Vous êtes sur la page 1sur 17

Nat Hazards (2013) 68:837–853

DOI 10.1007/s11069-013-0659-1

ORIGINAL PAPER

Assessment of the statistical earthquake hazard


parameters for NW Turkey

Banu Akol • Tolga Bekler

Received: 18 June 2012 / Accepted: 21 March 2013 / Published online: 1 April 2013
Ó Springer Science+Business Media Dordrecht 2013

Abstract We performed a probabilistic analysis of earthquake hazard input parameters,


NW Turkey covers Gelibolu and Biga Peninsulas, and its vicinity based on four seismic
sub-zones. The number of earthquakes with magnitude M C 3.0 occurred in this region for
the period between 1912 and 2007 is around 5130. Four seismic source sub-zones were
defined with respect to seismotectonic framework, seismicity and fault geometry. The
hazard perceptibility characterization was examined for each seismic source zone and for
the whole region. The probabilities of earthquake recurrences were obtained by using
Poisson statistical distribution models. In order to determine the source zones where strong
and destructive earthquakes may occur, distribution maps for a, b and a/b values were
calculated. The hazard scaling parameters (generally known as a and b values) in the
computed magnitude–frequency relations vary in the intervals 4.28–6.58 and 0.59–1.13,
respectively, with a RMS error percentage below 10 %. The lowest b value is computed for
sub-zone three indicating the predominance of large earthquakes mostly at Gelibolu
(Gallipoli) and north of Biga Peninsula (southern Marmara region), and the highest b value
is computed for sub-zone two Edremit Bay (SW Marmara region). According to the
analysis of each seismic sub-zone, the greatest risk of earthquake occurrence is determined
for the triangle of Gelibolu–Tekirdağ western part of Marmara Sea. Earthquake occurrence
of the largest magnitude with 7.3 within a 100-year period was determined to be 46 %
according to the Poisson distribution, and the estimated recurrence period of years for this
region is 50 ± 12. The seismic hazard is pronounced high in the region extending in a
NW–SE direction, north of Edremit Bay, west of Saros Bay and Yenice Gönen (southern
Marmara region) in the south. High b values are generally calculated at depths of 5–20 km
that can be expressed as low seismic energy release and evaluated as the seismogenic zone.

Keywords Northwest Anatolia  Seismic hazard  Poisson distribution

B. Akol  T. Bekler (&)


Department of Geophysics Engineering, Faculty of Engineering, Çanakkale Onsekiz Mart University,
17100 Çanakkale, Turkey
e-mail: tbekler@comu.edu.tr; banu_akol@hotmail.com

123
838 Nat Hazards (2013) 68:837–853

1 Introduction

Even today, it is still uncertain where, when and at what magnitude an earthquake will
occur. In seismically active regions, the evaluation of the probability of earthquake
occurrence and their recurrence periods is very important for selection of settlement areas.
The seismic hazard is defined as the potential of an earthquake to cause loss of life and
property in a certain region within a certain period of time. The seismic hazard estimation,
on the other hand, involves the evaluation of the seismicity within a set of seismogenic
zones, taken into consideration by using best suitable models among which the Poisson
distribution is the most commonly used statistical models. The basic equation
‘‘log N = a – b M,’’ known as Gutenberg–Richter (G–R) (1944), states that the logarithm
of the number of earthquakes is linearly dependent on the magnitude of the earthquakes.
A few studies consider significance and estimation for the investigated area. The general
statements point out that the a value is a measure of seismic activity that depends on size of
the area, observation period length, largest seismic magnitude and the stress level of the
area (Allen 1986; Khan et al. 2010). Some studies presented that calculated values vary
between 0.33 and 1.13 through the western part of Anatolia toward the Hellenic arc
(Yılmaztürk and Burton 1999).
On the other hand, there are a quite number of studies on the variation of b value. In
studies conducted in some regions that have high seismic activity, and that contain the
foremost effective faults, it is stated that the b value varies between 0.55 and 1.19
according to the type of tectonism and the number of earthquakes (Miyamura 1962;
Hatzidimitriou et al. 1985; Wang 1988; Tsapanos 1990; Tsapanos and Papazachos 1998;
Olsson 1999; Bayrak et al. 2002).
In some early studies, rock properties, stress conditions and increase in ductility affect
the b value variations; strain accumulation leads to low b values (Miyamura 1962; Allen
et al. 1965; Mogi 1967; Scholz 1968). Additionally, Weeks et al. (1978) demonstrated, as a
result of the experiments they performed on rocks, that the value of b decreases before an
earthquake occurs. Thus, b value is known as a tectonic recognizing parameter, and its
value is taken as an indicator of seismic activity and changes with respect to the seis-
motectonics. The decrease in the b value points to an energy pile up in those regions and
indicates that a major earthquake may occur. In terms of the value, Bath (1983) correlated
the b values with perceptibility and computed its average value as 0.84 for earthquakes in
Sweden (low seismicity area), while an average b value of 0.95 is found for Central
California (high seismicity area) where the San Andreas fault zone creates earthquakes
(Shi and Bolt 1982). Recent studies indicate that the b value strongly depends on physical
properties of rocks and tectonic implications. Mori and Abercrombie (1997) stated that
b values have significant variation for different tectonic regions. An area with a volcano-
tectonic system with b values less than 0.6 is interpreted as an indication of a relatively
higher stress regime (Farrell et al. 2009). This case can be associated with the current study
region which has a similar tectonic framework. The results of Giardina (1977) and
Brumbaugh and Evanzia (2010) for northern Arizona where b values are less than 1.0
suggest a somewhat higher stress level. In contradiction to Brumbaugh and Evanzia (2010)
results, for the smaller area where Giardina (1977) found a b value of 1.1, there were larger
numbers of smaller earthquakes than predicted by the G–R approach and also a lower
stress level.
Computing the modal values, namely a/b values rather than a and b values separately,
provide better understanding of zones of low and high seismic activity. This innovation
was developed to understand potential sources of earthquakes and to evaluate relative

