as they

remain more or less the same in all inversions.

Based on Fig. 8d, which is our proposed model for

the 6 July 2003, Mw 5.7 Saros earthquake, the major

slip patch (asperity) is ~45 km2. The slip is mainly

concentrated in an area with an along-strike dimension

of hardly 10 km and an along-dip dimension of ap-

proximately 6 km. The slip, averaged across the rup-

tured fault surface, is ~24 cm, which equals the value

computed from the empirical relation of Somerville et

al. (1999). A peak slip value of about 106 cm was

obtained very close to the adopted hypocenter. Gen-

erally most of the slip is concentrated close to and

around the hypocenter and at depths larger than 15 km

and up to 20 km.

5. Discussion

Kurt et al. (2000) describe the Saros depression as

a typical negative flower structure and therefore as-

sume that all fault traces are joined into a single one

at a certain depth. However, they also state that the

fault line that limits the depression to the south,

marking the boundary of the Gelibolu shelf, is the

main fault line and constitutes the western continua-

tion of the Ganos fault. Morphological analysis of

this southern branch shows a substantial amount of

vertical offset and therefore implies that normal

mechanism is an essential component of the Saros

fault activity. The fault mechanism solutions of some

past large events show a clear strike-slip mechanism

with minor extensional component (Taymaz et al.,

1991).

The present data, which are unprecedented in terms

of resolution, indicate that recent seismic activity is not

clustered around the southern border of the Saros de-

pression, but is concentrated on its northern border. The

fault plane solutions are mostly pure right lateral strike-

slip and clearly aligned with the axis of the depression.

The majority of the fault-plane solutions show minor

normal component with right-lateral fault planes dip-


ping to the south, which also implies that the activity is

related to the northern border.

A relatively deep seismogenic zone is also sug-

gested, which reaches to the depth of 20 km and

clearly indicates that the activity is concentrated well

below the branching point of the flower structure

inferred by Kurt et al. (2000). Furthermore, the inver-

sion of the STFs of the largest event of the 2003 Saros

Gulf sequence (6 July 2003, Mw 5.7) revealed that the

slip during this earthquake was confined at depths

greater than 12 km. The depth of the seismogenic

zone is characteristically similar to that observed at

the western end of the Marmara Sea, where accurate

and long term observations showed that the seismo-

genic zone extends to a depth of 25 km (Ozalaybey et

al., 2003).

It is worthwhile to note that the detailed analysis

of the July 2003 swarm leads to a conclusion that the

presently active seismic regime in the Gulf of Saros

is similar to the one in the Sea of Marmara. This

means that, although diffuse deformation occurs in

the south of a stable shelf, the principle deformation

zone consists of a single major strike-slip fault, which

lies close to the northern margin of the Saros depres-

sion and follows the deepest parts of this depression.

Thus, this fault is most probably identified as the

continuation of the northern branch of the NAF into

the Saros Gulf and North Aegean Trough as sug-

gested by regional tectonic models proposed by Kree-

mer et al. (2004). The data presented in this paper,

which are however constrained by the analysis of a

temporary swarm activity, provide clear evidence for

such conclusion.

A final consideration is given to the argument of

whether this well located swarm activity may mark

the western limit of the 1912 rupture of the Ganos

earthquake. This earthquake is given an Ms (=Mw)

magnitude of 7.4 or 7.3 (Ambraseys and Finkel,

1987; Ambraseys and Jackson, 2000, respectively).

