OBS Seismicity Monitoring in Marmara

Geometry and segmentation of the North Anatolian
Fault beneath the Marmara Sea, Turkey,
deduced from long-term ocean bottom
seismographic observations
Y. Yamamoto1
, N. Takahashi1
, A. Pinar2
, D. Kalafat2
, S. Citak1
, M. Comoglu2
, R. Polat2
,
and Y. Kaneda3
1
Research and Development Center for Earthquake and Tsunami, Japan Agency for Marine-Earth Science and Technology,
Yokohama, Japan, 2
Kandilli Observatory and Earthquake Research Institute, Boğaziçi University, Istanbul, Turkey, 3
Institute
of Education, Research and Regional Cooperation for Crisis Management Shikoku, Kagawa University, Kagawa, Japan
Abstract Both the geometry and the depth of the seismogenic zone of the North Anatolian Fault under
the Marmara Sea (the Main Marmara Fault (MMF)) are poorly understood, in part because of the fault’s
undersea location. We recorded 10 months of microseismic data with a dense array of ocean bottom
seismographs and then applied double-difference relocation and 3-D tomographic modeling to obtain
precise hypocenters on the MMF beneath the central and western Marmara Sea. The hypocenters show
distinct lateral changes along the MMF: (1) both the upper and lower crust beneath the Western High are
seismically active and the maximum focal depth reaches 26 km; (2) seismic events are confined to the upper
crust beneath the region extending from the eastern part of the Central Basin to the Kumburgaz Basin;
and (3) the magnitude and direction of dip of the main fault change under the Central Basin, where there is
also an abrupt change in the depth of the lower limit of the seismogenic zone. We attribute this change to a
segment boundary of the MMF. Our data show that the upper limit of the seismogenic zone corresponds
to sedimentary basement. We also identified several seismically inactive regions within the upper crust
along the MMF; their spatial extent beneath the Kumburgaz Basin is greater than beneath the Western High.
From the comparison with seafloor extensometer data, we consider that these regions might indicate zones
of strong coupling that are accumulating stress for release during future large earthquakes.
1. Introduction
From its junction with the East Anatolian Fault at the Karliova triple junction in eastern Turkey, the North
Anatolian Fault (NAF) extends 1600 km westward across northern Turkey, the Marmara Sea, and the
Aegean Sea (Figure 1a). Slip to accommodate right-lateral motion between the Anatolia and Eurasia plates
is about 25 mm/yr [e.g., Reilinger et al., 2006]. During the past 77 years, a sequence of devastating earthquakes
of surface-wave magnitude (Ms) > 7 has progressed westward along the NAF, starting with the 1939 Erzincan
earthquake (Ms = 7.9) in eastern Turkey and culminating in the 1999 Izmit-Golcuk (Ms = 7.7) and 1999 Duzce
(Ms = 7.4) earthquakes in the eastern Marmara region. Such a sustained series of releases of seismic moment
on a major transform plate boundary is unique not only in the eastern Mediterranean region but also world-
wide. Considering that the 2014 Aegean Sea earthquake (Mw = 6.9) also occurred on a segment of the NAF,
the seismic gap, only part of the 1600 km long NAF that has not ruptured during this period, lies beneath
the Marmara Sea.
Paleoseismological studies and historical records indicate that several large earthquakes have occurred
beneath the Marmara Sea [Fraser et al., 2010; Meghraoui et al., 2012; Drab et al., 2015; Parsons, 2004].
Although the exact locations of pretwentieth century earthquakes are not known [e.g., Pondrad et al.,
2007], at least two earthquakes of M > 7 in 1766 have been recognized as under the Marmara Sea earth-
quakes. Moreover, Armijo et al. [2005] proposed that the 1912 Ganos earthquake (Ms = 7.3) ruptured not only
the onshore region, as considered by previous researchers [e.g., Ambraseys, 2002], but also NAF segments
under the western Marmara Sea. Although several seismological and geodetic studies have attempted to
clarify the current stick-slip status along the NAF beneath the Marmara Sea [e.g., Bohnhoff et al., 2013;
Ergintav et al., 2007, 2014; Schmittbuhl et al., 2015], there is consensus only on the existence of shallow
coupling on the Prince Islands segment (Figure 1b).
YAMAMOTO ET AL. GEOMETRY AND SEGMENTATION OF MMF 2069
PUBLICATIONS
Journal of Geophysical Research: Solid Earth
RESEARCH ARTICLE
10.1002/2016JB013608
Key Points:
• Long-term ocean bottom seismic data
identified a segment boundary of the
Main Marmara Fault beneath the
central basin of the Marmara Sea
• The seismogenic zone of the Main
Marmara Fault locates beneath the
sedimentary basement
• Zones of no seismicity within the
upper crust appear to indicate locked
sections of the Main Marmara Fault
Supporting Information:
• Supporting Information S1
Correspondence to:
Y. Yamamoto,
yamamotoy@jamstec.go.jp
Citation:
Yamamoto, Y., N. Takahashi, A. Pinar,
D. Kalafat, S. Citak, M. Comoglu, R. Polat,
and Y. Kaneda (2017), Geometry and
segmentation of the North Anatolian
Fault beneath the Marmara Sea, Turkey,
deduced from long-term ocean bottom
seismographic observations, J. Geophys.
Res. Solid Earth, 122, 2069–2084,
doi:10.1002/2016JB013608.
Received 29 SEP 2016
Accepted 7 FEB 2017
Accepted article online 9 FEB 2017
Published online 6 MAR 2017
©2017. American Geophysical Union.
All Rights Reserved.