123
Nat Hazards (2013) 68:837–853 839

hazard levels of sub-zones (Bayrak et al. 2002). Consideration of the modal values to
evaluate the seismic hazard in Biga Peninsula (NW Turkey) is the basis of the current
study.
In the last decade, earthquake hazard studies have been a remarkably enticing subject
due to the high seismicity of the North Anatolian Fault (NAF) zone and its strands.
Earthquake perceptibility studies are an exceptional hazard assessment tool in addition to
G–R parameters. Öztürk et al. (2008) performed an earthquake perceptibility study using
Gumbel’s statistical theory of extreme values in Turkey and the surrounding areas. They
estimated the most probable earthquake magnitude will exceed 7.5 over a 100-year period.
They also determined the risk of the occurrence of an earthquake with a magnitude of 7.0
in southern strand of the NAF to be 93 % over 100 years. By using Poisson distribution,
the occurrence probabilities of an earthquake with a magnitude greater than Ms C 7 in the
next 100 years were calculated to be 60.3 % for Tekirdağ in the north, 50.9 % for Saros
Bay and 57.3 % for the Biga Peninsula (Sayıl and Osmanşahin 2005).
An area covering the southern part of the Marmara region and the northern part of the
Aegean region in NW Turkey is taken into consideration in this study. The examined
region is dominated by active normal and oblique slip faults, with locally high rates of
moderate magnitude earthquakes, and shows evidence of spatial and temporal partitioning
of deformation and displacement during transtension (Gürer et al. 2006). Thus, the
southwest Marmara region is considered to be an ideal one for the current study of hazard
parameters in consequence of deformation in western Anatolia since active faults are
presented and mapped in considerable detail. A comprehensive recent data and detailed
tectonic features of the seismic sources make this study more prominence for especially
NW Turkey by comparing the same statistical approaches used by Yılmaztürk and Burton
(1999).
Differences in the results of previous research indicate that the region presents complex
behavior, and additional studies are needed to explain the hazard assessment. Few studies
reflect the earthquake perceptibility and hazard status due to the different fault composi-
tions. In this study, so-called essential parameters for the hazard assessment, a, b and
a/b values for northwest Anatolia and its surroundings are calculated in order to provide
simplified evaluation of the probabilistic hazard maps of the investigated region.

2 Seismotectonic framework

The northward movement of the African and Arabian plates affects the tectonic complexity
of Turkey. This general setup yielded the most remarkable left lateral fault system, the
North Anatolian Fault Zone (NAFZ). The Marmara region, between the Bosphorus in
Istanbul and the Dardanelles in Çanakkale, belongs to the NAFZ. Recent GPS measure-
ments indicate that the Anatolian block is moving coherently away from the Eastern
Anatolian constructional boundary to the South Aegean consuming boundary (Oral et al.
1995). This movement causes large earthquakes which occur along the northern boundary
of the Anatolian block and also causes large earthquakes, especially in western Anatolia
due to the diffuse deformation in the block. The NAFZ is one of the well-known fault lines
in the world in terms of earthquake activity and also due to its similarity to the San Andreas
fault zone in North America.
In the west of the NAFZ, boundary is diffused and there is a wide zone composed of
strike-slip faults that lie sub-parallel to each other. The northern border of this diffuse zone
is the Marmara fault lying almost along the axis of the three main deep basins of the

123
840 Nat Hazards (2013) 68:837–853

Marmara Sea (Arpat and Şentürk 2000; İmren et al. 2001). There are a couple of right
lateral fault zones that are ENE–WSW and NE–SW trending passing through the Biga
Peninsula in the southern branch of the NAFZ. Figure 1 summarizes the main tectonic
elements of the Anatolian and Aegean blocks. Even though in northwest Anatolia the
majority of the western movement of the Anatolian block with respect to the Eurasian plate
is covered along the Marmara fault with a speed of 20 mm/year, right lateral slip with
velocities up to 5 mm/year is modeled by using GPS measurements on the sub-parallel
faults in the south (Flerit et al. 2003; Pondard et al. 2007). It is also very common to find
thermal sources along the fault zones where major earthquakes occur (Pfister et al. 1997;
Öcal et al. 1968; McKenzie 1972) in the Biga Peninsula (Ambraseys and Finkel 1991).
More specifically, extensions of the NAFZ are traced in the Biga Peninsula with dif-
ferent lineations and length of segmentation. The latter of course prepossesses the current
study in considering earthquake size. Gaziköy, Etili, Biga, Sarıköy and Yenice fault zones
are known active members of NAFZ in the south. The Saros–Gaziköy fault which is right
lateral caused the 7.3 magnitude Mürefte–Şarköy earthquake in 1912. The Çan–Biga fault
zone where a 6.4 magnitude earthquake occurred in 1935 is comprised of many northeast–
southwest directional fault segments. The length observed on the surface of the Yenice–
Gönen fault which caused the Ms = 7.2 magnitude earthquake in 1953, known as the
Yenice–Gönen earthquake, is about 50 km (Herece 1990). The separation of the different
faults which have generated large magnitudes in the Biga Peninsula allows us to divide this
region into four sub-zones.

3 Data and methods

39°–41°N latitudes and 25°–28°E longitudes form the borders of the studied area for
Çanakkale (center of investigated area) and its surroundings. The earthquake catalog
information is compiled from Boğaziçi University Kandilli Observatory and Earthquake
Research Institute (Kalafat et al. 2007), International Seismological Center (ISC) and USA
Advance National Seismic System composite catalogs. A combined different earthquake
catalogs mainly provide both data quality and magnitude completeness for a given period
in constituting a homogeneous database.
Sub-zone selection is completely tied to a typical scale related to the spatial structure of
the fault pattern dominating the zone and the tectonic conditions there in (Molchan et al.
1997). Seismic sub-source zones and boundaries are determined by two fundamental tools,
a seismicity profile and the homogeneous tectonic structure of the region under consid-
eration. Some authors carried out earthquake hazard studies which considered the current
study region as a solitary source zone dominated by a specific faulting pattern that may be
responsible for this homogeneous area (Erdik et al. 1985; Bayrak et al. 2008).
The configuration of the source zones (Fig. 2) is chosen firstly by the areal distribution
of epicenters up to the end of 2007 and secondly by using dominant characteristics of the
faulting types, relevant tectonic and geological. From a general point of view, the geology
of the NW Turkey (forms of Gelibolu and Biga Peninsulas) is dominated by two main
features: (1) the widespread occurrence of Tertiary plutonic and associated volcanic rocks
(for the former always connected with metamorphic units), and the (2) the general NE–SW
trend of all the geological units underlined by mainly strike-slip, dextral faults (Beccaletto
and Jenny 2004).
Zone A1: Almost pure strike-slip faulting is observed, and commonly small size
earthquakes are occurred in the Saros trough, western edge of the NAF zone. The fault that