Although it is known that the 1912 earthquake rup-

ture involved the entire length of the Ganos fault on

land, the exact geometry and slip distribution is

poorly known since the rest of the rupture must be

located submarine. The 1999 Izmit earthquake (Mw

7.4), which is nearly similar in size to the 1912

earthquake, created approximately 145 km rupture

zone on land (Barka et al., 2002) and appears to

have extended ~30 km westwards, into the Sea of

Marmara (Cakir et al., 2003; Armijo et al., 2005),

resulting in an overall rupture length of more than

170 km. Altunel et al. (2004) studied co-seismic and

cumulative slip distribution on the ruptured segment

of 1912 earthquake. They obtained right-lateral dis-

placement values of 3.5–4 m observed at several

localities on land. Using the analogy to the Izmit

earthquake and large slip values observed at the

termination of the Ganos fault on land, they con-

cluded that the rupture should extend at least 20 km

into the Gulf of Saros and 30 km into the Sea of

Marmara. Armijo et al. (2005) provide submarine

observations according to which the eastern continu-

ation of the 1912 surface rupture into the Sea of

Marmara is of the order of 60 km. On the other

hand, it is well recognized that consistent seismic

quiescence has been established along the on land

rupture length of the 1912 Ganos earthquake and

within the eastern part of the Gulf of Saros, based

on the seismicity data since this event (NEIC seis-

micity). This ~100 km long seismic gap is clearly

reflected in the present regional seismicity map pre-

viously shown (see Fig. 3d). The aftershock sequence

that we analyze here is located at the exit of the

Saros Gulf along the line of the seismic gap. As-

suming that the observed seismic gap in this area

covers part of the 1912 earthquake rupture, then

based on the high quality data that we used, recorded

both in Turkey and Greece, we speculate that the

northeastern termination of this sequence (at approx-

imately 26.38 E) marks the western termination of

this rupture into the Saros Gulf. Thus, we infer a

40–50 km long submarine segment in the Saros

Gulf, which along with the 50 km long on land

Ganos segment (Ambraseys and Jackson, 2000) and

the 60 km long submarine segment in the Sea of

Marmara (Armijo et al., 2005) results in an overall

length of ~160 km during the 1912 event. This

length is comparable to the observed rupture length

of the 1999 Izmit earthquake.

Acknowledgements

Thanks are due to our colleagues John Latoussakis

and George Stavrakakis from the Institute of Geody-

namics of the National Observatory of Athens for pro-

viding part of the data used. Thanks are also due to

Doug Dreger from the University of California at Ber-

keley for his continuous help and support. Esen Arpat

and Dean Childs are thanked for reading the manuscript

and providing fruitful suggestions. This work was

funded by TUBITAK from the Turkish side and the

General Secretariat of Research and Technology (Min-

istry of Development) from the Greek side. Most of the

figures were produced by GMT software (Wessel and

Smith, 1998).

Appendix A (continued)