Current knowledge of the geometry of the NAF beneath the Marmara Sea is based mainly on bathymetry and
shallow onshore and offshore structural information. Since Pinar [1943] first proposed a single strike-slip fault
system (the Main Marmara Fault (MMF)) roughly bisecting the Marmara Sea, various other fault models have
been proposed [Yaltırak, 2002]. Le Pichon et al. [2001] identified the MMF as a single, buried strike-slip fault,
mainly from high-resolution bathymetry and shallow seismic reflection data, whereas Okay et al. [1999] and
Parke et al. [1999], both using the data set of same cruise as Le Pichon et al. [2001], proposed two different
models. In the Okay’s model, the MMF traces the southern end of three major basins, further south than in
the model by Le Pichon et al. [2001]. Parke et al.’s [1999] model consisted of en echelon faulting, with no
strike-slip fault in the central Marmara Sea. Armijo et al. [2002, 2005] identified earthquake scarps on the sea-
floor that suggest the presence of individual fault segments that are oblique to the regional east-west trend
of the MMF and interpreted some of them to be normal faults related to opening of the Marmara Sea.
However, the activities of each faults and their deep extent were unclear because of limitation of their data
set. Although several models for the MMF have been proposed or assumed [e.g., Pondrad et al., 2007; Hergert
et al., 2011; Oglesby and Mai, 2012; Aochi and Ulrich, 2015], none were based on observational evidence of
fault geometry within the upper and lower crust beneath the central and western Marmara Sea.
To investigate the potential for future earthquakes beneath the Marmara Sea [e.g., Bohnhoff et al., 2016; Murru
et al., 2016], a better understanding of the fault geometry is needed so that areas of strong coupling can be
Figure 1. (a) Regional map showing Marmara Sea study area (dashed rectangle) and surrounding plate boundaries [Bird, 2003]. The segments of the North Anatolian
Fault (NAF) and the years of historical large earthquakes on them since 1900 are color coded. The focal mechanisms shown are global centroid moment tensor
solutions [Dziewonski et al., 1981; Ekström et al., 2012] for the 1999 Izmit and Duzce earthquakes and the 2014 Aegean Sea earthquake. EU, Eurasia Plate; AN, Anatolian
Plate; AF, African Plate; HE, Hellenic Plate; AR, Arabian Plate. (b) Map of study area showing bathymetry and structural elements. Ocean bottom seismograph locations
shown as numbered inverted triangles are colored to differentiate two periods of observation: blue 10 months and purple 4 months. TB, Tekirdag Basin; WH,
Western High; CB, Central Basin; CH, Central High; KB, Kumburgaz Basin. The red lines are the seafloor traces of faults under the Marmara Sea [Armijo et al., 2005].
Journal of Geophysical Research: Solid Earth 10.1002/2016JB013608

identified. Several ocean bottom seismograph (OBS) data sets have been used to attempt to obtain precise
hypocenters to address these questions [Sato et al., 2004; Tary et al., 2011; Cros and Géli, 2013; Yamamoto
et al., 2015]. However, the observation periods and/or spatial extent of observation area of these studies were
not extensive enough to interpret the fault geometry beneath whole MMF. On the basis of land and ocean
floor seismographic data recorded over a period of 5 years, Schmittbuhl et al. [2015] concluded that the seis-
mogenic zone beneath the Marmara Sea is confined to the upper 16 km. However, their data included many
earthquakes for which hypocenters were determined without using OBS data (because their temporal dense
offshore observation periods of 1 to 3 months were too short, and their enabled cable-type OBSs were sparse
(~40 km intervals) and limited the functioned duration of 1 to 2 years); thus, their conclusions may not apply
to the whole of the Marmara Sea. It is therefore possible that earthquakes have occurred within the lower
crust at depths of 20 km or deeper, as suggested by other OBS observations [Sato et al., 2004; Tary et al.,
2011; Cros and Géli, 2013; Yamamoto et al., 2015].
To investigate fault geometry, fault segmentation, and the spatial extent of the seismogenic zone on the
basis of long-term microearthquake activity beneath the western and central Marmara Sea, we recorded
continuous OBS data for 10 months, as part of the “Marmara Disaster Mitigation” project. Since our target area
is a very narrow region around the MMF and we considered that the ambiguities in the onshore structure
increase the location error, we used only our OBS data. We then applied 3-D seismic tomography and
double-difference hypocenter relocation by cross correlation of travel time differences to the data we
acquired. In this paper, we present the results of our analyses in terms of seismic activity, fault geometry,
and fault segmentation beneath the western and central Marmara Sea.
2. Observations
In September 2014, we deployed 10 OBSs in the area extending from the Tekirdag Basin to the Central High
(blue triangles in Figure 1b). In March 2015, we extended the seismic array to the east and west by adding five
more OBSs (purple triangles in Figure 1b). By taking into account the considerably higher microearthquake
activity along MMF, we considered that this time period was sufficiently long to estimate the upper and lower
depth limits using microseismicity. This time period was also undisturbed by series of aftershock sequence,
since there were no large (M > 6) earthquakes during our observation period. The average separation
between stations was 10 km. The OBSs were deployed during cruises of DSV Alcatras. We used both free-fall
and pop-up type OBSs equipped with three-component 4.5 Hz geophones and hydrophones. OBS locations
on the seafloor were determined by triangulation. Clock accuracy of better than 0.05 s was determined by
calibration of the OBS clock with GPS time just before deployment and immediately after recovery. The sam-
pling interval was 100 Hz. All OBSs were operational throughout the observation periods and were recovered
in July 2015 by R/V Tübitak Marmara.
3. Analyses
First, we used the short-term average/long-term average ratio method to search for microearthquake events
in our continuous OBS records. We used both the WIN system [Urabe and Tsukada, 1992] and Seiscomp3
search tools (www.seiscomp3.org) and combined the resultant event lists. Next, we manually picked the first
arrivals of P and S waves with P wave polarities of first arrivals. The picking accuracy of arrival time was less
than 0.1 s. We then calculated initial hypocentral locations by using the HYPOMH program [Hirata and
Matsu’ura, 1987] and the 1-D velocity model of Yamamoto et al. [2015]. We determined S wave travel time
delays due to the low-velocity shallow sedimentary layer by using the travel time differences between the
P to S converted waves (PS waves) generated at the base of the sediment layer and the direct P waves, by
assuming a Vp/Vs ratio of 3 within the sediment layer (Figure 2). We thus identified 714 microearthquake
events in the region of the NAF beneath the Marmara Sea and used them as the initial hypocenters for fol-
lowing tomographic study and determined their magnitudes from the maximum amplitude of the vertical
wave components [Watanabe, 1971].
To account for local-scale heterogeneity of the velocity structure, we applied double-difference (DD) 3-D seis-
mic tomography to our data by using tomoFDD software [Zhang and Thurber, 2006]. During tomographic
inversion, the DD data (travel time difference between two events separated within 10 km at one OBS) were
calculated from manually picked arrivals. Horizontal and vertical grid nodes for expression of velocity field