123
Nat Hazards (2013) 68:837–853 841

Fig. 1 Map ‘‘a’’ showing the major tectonic boundaries of the broader Anatolian block (modified after
Papazachos and Kiratzi 1996; Barka and Reilinger 1997). The major faults and extensions are also shown
with black lines and dashed arrows. Large arrows show the westward motion of the Anatolian block and
southwest motion of the southern Aegean, both relative to Eurasia. Shaded-relief map ‘‘b’’ shows the
simplified tectonics of the studied area. The shadow area indicates the approximate boundary of the Biga
Peninsula. Blue lines denote active faults suggested by Barka (1992, 1997)

123
842 Nat Hazards (2013) 68:837–853

Fig. 2 The distributions of the epicenter in four local seismic sub-zones and depths of the earthquakes in
the study region with magnitudes of M C 3 according to longitude and latitude on a simplified tectonic map
of western Marmara region

crosses the Gelibolu Peninsula, and which was responsible for the August 09, 1912,
Mw = 7.3 earthquake (Ambraseys and Jackson 2000) is a continuation of the ENE trend of
the Saros trough. This earthquake was the most striking one that surface rupture in 1912
also showed a combination of right lateral and normal slips (Taymaz et al. 2004), which is
reflected in the topography on this source zone. Thick sedimentary succession covers this
source zones along NE–SW direction. The largest aftershock of Ms = 6.8 occurred on
September 13, 1912 (Ambraseys and Finkel 1987). This zone has a wide range of small-
and moderate-sized earthquakes except a few of big events such as August 20, 1917
(Ms = 6.0).
Zone A2: The Yenice–Gönen fault is one of the most important active faults in the
source zone, and a very destructive earthquake Mw = 7.1 on March 18, 1953 (McKenzie
1972), occurred on the Yenice–Gönen fault, causing 263 losses. Strike slip with minor
normal composition of fault is responsible for the seismic activity in Biga Peninsula. As for
instrumental period, Çan–Biga (Ms = 6.4) in 1935 and Biga (Ms = 6.0) in 1983 are some
of the remarkable events on this zone. On the other hand, very narrow ophiolite-meta-
morphic assemblage seems a frontier between sedimentary succession of Gelibolu Pen-
insula at the north and large metamorphic and granite group of Biga Peninsula at the south.
Zone A3: The zone is bounded by the region between Lesvos Island, and the Edremit
Bay is one of the most tectonically active areas in the Northern Aegean Sea. The southern
strand of the NAF zone has some characteristic features of a large zone consisting of some
parallel faults seen on this source zone. NNE–SSW- and NW–SE-trending faults and the

123
Nat Hazards (2013) 68:837–853 843

associated some of large grabens are the most remarkable feature. The October 6, 1944
(Ms = 6.8), earthquake produced tsunami waves and damaged some coastal localities in
the Edremit Bay on a local scale (Altınok et al. 2012). At the western edge of this zone, on
December 19, 1981, a strong earthquake of magnitude Ms = 7.2 in the sea west of island
Lesbos (Kiratzi et al. 1991; Papazachos et al. 1984).
Zone A4: Two devastating earthquakes occurred in this zone: One is November 18,
1919 (Ms = 6.9), and the other one is September 22, 1939 (Ms = 6.6). The information on
both earthquakes is limited. The 1919 event ruptured a fault differing in strike by an E–W
normal fault in a part of Aegean extension (Nalbant et al. 1998). The shear deformation is
significantly different in this source zone related to complex composition of faulting
pattern.
Clearly, the classified earthquake catalog and well-defined faulting segmentation in the
study region make the selection of uncomplicated boundaries of each sub-zone much
easier. The distributions of the earthquakes with magnitudes of M C 3.0 that have taken
place in the studied region between the years 1912 and 2007 and the selected seismic sub-
zones are given in Fig. 2. In order to achieve homogeneity, all the magnitudes have been
examined as surface wave magnitude (Ms). We preferred required empirical relationship
between Ms and other magnitudes proposed by Kalafat et al. (2007). The earthquake
catalog includes magnitudes and focal depth information generally completed after 1970.
A few of the magnitudes larger than 4.0 have been supplemented by combining ISC
catalog with the National one (Kandilli Observatory and Earthquake Research Institute)
before 1970. The smallest magnitude from which earthquakes were constantly reported in
the catalogs of examined period has been chosen as a minimum magnitude (Mmin = 3.0) in
our case for all sub-regions. Maximum magnitude value (Mmax) is magnitude of the largest
earthquake that occurred in any sub-zone.
The G–R distribution applies to global catalogs of earthquakes or small geographical
regions and estimates the number of earthquakes greater than or equal to the magnitude
M that can be expected in some time period for a selected region, if accurate values of the
a and b parameters are known. Therefore, the G–R distribution plays a major role in
earthquake forecasting and in subsequent earthquake hazard modeling (Smyth and Mori
2009). The possible variations of the a and b values are important in evaluating the seismic
hazard mainly correlated with seismicity and overall tectonic framework. One of the
seismicity parameters is the a value, which depends on the observation period, the order of
the region and the seismic activity, and defines a mean annual seismic activity index. The
b value varies according to the tectonic characteristics of the region with respect to the
fault geometry by means of earthquake size (Okal and Romanowicz 1994).
Spatial variations of a and b values are estimated from magnitude frequency relation-
ships for the whole region (WR) and for each sub-zone individually. The WR (Fig. 2) was
divided into 35 cell grids for well-defined resolution. Each grid has a dimension of
0.5° 9 0.5°. The entire area was re-gridded by 0.1° 9 0.1° to enhance the image quality
taking account of number of events.
For each grid, a minimum number of events, Emin with M C Mc, are required in order to
estimate a reliable a and b values. For each grid defined in each zone that contains less than
Emin events, the values are not calculated. Nmin is erratically set to 30, because below this
value the uncertainty in b increases rapidly (Schorlemmer et al. 2004). Here, Mc magnitude
completeness threshold for the investigated area is in range 3.0 B Mc B 4.2 of the studied
time period. The catalog was cut out based on this interval, and the rest of events was then
used to calculate a and b values to verify that Mc is consistent throughout the time period
of the catalog as well as all over the WR. The dimension of the WR under the present study