Date Origin Lat (8N) Lon (8E) Depth (km) Mw

20030707 164741.8 40.363 26.140 15.0 2.7


Appendix A. Hypocenter parameters of the 101

located earthquakes of the 2003 Saros sequence

Date Origin Lat (8N) Lon (8E) Depth (km) Mw

20030706 191028.0 40.427 26.103 17.5 5.7

20030706 192451.4 40.448 26.131 17.6 3.6

20030706 192622.7 40.400 26.028 12.6 3.6

20030706 193950.7 40.411 25.996 18.5 4.2

20030706 194108.4 40.411 26.098 15.5 3.6

20030706 195837.5 40.407 26.165 19.1 2.8

20030706 200244.7 40.430 26.120 17.6 3.4

20030706 201015.6 40.439 26.108 16.4 5.3

20030706 201552.2 40.429 26.144 16.1 3.3

20030706 201953.9 40.431 26.116 17.8 3.2

20030706 202418.5 40.381 25.929 8.4 2.8

20030706 202642.5 40.405 25.964 10.0 3.3

20030706 202949.2 40.362 25.977 11.1 2.8

20030706 203206.2 40.370 25.971 15.0 2.8

20030706 204853.3 40.406 26.006 19.6 4.7

20030706 220548.5 40.403 25.989 16.3 4.2

20030706 222031.4 40.362 25.972 11.3 2.6

20030706 223940.5 40.239 26.001 20.9 2.8

20030706 224208.7 40.405 25.940 13.3 4.4

20030706 224604.2 40.399 25.968 10.4 3.0

20030706 225209.1 40.436 26.129 8.9 2.3

20030706 232719.1 40.369 25.987 12.9 2.3

20030706 233021.2 40.398 26.015 14.4 2.6

20030706 234720.1 40.358 26.001 13.9 2.6

20030706 235454.3 40.394 25.967 13.7 3.2

20030707 002407.3 40.381 25.921 9.5 3.4

20030707 004815.2 40.393 25.967 13.8 3.5

20030707 005535.1 40.410 26.013 9.9 2.7

20030707 011313.4 40.420 26.128 15.3 2.3

20030707 013639.6 40.392 25.945 10.9 2.7

20030707 030543.50 40.392 26.036 17.0 2.8

20030707 031625.3 40.382 26.056 17.9 2.4

20030707 034824.9 40.484 26.134 9.4 2.4

20030707 042407.4 40.371 25.979 10.5 2.4

20030707 062939.7 40.496 26.272 6.3 2.4

20030707 071011.6 40.433 26.068 14.9 2.6

20030707 071202.5 40.383 25.924 12.8 3.4

20030707 071503.1 40.383 25.909 8.9 3.1

20030707 095911.0 40.392 26.158 17.0 2.5

20030707 104548.6 40.409 26.159 12.4 3.0

20030707 124932.2 40.431 26.193 16.2 2.9

20030707 140802.0 40.403 26.031 13.0 2.8

20030707 142706.5 40.471 26.230 10.3 2.4

20030707 143056.5 40.407 26.202 12.3 2.2

20030707 151355.9 40.343 26.176 18.2 2.3

20030707 151649.6 40.403 26.188 10.4 2.7

20030707 161738.7 40.403 25.921 12.7 3.4

20030707 164402.1 40.398 26.182 17.8 3.0

20030707 164540.3 40.386 25.925 13.1 3.4

20030707 165339.3 40.399 26.186 18.1 2.7

20030707 185029.1 40.371 25.928 10.9 2.7

20030707 195745.3 40.396 25.897 10.7 3.3

20030707 211828.0 40.400 26.021 14.8 2.7

20030707 214113.2 40.416 26.177 15.0 2.4

20030707 232819.4 40.436 26.134 13.5 2.6

20030707 234430.3 40.410 26.093 12.6 2.5

20030708 012445.1 40.397 25.957 10.6 2.6

20030708 043123.8 40.411 26.177 15.4 3.3

20030708 072856.0 40.430 26.201 15.8 2.6

20030708 101231.5 40.447 26.056 8.3 2.5

20030708 101453.7 40.402 26.006 14.5 3.0

20030708 172054.6 40.418 26.232 13.6 2.3

20030708 192936.3 40.429 26.218 14.0 2.5

20030708 205123.2 40.397 26.072 18.2 2.4

20030708 234227.5 40.416 26.175 17.2 3.1

20030709 000651.1 40.389 25.940 9.4 2.4

20030709 000758.2 40.400 25.920 10.3 3.4

20030709 002116.0 40.391 26.165 17.9 2.6

20030709 002306.6 40.387 25.916 10.2 2.9

20030709 085042.2 40.219 26.052 26.8 2.8

20030709 205131.6 40.440 26.158 15.5 2.8

20030709 220157.5 40.385 25.913 12.6 3.8

20030709 220849.5 40.386 25.902 11.9 4.1

20030709 223140.7 40.388 25.912 15.8 4.7