were at 10 km and 2–10 km intervals, respectively (Figure 3a). We incorporated regional variations of the
thickness of the sediment layer [from Bayrakci et al., 2013] and depth to the Moho discontinuity [from
Bécel et al., 2009] in the initial velocity model (Figure 3b). The initial Vp/Vs ratio was set at 1.73 for the entire
depth of the model. After 20 iterations, the RMS travel time residuals had decreased from 0.28 to 0.15 s for
P waves and from 0.95 to 0.30 s for S waves. The average horizontal and vertical location errors of the
calculated hypocenters were about 0.46 km and 0.57 km, respectively. Finally, we ran an additional 22
iterations of DD relocation by using DD data obtained by the cross-correlation method under the obtained
3-D velocity model from our tomographic inversion. We used waveforms ﬁltered in the 4–8 Hz frequency
band of vertical and horizontal components for P and S, respectively. The total length of computing
the cross-correlation values was 3 s (300 samples), and the start of master waveform was 1 s before the
Figure 2. Examples of observed waveforms. All data were normalized but not ﬁltered. V and H1 indicate the vertical and
horizontal components, respectively.

Figure 3. Settings for tomography. (a) Map view of initial hypocenters (circles; color scale indicates the focal depth), OBSs (inverted triangles), and grid nodes
(crosses). The red lines are the fault traces on the ﬂoor of the Marmara Sea [Armijo et al., 2005]. (b) Vertical cross section of the initial P wave velocity model
(contour interval = 0.25 km/s) along line Y = 5 km with all initial hypocenters projected onto the cross section.

manually picked P or S arrivals. The maximum distance between master and slave events was 10 km, and we
set the computed maximum correlation function value as the weight of data. The average horizontal and ver-
tical location errors of the calculated hypocenters were improved to both about 0.20 km, less than half of
the result of before the DD relocation. The information of all relocated earthquakes was summarized in
Table S1 in the supporting information.
4. Results and Margins of Error
4.1. Hypocenter Locations
The distributions of the final relocated microearthquake hypocenters projected onto cross sections of the P
and S wave velocity models showed clear lateral and vertical variations (Figure 4a). The most remarkable fea-
ture is a sharp change of the lower limit of the seismogenic zone beneath the Central Basin. In the western
half of the profile (X < 0) it is at about 26 km depth, whereas in the eastern half (X > 0) it is at about 12 km
depth. The upper limit of the seismogenic zone also shows lateral changes. Beneath the Western High there
is no microearthquake activity from the seafloor to 8 km depth, but under the eastern Central Basin the upper
limit is at only about 5 km depth.
Although the accuracy of our relocations was about 0.2 km, the results are also dependent on the initial velo-
city model and station corrections used. We therefore tested the 1-D initial velocity models of Tary et al.
[2011], Karabulut et al. [2011], and Yamamoto et al. [2015]. Because Tary et al. [2011] and Karabulut et al.
[2011] did not apply station corrections, we tested their original settings with and without station corrections.
Comparison of the results for the various models showed average differences in both the horizontal and ver-
tical directions of less than 0.3 km when the station corrections were taken into account. We also tested the
use of Vp/Vs ratios of 2 and 4 for the calculation of station corrections. For Vp/Vs ratios of 2 and 4, the relocated
hypocenters were 0.82 km deeper and 0.87 km shallower, respectively, than the original result (Vp/Vs = 3).
Thus, we concluded that the maximum errors of relocation depend on initial velocity model and assumption
of Vp/Vs for the sediment layer; these maximum errors were 0.3 km (horizontal) and 1.2 km (vertical; summa-
tion of 0.3 km and 0.82 or 0.87 km). On the other hand, without the station corrections, focal depths tend to
deepen 1.7 km and 3.8 km on average for Tary et al. [2011] and Karabulut et al. [2011], respectively. This means
that the station corrections for S wave arrivals are important to obtain precious hypocenter locations.
4.2. Velocity Structure
We used a checkerboard test to assess the spatial resolution of the velocity structure we obtained. We
assumed velocity variations of ±5% to be anomalous and calculated synthetic travel times with standard
deviations of 0.1 s for P waves and 0.2 s for S waves. After testing several scales of checkerboard pattern,
we concluded that the scale of resolved features was 20 km horizontally and 5 km vertically (Figure 4b). By
comparing the checkerboard test result with the derivative weight sum (DWS) distribution [Thurber and
Eberhart-Phillips, 1999], we define the area of adequate resolution to be that defined by DWS values >1000
for P waves and >500 for S waves. These results indicate that the P wave model was more reliable than
the S wave model.
Both the P and S wave velocity structures show a thickening of the low-velocity zone beneath the Central
Basin, corresponding to a thick sediment layer identified by Bayrakci et al. [2013]. The average vertical P wave
velocity is almost the same as that of Bécel et al. [2009], although for our model, the P wave velocity within the
lower crust (~6.4 km/s) is slower than theirs (6.7 km/s). There is a zone of high S wave velocity near the lower
limit of seismicity beneath the Western High that is not replicated in the P wave velocity model. The lack of
resolution of the velocity models at similar depths on the eastern side of the model (Figure 4b) prevented
further consideration of this feature. A regional-scale tomographic study incorporating both onshore and
offshore data may extend the resolution of the model in this area.
5. Discussion
5.1. Comparison With Land-Based Earthquake Catalogue
We compared our results with the hypocenters of microseismic events recorded during the period of
our study by land-based observations at Kandilli Observatory and Earthquake Research Institute (KOERI)
(Figure 5a). Only 95 of the events from the KOERI catalogue were near the MMF, which is less than 15% of

Figure 4. Results of tomographic inversion. (a) Cross-sectional views along Y = 5 km and 5 km (locations in Figure 3a) showing relocated hypocenters and results of
velocity inversions (Vp and Vs). Contour interval is 0.25 km/s for both Vp and Vs models. Hypocenters within 5 km on either side of the proﬁles (black dots) are
projected onto the section. The faded colors indicate the unresolved areas, where derivative weight sums (DWS) [Thurber and Eberhart-Phillips, 1999] are less than
1000 and 500 for Vp and Vs, respectively. Other symbols are the same as in Figure 3. (b) Results of the checkerboard test with DWS isovalues of 1000 and 500
shown for Vp and Vs models, respectively. Other symbols are the same as in Figure 4a.

