Active seismotectonics of the East Anatolian Fault

SUMMARY

The East Anatolian Fault (EAF) is a 700-km-long left-lateral transform fault located between the Anatolian and Arabian plates. The proximity of the Euler Pole to the Arabia–Anatolia Plate boundary leads to rapid changes in plate velocity along the boundary, which is manifested by the decreasing slip rates from east (10 mm yr^–1) to west (∼1–4 mm yr^–1). The EAF displays heterogeneous seismicity patterns with seismic gaps, localized clusters and broad diffuse zones. In this study, in order to understand the origin of these complexities and quantify the seismic hazard along the EAF, we present an improved seismicity catalogue with more than 26 000 earthquakes and 160 focal mechanisms from regional moment tensor inversion between 2007 and 2020. The focal mechanisms and seismicity show that the EAF dips towards north and forms a well-defined plate boundary in the east between Palu and Çelikhan with almost pure left-lateral motion. Further west, the boundary becomes broader with activity along subparallel faults. Focal mechanisms show heterogeneous stress orientations in consistence with geodetically determined strain rate field. The stress orientations show a transition from strike-slip to extension towards the west of Çelikhan. Amongst all segments of EAF, the Pütürge segment, which holds the near-repeating earthquakes in the vicinity of the nucleation of the 2020 M_w 6.8 earthquake, is distinguished with its steady and high rate of seismicity. Further east, the neighbouring Palu segment is characterized by several distinct moderate earthquakes. We do not observe any change in the seismicity rate on these segments of the EAF following large earthquakes. In order to quantify the seismic hazard along the EAF, we calculate the recurrence time and maximum magnitude for each segment by using an extended seismicity catalogue of 150 yr including the large historical earthquakes and the geodetic strain rate. The results show ∼150 yr recurrence time with M_max∼6.7–7.0 along the seismically active Palu and Pütürge segments on the east, while relatively silent western segments yield longer recurrence times; 237–772 for Pazarcık and 414–917 for Amanos segments with slightly larger magnitudes (M_max ∼7–7.4). We infer that the seismicity patterns and strain-rate field along the EAF are shaped by several factors such as strong geometrical irregularities, heterogeneous coupling and complex plate motion leading to rapid change of fault slip rate.

Creep and deformation, Seismic cycle, Earthquake hazards, Earthquake source observations, Seismicity and tectonics

1 INTRODUCTION

The East Anatolian Fault (EAF) is a ∼700-km-long left-lateral strike-slip fault along Anatolia–Arabia Plate boundary (Arpat & Şaroğlu 1972; Arpat & Şaroğlu 1975; Hempton et al. 1981) (Fig. 1). The EAF lies along the southern boundary of Anatolia and together with Bitlis-Zagros Fold and Thrust Belt and the Caucasus Fold and Thrust Belt accommodates the relative motion between Eurasia and Arabia (McKenzie 1972; Şengör et al. 1985; Le Pichon & Kreemer 2010).

Figure 1.

Tectonic setting of the study region with shaded topography (http://dds.cr.usgs.gov/srtm/version2_1/SRTM30/). North direction is indicated on the top left. Thick and thin red lines show the East and North Anatolian Faults, respectively. Secondary faults are shown by thin black lines (Emre et al. 2013). Black arrows on the fault traces indicate the relative motion of the faults. White arrows indicate the GPS velocities with Arabia fixed reference frame (Reilinger et al. 2006). (Inset) Figure shows the major plate boundaries surrounding the study region. Red arrows represent direction of the plate motions of Anatolia and Arabia. Hellenic and Cyprus Arcs are shaded in red. Blue triangle represents the position of the Euler pole from Bletery et al. (2020). Study region is bounded by the red box. EAF: East Anatolian Fault. NAF: North Anatolian Fault. LH. Lake Hazar. BZFTB: Bitlis-Zagros Fold and Thrust Belt. T: Türkoğlu. MOF: Malatya-Ovacık Fault. SF: Sürgü Fault. AF: Amanos Fault. YsF: Yesemek Fault. YF: Yumurtalık Fault. TF: Toprakkale Fault. NFZ: Narlı Fault Zone. DSF: Dead Sea Fault. VFZ: Varto Fault Zone. For simplicity, we refer to the SW and NE of the EAF in the text as west and east, respectively. Inset: K: Kahramanmaraş Junction. Ç: Çelikhan. Krl: Karlıova, CA: Cyprus Arc. HA: Hellenic Arc.

Several geological and geodetic studies estimated the slip rates on the segments along the EAF and its subparallel segments (e.g. McClusky et al. 2000; Reilinger et al. 2006; Yılmaz et al. 2006; Meghraoui et al. 2011; Duman & Emre 2013). Geological and geomorphological studies indicate that the slip rate systematically decreases from ∼10 mm yr^–1 near Karlıova to ∼4 mm yr^–1 near Türkoğlu (Koç & Kaymakçı 2013; Mahmoud et al. 2013; Bayrak et al. 2015; Aktuğ et al. 2016; Fig. 1). Further west, the slip rate decreases further down to ∼2.5 mm yr^–1 on the main fault and to ∼1 mm yr^–1 on subparallel faults (Herece 2008; Duman & Emre 2013; Gülerce et al. 2017; Yönlü et al. 2017). The slip rates estimated from Holocene offsets are compatible with the geodetic fault slip rates (Duman & Emre 2013).

Devastating earthquakes occurred on the EAF during both historical and instrumental periods; The 1789 (M7.2), 1874 (M7.1) and 1875 (M6.7) earthquakes ruptured the Palu, the section beneath the Lake Hazar and Pütürge Segments (Fig. 2) (Ambraseys & Finkel 1995; Tan et al. 2008; Duman & Emre 2013; Köküm & Özçelik 2020). The Palu segment released significant portion of the accumulated strain after the 2010 M_w 6.0 and 2011 M_w 5.4 events. Similarly, the 2020 M_w 6.8 Sivrice earthquake ruptured ∼45 km of ∼96-km-long Pütürge segment (Duman & Emre 2013; Konca et al. 2021). To the west of Çelikhan, the locations and rupture extents of the 1795 (M7.0) event on the Pazarcık 1893 (M7.1) event on the Erkenek 1872 (M7.2) event on the Amanos segments are poorly known (Duman & Emre 2013, Fig. 2). The rupture lengths and terminations are usually associated with segments delimited by the releasing and restraining bends observed on the surface morphology (Ambraseys 1989; Ambraseys & Jackson 1998; Tan et al. 2008; Palutoğlu & Şaşmaz 2017).

Figure 2.

Fault segmentation based on the historical earthquakes and geology along the EAF and the faults further west of KMTJ (Duman & Emre 2013). The major structural features, for example releasing and restraining bends, are also shown. The coloured bands indicate the segments of the EAF: Palu (blue), Pütürge (red), Erkenek(green), Pazarcık (orange), Amanos (cyan) and Karlıova (grey). The basins are shaded in yellow. The locations of historical earthquakes are from Ambraseys (1989), Ambraseys & Jackson (1998) and Tan et al. (2008). FC: Fault Complex, RB: Releasing Bend, RtB: Restraining Bend; DRrB: Double Restraining Bend; PB: Paired Bend. Ulv: Uluova. NAF: North Anatolian Fault.

Seismic observations during the instrumental period had limited resolution to characterize the kinematics and the geometry of the segmentation until recently. With the availability of high quality digital seismic stations, the seismicity distribution provided further evidence of fault segmentation and diversity of the fault mechanism solutions of small and moderate size earthquakes.

The studies of earthquake source mechanisms show that the faulting along eastern part of the EAF is dominated by a stress regime with horizontal maximum and minimum compressional stresses, characterized by pure left-lateral strike-slip faulting (Lyberis et al. 1992; Kiratzi 1993). Taymaz et al. (1991) studied five large to moderate size earthquakes between 1964 and 1986 which yield strike-slip solutions with reverse and normal components. Tan et al. (2011) obtained the focal mechanisms of 2010 M_w 6.0 Palu-Kovancılar earthquake and 16 aftershocks within 20 km section of the Palu segment using both first motion polarities and local moment tensor inversion. Bulut et al. (2012) provided focal mechanism solutions of 40 earthquakes along the EAF with M > 4 from full waveform inversion, indicating mostly left-lateral strike-slip and a few thrust and normal mechanisms on the Palu and Pütürge segments. Birsoy (2018) studied the southwestern section of Çelikhan and presented 29 focal mechanisms with M > 3.6 having oblique left-lateral strike-slip as well as normal mechanism solutions. Cambaz & Mutlu (2016) presented left-lateral strike-slip solutions of 15 earthquakes with M > 4.5 along the EAF.