20030709 223708.6 40.370 25.882 9.5 3.4

20030710 000504.0 40.368 25.913 11.9 3.0

20030710 012617.7 40.387 25.912 16.5 4.4

20030710 073345.8 40.365 25.902 10.6 2.4

20030710 132533.4 40.381 25.897 10.2 3.4

20030710 151028.9 40.435 26.162 10.7 2.1

20030710 201445.8 40.386 25.893 10.1 3.1

20030710 203052.1 40.383 25.856 6.7 2.9

20030711 025628.2 40.397 25.877 7.0 2.7

20030711 072248.7 40.421 26.165 15.3 3.2

20030711 235114.3 40.191 25.283 21.0 3.8

20030713 020640.3 40.395 26.135 11.9 2.9

20030713 062015.8 40.397 25.921 10.8 3.0

20030713 063208.1 40.389 25.923 13.9 4.0

20030713 101250.8 40.431 26.064 18.1 3.0

20030715 214938.9 40.396 26.145 10.5 3.0

20030716 100248.4 40.261 26.195 22.1 2.7

20030716 162439.8 40.180 25.293 19.3 3.5

20030718 054407.2 40.394 25.962 14.0 3.8

20030718 125212.2 40.436 26.116 18.7 3.3

20030723 193706.0 40.437 26.153 16.8 2.8

20030805 034844.9 40.423 26.004 10.5 2.8

20030810 083635.3 40.314 25.901 17.3 2.8

20030811 231429.9 40.401 26.208 11.0 2.6

20030831 075056.7 40.415 25.972 17.0 4.0

20030914 091526.7 40.365 25.980 13.1 2.8


Appendix B

Focal mechanisms of the earthquakes listed in Table 2. For each station the vertical, tangential and radial

components are shown. Observed waveforms are marked with solid lines and synthetics with dashed lines. Station

names and variance reduction are listed below each waveform. At the right part of the figure the parameters of the

focal mechanism and the moment magnitude are shown along with the percentage of the double couple (DC) and of

the compensated linear vector dipole (CLVD) for each earthquake.






References

Altunel, E., Meghraoui, M., Akyuz, H.S., Dikbas, A., 2004. Char-

acteristics of the 1912 co-seismic rupture along the North Anato-

lian Fault Zone (Turkey): implications for the expected Marmara

earthquake. Terra Nova 16, 198–204.

Ambraseys, N.N., 1990. Engineering seismology. Earthquake Eng.

Struct. Dyn. 17, 1–105.

Ambraseys, N.N., Finkel, C.F., 1987. The Saros–Marmara earthquake

of 9 August 1912. Earthquake Eng. Struct. Dyn. 15, 189–211.

Ambraseys, N.N., Finkel, C.F., 1991. Long-Term Seismicity of Istan-

bul and the Marmara Region. Eng. Seismol. Earthquake Rep., vol.

91/8. Imperial College.

Ambraseys, N.N., Jackson, J., 2000. Seismicity of the Sea of Marmara

(Turkey) since 1500. Geophys. J. Int. 141, F1–F6.

Armijo, R., Pondard, N., Meyer, B., Ucarkus, G., Mercier de Lepinay,

B., Malavieille, J., Dominguez, S., Gustcher, M.-A., Schmidt, S.,

Beck, C., Cagatay, N., Cakir, Z., Imren, C., Eris, K., Natalin, B.,

Ozalaybey, S., Tolun, L., Lefevre, I., Seeber, L., Gasperini, L.,

Rangin, C., Emre, O., Sarikavak, K., 2005. Submarine fault scarps

in the Sea of Marmara pull-apart (North Anatolian Fault): impli-

cations for seismic hazard in Istanbul. Geochem. Geophys. Geo-

syst. 6, Q06009. doi:10.1029/2004GC000896.

Barka, A., Akyuz, H.S., Altunel, E., Sunal, G., Cakir, Z., Dikbas, A.,

Yerli, B., Armijo, R., Meyer, B., de Chabalier, J.B., Rockwell, T.,

Dolan, J.R., Hartleb, R., Dawson, T., Christofferson, S., Tucker,


A., Fumal, T., Langridge, R., Stenner, H., Lettis, W., Bachhuber,

J., Page, W., 2002. The surface rupture and slip distribution of the

17 August 1999 Vzmit earthquake (M 7.4), North Anatolian Fault.

Bull. Seismol. Soc. Am. 92 (1), 43–60.