123
844 Nat Hazards (2013) 68:837–853

is selected to have a sufficient number of events representing the a and b values for each
grid. Note that a few grids having insufficient events were excluded from further analysis
and evaluation.
The parameters a and b values are calculated using maximum likelihood (ML)
approach. ML, in providing equal weighting to all earthquakes, is the statistically correct
approach for estimating b values rather than linear least square fitting (Aki 1965; Utsu
1965). The ML estimation of b value has some advantages over the standard linear least-
squares calculation, because it weights the occurrence of each earthquake equally. Since
the linear G–R relationship is graphed with the log of the number of earthquakes, a simple
linear regression weights the data points representing fewer higher magnitude earthquakes
as much as those representing the many more earthquakes at lower magnitudes when
determining the linear fit.
In the context of goodness of fit, the following criteria are usually used to make
uncertainty analysis of hazard parameter (a and b values) maps, measures based on the
closeness of the simulated (calculated) values to the observed earthquake catalog database.
Although alternative measures may be defined and used to quantify the merit of different
uncertainty analysis, the following methods are most popularly employed.
To compare the closeness of modeled volumes to the observed volumes, the percentage
root mean square error (%RMSE) is defined as follows in Eq. 1:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
P a a
Vcal  Vobs 100  n
%RMSE ¼ :P 2 ð1Þ
n Vobs
a a
Here, Vcal and Vobs are the calculated and observed database on link hazard parameters
(a and b values), respectively, n is the total number of observation. This statistical
approach is used to indicate the quality of fitting level. The smaller the RMS value
obtained for a set of parameters, the closer the modeled parameters to the (target) realistic
ones. That is, the parameters much likely represent the real case and their use provides a
better understanding for the region under consideration. The RMS value also helps define
confident limit check for each parameter recovered. A simple approach is to relax the RMS
value, say 10 %, and find the family of each parameter by trial and error or using the most
square approach responds of which provides the RMS below or equal to relaxed RMS
value (Meju and Hutton 1992). Then, the extreme values in each family may be taken as
lower and upper bounds of the related parameters. Despite relaxing the RMS, if the range
remains narrow, it means that the data have enough information content to solve this
parameter, and it can be used with a high degree of confidence. On the other hand, if the
variation range is large, then the parameter provides less meaningful information for the
region. At the depth of the high seismicity zone, the number of events with small and
moderate size magnitudes is higher than those with big magnitudes (Jimenez et al. 2005).
This is probably due to the high heterogeneity of the crust and is in agreement with
previous research, which showed that magnitudes of the most earthquakes are below
M = 4.0 (Sounders et al. 1998; Gürer and Bayrak 2007).

4 Analysis of earthquake hazard parameters

Some statistical models have been frequently preferred to represent the process of earth-
quake occurrence in time domain. One of the known ones is the Poisson model which
assumes spatial and temporal independence of all earthquakes including great earthquakes

123
Nat Hazards (2013) 68:837–853 845

(Patwardhan et al. 1980). The temporal formations of earthquakes are most frequently
defined by the Poisson model which forms an important framework for the evaluation of
probabilities. The Poisson process produces a random variable that defines the number of
occurrences of an event in a certain region within a specified time interval. The features of
the Poisson process have been listed below (Kramer 1996):
1. The number of occurrences within a certain time interval is independent of the number
of occurrences in another time interval.
2. The probability of occurrence within a very short time interval is directly proportional
to the length of this interval.
3. The probability of the occurrence of more than one event within a very short time
interval may be neglected.
These features show that the Poisson process is random and that the time, magnitude or
location of the event is independent.
According to the Poisson distribution, the distribution of the time required until the next
earthquake is not affected by the time that has passed since the previous earthquake and the
statistical data show that the Poisson model is valid for medium and high magnitude
earthquakes (Feller 1968; Tabban and Gençoğlu 1975). The risk of an earthquake with a
magnitude of M to occur in a region within an observation interval of ‘‘T’’ years can be
calculated by using the equations given below:
a0 ¼ a  logðb ln 10Þ ð2Þ
a1 ¼ a  log T ð3Þ
a01 0
¼ a  log T ð4Þ
a01 bM
nðM Þ ¼ 10 ð5Þ
Tr ¼ 1=nðM Þ ð6Þ

R1 ¼ 1  en ð M Þ ð7Þ

Rt ¼ 1  etnðMÞ ð8Þ
In these equations, n(M) represents the number of earthquakes in a year with magnitude M;
T represents the time interval; a0 , a1, a10 represent the earthquake hazard parameters; Tr
represents the recurrence period; R1 represents the annual risk and Rt represents the risk for
t years.

5 Results

For the overall region, the probability of the occurrence of an earthquake with a magnitude
of M C 7.0 within the next 100 years is determined to be 63 % according to the Poisson
distribution. The recurrence period of an earthquake with a magnitude of 7.2 was deter-
mined to be 145 ± 25 years according to Poisson distribution. In every case, the Poisson
distribution gives safer (more conservative) earthquake hazard results (Erdik et al. 2004;
Jordanovski and Todorovska 1995). Statistical data show that the Poisson model is valid
for mid- and high-range earthquakes (Gençoğlu 1972; Tabban and Gençoğlu 1975), so
Poisson distribution results should be taken into account, especially for earthquakes with
magnitudes greater than 6.0.