Recently, the studies of the 2020 M_w 6.8 Sivrice earthquake and its aftershocks improved the understanding of the structure and the behaviour of the Pütürge segment. Pousse-Beltran et al. (2020) presented some of the aftershock source mechanisms with normal component on the western part of the 2020 M_w 6.8 rupture. They also suggested that the western segment of coseismic slip model displays a small amount of normal component (rake angle of −18). Both Pousse-Beltran et al. (2020) and Konca et al. (2021) suggested that the dip angle is shallower (55°–64°) to the west of the Pütürge segment compared to the eastern part (77°–80°). Furthermore, Gallovic et al. (2020) decomposed the 2020 M_w 6.8 event into three subevents, the westernmost of which has a significant thrust component. The recent studies showing heterogeneous interseismic fault coupling (Blettery et al. 2020) and coseismic slip distribution of the 2020 earthquake with distinct patches show that the significant heterogeneities exist both in interseismic behaviour and geometry along the Pütürge segment.

While the majority of the reported focal mechanisms display left-lateral strike slip motion, associated with the main fault, some mechanisms with normal and thrust components also exist along the EAF. However, the sparsity of the earthquake source mechanisms with non-strike slip solutions prohibits finding a mechanics how these variations are related to the main fault. Therefore, understanding the cause of these oblique mechanisms and the cause of the heterogeneities in the associated stress field requires a greater number of earthquake source mechanism solutions.

The EAF displays seismicity patterns with gaps, localized clusters and sections with diffuse activity. This complexity partly results from the geometry and the direction of the plate motion. The Euler Pole of the Arabian–Anatolian motion is located ∼13° away from the EAF towards SW, implying that the fault zone is possibly subjected to a rapid change in the direction of the relative motion between the two plates (Fig. 1 inset; Reilinger & McClusky 2011; Bletery et al. 2020).

Considering the proximity of the Euler Pole to the fault zone, geometrical complexities and irregularities in addition to the complex plate boundaries which delimit the fault zone at both ends, it is unlikely that a pure left-lateral motion along the EAF would be maintained. In this study, we try to improve our understanding of the seismic behaviour of the EAF with new seismic observations. The new seismicity and focal mechanism catalogues lead to a better characterization of the fault kinematics and more accurate description of the geometry. Based on the distribution of different types of focal mechanisms, seismicity, cumulative moment release along with the results of the stress inversion and comparison with the previous study of the strain-rate field from Weiss et al. (2020), we present an improved view of the active seismotectonics and how the aforementioned complex motion is accommodated along EAF. We finally use the geodetic strain rate field, fault coupling and the long term seismicity to infer maximum magnitudes and recurrence periods of earthquakes along the segments of the EAF.

2 TECTONIC EVOLUTION OF THE EAF AND SEGMENTATION

The EAF evolved as a consequence of the final stage of Eurasia–Arabia collision during Middle Late Miocene (16–20 Ma) (Şengör & Yılmaz 1981; Dewey et al. 1986; Hempton 1987; Yılmaz 1993; Robertson 2000). The timing of transition from palaeotectonics with thrust and reverse faulting to neotectonics with strike-slip faulting is still under debate. While early studies suggest that the transition started in Late Serravalian (∼11 Ma, Şengör et al. 1985; Şaroğlu & Yılmaz 1986), later studies argued that the EAF evolved in Late Miocene–Early Pliocene (∼5 Ma) due to detachment of the slab under Eastern Anatolia (Faccenna et al. 2006). Geological and geodetical slip rates along with the offset of volcanism imply that the left-lateral offsets along the EAF initiated no earlier than Late Pliocene (3 Ma, Şaroğlu et al. 1992; Westaway & Arger 2001; Allen 2010).

The EAF was exposed to several tectonics inversions between Middle Miocene (∼12 Ma) and Early Pliocene (3 Ma) resulting in changes of tectonic regime, basin type and deformation pattern (Koçyiğit et al. 2001). During the Neogene deformation, the compressional Miocene structures with favourable orientation were reactivated and the EAF was transformed into its current geometry with several segments and geometrical complexities (Hempton et al. 1983; Westaway & Arger 1996; Duman & Emre 2013.

The EAF is bounded by the Karlıova basin to the east, where the relative motion between Eurasia and Arabia is distributed among the NAF, EAF and the Varto Fault Zone (VFZ, Şengör 1979; Şaroğlu 1985; Şengör et al. 1985). In the west, termination of the EAF is still debated. Early studies on the EAF suggested that the main fault continues towards Cyprus Arc where the convergence is accommodated between Nubia and Anatolia (McKenzie 1972; Dewey et al. 1973). Several studies argued that the EAF joins Dead Sea Fault (DSF) at Kahramanmaraş Triple Junction (KMTJ), near Türkoğlu (e.g. Şengör 1979; Jackson & McKenzie 1984; Hempton 1987; Muehlberger & Gordon 1987; Barka & Kadinsky-Cade 1988). Other studies claimed that the EAF extends further south and meets the DSF in Hatay-Adana-Cilicia basin (e.g. Arpat & Şaroğlu 1975; Chorowicz et al. 1994; Duman & Emre 2013; Şengör et al. 2019). In this study, we adopt the convention of Duman & Emre (2013) for the segment boundaries and names along the EAF, and include the segments further west of the KMTJ into the analysis.

The EAF is characterized by several extensional and compressional geometric complexities such as the Göynük paired bend, the Gökdere restraining bend, the Lake Hazar releasing bend, the Gölbaşı releasing step over, the Türkoğlu releasing step over, the Nurhak Fault Complexity, the Göksun releasing bend and the Delihalil releasing bend from east to west (Duman & Emre 2013; Fig. 1). The releasing bends are accompanied by pull-apart basins such as Bingöl, Karakoçan, Kovancılar, Palu-Uluova, Lake Hazar, Malatya and Gölbaşı from east to west (e.g. Lyberis et al. 1992; Westaway 1994; Duman & Emre 2013).

The most recent work on the segmentation along the EAF based on field observations was presented by Duman & Emre (2013) where they divide the main strand of the EAF into seven segments. In addition, they also considered the subparallel faults that split from the EAF around the Sürgü Fault as the northern strand of the EAF (Fig. 2). The main fault, from east to west, comprises of the Karlıova, Ilıca, Palu, Pütürge, Erkenek, Pazarcık and Amanos segments. The northern strand consists of Sürgü, Toprakkale and Yumurtalık segments. In their fault map, the Sürgü Fault splays from the main fault, near Çelikhan and extends towards west for ∼30 km.

3 DATA AND METHODS

3.1 Data

We compiled P, S phase picks and waveforms of 13 yr (01.01.2007–31.12.2019) for the study region. We compiled the phase picks from the bulletins of Kandilli Observatory and Earthquake Research Institute (KOERI:,2001; http://www.koeri.boun.edu.tr), Disaster and Emergency Management Authority (AFAD, 1990; https://tdvms.afad.gov.tr) and The Scientific and Technological Research Council of Turkey (TUBITAK-MRC, 2020; https://mam.tubitak.gov.tr/en). We obtained a catalogue containing more than 26 000 events. The station distribution in the region changed significantly during 13 yr of observation period, hence the accuracy of the earthquake locations was not uniform throughout the observation period. The highest accuracy was achieved during the time period between 2007 and 2013, when a dense array of stations was installed in order to monitor the activity along the EAF as part of the TURDEP project (İnan et al. 2007) operated by TUBITAK-MRC (Fig. S1). As a result, the detection threshold and location uncertainties decreased for the observation period between 2007 and 2013. We built a waveform database from the continuous data using the three networks to revise the phase readings with large residuals and for the focal mechanism determination. The long term seismicity catalogue between 1905 and 2019 is obtained from KOERI. The locations and magnitudes of the historical earthquakes are from Duman & Emre (2013) and Köküm & Özçelik (2020).

3.2 1-D velocity model and relocation

For earthquake locations, we used HYPOCENTER location software (Lienert & Havskov 1995). The initial velocity models which were used by the two national agencies to locate earthquakes had large average residuals for the 26 000 events (∼0.86 s). We improved the initial crustal model using available tomographic models in the region (Delph et al. 2015) and average Vp/Vs ratios (1.74 km s^–1) obtained from the seismic phase readings (Fig. S5).

A single 1-D velocity model may not be representative for ∼600-km-long fault. However, the number of earthquakes in the west of Çelikhan with magnitudes sufficiently large to be detected by the seismic network were not enough to reliably constrain local crustal model and the station residuals. Therefore, a representative 1-D velocity model for the region is computed from 700 selected earthquakes of the catalogue using VELEST (Kissling et al. 1994) inversion code. The selection criteria for 700 events is based on the number of phase readings (>25) and azimuthal gap (<80°, Fig. S2, Table S1). The deviations from 1-D velocity model are accounted in the station corrections. After the relocation using the new velocity model and station corrections, the mean RMS of the initial catalogue is reduced from ∼0.86 to ∼0.2 s (Fig. S3). Pousse-Beltran et al. (2020) presented four different 1-D crustal models for four different segments of the EAF. We compared the 1-D velocity model of this study with their four models and tested using waveform inversion to determine source mechanisms of four earthquakes for the four segments of the fault (Figs S6a–d). The inversion results are similar with variance reductions slightly better with the single 1-D model of this study.