Benetatos, C., Roumelioti, Z., Kiratzi, A., Melis, N., 2002. Source

parameters of the M 6.5 Skyros Island (North Aegean Sea)

earthquake of July 26, 2001. Ann. Geophys. 45 (3/4), 513–526.

Cakir, Z., de Chabalier, J.-B., Armijo, R., Meyer, B., Barka, A., Peltzer,

G., 2003. Coseismic and early postseismic slip associated with the

1999 Izmit earthquake (Turkey), from SAR interferometry and

tectonic field observations. Geophys. J. Int. 155, 93–110.

Clayton, R.W., Wiggins, R.A., 1976. Source shape estimation and

deconvolution of teleseismic body waves. Geophys. J. R. Astron.

Soc. 47, 151–177.

Dreger, D., 1994. Empirical Green’s function study of the January 17,

1994 Northridge, California earthquake. Geophys. Res. Lett. 21,

2633–2636.

Dreger, D., 2002. Manual of the Time-Domain Moment Tensor

Inverse Code (TDMT_INVC), Release 1.1. Berkeley Seismolog-

ical Laboratory. 18 pp.

Dreger, D., Helmberger, D., 1990. Broadband modeling of local

earthquakes. Bull. Seismol. Soc. Am. 80, 1162–1179.

Dreger, D., Helmberger, D., 1991. Complex faulting deduced from

broadband modelling of the 28 February 1990 Upland earthquake

(ML=5.2). Bull. Seismol. Soc. Am. 81, 1129–1144.

Dreger, D., Helmberger, D., 1993. Determination of source para-

meters at regional distances with single station or sparse network

data. J. Geophys. Res. 98, 8107–8125.

Hartzell, S., 1978. Earthquake aftershocks as Green’s functions. Geo-

phys. Res. Lett. 5, 1–4.

Kiratzi, A., 2002. Stress tensor inversions along the westernmost

North Anatolian Fault Zone and its continuation into the North

Aegean Sea. Geophys. J. Int. 151, 360–376.

Kissling, E., Ellsworth, W.L., Eberhart-Phillips, D., Kradolfer, U.,

1994. Initial reference models in local earthquake tomography. J.

Geophys. Res. 99 (B10), 19635–19646. doi:10.1029/93JB03138.

Kreemer, C., Chamot-Rooke, N., Le Pichon, X., 2004. Constraints on

the evolution and vertical coherency of deformation in the North-

ern Aegean from a comparison of geodetic, geologic and seismo-

logic data. Earth Planet. Sci. Lett. 25, 329–346.

Kurt, H., Demirbag, E., Kuscu, I., 2000. Active submarine tectonism

and formation of the Gulf of Saros, Northeast Aegean Sea,

inferred from multi-channel seismic reflection data. Mar. Geol.

165, 13–26.

Lee, W.H.K., Lahr, J.C., 1975. HYPO71: a computer program for

determining hypocenter, magnitude and first motion pattern of

local earthquakes. USGS Open File Rep. 75–311, 1–116.

Le Pichon, X., Sengor, A.M.C., Demirbag, E., Rangin, C., Imren, C.,

Armijo, R., Gorur, N., Cagatay, N., Mercier de Lepinay, B.,

Meyer, B., SaatcVlar, R., Tok, B., 2001. The active main Marmara

Fault. Earth Planet. Sci. Lett. 92 (4), 595–616.

Mori, J., Hartzell, S., 1990. Source inversion of the 1988 Upland

earthquake: determination of a fault plane for a small event. Bull.

Seismol. Soc. Am. 80, 278–295.

Novotny, O., Zahradnik, J., Tselentis, G.-A., 2001. North-western

Turkey earthquakes and the crustal structure inferred from surface

waves observed in Western Greece. Bull. Seismol. Soc. Am. 91,

875–879.

Okay, A., Demirbag, E., Kurt, H., Okay, N., Kuscu, I., 1999. An

active, deep marine strike-slip basin along the North Anatolian

fault in Turkey. Tectonics 18, 129–147.