Figure 5. Comparison of hypocenters determined in our study with those determined from land-based observations of the Kandilli Observatory and Earthquake
Research Institute (KOERI). (a) Map view and vertical profile. The orange and gray circles indicate our results and those of KOERI, respectively. Magnitudes were
from KOERI catalogue. On the vertical profile, the pink numbers (1 to 4) indicate the examples of earthquake clusters we identified. Other symbols are the same as in
Figure 3. (b) Comparison of our observed magnitudes by using Watanabe [1971] and those from the KOERI catalogue. (c) Histogram of magnitude for all detected
event (red bars) and for event listed in KOERI catalogue (blue). The black line shows the cumulative number.

those we identified. We also found that the relocated microearthquakes show numerous seismic clusters.
Even though the hypocenters from the KOERI catalogue show considerably more scatter than the OBS data,
they tend to lie within similar regions of limited lateral extent (e.g., numbered 1 to 4 in Figure 5a). Because the
waveforms of events in same cluster are very similar to those recorded at nearby OBSs (Figure 6), we conclude
that our relocations are better constrained than those of the KOERI catalogue.
We recognized some major differences between the hypocenters we determined and those of KOERI. Our
results indicate that the active fault in the Tekirdag Basin is a branch fault that is northwest of the MMF, with
little activity on the MMF between 27.5°E and 27.65°E, whereas most earthquakes in the region occur on the
MMF according to the KOERI catalogue. Besides, beneath the Western High, most of the hypocenters we
determined are deeper than 15 km, whereas the KOERI catalogue shows the vertical extent of seismic activity
to be from 4 to 23 km depth. We think that the focal depth could not be well constrained by using permanent
seismic network since the Western High is far from (~20 km) nearby cable-type OBSs.
Comparison of the event magnitudes we estimated and those of the local magnitude (ML) from KOERI cata-
logue (Figure 5b) show that, in general, our estimates are similar to or smaller than those of KOERI; the
maximum difference is about 1 unit of magnitude. In addition, Schmittbuhl et al. [2015] pointed out that
the magnitude of KOERI catalogue is biased for low magnitude. Thus, we conclude that the magnitudes we
estimated from our OBS data are suitable only for qualitative application. Figure 5c shows the histogram of
magnitude determined from OBS record for all events, whose magnitude was calculated (red) and listed event
in KOERI catalogue (blue outline). Although we could not obtain the magnitude of 29 events due to noisy
record, this indicates that the most of the newly detected events have a magnitude lower than 1.5, and
almost all event whose magnitude is larger than 2 was listed in KOERI catalogue. The above comparisons sug-
gest that a dense OBS network such as ours can provide effective microearthquake monitoring in this region.
5.2. Segmentation of Main Marmara Fault
Here we consider the geometry and segmentation of the MMF on the basis of our microearthquake data. The
depth of the lower boundary of the seismogenic zone changed sharply (by about 10 km; see section 4.1) at
about 28.05°E beneath the Central Basin (Figure 4). Considering the error in our estimates of focal depth
(1.2 km; section 4), this change of depth is plausible and may represent a segment boundary. However, to
precisely define the segments of the MMF, we need to eliminate events on the many subfaults beneath
the Marmara Sea [e.g., Armijo et al., 2005; Le Pichon et al., 2001]. To achieve this, we assumed the MMF to
be a right-lateral strike-slip fault, as indicated by surface geodetic observations [e.g., Reilinger et al., 2006].
Because it is difficult to constrain the focal mechanisms of individual microearthquakes, we calculated com-
posite focal mechanisms for clusters of seismic events that we identified as follows. We first included all
groups of events separated horizontally by less than 1 km into single clusters and then defined subclusters
within them when they separated vertically larger than 2 km. We defined 51 seismic clusters (stars in
Figure 7a) for which we then calculated composite focal mechanisms by using the FOCMEC program
[Snoke, 2003]. We first set the search windows of B axis plunge into three range during focal mechanism cal-
culation: (1) 0 to 30°, assuming normal or reverse fault; (2) 30 to 60°, assuming mixed mechanism; and (3) 60
to 90°, assuming strike-slip fault. Then, we compared the error values among them. Finally, we selected all of
the clusters for which the error value was minimum for (3) and the result shows right-lateral strike-slip com-
posite focal mechanism as the seismic activity on MMF (yellow stars in Figure 7a) and excluded all others.
As a result, we selected 13 clusters that we consider to be representative of seismic activity on the MMF
(Figure 7c). The breakdown of other 38 clusters were as follows: 6 clusters were dip slip, 5 clusters were mixed,
and other 27 were not constrained (purple, blue, and green stars in Figure 7a, respectively). Examination of
the alignment of hypocenters and the seafloor trace of the MMF in cross-sectional views (Figure 7c) indicates
that west of 28°E the MMF fault appears to be near vertical to northward dipping (average about 85°N),
whereas east of 28.10°E it appears to dip about 80° southward. These conflicting dips provide strong evi-
dence of a fault segment boundary between 28.0°E and 28.10°E. The abrupt change of the lower limit of
the seismogenic zone at 28.05°E also supports the presence of a segment boundary beneath the Central
Basin in the region around 28.05°E. Moreover, if the 1912 earthquake (Figure 1a) ruptured both onshore
and offshore, as proposed by Armijo et al. [2005], the eastern end of that rupture zone might correspond
to that segment boundary, and perhaps, this segment boundary marks the western boundary of the seismic
gap beneath the Marmara Sea, in which there have been no earthquakes since 1766.

Figure 6. Three-component waveforms of selected event clusters (1 to 4 shown in Figure 5a). (a) Cluster 1 at OBS 14, (b) cluster 2 at OBS 06, (c) cluster 3 at OBS 08,
and (d) cluster 4 at OBS 09. V, vertical component; H1 and H2, horizontal components. P wave arrivals, average arrival time of Ps converted wave, and expected
time for reﬂection wave at the sea surface were also shown on the top of each ﬁgure.