123
846 Nat Hazards (2013) 68:837–853

First, Poisson distribution was used to evaluate earthquake hazard parameters for the
studied region and the seismic sub-zones given in Table 1. The highest b value is in Zone
A2 (b = 1.0821 ± 0.11), the lowest a value is in Zone A3 (b = 0.5941 ± 0.045).
Additionally, the largest a/b ratio (7.22) is determined for Zone A3. Comparison of the
magnitudes of earthquakes that took place in these zones indicates that the largest earth-
quake that occurred in Zone A4 was of 5.9 magnitude, whereas in Zone A3, earthquakes
with magnitudes of 6.3, 6.4, 7.2 and 7.3 have been observed.
The recurrence probability of an earthquake with magnitude 6.0 in the WR (consists of
all source zones) over 30 years is calculated to be 84 %, whereas the recurrence period is
calculated to be 16.12 ± 5.2 years. The recurrence probability of an earthquake with a
magnitude of 7.0 in 30 years is 63 %, whereas its recurrence period is determined to be
101 ± 23 years (Table 2; Fig. 3a).
For Zone A1 and A2, the occurrence probability of an earthquake with a magnitude of
6.0 in 100 years is determined to be 39 and 29 %, and its recurrence period is
199.6 ± 23.2 and 285.32 ± 31.3, respectively. The small differences are in perceptibility
percentages due to the seismic activity that is rather low in zone A1. For Zone A2, the
occurrence probability of an earthquake with a magnitude of 7.2 in 100 years is 53 %, and
its recurrence period is 133.83 ± 41.2 years. The zone A3, which is neighborhood to A2,
has almost closer hazard values. Whereas for Zone A4 where extensional type of tectonic
regime exists, the occurrence probability of an earthquake with a magnitude of 6.9 in
100 years is 63 %, and its recurrence period is 101.02 ± 21.7 years.
The cumulative distribution of b values is non-uniformly distributed over the entire depth
range; the lowest value is found at a depth of 18 km in Zone A1 and 32 km in Zone A2. The
change in slope of the distribution might be due to variation in the crustal structure. A gentle
change in rheology suggests that the crust may be divided into two zones: the mid- and lower crust.
The values of b are high at an average depth of 0–20 km (Fig. 4). The region at these depths can be
evaluated as the seismogenic zone. Below an average depth of 30 km, the number of earthquakes
is very low. This result is in accordance with the average crustal thickness of the Marmara Region
(30 ± 2 km) which displays great heterogeneity. Ulugergerli et al. (2007) also proposed a similar
result for western Anatolia using magnetotelluric and seismological data.
A map of estimated a values (Fig. 5) shows a relative high over the area. This might be
associated with higher stress. This seems to dominate along the SW–NW direction. The
a value contours follow the trend of major regional tectonic structures. Brittle failure of the
weaker crust at lower stress levels along this direction might account for higher a values
(Kumar et al. 2011).
High seismicity with high b values is commonly associated with a high degree of crustal
heterogeneity and corroborates the above observation (Watson et al. 2002). Crustal

Table 1 Earthquake hazard parameters with RMS error (RMSE) percentage (Poisson distribution) for the
studied region and each seismic sub-zone
Earthquake hazard parameters a %RMSE b %RMSE a0 a1 a10 a/b

Seismic zones WR 5.85 9 0.80 8 5.59 3.84 3.58 7.34


A1 5.77 8 0.95 6 5.43 3.76 3.43 6.06
A2 5.39 10 0.80 10 5.13 3.38 3.12 6.80
A3 4.28 10 0.59 9 4.15 2.27 2.14 7.21
A4 6.14 6 1.08 6 5.75 4.13 3.74 5.68
WR whole region

123
Nat Hazards (2013) 68:837–853 847

Table 2 The seismic risk and recurrence periods obtained for the each source zone (SZ) cited in Fig. 2 by
using a Poisson distribution approach
Ms-SZ n(M) R1 R10 R30 R50 R75 R100 Tr

6.0-A1 0.0050 0.004 0.05 0.13 0.22 0.31 0.39 199.60


6.0-A2 0.0030 0.003 0.03 0.10 0.16 0.23 0.29 285.32
6.4-A2 0.0215 0.020 0.19 0.47 0.65 0.80 0.88 46.400
6.6-A4 0.0298 0.030 0.26 0.59 0.77 0.89 0.95 33.580
6.8-A1 0.0340 0.014 0.13 0.35 0.50 0.61 0.71 90.980
6.8-A3 0.0350 0.014 0.13 0.33 0.48 0.61 0.70 92.500
6.9-A4 0.0335 0.013 0.11 0.28 0.41 0.52 0.63 101.02
7.2-A2 0.0072 0.007 0.07 0.20 0.32 0.43 0.53 133.83
7.2-A3 0.0072 0.007 0.07 0.19 0.30 0.41 0.51 138.60
7.3-A1 0.0074 0.007 0.06 0.16 0.25 0.35 0.43 175.22

Fig. 3 The magnitude–seismic risk values determined by using Poisson distribution for WR and each sub-
zone with error bars

complexity, including fault zones, is very common features of northwest Turkey. Based on
the distribution (Fig. 6), overall b values are widely spread over northwest Turkey. The
minimum and maximum estimated b values in the study area are 0.65 ± 0.058 and
1.55 ± 0.132, respectively. Seismic source zones of Zone A1 and Zone A2 with high
b values (Fig. 6) in some pockets in the western part of the study area could be related to
the strong regional tectonics.
Öztürk et al. (2008) scrutinized earthquake hazard parameters distributed in Turkey and
the adjacent areas by examining 24 seismic sub-regions. One of their sub-regions overlaps
with one in the current study area. In the previous study the most probable maximum
magnitudes, the mean recurrence periods (in years) and the probabilities over different

123
848 Nat Hazards (2013) 68:837–853

Fig. 4 The b value depth [a WR; b Zone A1;cZone A2;d Zone A3; eZone A4]

time periods for a given magnitude were computed in order to estimate the seismicity of
the 24 seismic regions by utilizing the Gumbel I asymptotic distribution. The earthquake
hazard parameters of a and b values were estimated at 2.43 and 0.64, respectively. For the
overall region of the current study, those values are 2.16 and 0.54. The difference is due
mainly to widely separated fault zones and their seismotectonic properties.
A hazard study was performed by Sayıl and Osmanşahin (2008) for western Turkey.
One of their sub-regions is located southwest of the Marmara region and almost overlaps
with Zone A2 in the current study. The hazard parameters of a and b values were 3.39 and