Ozalaybey, S., Karabulut, H., Ergin, M., Aktar, M., TapVrdamaz, C.,

Bicmen, F., Yoruk, A., 2003. Seismogenic zones of Marmara Sea:

recent seismicity and focal mechanism solutions. EGS-AGU-EUG

Joint Assembly, Nice, France, 6–11 April 2003.

Papazachos, B., Papazachou, C., 1997. The Earthquakes of Greece.

Ziti Publ. Co, Thessaloniki, Greece. 356 pp.

Pasyanos, M., Dreger, D., Romanowicz, B., 1996. Towards real-time

determination of regional moment tensors. Bull. Seismol. Soc.

Am. 86, 1255–1269.

SaatcVlar, R., Ergintav, S., Demirbag, E., Inan, S., 1999. Character

of active faulting in the North Aegean Sea. Mar. Geol. 160,

339–353.

Saikia, C., 1994. Modified frequency–wave number algorithm for

regional seismo-grams using Filon’s quadrature; modelling of

Lg waves in eastern North America. Geophys. J. Int. 118,

142–158.

SakVnc, M., Yaltirak, C., Oktay, F.Y., 1999. Palaeogeographical evo-

lution of the Thrace Neogene Basin and the Tethys–Paratethys

relations at northwestern Turkey (Thrace). Palaeogeogr. Palaeo-

climatol. Palaeoecol. 153, 17–40.

Somerville, P., Irikura, K., Graves, R., Sawada, S., Wald, D., Abra-

hamson, N., Iwasaki, Y., Kagawa, Smith, N., Kowada, A., 1999.

Characterizing crustal earthquake slip models for the prediction of

strong ground motion. Seismol. Res. Let. 70, 59–80.

Taymaz, T., Jackson, J., McKenzie, D., 1991. Active tectonics of the

north and central Aegean Sea. Geophys. J. Int. 106, 433–490.

Tuysuz, O., Barka, A.A., Yigitbas, E., 1998. Geology of the Saros

graben and its implications for the evolution of the North Anato-

lian fault in the Ganos–Saros region, North-western Turkey. Tec-

tonophysics 293, 105–126.

YaltVrak, C., Alpar, B., 2002. Kinematics and evolution of the north-

ern branch of the North Anatolian Fault (Ganos Fault) between the

Sea of Marmara and the Gulf of Saros. Mar. Geol. 190, 351–366.

YaltVrak, C., Alpar, B., Yuce, H., 1998. Tectonic elements controlling

the evolution of the Gulf of Saros (Northeastern Aegean Sea).

Tectonophysics 300, 227–248.

YaltVrak, C., Alpar, B., Sakinc, M., Yuce, H., 2000a. Origin of the

Strait of Canakkale (Dardanelles): regional tectonics and the

Mediterranean–Marmara incursion. Mar. Geol. 164, 139–156

(with Erratum 167, 189–190).

YaltVrak, C., Sakinc, M., Oktay, F.Y., 2000b. Westward propagation of

North Anatolian fault into northern Aegean: timing and kinemat-

ics. Comment, Geol. 28, 187–188.

Waldhauser, F., Ellsworth,W.L., 2000. A double-difference earthquake

location algorithm: method and application to the Northern Hay-

ward fault, California. Bull. Seismol. Soc. Am. 90, 1353–1368.

Wessel, P., Smith, W.H.F., 1998. New improved version of the

Generic Mapping Tools released, EOS Trans. AGU 79, 579.

Zahradnik, J., 2002. The weak motion modelling of the Skyros island,

Aegean Sea, Mw=6.5 earthquake of July 26, 2001. Stud. Geo-

phys. Geod. 46, 753–771.

Documents

A source study of the 6 July 2003 (M 5.7) earthquake ...users.auth.gr/~zroum/Publications/Papers/Saros.pdf · the Ganos fault closer to the northern margin of the trough. The aim