Figure 7. Seismic clusters and their composite focal mechanisms. (a) Map showing composite hypocenters of seismic clusters defined in this study. The yellow stars
indicate the clusters with right-lateral strike-slip focal mechanisms, purple indicate the clusters with normal or reverse focal mechanisms, and blue indicate the
clusters with mixed-type focal mechanisms. The green stars indicate the clusters with having possibilities of all focal mechanism types. Other symbols are the same as
in Figure 3. (b) Epicenters (red dots) of 13 clusters with right-lateral strike-slip focal mechanisms (yellow stars in Figure 7a). The white line indicates the connected
horizontal projection of extended location of lines between the seafloor trace of MMF and 13 clusters at 25 km depth. (c) Plan and roughly N-S cross-sectional
views showing composite focal mechanisms of four selected clusters. The small red dots are the individual events in the four clusters. In the cross sections, micro-
seismic events within 2.5 km either side of profiles are also projected onto the section; the red hexagons on the seafloor mark the seafloor trace of the MMF.

5.3. Constraints on the Seismogenic Zone
To further investigate the relationship of the distribution of the microearthquakes we recorded to the MMF,
we considered only the relocated hypocenters that we determined to be on the MMF. We discriminated
on-fault and off-fault events on the basis of dip of the fault plane by considering both the composite focal
mechanism of the event and the dip of the fault determined from the spatial relationship between the
hypocenter and the seafloor trace of the MMF. Based on estimated dip angle and their variation in previous
section, we assumed that the plausible location of MMF was within 10° of both side of average dip angle (i.e.,
viewing angle is 20°). We thus selected events west of 28.05°E with fault planes dipping either at greater than
75°N or greater than 85°S and events east of 28.05°E with fault planes dipping at greater than 70°S. These
criteria resulted in the selection of 476 on-fault events (Figure 8).
The upper limit of seismicity on the profile along latitude 40.8218°N becomes shallower east of 28.05°E
(Figure 8b). Comparison of our profile with a detailed shallow 3-D tomographic model based on active-source
OBS observations [Bayrakci et al., 2013] showed that the lateral variations of the upper limit of seismicity
clearly correspond to variations in the thickness of the sedimentary layer. Thus, we concluded that the seis-
mogenic zone in this region is within the upper crust, beneath the sedimentary basement. Although we have
observed a few events whose focal depths were shallower than 5 km, all of them were recognized as off-fault
earthquake of MMF (Figures 8b and 8c).
Cros and Géli [2013], however, identified shallow aftershocks of a Mw 5.2 earthquake that occurred on 25
July 2011; the aftershocks were at 2–6 km depth within the sedimentary layer on the Western High.
They attributed the aftershocks to release of gas from a gas hydrate reservoir 2–4 km below seafloor.
Schmittbuhl et al. [2015] also showed that most earthquakes beneath the Western High were at 2–8 km depth.
However, the major periods of shallow seismic activity seem to be only after earthquakes of ML > 4
[Schmittbuhl et al., 2015, Figure 4]. Because we recorded no earthquakes of ML > 4, and no events within
the sediment layer of the Western High, we consider that microearthquakes identified in the sedimentary
layer by other researchers may be aftershocks triggered by moderate earthquakes in upper crust beneath
the Western High, as suggested by Cros and Géli [2013]. Moreover, the higher density of our OBS network,
compared to those of earlier studies, would be expected to provide more accurate hypocenters and would
have detected such shallow earthquakes had they occurred during our observation period. It is also note-
worthy that at the time of the Mw 5.2 earthquake, the OBSs deployed by Cros and Géli [2013] above the
Western High were not functional.
Our data indicate that the lower limit of the seismogenic zone is deeper than the Conrad discontinuity of
Bécel et al. [2009] under the western Central Basin but shallower than the Conrad discontinuity under the
eastern Central Basin. Because the deepest of the events we identified was shallower than 26 km and the
Moho discontinuity in this region was estimated at 26–27 km depth [Bécel et al., 2009], we concluded that
none of those events occurred in the mantle. Although Schmittbuhl et al. [2015] suggest that the lower limit
of seismogenic zone along the MMF is at 16 km depth, we recorded several events at 20 km depth or greater,
as was did several previous studies using OBS data [Sato et al., 2004; Tary et al., 2011; Cros and Géli, 2013;
Yamamoto et al., 2015]. We are confident of the accuracy of the hypocenters we estimated at depths of
20 km or more (see section 4.1), and, on the basis of our data, we conclude that the seismogenic zone is
confined to the upper crust under the eastern Central Basin but extends across both the upper and lower
crust under the western Central Basin.
Although the lower crust is generally considered to be ductile and aseismic, there is much observational evi-
dence of seismic activity within it [e.g., Simpson, 1999, and references therein]. Several explanations have
been proposed for such earthquakes. They may occur in areas where the lower crust is cooler than normal
[Doser and Yarwood, 1994], where the lower crust is drier and more mafic than normal [Shudovsky, 1985],
or where there are mantle-derived fluids under high pore pressure [Reyners et al., 2007]. In the Marmara
Sea region, helium isotope studies have identified mantle-derived fluid in the MMF [Burnard et al., 2012]
and noted high 3
He/4
He ratios (>1) along the MMF between the western Central Basin and the eastern
Tekirdag Basin that correspond to the areas of seismicity we identified in the lower crust. We consider that
the seismicity within the lower crust might be related to the presence of fluids derived from the mantle,
although the two other explanations cannot be ruled out. On the other hand, our tomographic image
has insufficient spatial resolution for evaluation whether fluids exist along the MMF in lower crust or not.

Figure 8. On-fault earthquake events. (a) Map view of distribution of on-fault relocated hypocenters. The gray circles are the off-fault earthquakes; other symbols are
the same as in Figure 3. Magnitudes were from this study. (b) E-W vertical profile of hypocenter distributions along latitude 40.8218°N (dashed line in Figure 8a). The
background colors show the P wave velocity profile extracted along latitude 40.8218°N (same line as shown in Figure 13 of Bayrakci et al. [2013]) from our tomo-
graphic model. The blue line indicates the depth to sedimentary basement [Bayrakci et al., 2013]; the upper and lower black dashed lines indicate the Conrad
and Moho discontinuities [Bécel et al., 2009], respectively. The red dashed rectangles A to E indicate the on-fault areas of low seismicity. The red inverted triangles
indicate the seafloor extensometer observation points (eastern point [Sacik et al., 2016] and western point [Yamamoto et al., 2016]). (c) N-S vertical profiles of
hypocenters for intervals 27.6 to 27.84°E, 27.84 to 28.05°E, and 28.18 to 28.50°E. Other symbols are the same as in Figure 7c. The shaded purple indicates another
fault-like structure from the hypocenter distribution under the western Central Basin.