123
Nat Hazards (2013) 68:837–853 849

Fig. 5 Variation of a values in the study area with RMSE contours. Solid red line represents 10 %
threshold of CI. Dashed redline and black line show the above and the below the level, respectively

Fig. 6 Variation of b values in the study area with RMSE contours. Solid red line represents 10 %
threshold of CI. Dashed redline and black line show the above and the below the level, respectively

0.43 from Sayıl and Osmanşahin (2008), respectively. In this study, those values are
calculated as 3.05 and 0.85 for Zone A2. Large number of data and magnitude threshold
(M C 4) are the factors in the difference between the results. Epicentral distribution and
the time period of earthquake catalogs used in this study are quite sufficient in order to
evaluate seismicity and further detailed hazard perceptibility analysis comparing with
previous studies (Yılmaztürk and Burton 1999; Bayrak et al. 2005).
This presentation of earthquake hazards is also a powerful tool that provides detail of
the zones of low and high seismic activity, and it may be used as a measure of potential
non-defined or suspicious faults and relative hazard levels. For this purpose, low a/b values

123
850 Nat Hazards (2013) 68:837–853

Fig. 7 Variation map of a/b values in the study area with RMSE contours. Solid red line represents 10 %
threshold of CI. Dashed redline and black line show the above and the below the level, respectively

were obtained for the center and around Çanakkale province (Fig. 7). The region of low
RMS error percentage represents the best-fit solution to evaluate the confidence levels
(CIs) of hazard parameters. Figures 5, 6 and 7 show the RMS error percentage contour
diagrams that have been calculated using the relationship described in Eq. (1). Large and
destructive earthquakes yield high a/b values. Small b values are also observed in some
regions in Turkey (Bayrak et al. 2005). When we look at our seismic regions from such a
perspective, we see the largest a/b value (a/b = 7.2119) and the lowest b value
(b = 0.5941 ± 0.057) in Zone A3. This result is in accordance with the fact that earth-
quakes with magnitudes of 6.3, 6.4, 7.2 and 7.3 have taken place in Zone A3, whereas the
a/b value at Zone A2 is 6.7951 ± 0.70. The same value is 5.6785 ± 0.45 for Zone A2.
When we look at the highest magnitude earthquakes that have taken place in these regions,
magnitudes of 6.0, 6.6, 6.8 and 7.0 are known from Zone A2, and 5.0, 5.4, 5.5 and 5.9 from
Zone A4.

6 Conclusion

When we analyze the a/b modal distribution maps for the b value, the tectonic parameter of
the region, comparatively higher a/b values, is observed together with low b values. These
regions are generally in the vicinity of fault zones and point to regions where high mag-
nitude earthquakes occurred. When earthquakes with magnitudes of M C 3.0 occurring
within the last 100 years are analyzed, Saros–Gaziköy and Yenice–Gönen faults are
located where the b value has decreased relatively and where the risk of earthquake
occurrence has increased. These faults form western continuation of the main NAFZ which
has caused high magnitude earthquakes in the past. Low a/b values have been obtained in
the center of Çanakkale and its surrounding area. The b values range between 1.0 and 1.15
in this region with 90 % CI. High magnitude earthquakes have not been observed in the
past. However, this does not mean that it will not be affected by large magnitude and
destructive earthquakes that might occur along the faults. The occurrence probabilities and

123
Nat Hazards (2013) 68:837–853 851

recurrence periods obtained, especially for earthquakes with magnitudes of seven and
above, by using Poisson distributions support previous studies on this region, whereas the
probability of occurrence of an earthquake with a magnitude of seven and above is 90 %
for the next 100 years, and its recurrence period is 50-year average. The probability of
occurrence was determined to be 65 % by using the Poisson distribution, whereas the
recurrence period is calculated to be 100 years. When the values of tectonism and seis-
micity compatible a, b and a/b value distributions are analyzed, it is seen that regions with
large magnitude earthquakes and weak zones are determined. Accordingly, north of the
Gulf of Edremit, offshore Tekirdağ-NW–SE extension region, between Yenice–Gönen and
west of the Gulf of Saros are regions with high risk of earthquake occurrence.
Another region which is shown to have lower b and higher a/b values on maps and for
which a fault zone is not known in the region north of the Gulf of Edremit. An earthquake
with high magnitude has not been observed during the years of 1912–2007. Also, a fault
zone is not observed in this region on the active fault map. Lower a/b values and higher
b values have been obtained in the western part of Edremit Bay. When the general geology
of the region is considered, this may be due to the Kestanbol granites, north of Edremit
Bay, that have a sturdier lithology. The investigated region of the current study is heavily
tectonized. Ophiolite and metamorphic assemblage may be one of the factors associated
with high seismicity and may play a role in value of earthquake hazard parameters.

Acknowledgments We would like to thank the Boğaziçi University Kandilli Observatory and Earthquake
Research Institute UDIM for supplying the earthquake catalog. We would like to thank Dr. E. U. Ulugergerli
(Petronas Research and Scientific Service) for his valuable critiques. The software used in this study was
made available by DEİVİL (ÇOMÜ Earthquake Observation and Data Processing Laboratory).