We adopt the magnitudes of the earthquakes from the reported values of the two national agencies. Two agencies (KOERI and AFAD) use moment magnitude scale for the earthquakes with M > 3.5. The magnitudes of the smaller earthquakes in both catalogues may differ. Therefore, we used the empirical relationships from Tan (2021) which are generated from general orthogonal regression in order to estimate the equivalent moment magnitudes. The magnitude differences of two catalogues from two agencies are mostly due to the use of different stations. We observed that 75 per cent of the magnitudes of the same events reported by two agencies have differences less than ±0.2. Therefore, we used the magnitudes reported by AFAD for common events as they have better station coverage in the region.

Fig. S7 shows the statistics of the final catalogue. The magnitude completeness of the catalogue is calculated as ∼2.5. The mean horizontal location uncertainty is less than 2.0 km in NS and EW directions. The mean of the depth uncertainty is ∼3.5 km, varying between 2.0 and 8 km. The azimuthal gap varies between 90 and 150 for most of the earthquakes.

3.3 Earthquake focal mechanisms

The earthquake source mechanisms were computed from the inversion of regional waveforms in the study region using Cut-and-Paste method (CAP, Zhao & Helmberger 1994; Zhu & Helmberger 1996; Zhu & Ben Zion 2013). Three component seismic waveforms are split into five time windows. The time window is 35 s for vertical and radial P_nl and 70 s for the three components of S-surface wave windows. The Green's function errors are accounted for allowing time-shifts between the data and synthetic waveforms. Each time window is filtered by different bandpass filters and weighted separately in order to compute the synthetic waveforms. In addition, a distance range scaling factor is introduced which reduces the effect of geometrical decay with distance in the misfit function. Finally, the best-fitting fault parameters and moment magnitudes are obtained from a grid-search over strike, dip, rake and moment magnitude.

The time shifts are allowed up to 3 s for P_nl and 5 s for S-surface wave windows. The frequency band for the P-wave and S-wave windows are chosen as 0.04–0.1 Hz and 0.04–0.08 Hz, respectively. Waveform fits with correlation coefficient lower than 0.5 are removed manually after an initial inversion step. Fig. S8 shows examples of waveform fits. 1-D Green's functions were calculated by using the frequency-wave number algorithm of Zhu & Rivera (2002) using the 1-D velocity model obtained in this study.

The solutions with variance reduction greater than 50 per cent are included in the final focal mechanism catalogue (Fig. S9, Table S2). 16 of the computed earthquake mechanisms with M_w > 4.5 is compared with the Global CMT solutions of the same events (Dziewonski et al. 1981; Ekström et al. 2012, Fig. S10). This comparison confirms the reliability of the source mechanisms solutions for 160 earthquakes with magnitude M_w ≥ 3.5 between 2007 and 2020.

3.4 Stress inversion

We compute the stress orientations from focal mechanisms using the method of Vavrychuk (2014). The stress inversions include uncertainties due to several factors such as insufficient number of mechanisms and selection of fault planes. However, reasonable uncertainties may exist for the estimate of long term fault behaviour even large number of earthquake mechanisms are used when a limited time interval is considered. We acknowledge that this technique is based on the assumptions of uniform tectonic stress, pre-existing faults, slip vectors and shear tractions being parallel on the fault plane.

3.5 Estimation of maximum magnitude (M_max) and average recurrence period (T_r)

Molnar (1979) used the Gutenberg–Richter relationship and slip rate on a fault to estimate the recurrence time (T_r) and maximum magnitude (M_max) along a fault zone. Recently, Stevens & Avouac (2021) proposed a method to predict M_max and T_r by equating the geodetically accumulated strain to the one released by earthquakes considering the Gutenberg–Richter law and applied this method to India–Asia collision zone.

The availability of the seismicity catalogue, the fault coupling distribution and the detailed strain-rate field from geodetic data allow us to perform similar analysis along the segments of the EAF. We estimate the M_max and T_r for each segment using the strain rate field (Weiss et al. 2020) and the long-term seismicity of the EAF following Stevens & Avouac (2021). This method assumes that the geodetic strain rates are stationary through time and the moment is conserved

The moment build-up rate is:

$$\begin{eqnarray} \dot{m} = {c_{\rm g}} \cdot \mu \cdot {T_{\rm s}} \cdot A \cdot \dot{ \in }, \end{eqnarray}$$

(1)

where μ is the rigidity (35 GPa), A is the area where the strain rate is computed, T_s is the seismogenic thickness and c_g is the geometric factor and |$\dot{ \in }$| is the second invariant of the strain rate. The value of c_g depends on the orientation and dip angle of the fault in order to take into account the partitioned horizontal strain rate onto the fault plane. It is computed using |${c_{\rm g}} = 1/[\sin \delta \cdot \cos \delta ]$|⁠, where δ is the dip angle.

In order to determine the Gutenberg–Richter parameters, in addition to the 13-yr catalogue analysed as part of the study, we also utilize the long-term seismicity catalogue between the years of 1905 and 2019 from KOERI. The magnitude completeness of the long term catalogue of the EAF is determined as M = 4.5. We further expand this catalogue by including the historical earthquakes (Duman & Emre 2013; Köküm & Özçelik 2020) and calculate the a and b values. Once the moment build-up rate and Gutenberg–Richter parameters are calculated,

$$\begin{eqnarray} {M_{\max }} = \frac{1}{{{{3}/\!{{{2^{ - b}}}}}}}\left( {{{\log }_{10}}\left( {1 - \frac{{2b}}{3}} \right) + {{\log }_{10}}(\alpha \cdot \dot{m}) - 9 - a} \right), \end{eqnarray}$$

(2)

where α is a constant related to the interseismic coupling which gives the ratio of the seismically released moment rate to the accumulated moment rate. The value of α is determined by calculating the average of the coupling coefficient distribution for each segment from Bletery et al. (2020) and assumed to be 0.4 and 0.3 for Palu and Pütürge segments, respectively. For the west of KMTJ, where fault coupling is not available, α is assumed to be 0.5. The recurrence time is calculated using M_max,

$$\begin{eqnarray} {T_{\rm r}} = 1/{10^{(a - b\,{M_{\max }})}}, \end{eqnarray}$$

(3)

4 RESULTS

4.1 Seismicity distribution along EAF

Fig. 3 shows the seismicity distribution of relocated events (∼26 000) with the fault geometry adopted from Emre et al. (2013). Along the EAF, the seismicity rate decreases dramatically from east to west. The seismic activity is relatively localized and follows the surface trace of the EAF between Çelikhan and Palu. The activity on this section mostly occurs on the northern part of the fault trace but several clusters are also observed on the southern part of the fault. To the west of Çelikhan, the seismic activity is low and the majority of the activity takes place far from the surface trace of the EAF (e.g. Yumurtalık-Toprakkale-Narlı Faults, Fig. 3).

Figure 3.

(a) Map view and (b) depth section of the seismicity along the EAF during the period 2007–2019. Earthquakes greater than M2.5 around the fault trace are shown by coloured circles colours based on the segment that they belong. The depth section includes seismicity ±15 km around the fault trace at surface. Faults are shown by black lines (Emre et al. 2013). Numbers on the map are the ID's of the events with M > 5 which are represented by yellow stars. Red dashed lines indicate segment boundaries as introduced in Fig. 2. White stars represent the locations of the near-repeating earthquakes (Konca et al. 2021). KMTJ: Kahramanmaraş Triple Junction. LH: Lake Hazar. BZFTB: Bitlis-Zagros Fold and Thrust Belt. SF: Sürgü Fault. AF: Amanos Fault. YsF: Yesemek Fault. YF: Yumurtalık Fault. TF: Toprakkale Fault. NFZ: Narlı Fault Zone.

In order to explore the seismicity rate variations along the EAF, the activity was split into five pieces relying on the segmentation by Duman & Emre (2013) based on field observations, surface offsets and historical earthquakes. These segments are from east to west: Palu, Pütürge, Erkenek, Pazarcık, Amanos. On Fig. 3, the seismic activity with events M > 2.5 are displayed with different colours for the five segments. We computed the magnitude completeness and b-values for each segment. The completeness is similar for all segments as M_c = ∼2.5 and b-values are varying between 0.84 and 1.3 (Table 1, Fig. S11).

Table 1.

A summary of geodetic and seismologic parameters on the five segments of EAF. NOE: Number of events. Geodetic strain accumulation rates are from Reilinger et al. (2006). The second row with * is obtained from the rate of the earthquakes with M_c = 2.5 per year between 2007 and 2019. The dip angles of Amanos and Pazarcık are not estimated in this study and assumed to be 90°.