Although the presence of fluids in lower crust should decrease both P and S wave velocities, our S wave velo-
city near the seismic zone within the lower crust shows high velocity (Figure 4), whereas P wave velocity
remains relatively low. To confirm the presence of fluids along the fault plane, we should obtain more
detailed and deeper structural information by using additional data set, such as land station and other
OBS data.
5.4. Possible Locked Zones Along the Main Marmara Fault
Bohnhoff et al. [2013], in a study of microearthquake activity on the Prince Island segment of the MMF west of
the epicenter of the 1999 Izmit earthquake, interpreted a seismically quiescent zone on the MMF near
Istanbul to represent a patch of strong coupling. Schmittbuhl et al. [2015] proposed the existence of locked
patches corresponding to inactive seismicity zones, mainly in the upper crust, on the MMF beneath the east-
ern Marmara Sea. On the basis of magnitude-frequency b values, Schmittbuhl et al. [2015] proposed that
creep is the dominant slip mechanism on the MMF under the western Marmara Sea. Although we could
not consider b values because of uncertainties in our magnitude estimations (Figure 5b), our results show
seismicity inactive zones beneath the Marmara Sea (marked A to E in Figure 8b). If these zones correspond
to locked patches on the MMF, the eastern side of the section may be more strongly locked than those on
the western side since the eastern side have larger seismically inactive area than the western side.
Recently, seafloor extensometer observations were conducted along the MMF. Extensometer data at about
28.5°E, immediately above patches D and E, indicate almost complete coupling or very slow creep (less than
6 mm/yr) [Sakic et al., 2016], whereas extensometer data above patch B indicates creep at 8–11 mm/yr, sug-
gesting partial locking in the upper crust [Yamamoto et al., 2016]. Although Ergintav et al. [2014] proposed
that creep is dominant in the eastern half of the section on the basis of an onshore GPS study, we consider
that seafloor extensometer observations provide direct evidence of the coupling status of the MMF that is
consistent with our microseismicity data. From this coincidence, we propose that these seismically inactive
zones might be accumulating the strain toward the next large earthquakes. However, we recognize that
the observation periods of both of the extensometer studies were considerably shorter than the GPS studies,
and both are much shorter than the recurrence intervals of large earthquakes.
5.5. Limitations of This Study and Further Research
We considered a relatively simple fault model to represent a complex fault system in a study that was based
predominantly on only 10 months of OBS observations. To better constrain the geometry of the fault system,
the many branch faults, including normal faults related to pull-apart activity [e.g., Armijo et al., 2005], must be
taken into account. We also noted another fault-like structure from the hypocenter distribution under the
western Central Basin (shaded purple in Figure 8c) that deserves further attention. These aligned hypocenter
distributions were not found beneath the Western High and Kumburgaz Basin (Figure 8c); they might relate
to pull-apart of the Central Basin. In addition, it is possible that some of the event clusters we attributed to
activity on the MMF were generated by other fault systems, in which case the no seismicity regions along
the MMF may be more extensive than we have suggested here. Longer-term seafloor seismic and geodetic
observations are needed to further clarify the geometry of the MMF and spatial and temporal changes of its
coupling status.
6. Conclusions
We recorded OBS data over a period of 10 months and used precise hypocenter relocation coupled with a 3-D
velocity structure model to investigate the geometry of MMF under the Marmara Sea. We identified 714
events close to the MMF, which is about 7 times the number of events identified from land-based seismic
data over the same period. Lateral variations of the distribution of hypocenters along the MMF clearly show
the presence of a fault segment boundary beneath the Central Basin, where both the lower limit of seismo-
genic activity and the dip of the fault change. Comparison of our hypocentral distribution with that derived
from an analysis of previous active source OBS data suggests that the upper limit of the seismogenic zone is
at the top of the upper crust. Seismicity that we observed within the lower crust beneath the Western High
might be related to upwelling of fluid from the mantle. We also identified areas of low seismicity along the
MMF, for which different creep rates have been estimated from seafloor geodetic data. We interpret these
areas to represent locked zones along the MMF. More detailed investigations based on longer periods of
observation data are needed to further clarify the frictional status along the MMF.