References

Aki K (1965) Maximum likelihood estimate of b in the formula log N = a - bM and its confidence limits.
Bull Earthquake Res Inst Uni Tokyo 43:1647–1650
Allen JRL (1986) Earthquake magnitude-frequency, epicentral distance, and soft-sediment deformation in
sedimentation basins. Sedimen Geol 46:67–75
Allen CR, Amand P, Richter CF et al (1965) Relation between seismicity and geological structure in the
southern California region. Bull Seismol Soc Am 55:52–797
Altınok Y, Alper B, Yaltırak C, Pınar A, Özer N (2012) The earthquakes and related tsunamis of October 6,
1944 and March 7, 1867; NE Aegean Sea. Nat Hazards 60:3–25
Ambraseys NN, Finkel C (1987) The Saros–Marmara earthquake of 9 August 1912. J Earthquake Eng Struct
Dyn 15:189–212
Ambraseys NN, Finkel CF (1991) Long-term seismicity of Istanbul and of the Marmara Sea region. Terra
Nova 3:527–539
Ambraseys NN, Jackson JA (2000) Seismicity of the Sea of Marmara (Turkey) since 1500. Geophys J Int
141(3):F1–F6
Arpat E, Şentürk K (2000) Marmara Denizi’nin Gelişimi. Marmara Denizi 2000 Sempozyumu (11-12
Kasım 2000, İstanbul) (in Turkish) 231–237
Barka A (1992) The North Anatolian fault. Ann Tecton 6:164–195
Barka A (1997) Neotectonics of the Marmara Sea. In: Schindler C, Pfister M (eds) Active tectonics of
northwest Anatolia: the Marmara project. Verlag der Fachvereine, Zurich, pp 55–87
Barka A, Reilinger R (1997) Active tectonics of the eastern Mediterranean region deduced from GPS,
neotectonic, and seismicity data. Annali Geofisica 40:587–610
Bath M (1983) Earthquake magnitude-recent research and current trends. Earth Sci Rev 17:315–398
Bayrak Y, Yılmaztürk A, Öztürk S (2002) Lateral variations of the model (a/b) values for the different
regions of the world. J Geodyn 34:653–666
Bayrak Y, Yılmaztürk A, Öztürk S (2005) Relationships between fundamental seismic hazard parameters
for the different source regions in Turkey. Nat Hazards 36:445–462

123
852 Nat Hazards (2013) 68:837–853

Bayrak Y, Öztürk S, Koravos G et al (2008) Seismicity assessment for different regions in and around
Turkey based on Instrumental data: gumbel first asymptotic distribution and Gutenberg–Richter
cumulative frequency law. Nat Hazard Earth Syst Sci 8:109–122
Beccaletto L, Jenny C (2004) Geology and correlation of the Ezine zone: a Rhodope fragment in NW
Turkey? Turkish J Earth Sci 13:145–176
Brumbaugh DS, Evanzia DA (2010) A preliminary seismic hazard study in Northern Arizona: another look
at the b-value. AGU, Fall Meeting 2010, abstract: S33B-2080
Erdik M, Doyuran V, Akka N et al (1985) Probabilistic assessment of the seismic hazard in Turkey.
Tectonophysics 117(3–4):295–344
Erdik M, Demircioğlu M, Sesetyan K et al (2004) Earthquake hazard in Marmara Region, Turkey. Soil Dyn
Earthq Eng 24(8):605–631
Farrell J, Husen S, Smith RB (2009) Earthquake swarm and b-value characterization of the Yellowstone
volcano-tectonic system. J Volcanol Geoth Res 188:260–276
Feller W (1968) An introduction to probability theory and its applications, vol 1. Wiley, USA, pp 156–159
Flerit F, Armijo R, King GCP et al (2003) Slip partitioning in the Sea of Marmara pull-apart determined
from GPS velocity vectors. Geophys J Int 154:1–7
Gençoğlu S (1972) Kuzey Anadolu Fay Hattının Sismisitesi ve Bu Zon Üzerinde Sismik Risk Çalışmaları.
Kuzey Anadolu Fayı ve Deprem Kuşağı sempozyumu, MTA. Enstitüsü Ankara (in Turkish)
Giardina S (1977) A regional seismic evaluation of Flagstaff. Ariz Assoc Eng Geol Bull 14(2):89–103
Gürer A, Bayrak M (2007) Relation between electrical resistivity and earthquake generation in the crust of
West Anatolia, Turkey. Tectonophysics 445:49–65
Gürer ÖF, Sangu E, Özburan M (2006) Neotectonics of the SW Marmara region, NW Anatolia, Turkey.
Geol Mag 143:229–241
Gutenberg B, Richter CF (1944) Frequency of earthquakes in California. Bull Seismol Soc Am 34:185–188
Hatzidimitriou PM, Papadimitrou EE, Mountrakis DM et al (1985) The seismic parameter b of the fre-
quency-magnitude relation and its association with geological zones in the area of Greece. Tectono-
physics 120:141–151
Herece E (1990) 1953 Yenice-Gönen Deprem Kırığı ve Kuzey Anadolu Fay Sisteminin Biga Yar-
ımadası’ndaki Uzantıları. MTA Dergisi 111:47–59 (in Turkish)
İmren C, Le Pichon X, Rangin C et al (2001) The North Anatolian Fault within the Sea of Marmara: a new
interpretation based on multi-channel seismic and multi-beam bathymetry data. Earth Planet Sci Lett
186:143–158
Jimenez A, Posadas AM, Marfil JM (2005) A probabilistic seismic hazard model based on cellular automata
and information theory. Nonlinear Proc Geoph 12:1–16
Jordanovski LR, Todorovska MI (1995) Earthquake source parameters for seismic hazard assessment: how
to obtain them from geologic data, historic seismicity and relative plate motions. In: Duma G (ed)
Proceedings of 10th European conference earthquake engineering, Aug 28–Sept 2, 1994, Vienna,
Austria special theme session S01.2, Source mechanism Balkema Rotterdam 4, pp 2561–2566
Kalafat D, Güneş Y, Kara M, Deniz P, Kekovalı K, Kuleli HS, Gülen L, Yılmazer Y, Özel NM (2007) A
revised and extended earthquake catalogue for Turkey since 1900, M C 4.0. Bogaziçi University
Publication, pp 1–553 ISBN 978-975-518-218-0
Khan PK, Ghosh M, Chakraborty PP et al (2010) Seismic b-value and the assessment of ambient stress in
Northeast India. Pure Appl Geophys 168:1693–1706
Kiratzi AA, Gregory S, Wagner S et al (1991) Source parameters of some earthquakes in aegean determined
by body waveform inversion. Pure Appl Geophys 135:515–527
Kramer SL (1996) Geotechnical earthquake engineering. Prentice-Hall, USA, p 653
Kumar PK, Ghosh M, Pratim P et al (2011) Seismic b-value and the assessment of ambient stress in
Northeast India. Pure Appl Geophys 168(10):1693–1706
McKenzie DP (1972) Active tectonics of the Mediterranean region. Geophys J R Astron Soc 30:109–185
Meju MA, Hutton VRS (1992) Iterative most-squares inversion: application to magnetotelluric data. Geo-
phys J Int 108:758–766
Miyamura S (1962) Magnitude-frequency relations and its bearing to geotectonics. Proc Jpn Acad 38:27–30
Mogi K (1967) Earthquakes and fractures. Tectonophysics 5:35–55
Molchan G, Kronrod T, Giuliano F (1997) Multi-scale seismicity model for seismic risk. Bull Seismol Soc
Am 87(5):1220–1229
Mori J, Abercrombie RE (1997) Depth dependence of earthquake frequency-magnitude distributions in
California: implications for the rupture initiation. J Geopyhys Res 102:15081–15090
Nalbant S, Hubert A, King GCP (1998) Stress coupling between earthquakes in northwest Turkey and the
north Aegean sea. J Geophys Res 103:24469–24486