	Palu	Pütürge	Erkenek	Pazarcik	Amanos
b-Value (2007–2019)	0.84	0.91	1.30	1.10	1.10
*b-Value (Since ∼largest reported eq.)	0.90	0.96	0.88	0.94	0.92
NOE with M2.5+ (2007–2019)	224	225	52	70	39
NOE with M5+ (2007–2019)	6	2	-	-	-
Relative Background Seismicity rate (2007–2019)	7	18	6	3	1
Seismic Slip Rate (mm yr–1) (2007–2019)	∼10	∼2.5	<1	<1	<1
Geodetic strain accumulation rates (mm yr–1) (2007–2019)	10	10	6.5	4	3
Estimated Average Dip Angle	65	70	57	-	-
M of Largest eq. recorded (last 50 yr)	6.0	6.0	5.7	4.7	5.0
M of Largest eq. recorded (last 25 yr)	6.0	6.8	5	4.6	4.4
Largest eq. reported (Latest historical eq.)	7.2	6.8	7.1	7.0	7.2
Mmax	6.8–7.0	6.8–7.0	-	7.0–7.3	7.2–7.4
Tr	108–228	100–230	-	237–772	414–915
Log(MBR) (dyn cm–1)	25.0489	25.0034	24.6319	24.7816	24.8902
Log(MRR) (dyn cm–1)	25.2112	25.9285	24.7167	23.9697	24.1088

The easternmost Palu segment, ∼77 km long, has two distinct clusters of activity at the two ends. The largest events in both clusters have magnitudes 6.0 and 5.4 (Fig. 3, events with numbers 3 and 10, respectively). M_w 6.0, Palu (Kovancılar) earthquake occurred within the eastern cluster in 2010 and is followed by an aftershock activity which dominates the seismicity to the east of the Palu segment (Tan et al. 2011). The largest aftershock with M_w 5.8 occurred on the same day. In the west of Palu segment, M_w 5.4 earthquake occurred in 2011. The aftershock activity of 2011 earthquake forms the major part of the western cluster. Fig. 4 shows the temporal evolution of these clusters with the depth distribution of the earthquakes.

Figure 4.

Depth and space-time distribution of the seismicity along the five domains of the EAF: from top to bottom, Palu, Pütürge, Erkenek, Pazarcık and Amanos Segment. Five segments are introduced with different colours. (a) Grey bars show normalized depth-frequency distribution of the complete catalogue in each segment. (b) Coloured circles represent space-time distribution of the complete catalogue in each segment and scaled by magnitude. (c) Grey bars show normalized depth-frequency distribution of the declustered catalogue in each segment. (d) Coloured circles represent space–time distribution of the declustered catalogue in each segment and scaled by magnitude.

The Pütürge segment, which extends 96 km between the Lake Hazar releasing bend and the Yarpuzlu restraining double bend near Çelikhan, has the highest seismicity rate amongst the segments of the EAF (Fig. 3). The seismic activity is not clustered but spread uniformly both in space (Fig. 3) and time (Fig. 4). Two M_w 5.2 earthquakes occurred on this section in 2007 (Şentürk et al. 2019) and 2019 (Fig. 3, Events with numbers 1 and 13). The epicenter of 2019 event is located within the rupture zone of the 2020 M_w 6.8, Sivrice earthquake, hence it can be considered as an early foreshock (Pousse-Beltran et al. 2020). In fact, it is worthwhile mentioning that the 2020 rupture was preceded by an accelerated seismic activity (Fig. 5). Moreover, Konca et al. (2021) shows that this segment has multiple long-term and short-term near-repeating earthquake clusters during the 10 yr of observation period (2007–2016) right below the eastern tip of the coseismic rupture zone of the 2020 earthquake (white stars on Fig. 3). The near repeating earthquakes tend to have very similar waveforms (correlation coefficient of 95 per cent and higher) which come from very narrow zones (∼100 m), but the magnitudes and time intervals vary significantly so they differ from fully repeating earthquakes which have similar magnitude and time intervals (Konca et al. 2021; Shaddox et al. 2021). The observed near repeating events imply seismic creep on the east of the nucleation zone of the 2020 event (Konca et al. 2021) in consistence with the inferred heterogeneous interseismic coupling along this segment (Bletery et al. 2020).

Figure 5.

Temporal distribution of seismicity along the segments of the EAF. (a) Coloured lines are the cumulative number of earthquakes for each segment between 2007 and 2019. Each curve is normalized with the maximum value of the Palu segment. (b) The seismicity rates from the number of earthquakes per day using a Gaussian operator over a time span of 60 d. (c) The declustered cumulative seismicity for each segment. Each curve is normalized with the maximum value of the Pütürge segment. (d) The seismicity rates from the number of earthquakes in declustered catalogue per day using a Gaussian operator over a time span of 60 d. Dashed black lines mark date of earthquakes with M > 5 (events with numbers in Fig. 3). Colour legend for the segments is at the upper left corner of the panel.

To the west of Çelikhan, the seismicity rate drops significantly around the Nurhak Complex where the Sürgü Fault and the Bitlis Fold and Thrust Belt connects to the ∼62-km-long Erkenek segment. The Erkenek segment has low seismicity rate with no earthquakes with M > 3.7 during the observation period. The seismic activity along the Pazarcık segment (∼82 km long) is restricted to a single cluster with several events of M < 4.5 continuously active through the observation period (Fig. 4). Further west of the Pazarcık segment, the seismicity is mostly localized on the Narlı Fault Zone to the south and along the north of the İskenderun Bay, on Yumurtalık and Toprakkale Faults (Fig. 4).

4.2 Spatiotemporal distribution of declustered seismicity

In this section, we study the spatiotemporal distribution of seismicity, and explore its behaviour within each segment and relation to the changes of plate motion, strain field and segmentation. However, since the observation period is much shorter than the full seismic cycle, the observed pattern of seismicity might be dominated by transients, such as aftershock sequences and swarms (Zhuang et al. 2002). Therefore, in order to obtain the background rate, the seismicity data must be filtered to remove earthquakes due to the coseismic stress changes and post-seismic slip (Marsan et al. 2017).

We declustered the seismicity catalogue using the method of Marsan et al. (2017) which was modified from Zhuang et al. (2002, 2004). Fig. 4 shows the distribution of seismicity before and after declustering as a function of time in addition to the depth distribution for each segment (Fig. 4). The average seismogenic thickness is ∼15 km between the depths of 5 and 20 km which is in agreement with previous studies (Tan et al. 2011; Pousse-Beltran et al. 2020).

The seismicity of the Palu segment is dominated by two events; 2010 M_w 6.0 and 2011 M_w 5.4 earthquakes. These two earthquakes are located 30 km apart from each other and the later event occurred with a delay of 1 yr. Along the Palu segment, seismicity rate after declustering is low, indicating that tectonic stress is released by these two earthquakes.

The Pütürge segment, which holds the largest recorded earthquake (M_w 6.8 in 2020) in the study area, has the most interesting behaviour. It has the highest seismicity rate of all segments with no earthquake with M > 4.5 except for two M_w 5.2 events occurred in 2007 and 2019. The event distribution is quite uniform throughout time and the peak of the depth distribution of events is located at depths ∼10 km (Fig. 4). After declustering, the number of earthquakes still stays high indicating that the pattern is not dominated by main shock–aftershock sequences.

The other 3 segments, Erkenek, Pazarcık and Amanos, have relatively low seismicity levels. The magnitudes of the events do not exceed M4 on these segments and seismicity patterns before and after declustering are similar (Fig. 4). While the Erkenek segment shows a diffuse seismicity pattern with low aftershock activity, the seismicity along the Pazarcık segment occurs within a narrow zone, located at the intersection of a complex fault network including Sürgü-Çardak faults, Dead Sea fault and the Cyprus Arc.

4.3 Cumulative seismicity and seismic interactions along the EAF

Studies of the 1999 M_w 7.4 İzmit and M_w 7.1 Düzce earthquakes show that the stress changes due to the earthquakes increased the seismicity rate of segments that are as far as 200 km away from the mainshock (e.g. Durand et al. 2013). The cumulative seismicity and seismicity rates obtained here for the EAF provide insights on whether similar interactions occur along the EAF after moderate or large earthquakes (Fig. 5). Coseismic and post-seismic responses of two earthquakes on the Palu segment are the largest surge of seismicity observed among all segments within the observation period. Interestingly none of the other segments had any significant changes on the seismicity rates, dynamically or statically due to the 2010 and 2011 earthquakes. We also marked some large regional earthquakes (e.g. 2011 M_w 7.1 Van earthquake about ∼400 km from the EAF) on Fig. 5 after which no detectable change is observed on the seismicity rates.

The cumulative number of earthquakes with M > 2.5 is computed in each cluster and displayed in Fig. 5(a). The traces are interpolated in time in order to have a uniform sampling. We normalized the time series with the segment which has the largest cumulative number of events (Palu). We also calculated the seismicity rates from the number of earthquakes per day with a Gaussian smoothing operator over a time span of 60 d (Fig. 5b).