References
Ambraseys, N. N. (2002), The seismic activity of the Marmara Sea region over the last 2000 years, Bull. Seismol. Soc. Am., 92(1), 1–18,
doi:10.1785/0120000843.
Aochi, H., and T. Ulrich (2015), A probable earthquake scenario near Istanbul determined from dynamic simulations, Bull. Seismol. Soc. Am.,
105(3), 1468–1475, doi:10.1785/0120140283.
Armijo, R., B. Meyer, S. Navarro, G. King, and A. Barka (2002), Asymmetric slip partitioning in the Sea of Marmara pull-apart: A clue to
propagation processes of the North Anatolian Fault?, Terra Nova, 14, 80–86.
Armijo, R., et al. (2005), Submarine fault scarps in the Sea of Marmara pull-apart (North Anatolian Fault): Implications for seismic hazard in
Istanbul, Geochem. Geophys. Geosyst., 6, Q06909, doi:10.1029/2004GC000896.
Bayrakci, G., M. Laigle, A. Bécel, A. Hirn, T. Taymaz, S. Yolsal–Çevikbilen, and SEISMARMARA team (2013), 3-D sediment–basement
tomography of the northern Marmara trough by a dense OBS network at the nodes of a grid of controlled source profiles along the North
Anatolian Fault, Geophys. J. Int., 194, 1335–1357, doi:10.1093/gji/ggt211.
Bécel, A., et al. (2009), Moho, crustal architecture and deep deformation under the North Marmara Trough, from the SEISMARMARA Leg 1
offshore–onshore reflection–refraction survey, Tectonophysics, 467, 1–21, doi:10.1016/j.tecto.2008.10.022.
Bird, P. (2003), An updated digital model of plate boundaries, Geochem. Geophys. Geosyst., 4(3), 1027, doi:10.1029/2001GC000252.
Bohnhoff, M., F. Bulut, G. Dresen, P. R. Malin, T. Eken, and M. Aktar (2013), An earthquake gap south of Istanbul, Nat. Commun., 4, 1999,
doi:10.1038/ncomms2999.
Bohnhoff, M., P. Martínez-Garzón, F. Bulut, E. Stierle, and Y. Ben-Zion (2016), Maximum earthquake magnitudes along different sections of
the North Anatolian Fault zone, Tectonophysics, 674, 147–165, doi:10.1016/j.tecto.2016.02.028.
Burnard, P., S. Bourlange, P. Henry, L. Geli, M. D. Tryon, B. Natal’in, A. M. C. Sengor, M. S. Ozeren, and M. N. Cagatay (2012), Constraints on fluid
origins and migration velocities along the Marmara Main Fault (Sea of Marmara, Turkey) using helium isotopes, Earth Planet. Sci. Lett.,
341–344, 68–78.
Cros, E., and L. Géli (2013), Characterisation of microseismicity in the Western Sea of Marmara: Implications in terms of seismic monitoring.
[Available at http://doi.org/10.13155/38916.]
Doser, D. I., and D. R. Yarwood (1994), Deep crustal earthquakes associated with continental rifts, Tectonophysics, 229(1–2), 123–131.
Drab, L., A. Hubert-Ferrari, S. Schmidt, P. Martinez, J. Carlut, and M. E. Ouahabi (2015), Submarine paleo-earthquake record of the
Cinarcik segment of the North Anatolian Fault in the Marmara Sea (Turkey), Bull. Seismol. Soc. Am, 105, 622–645, doi:10.1785/
0120130083.
Dziewonski, A. M., T.-A. Chou, and J. H. Woodhouse (1981), Determination of earthquake source parameters from waveform data for studies
of global and regional seismicity, J. Geophys. Res., 86, 2825–2852, doi:10.1029/JB086iB04p02825.
Ekström, G., M. Nettles, and A. M. Dziewonski (2012), The global CMT project 2004–2010: Centroid-moment tensors for 13,017 earthquakes,
Phys. Earth Planet. Inter., 200–201, 1–9, doi:10.1016/j.pepi.2012.04.002.
Ergintav, S., U. Doğan, C. Gerstenecker, R. Çakmak, A. Belgen, H. Demirel, C. Aydın, and R. Reilinger (2007), A snapshot (2003–2005) of the 3D
postseismic deformation for the 1999, Mw = 7.4 Izmit earthquake in the Marmara region, Turkey, by first results of joint gravity and GPS
monitoring, J. Geodyn., 44, 1–18.
Ergintav, S., R. E. Reilinger, R. Çakmak, M. Floyd, Z. Çakır, U. Doğan, R. W. King, S. McClusky, and H. Özener (2014), Istanbul’s earthquake hot
spots: Geodetic constraints on strain accumulation along faults in the Marmara seismic gap, Geophys. Res. Lett., 41, 5783–5788,
doi:10.1002/2014GL060985.
Fraser, J., K. Vanneste, and A. Hubert-Ferrari (2010), Recent behavior of the North Anatolian Fault: Insights from an integrated
paleoseismological data set, J. Geophys. Res., 115, B09316, doi:10.1029/2009JB006982.
Hergert, T., O. Heidbach, A. Becel, and M. Laigle (2011), Geomechanical model of the Marmara Sea region—I. 3-D contemporary kinematics,
Geophys. J. Int., 185(3), 1073–1089, doi:10.1111/j.1365-246X.2011.04991.x.
Hirata, N., and M. Matsu’ura (1987), Maximum-likelihood estimation of hypocenter with origin time eliminated using nonlinear inversion
technique, Phys. Earth Planet. Inter., 47, 50–61.
Karabulut, H., J. Schmittbuhl, S. Özalaybey, O. Lengliné, A. Kömeç-Mutlu, V. Durand, M. Bouchon, G. Daniel, and M. P. Bouin (2011), Evolution
of the seismicity in the eastern Marmara Sea a decade before and after the 17 August 1999 Izmit earthquake, Tectonophysics, 510, 17–27,
doi:10.1016/j.tecto.2011.07.009.
Le Pichon, X., et al. (2001), The Main Marmara Fault, Earth Planet. Sci. Lett., 192, 595–616.
Meghraoui, M., M. E. Aksoy, H. S. Akyüz, M. Ferry, A. Dikbaş, and E. Altunel (2012), Paleoseismology of the North Anatolian Fault at Güzelköy
(Ganos segment, Turkey): Size and recurrence time of earthquake ruptures west of the Sea of Marmara, Geochem. Geophys. Geosyst., 13,
Q04005, doi:10.1029/2011GC003960.
Murru, M., A. Akinci, G. Falcone, S. Pucci, R. Console, and T. Parsons (2016), M ≥ 7 earthquake rupture forecast and time dependent probability
for the sea of Marmara region, Turkey, J. Geophys. Res. Solid Earth, 121, 2679–2707.
Oglesby, D. D., and P. M. Mai (2012), Fault geometry, rupture dynamics and ground motion from potential earthquakes on the North
Anatolian Fault under the Sea of Marmara, Geophys. J. Int., 188, 1071–1087, doi:10.1111/j.1365-246X.2011.05289.x.
Okay, A., E. Demirbağ, H. Kurt, N. Okay, and I. Kuşçu (1999), An active, deep marine strike-slip basin along the North Anatolian Fault in Turkey,
Tectonics, 18, 129–147, doi:10.1029/1998TC900017.
Parke, J. R., T. A. Minshull, G. Anderson, R. S. White, D. McKenzie, I. Kuscu, J. Bull, N. Go rür, and A. M. C. Sengor (1999), Active faults in the Sea of
Marmara, western Turkey, imaged by seismic reflection profiles, Terra Nova, 11, 223–227.
Parsons, T. (2004), Recalculated probability of M ≥ 7 earthquakes beneath the Sea of Marmara, Turkey, J. Geophys. Res., 109, B05304,
doi:10.1029/2003JB002667.
Pinar, N. (1943), Marmara Denizi Havzasinin Sismik Jeoloji ve Meteorolojisi, PhD Thesis, Institut de Géologie, Institut de Physique Générale de
ľUniversité ďIstanbul, Kenan Matbaasi, Istanbul, 64 pp.
Pondrad, N., R. Armijo, G. C. P. King, B. Meyer, and F. Flerit (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.
Reilinger, R., et al. (2006), GPS constraints on continental deformation in the Africa–Arabia–Eurasia continental collision zone and
implications for the dynamics of plate interactions, J. Geophys. Res., 111, B05411, doi:10.1029/2005JB004051.
Reyners, M., D. Eberhart-Phillips, and G. Stuart (2007), The role of fluids in lower crustal earthquakes near continental rifts, Nature, 446,
1075–1078, doi:10.1038/nature05743.
Sakic, P., et al. (2016), No significant steady state surface creep along the North Anatolian Fault offshore Istanbul: Results of 6 months of
seafloor acoustic ranging, Geophys. Res. Lett., 43, 6817–6825, doi:10.1002/2016GL069600.
Acknowledgments
We thank the captains and crews of DSV
Alcatras and R/V Tübitak Marmara. We
also thank Satoshi Shimizu, Takuya
Maekawa, Seiichi Mori, Kaoru Tsukuda,
Ozkan Cok, Murat Suvarikli, Ibrahim
Zafer Ogutcu, and Suleyman Tunc for
the preparation of OBSs and onboard
operations and Takane Hori, Ryosuke
Ando, and Hiroaki Yamanaka for their
fruitful discussions about our project. All
figures were created using Generic
Mapping Tools [Wessel and Smith, 1991].
We gratefully acknowledge Editor
Martha Savage and Associate Editor
Mladen Nedimović for their support.
Comments and suggestions from three
anonymous reviewers were helpful to
greatly improve our manuscript. OBS
observations were conducted under the
Marmara Disaster Mitigation (MarDIM)
project, formally known as the
“Earthquake and Tsunami Disaster
Mitigation in the Marmara Region and
Disaster Education in Turkey” project.
MarDIM receives financial support from
the Japan International Cooperation
Agency, Japan Science and Technology
Agency, and the Ministry of
Development in Turkey. The hypocenter
catalogue of KOERI was obtained from
KOERI seismic network (doi:10.7914/SN/
KO). Please contact Y.Y. for any requests
for data and other information.