123
Nat Hazards (2013) 68:837–853 853

Öcal N, Uçar SB, Taner D (1968) Manyas Karacabey Depremi. 6 Ekim 1965 Milli Eğitim Bakanlığı.
İstanbul Kandilli Rasathanesi Sismoloji Yayınları 11:3–43 (in Turkish)
Okal EA, Romanowicz BA (1994) On the variation of b-values with earthquake size. Phys Earth Planet
Interiors 87:55–76
Olsson R (1999) An estimation of maximum b-value in the Gutenberg–Richter relation. J Geodyn
27:547–552
Oral MB, Reilinger RE, Toksöz MN et al (1995) Global positioning system offers evidence of plate motions
in eastern Mediterranean. EOS AGU 76:9–11
Öztürk S, Bayrak Y, Çınar H et al (2008) A quantitative appraisal of earthquake hazard parameters com-
puted from Gumbel I method for different regions in and around Turkey. Nat Hazards 47:471–495
Papazachos CB, Kiratzi A (1996) A detailed study of the active crustal deformation in the Aegean and
surrounding area. Tectonophysics 253:129–153
Papazachos CB, Kiratzi A, Voidomatis Ph et al (1984) A study of the December 1981–January 1982 seismic
activity active crustal deformation in the Aegean and surrounding area. Tectonophysics 253:129–153
(in Northern Aegean Sea. Boll. Di Geof. Teor. Ed Appl. XXVI: 101–113)
Patwardhan SA, Kulkarni RB, Tocher D (1980) A semi-Markov model for characterizing recurrence of great
earthquake. Bull Seismol Soc Am 70:323–347
Pfister M, Balderer W, Greber, et al (1997) Synthesis of the MARMARA poly-project, active tectonics of
northwestern Anatolia-the Marmara poly-project. In: Schindler C, Pfister M (ed), pp 539–565
Pondard N, Armijo R, King GCP et al (2007) Fault interactions in the Sea of Marmara pull-apart (North
Anatolian Fault): earthquake clustering and propagating earthquake sequences. Geophys J Int
171:1185–1197
Sayıl N, Osmanşahin I (2005) An Application of the time—and magnitude-predictable model for the
longterm earthquake prediction in the Western Anatolia. In: International symposium on the geody-
namics of Eastern Mediterranean: active tectonics of the Aegean region, Kadir Has University,
Istanbul, 15–18 June 2005
Sayıl N, Osmanşahin İ (2008) An investigation of seismicity for western Anatolia. Nat Hazards 44:51–64
Scholz CH (1968) The frequency-magnitude relation of microfracturing in rock and its relation to earth-
quakes. Bull Seismol Soc Am 58:399–415
Schorlemmer D, Weimer S, Wyss M (2004) Earthquake statistics at Parkfield: 1. Stationarity of b values.
J Geophys Res 109:B12307
Shi Y, Bolt BA (1982) The standard error of the magnitude-frequency b-value. Bull Seismol Soc Am
72:1677–1687
Smyth C, Mori J (2009) Temporal variations of the Gutenberg–Richter distribution prior to the Kobe
earthquake. Ann Disaster Prev Res Inst 52B:255–261
Sounders P, Priestley K, Taymaz T (1998) Variations in the crustal structure beneath western Turkey.
Geophys J Int 134:373–389
Tabban A, Gençoğlu S (1975) Earthquake and its parameters. Bull Earthq Res Inst Turkey 11:7–83
Taymaz T, Tan O, Yolsal S (2004) Active tectonics of Turkey and surroundings: seismic risk in the
Marmara sea region. In: Fujii N, Kasahara J, Higashihara H, Ogawa K. (eds) The proceedings of 1st
international workshop on active monitoring in the solid earth geophysics (IWAM04), extended
abstract book pp 110–115
Tsapanos TM (1990) b-Values of two tectonic parts in circum-Pacific belt. Pure Appl Geophys 134:229–242
Tsapanos TM, Papazachos BC (1998) Geographical and vertical variation of the earth’s seismicity. J Seis-
mol 2:183–192
Ulugergerli EU, Seyitoğlu G, Başokur AT et al (2007) The geoelectrical structure of northwestern Anatolia,
Turkey. Pure Appl Geophys 164:999–1026
Utsu T (1965) A method for determining the value of b in a formula log n = a - bM showing the
magnitude-frequency relation for earthquakes. Geophys Bull Hokkaido Uni 13:99–103
Wang JH (1988) b values of shallow earthquakes in Taiwan. Bull Seismol Soc Am 78(3):1243–1254
Watson HR, Sharma ML, Khan PK et al (2002) Analysis of aftershocks of the Chamoli Earthquake of March
29, 1999 using broadband seismic data. J Him Geol 23:7–18
Weeks J, Lockner D, Byerlee J (1978) Change in b-values during movement on cut surfaces in granite. Bull
Seismol Soc Am 68:333–341
Yılmaztürk A, Burton PW (1999) An evaluation of seismic hazard parameters in southern Turkey. J Seismol
3:61–81

123

Vous aimerez peut-être aussi