Fig. 5(c) shows cumulative number of events after declustering. The cumulative number of earthquakes after declustering along the Pütürge segment is the largest and does not significantly change compared to the other segments. The rest of the segments similarly show much lower cumulative number of events with respect to the Pütürge segment (Table 1).

4.4 Cumulative moment release and slip

We calculated the cumulative seismic moment release and slip rate along the EAF in order to make a first order estimation of relative seismic coupling. The seismic slip rate over a fault zone can be assessed from the seismic moment rate deduced from the seismicity catalogue providing position and magnitude of each event. Wdowinski (2009) used this approach for the San Jacinto Fault and concluded that the seismicity rate is correlated with the geodetic rate at the lower section of the seismogenic crust, at depths of 10–17 km, that is near the transition between the locked shallow crust and creeping sections at greater depths. The cumulative seismic slip rate is estimated as: |$v = \frac{1}{{\mu A\Delta t}}\sum\nolimits_{\Delta t} {{M_0}} $|⁠, where µ is the shear modulus, A is the fault area and t is the duration of the catalogue. The seismic moment M₀ for each event is deduced from the magnitude M using Kanamori's law (Kanamori & Anderson 1975): log (M₀) = 1.5M + 9.1.

We applied the same approach to obtain the cumulative seismic slip along the EAF, using a shear modulus of 35 GPa and an average density of 2800 kg m^–3. Fig. 6 shows the deduced cumulative moment release along the fault. The profile of the event rate per year computed every 15 km along the EAF is also plotted (Fig. 6a). A quantitative measure of the slip rate is difficult to estimate in each domain as the observation period is rather short (13 yr) compared to the typical duration of the seismic cycle. Nevertheless, the cumulative seismic slip rate and moment release are larger along the Palu and Pütürge segments (Fig. 6b). Event rate in the Pütürge segment is uniform and high both in space and time during the observation period and the total average slip due to seismicity is ∼1 cm for 13 yr. To the west of the EAF, the cumulative seismic slip reduces gradually similar to the decreasing fault slip rate determined from GPS observations (Table 1).

Figure 6

(a) Total moment release map between 2007 and 2019. Faults are shown by red lines (Emre et al. 2013). (b) Cumulative slip (blue) and event rate (red) along strike EAF, computed with 20 km width along the fault. The average seismic slip rate is estimated from the method proposed by Wdowinski (2009) using a shear modulus 35 GPa and a fault width of 15 km. (c) Cumulative moment release between 2007 and 2019. Dashed red lines mark the segment boundaries. Dashed black line bounds the area where the near-repeaters occur (Konca et al. 2021). The locations of the largest geometric complexities are marked by arrows. LHBR: Lake Hazar Releasing Bend. GRtB: Gökdere Restraining Bend.

4.5 Earthquake source mechanisms, geometry and type of faulting

We obtained a total of 160 focal mechanism solutions with M > 3.5 between 2007 and 2020 (Fig. 7) and combined them with the previously determined solutions, including the GCMT catalogue (Dziewonski et al. 1981; Taymaz et al. 1991; Örgülü et al. 2003; Tan & Taymaz 2006; Ekström et al. 2012). Some of the focal mechanisms solutions, which are also provided by previous studies such as Tan et al. (2011), Cambaz & Mutlu (2016), Gallovic et al. (2020), Pousse-Beltran et al. (2020) and by different agencies, are compatible with the results from this study. The comprehensive focal mechanism catalogue, consisting of 214 earthquakes, is utilized to investigate the kinematics of faulting, geometry and segmentation of the EAF.

Figure 7.

(a) Moment tensor solutions of 160 events with magnitudes M > 3.7 obtained in this study. ID number of each solution is above the beach ball and the source parameters are presented in Table S2. Blue ellipses mark the geometrical complexities in Fig. 2. Black boxes bound the area of each depth cross section in (b). (b) Fault perpendicular cross sections of the seismicity and dip of the fault plane solutions. The dip variations are displayed, from west to east: Box 1: Pazarcık segment. Box 2: Erkenek segment. Box 3–4–5–6: Pütürge segment. Box 7–8: Palu segment. The earthquakes projected to the profiles within ±15 width in related boxes in (a). Box numbers are at the upper left-hand corner of each depth profile. Black lines indicate the dip angles for the selected fault plane parallel to EAF in each box in (a). Solid red lines show the location of the EAF. Dashed red lines show the average dip angle from the earthquake source mechanisms. The average dip angle is at the lower right-hand corner of each cross section panel.

In order to determine the type of faulting along the EAF, we classified the earthquake focal mechanism catalogue based on double-couple source type (Kavarina et al. 1996, Kagan 2005; Álvarez-Gomez 2014,Fig. S12). 25 per cent of the focal mechanisms consists of non-strike slip solutions [normal (22 per cent); reverse (3 per cent)]. 22 per cent of the focal mechanisms displays pure strike-slip motion, while 20 per cent are strike-slip with normal, strike-slip with reverse (17 per cent), reverse with strike-slip (8 per cent) and normal with strike-slip component (8 per cent, Fig. S13).

We identified a large number of left-lateral strike-slip mechanisms along the Palu segment consistent with the orientation of the main fault (Fig. 7a). The segment is characterized by left-lateral strike-slip mechanisms with significant reverse component and reverse mechanisms with left-lateral strike-slip components. Most of the mechanisms with reverse component takes place near the Gökdere restraining bend on the eastern tip of the Palu segment (Fig. 7a-box 8). The western cluster near Lake Hazar consists of mostly pure strike-slip mechanisms (Fig. 7a-box 7).

To the west of the Lake Hazar along the Pütürge segment, the source mechanisms are dominantly left-lateral strike-slip with normal component and normal mechanisms. Near the Lake Hazar releasing bend, there is a significant number of normal and normal with strike-slip mechanisms (Fig. 7a-Box 5–6). The mechanisms and dip angles of the common earthquakes of this study and the recent studies such as Pousse-Beltran et al. (2020) and Gallovic et al. (2020) are compatible. Type of faulting towards the Yarpuzlu restraining bend near Çelikhan are mostly left-lateral strike-slip and some oblique thrust mechanisms are also observed (Fig. 7a-Boxes 3–4).

The Erkenek segment exhibits complex faulting compared to the neighbouring segments, with several reverse mechanisms along with a single pure left-lateral strike-slip mechanism (Fig. 7a-Box 2). The focal mechanisms to the north of the Erkenek segment show left-lateral strike slip motions with reverse component, related to the Sürgü Fault.

A shallow cluster located at the eastern tip of the Pazarcık segment is characterized by pure left-lateral strike-slip motion (Fig. 7a-Box 1-Fig. 7b-cross section 1). Further west of KMTJ, the mechanisms show NE–SW normal faulting.

Although N–S seismicity profiles show no apparent dip of the fault, the earthquake locations are to the north of the surface trace of the fault, indicating a north dipping fault. The earthquake focal mechanisms also show that the EAF dips to the north and the dip angle along the EAF is changing. We estimated an average dip angle for 8 cross section in Fig. 7(a) from the focal mechanisms shown in the black boxes. The number of focal mechanisms included in the dip angle estimation in Boxes 3–4-5–7 are insufficient to infer neither the geometry nor the type of faulting. The cross sections 1-2-6-8 include sufficient numbers of earthquake focal mechanisms solutions, hence more suitable to make inferences on the geometry of the EAF.

Along the Palu segment the average dip calculated from focal mechanisms is ∼65° which is compatible with the dip of the seismicity below ∼7.5 km depth (Fig. 7b-cross section 8). Along the Pütürge segment, from Lake Hazar towards the Yarpuzlu releasing bend, the average dip of the fault is ∼70° (cross sections 4-5-6) and becomes shallower with a dip angle of ∼40° below 7.5 km (cross section 3), as the fault reaches to Çelikhan (Fig. 7b). To the west, along the Erkenek segment, the fault is steeper compared to its eastern part with an average dip angle of 57° below 7.5 km (cross section 2).

4.6 Stress orientations from focal mechanisms

We analyse the spatial variations of the stress orientations along the EAF using earthquake source mechanisms using the method of Vavrychuk (2014). This method solves for the stress orientations from focal mechanisms using an iterative method by differentiating the fault plane and the auxiliary using the fault instability constraint (Lund & Slunga1999) and assuming that the trace off the stress tensor is zero (Michael 1984).

For the stress inversion, we utilize the earthquakes within 20 km distance from the main fault, except for the eastern tip of the Pütürge segment where the earthquakes to the south of the fault might represent secondary faulting or geometrical complexities close the EAF (Şentürk et al. 2019). In order to infer regional changes in the stress field, we divide the EAF zone into 12 subregions and perform the stress inversion for each subregion (Fig. 8, Table S3).

Figure 8.