Sato, T., J. Kasahara, T. Taymaz, M. Ito, A. Kamimura, T. Hayakawa, and O. Tan (2004), A study of microearthquake seismicity and focal
mechanisms within the Sea of Marmara (NW Turkey) using ocean bottom seismometers (OBSs), Tectonophysics, 391, 303–314,
doi:10.1016/j.tecto.2004.07.018.
Schmittbuhl, J., H. Karabulut, O. Lengliné, and M. Bouchon (2015), Seismicity distribution and locking depth along the Main Marmara Fault,
Turkey, Geochem. Geophys. Geosyst., 17, 954–965, doi:10.1002/2015GC006120.
Shudovsky, G. N. (1985), Source mechanisms and focal depths of East African earthquakes using Rayleigh-wave inversion and body-wave
modelling, Geophys. J. R. Astron. Soc., 83, 563–614.
Simpson, F. (1999), Stress and seismicity in the lower continental crust: A challenge to simple ductility and implications for electrical
conductivity mechanisms, Surv. Geophys., 20, 201–227.
Snoke, J. A. (2003), FOCMEC: FOCal MEChanism determinations, in International Handbook of Earthquake and Engineering Seismology,
edited by W. H. K. Lee et al., pp. 1629–1630, Academic Press, Amsterdam.
Tary, J. B., L. Geli, P. Henry, B. Natalin, L. Gasperini, M. Comoglu, N. Cagatay, and T. Bardainne (2011), Sea-bottom observations from the
western escarpment of the Sea of Marmara, Bull. Seismol. Soc. Am., 101, 2, doi:10.1785/012000014.
Thurber, C. H., and D. Eberhart-Phillips (1999), Local earthquake tomography with ﬂexible gridding, Comput. Geosci., 25, 809–818.
Urabe, T., and S. Tsukada (1992), WIN—A workstation program for processing waveform data from microearthquake networks [in Japanese],
Prog. Abst. Seismol. Soc. Jpn., 2, P-41.
Watanabe, H. (1971), Determination of earthquake magnitude at regional distance in and near Japan [in Japanese with English abstract],
Zisin, 2(24), 189–200.
Wessel, P., and W. H. F. Smith (1991), Free software helps map and display data, Eos. Trans. AGU, 72, 441, doi:10.1029/90EO00319.
Yaltırak, C. (2002), Tectonic evolution of the Marmara Sea and its surroundings, Mar. Geol., 190, 493–529.
Yamamoto, R., M. Kido, Y. Ohta, N. Takahashi, Y. Yamamoto, D. Kalafat, A. Pinar, S. Ozeren, and Y. Kaneda (2016), Creep rate measurement and
fault modeling at the North Anatolian Fault, beneath the Sea of Marmara, Turkey, by means of acoustic ranging, paper presented at Japan
Geoscience Union meeting 2016, S-CG59-07, Chiba, Japan.
Yamamoto, Y., N. Takahashi, S. Citak, D. Kalafat, A. Pinar, C. Gurbuz, and Y. Kaneda (2015), Offshore seismicity in the western Marmara Sea,
Turkey, revealed by ocean bottom observation, Earth Planets Space, 67, 147, doi:10.1186/s40623-015-0325-9.
Zhang, H., and C. Thurber (2006), Development and applications of double-difference seismic tomography, Pure Appl. Geophys., 163,
373–403.

OBS Seismicity Monitoring in Marmara

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Similar to OBS Seismicity Monitoring in Marmara

Similar to OBS Seismicity Monitoring in Marmara (20)

More from Ali Osman Öncel

More from Ali Osman Öncel (20)

Recently uploaded

Recently uploaded (20)

OBS Seismicity Monitoring in Marmara