Stress inversion computed from the source mechanism solutions located in different regions along the EAF and surroundings (Table S3). The lower hemispheric projection of the P and T axes of the earthquake source mechanisms are shown outside of the map and named by region. σ₁, σ₂ and σ₃ are indicated by yellow boxes, circles and triangles, respectively. Maximum and minimum horizontal stress orientations are represented by black and magenta arrows, respectively while white arrows indicate the pure extension. Stress orientations are scaled by plunge angle.

Results indicate that the stress field orientations along the EAF is uniform between Palu and Çelikhan except for the Lake Hazar releasing bend. To the west of Çelikhan, stress orientations change in short distances around the İskenderun Bay. The σ₁ and σ₃ are horizontal along the Palu segment, indicating pure left-lateral strike-slip motion (Fig. 8, R1-Palu). The stress orientations at the eastern tip of the Pütürge segment show NW–SE extension (Fig. 8, R2-Pütürge). In the R3-Pütürge region σ₂ is vertical. The orientations of σ₁ and σ₃ in R4-Erkenek are horizontal and consistent with left-lateral motion. On the west of the Erkenek segment in R5-Pazarcık region σ₂ is steeper compared to its east. Extension is dominant in R6, R7-Toprakkale and R10-Amanos with σ₁ vertical (Fig. 8). R8-Yumurtalık, R9-Cyprus Arc and R11- Narlı are dominated by left-lateral strike-slip type stress orientations (Table S3).

4.7 Seismic potential of the segments of EAF

The reliable estimate of the seismic potential of a fault requires observations of the fault behaviour for multiple seismic cycles. Furthermore, the locations, extents and magnitudes of historical earthquakes are needed to be well-known which is hardly the case for the EAF. Therefore; we attempted to overcome the shortcomings of insufficient observation period and estimate seismic potential (M_max) of the EAF by utilizing the moment build-up rate (Fig. 9a) and a long-term seismicity catalogue following Stevens & Avouac (2021, Fig. 9b).

Figure 9.

The seismic potential of the EAF estimated from geodetic strain rates. (a) Strain magnitudes along the EAF (Weiss et al. 2020). The white contours indicate the moment build-up rates. Grey stars represent the locations of the historical earthquakes. (b) The Gutenberg–Richter relationship computed for each segment of the EAF. The blue and green circles represent short and long term catalogues, respectively. The red solid lines represent the GR fits of the merged catalogue with historical events. Blue and green solid lines represent the GR fits of short- and long-term catalogues, respectively. The a- and b-values are showed at the upper right corners with the corresponding colours of the catalogues. The comparison of the moment build-up rate (MBR) and the moment release rate (MRR) of each segment. (d) The black curve represents 150 yr of moment budget. The grey curve is the 150 yr of cumulative seismic moment release.

The seismogenic thickness is assumed to be 15 km and the geometric factor is determined for each segment using the dip angles in Table 1. The long-term seismicity catalogue shows that the seismicity rate of each segment of the EAF from the short term catalogue is comparable to the ones from the long term catalogue except for the Erkenek segment (Fig. 9b). Hence, the values of M_max and T_r cannot be obtained for the Erkenek segment.

The estimated value of M_max for the Palu and Pütürge segments is similar as ∼6.9 with a return period of ∼150. To the west, the magnitudes of largest recorded earthquake (M > ∼4) along Amanos and Pazarcık segments are much smaller than the magnitudes of the historical earthquakes. However, since the seismicity rates of the short term and the long term catalogues are compatible, we extrapolate the Gutenberg–Richter curve to the higher magnitudes and make a first order estimation of M_max and T_r for these two segments. Results show that an earthquake might occur on the Pazarcık segment with M_max = 7.3 in every ∼772 yr while the Amanos segment can produce an earthquake with M_max = 7.4 in every ∼915 yr.

The computed return periods and maximum magnitudes change significantly along the EAF. The higher recurrence times to the west of the fault zone can be partly due to the estimates of a and b values as the magnitudes of the historical events might be overstated. Another factor might be the value of α which determines the ratio of the seismically released moment. To the west of Çelikhan α is assumed to be 0.5. A value of α lower than 0.5 would lower the estimates of the M_max and T_r while a higher value would result in even longer recurrence times with larger magnitudes. However, the results show that the contrast of M_max and T_r between the eastern and western segments is maintained when the uncertainties in interseismic coupling and Gutenberg–Richter parameters are taken into account (Table 1).

Fig. 9(c) compares the geodetic moment build-up rate and the seismic moment release rate for each segment. It appears that the Palu, Pütürge and Erkenek segments release nearly the same amount of seismic moment as the moment build-up. In contrast, seismic moment release rate of the Pazarcık and Amanos segments are lower than the moment build-up rates.

In order to better understand the variations of M_max and T_r along the EAF, we calculated the moment budget of the five segments for the last 150 yr (accounting for strain and seismicity within 25 km from the fault and assuming a seismogenic thickness of 15 km, Fig. 9d). Result show that the cumulative seismic moment released during 150-yr period is insufficient to balance the moment budget from the strain rate along the Erkenek, Pazarcık and Amanos segments. To the east, the moment budget is relatively balanced on the Palu and Pütürge segments. The results imply that each segment of the EAF produces potential large earthquakes with increasing recurrence interval (T_r) towards west (Table 1).

5 DISCUSSION

5.1 Which factors control the seismicity and seismic slip rates along the EAF?

The seismic slip rates computed from the seismic activity rate between 2007 and 2019 decrease significantly from east to west (Fig. 5b). Geodetic observations also show lower strain accumulation in the western segments compared to the east of the EAF (2–10 mm yr^–1, Mahmoud et al. 2013). Although similar trends of decreasing geodetic and seismic slip rates from east to west are observed along the segments, the variations in seismic slip rates from east to west are much greater and more heterogeneous than the variations of the geodetic slip rates (Table 1).

Weiss et al. (2020) recently presented the strain rate field which is obtained from 5 yr of ascending and descending InSAR tracks together with the GNSS data. Fig. 10(a) compares the principle components of the geodetic strain rate tensor and stress tensors obtained from the earthquake focal mechanisms of this study. The second invariant of strain-rate field shows that the deformation directions and amplitudes change significantly at short length scales along the EAF, indicating the heterogeneity of the strain accumulation along the boundary (Fig. 10a). We observed that the stress directions determined from earthquake source mechanism solutions at the depth range of 5–20 km are consistent with the strain-rates calculated from geodetic measurements on the surface.

Figure 10.

(a) Second invariant of the strain rate tensor derived from a joint inversion of GNSS and InSAR (Weiss et al. 2020) and stress orientations shown in Fig. 8. Maximum and minimum horizontal stress orientations are presented by black and magenta arrows, respectively and scaled by plunge angle. White arrows indicate the pure extensional stress. Black and magenta bars represent the compressional and extensional principal strain directions, respectively. The EAF is shown by black lines (Emre et al. 2013). (b) Dilatation component of the strain rate tensor of Weiss et al. (2020) and the fault mechanism solutions represented by a coloured circle depending on the type of faulting on map and SS: Strike-slip. N–NS: normal, normal with strike-slip component, strike-slip with normal component. R–RS: reverse, reverse with strike-slip component, strike-slip with reverse component.

A good correlation is also observed between dilatation rates of geodetic observations and earthquake focal mechanism solutions. Earthquakes with normal components are concentrated in regions with dilatational strain rate while thrust earthquakes tend to occur in the regions with compressional strain rate, indicating that the seismic behaviour is related to the heterogeneous strain field (Fig. 10b).

We conclude that the heterogeneity of the seismicity distribution is related to the observed heterogeneity of the strain-rate field, which is unveiled by the consistency of the focal mechanisms, the strain rate variations and seismicity rates. In the following sections we discuss several possible factors contributing to the heterogeneity of the strain rate field and seismicity.

5.1.1 Geometrical complexities

The earthquakes on the EAF were previously suggested to occur on a steeply dipping fault planes (Duman & Emre 2013; Bulut et al. 2012). Recent studies of Pousse-Beltran et al. (2020) and Gallovič et al. (2020) suggested that the Pütürge segment dips 80° to the north. The improved seismicity catalogue of this study shows that most of the seismicity is located on the north of the surface trace of the EAF except for the splay faults such as the Adıyaman Fault near Lake Hazar (Şentürk et al. 2019). The possible dip angles from the new focal mechanism solutions indicate that the Palu and Pütürge segments dip to north with 65°–70° while the dip angle of the western segments are shallower with a greater uncertainty due to low number of earthquake source mechanisms.

The depths of earthquakes with the focal mechanism solutions used for the dip estimation are between 5 and 20 km (Fig. 7b). Assuming a constant dip angle along depth, the extrapolation of the fault towards shallower depths would yield an estimated surface trace which is further south of the actual surface trace of the EAF. These observations imply that the dip angle might be steepening at shallower depths (<5 km). The geological studies also observe steep fault scarps at the surface (Duman & Emre 2013). This depth variation of the fault dip might be a contributing factor for the lack of clear coseismic surface offsets in both historical and recent earthquakes along the EAF (Pousse-Beltran et al. 2020).

The geometrical irregularities have a significant effect on the complexity of the seismicity and type of faulting along the EAF. The largest seismic moment release is observed around the geometrical discontinuities of the Gökdere restraining bend to the east of the Palu segment and the Lake Hazar releasing bend to the west of the Pütürge segment (Fig. 6).

Near the Lake Hazar releasing bend, at the eastern tip of the Pütürge segment, σ₁ is almost vertical, in contrast with the neighbouring sections. This observation is consistent with the strain rate field indicating transition from strike-slip to extensional regime towards the west of Lake Hazar (Fig. 10a). This localized extension is possibly the result of transtension due to the Lake Hazar releasing bend and the Uluova Basin to the north of Lake Hazar. Further west, the orientation of σ₂ reverts back from horizontal to vertical indicating pure strike-slip motion in accordance with the EAF. The location of the transition from vertical σ₁ to horizontal σ₁ to the NE of the Pütürge segment is consistent with rupture termination of the 2020, M_w 6.8 Sivrice earthquake (Pausse-Beltran et al. 2020) near Lake Hazar and may have also played a role on the rupture extents of the historical events (Köküm & Özçelik 2020).

A large number of aftershocks of 2020, M_w 6.8 Sivrice earthquake, between Lake Hazar and Sivrice show left-lateral focal mechanism solutions with significant normal component (Fig. 7, Box 6). While the strike-slip with normal component mechanisms are observed during the 13 yr of the observation period, the seismic activity with normal mechanisms has increased following the 2020 M_w 6.8 rupture, near Lake Hazar on the western boundary of the Uluova Basin (Fig. 2). Hence, considering that the number of the normal mechanism solutions is low before the 2020 event, the increased seismic activity in this region with normal faulting is possibly a result of the favourable oriented faults activated by the 2020 earthquake.

The oblique left-lateral-reverse mechanism earthquakes along the Erkenek segment is in agreement with the compressional strain rate observed in the region (Fig. 10b). The surface expressions of the fault display dominant reverse faulting due to the Yarpuzlu restraining double bend where the Sürgü Fault, the Bitlis Fold and Thrust Belt meet with the EAF (Duman & Emre 2013). The strain rate field changes from compression to left-lateral strike-slip motion towards the west of the Erkenek segment. This is coherent with the focal mechanisms and the field observations which show pure left-lateral offset of 13 km (Şaroğlu et al. 1992).

Several studies suggest that the Sürgü Fault, which is the eastern section of the northern stand of the EAF, has left-lateral motion (e.g. Arpat & Şaroğlu 1975; Perinçek & Kozlu 1984; Duman et al. 2013, 2020), while others claim that the motion is right lateral (Koç & Kaymakçı 2013; Gülerce et al. 2017). The new focal mechanism solutions along the Sürgü Fault shows left-lateral strike slip motion with some reverse component. This is consistent with the mechanisms of two M > 5 earthquakes occurred in 1986 (Taymaz et al. 1991). The strain rate field also indicates left-lateral strike-slip motion with increasing compressional component towards the junction of the EAF and the Sürgü Fault. From these recent observations, we favour that the Sürgü Fault is a left-lateral strike-slip fault which is under compression, especially where it connects to the EAF.

To the west of KMTJ, the pure extensional regions are possibly related to the opening of the Adana-Cilicia Basin and the Karasu Basin (Fig. 10b). South of the İskenderun Bay is dominated by extensional stress regime with vertical σ₁, displaying EW extension which is consistent with the releasing bend in the Karasu Basin (Duman et al. 2013).

5.1.2 Creep and lack of fault interactions along the EAF

The high seismicity rate with diffuse distribution is commonly associated with weak interseismic coupling which might be accompanied by aseismic creep (e.g. Schmittbuhl et al. 2015; Métois et al. 2016). Recently, Blettery et al. (2020) suggested that the interseismic coupling along the Pütürge segment is heterogeneous, demonstrating a partially coupled seismogenic zone. In addition, near-repeating earthquakes are observed within ∼10 km of the fault zone along the same section of the fault (Konca et al. 2021). One way to explain relatively uniform high rate of seismicity (Fig. 6) is that these earthquakes occur due to small asperities in this partially or fully creeping segment. If there is continuous creep along this segment, the small asperities might generate a uniform spatiotemporal distribution rather than clustered seismicity as one would expect from the typical main shock aftershock sequences. Blettery et al. (2020) shows that the coupling is lower along the eastern sections of the segment towards Lake Hazar, while coupling is higher further west, which is consistent with the coseismic slip distribution of the M_w 6.8 2020 Elazığ Earthquake (Konca et al. 2021). A plausible explanation is that while the background seismicity reflects the partially coupled zones in creeping regions, near repeaters take place in the transition zone from low the high interseismic coupling.

Similarly, the 2010 M_w 6.0 and 2011 M_w 5.4 earthquakes occurred along the Palu segment, 30 km apart with ∼1 yr delay. These two events are responsible for most of the released seismic moment in this segment. We infer that the stress along the 30 km separation between these two earthquakes was not high enough for the whole segment to rupture during a single event. Bletery et al. (2020) model of the interseismic coupling shows weaker coupling in between these two rupture zones, which might explain the lack of interaction between the two zones. Furthermore, the background seismicity displays a homogeneous pattern which indicates that the seismic localizations at the tips of this segment are related to the aftershock activities.

The moderate size earthquakes on the EAF and large regional earthquakes did not significantly alter seismicity rates on the adjacent segments during the observation period. For instance, it would be expected that following the main shocks in 2010 and 2011 on the Palu segment, the seismicity on the neighbouring Pütürge segment be triggered due to static or dynamic stress changes (King et al. 1994; Stein et al. 1997; Luo & Liu 2010; Işık et al. 2017). However, we observe no triggered activity or interactions neither in coseismic nor in post-seismic periods. Similarly, 2020 M_w 6.8 earthquake ruptured 45 km section of the Pütürge segment and its aftershocks are almost confined to the Pütürge segment where the rupture occurred (Konca et al. 2021).

Unlike the EAF, seismicity rates along the NAF change following large earthquakes even at far distances (Durand et al. 2010). This observation may indicate that the segmentation along the EAF is much stronger with the associated fault complexity where its geometry changes both laterally and vertically at short length scales. Moreover, recent studies of Pausse-Beltran et al. (2020) and Gallovič et al. (2020) argued that the fault's low maturity has a significant effect on the quiescence of the EAF before the 2020, M_w 6.8 earthquake and its rupture extent. The lack of fault interactions along the EAF is possibly result from a combination of high segmentation, heterogeneous coupling and low maturity to transfer the stress along strike.

5.1.3 The role of the EAF along the Anatolia-Arabia boundary

Whether the Eastern Anatolia behaves like a rigid block or not is still debated (e.g. Cavalié & Jonsson 2014; Walters et al. 2014). While the plate motion on the EAF is localized to the NE of Çelikhan (Fig. 1; Cavalié & Jonsson 2014), from Çelikhan towards the SW, the motion on the EAF is distributed within a relatively wider zone along the Pazarcık and Amanos segments (Walters et al. 2014). To better understand the role of the EAF as the plate boundary fault between Anatolia and Arabia, the relative plate motion along the EAF is calculated with respect to the Euler Pole from Blettery et al. (2020) using the GPS velocities (Reilinger et al. 2006) and InSAR data. In Fig. S14, we compare the earthquake slip vectors with the computed relative motion assuming that the EAF represents the plate boundary. We selected the fault planes of the source mechanisms consistent with the mapped fault orientations. This comparison shows that along the Palu and Pütürge segments, the slip vectors are parallel to the trend of the main fault and coherent with the plate motion except at the Hazar releasing bend. This result is consistent with the geodetic study of Cavalié & Jonsson (2014) which shows that the EAF forms a clear plate boundary in this region.

However, to the west of Çelikhan, the plate motion becomes more oblique with some convergence (Fig. S14). In this region, the earthquake slip vectors are also complex, with varying directions of slip vectors and increased number of earthquakes with thrust components (Fig. 10). The earthquakes with thrust component both on the eastern Erkenek segment and the subparallel Sürgü Fault might be related to this change in the plate motion direction.

Further west, near KMTJ, the DSF, the transpressional eastern Cyprian arc and the EAF forms a diffuse zone of triple junction. Hence, the effects of this diffuse junction arc might be coming into play, leading to a complicated fault behaviour in the region (Fig. 7).

5.2 Changing recurrence periods and seismic potentials along the EAF

The EAF has been subjected to several large earthquakes both in historical and instrumental period and is still prone to large earthquakes, which raise the importance of seismic hazard assessment for the region. The shear strain is localized mainly between Palu and Çelikhan where the fault slip rate is relatively higher (∼10 mm yr^–1) and most of the seismic moment accumulated during the 13 yr of observation period is released along the same section (Fig. S15a). Similarly, the cumulative seismic moment release from the long term earthquake catalogue including the historical events reveals that the Palu and Pütürge segments take up the most of the released moment along the EAF (Figs 9 and S15b). Between Çelikhan and Karlıova, the moment budget obtained from the interseismic strain rate is balanced by the moment release where the 2020, M_w 6.8 Sivrice earthquake, 2010, M_w 6.0 and 2011, M_w 5.4 Palu earthquakes occurred. The time between the historical earthquakes and recorded earthquakes are consistent with the estimated T_r values for these segments which are ∼150 yr. To the west of Çelikhan, there is a significant moment deficit which build-up in the interseismic period despite lower fault slip rates. The low rate of seismicity result in longer and recurrence intervals (T_r) and higher estimates of M_max for the Pazarcık and Amanos segments.

5.3 The 2020 M_w 6.8 Sivrice earthquake and the active seismotectonics of the EAF

The 2020 M_w 6.8 Sivrice Earthquake is the largest event occurred on the EAF since the 19th century (Gallovic et al. 2020; Pousse-Beltran et al. 2020; Konca et al. 2021). The mainshock ruptured 45 km of the 95-km-long Pütürge segment with a rupture velocity of ∼2.5 km without a surface slip (Konca et al. 2021). The recent study of Konca et al. (2021) revealed that the 2020 M_w 6.8 Sivrice Earthquake was preceded by a M ∼ 5.4 foreshock which is close to the near-repeating earthquakes.

The characterization of the 2020 M_w 6.8 Sivrice earthquake is consistent with the implications of the active seismotectonics of the EAF. Similar to the 2010 and 2011 Palu events, although relatively larger, the 2020 Sivrice event did not rupture the whole Puturge segment and did not trigger significant seismic activity on the neighbouring segments (Konca et al. 2021). Recent studies revealed that the dip of the Pütürge segment gets shallower towards SE near Çelikhan (Pousse-Beltran et al. 2020; Konca et al. 2021), which is compatible with the shallow dip angle obtained from focal mechanisms in this study (Fig. 7). Together with the dip angle, the lack of surface slip, slow and complex rupture, the 2020 Sivrice event is a manifestation of the characteristics of the active seismotectonics of the EAF, within a transition from weak to high coupling along a creeping segment with near-repeating earthquakes.

6 CONCLUSION

The improved earthquake catalogue between 2007 and 2019 and 160 new focal mechanisms with M > 3.5 elucidate the present state of the seismotectonic behaviour of the EAF. The analysis of these observations indicate that the EAF dips to north ∼65°–70° and steepens at shallow depths (<5 km) between Palu and Çelikhan where the strain is localized within a narrow zone of ∼15 km. The rapidly changing geodetic strain accumulation rate and segmentation with several geometrical complexities result in heterogeneous seismic behaviour. The diversity of the focal mechanisms solutions is consistent with the changes in stress orientations and strain rate field.

The space-time distribution of the seismicity shows no detectable interactions between the segments during the observation period which is the consequence of highly segmented and heterogeneous fault zone. It is likely that this heterogeneity is due to the combination of several factors such as oblique plate motion, variations in seismic coupling and geometric complexities. Furthermore, the unique seismic characterization with the homogeneously distributed high seismic activity of the Pütürge segment, holding the M_w 6.8, 2020 event, is consistent with the previous findings of creeping zones with near-repeating earthquakes (Konca et al. 2021).

The variations in the geodetic and seismic moment release rates along with the reducing strain rates towards the west of Çelikhan result in changing magnitudes of potential large earthquakes and their recurrence periods. The results show the recurrence time of an M_max ∼ 6.7–7.0 earthquake on the Palu and Pütürge segments is ∼150 yr, while the western segments yield longer recurrence times; 237–772 for Pazarcık and 414–917 for Amanos segments with larger magnitudes (M_max ∼ 7.0–7.4). Although smaller in size, the EAF displays complex behaviour like other examples such as the Alpine Fault in New Zealand (e.g. Leitner et al. 2001; Smith et al. 2017), the San Andreas Fault (e.g. Hardebeck & Hauksson 2001; Templeton et al. 2008) and the North Anatolian Fault (Barka & Kadinsky 1988; Schmittbuhl et al. 2015).

SUPPORTING INFORMATION

Figure S1. Seismic stations used to locate the earthquakes and determine the source mechanisms. Inset figure: Red triangles show the stations installed after 2020, M_w 6.8 Sivrice earthquake. Yellow star shows the epicenter of the 2020, M_w 6.8 Sivrice earthquake mainshock. The fault traces are from Emre et al. (2013).

Figure S2. 700 events are selected to compute 1-D velocity model and station corrections using VELEST inversion code (Kissling 1994). The selected events have azimuthal gap <80° and are relocated using at least 25 phase readings.

Figure S3. Distribution of average root mean square (RMS) error of the P and S traveltime residuals of 26 000 events compiled in this study between 2007 and 2019.

Figure S4. The comparison of the 1-D velocity model obtained in this study and the 1-D velocity models from Pousse-Beltran et al. (2020). Red thick lines represent the P and S velocities obtained in this study.

Figure S5. Average Vp/Vs and crustal velocities of the seismogenic crust. (a) Red and blue dots show P and S wave first arrival traveltimes, respectively. S wave traveltimes are corrected with Vp/Vs = 1.74 which indicates the average value for the crust. The reducing velocity of 7.8 km s^–1 is used for the Pn velocity of the velocity model. (b) Same as (a) with the reducing velocity of 6.0 km s^–1 is used to show the average crustal velocity.

Figure S6. (d) The best-fitting focal mechanism solutions from the regional waveform inversion of selected earthquakes for 4 different regions. The event information is given on top of the plot. Left-hand panel: solution using 1-D crustal model of this study. Right-hand panel: solution using 1-D crustal model of Pousse-Beltran et al. (2020) for the four segments of the EAF. Velocity waveform data (black) and model fits (red) for the vertical and radial P waves are shown in first two columns. Vertical, radial and tangential S and surface wave velocity data (black) and model predictions (red) are shown in columns 3, 4 and 5, respectively. Station name, distance (km) and azimuth (degree) are shown on the left of the traces. Bandpass filters are applied between 0.04 and 0.1 Hz for the 35 s Pnl window and 0.04–0.08 Hz for the 70 s S-surface wave window. Grid search over strike, dip and rake angles with 5° intervals and moment magnitude with a step size of 0.1 were used to determine best-fitting focal mechanism. Maximum time shifts were chosen as 2 and 5 s for the Pnl and Surface wave components, respectively. Waveform fits with a correlation less than 50 were manually eliminated.

Figure S7. Statistics for the location accuracy of the seismicity catalogue between 01.01.2007 and 31.12.2019. (a) Horizontal uncertainties in NS direction, (b) horizontal uncertainties in EW direction, (c) depth uncertainty and (d) magnitude–frequency relation of the catalogue and the b-value.

Figure S8. Results of the best-fitting focal mechanism solutions from the regional waveform inversion of two earthquakes the study region. Velocity waveform data (black) and model fits (red) for the vertical and radial P waves are shown in first two columns. Vertical, radial and tangential S and surface wave velocity data (black) and model predictions (red) are shown in columns 3, 4 and 5, respectively. Station name, distance (km) and azimuth (°) are shown on the left of the traces. Bandpass filters are applied between 0.04 Hz and 0.1 Hz for the 35 s P_nl window and 0.04–0.08 Hz for the 70 s S-surface wave window. Grid search over strike, dip and rake angles with 5° intervals and moment magnitude with a step size of 0.1 were used to determine best-fitting focal mechanism. Maximum time shifts were chosen as 2 and 5 s for the P_nl and Surface wave components, respectively. Waveform fits with a correlation less than 50 were manually eliminated. Eliminated traces are shaded in grey. (a) Moment tensor solution and waveform fits of 02.03.2017, M 5.3 event for 5 km. (b) Depth grid search for the event in (a). The event information is given on top of the plot. (c) Moment tensor solution and waveform fits of 12.02.2020, M 3.65 event for 8 km. (d) Depth grid search for the event in (c). The event information is given on top of the plot.

Figure S9. (a) Histogram of number of stations used in moment tensor inversion. (b) Histogram of variance reduction of the moment tensor inversions.

Figure S10. Comparison of the focal mechanism solutions obtained in this study with the GCMT solutions between 2007 and 2020. Blue and red beach balls indicate focal mechanisms solutions from GCMT and this study, respectively.

Figure S11. Magnitude–frequency analysis and b-values of different segments of the East Anatolian Fault. (a) Palu Segment, (b) Pütürge Segment, (c) Erkenek Segment, (d) Pazarcık Segment and (e) Amanos Segment.

Figure S12. Diversity of the fault mechanism solutions along the East Anatolian Fault. Each focal mechanism is represented by a coloured square depending on the type of faulting on (a) map view and (b) depth sections. SS: Strike-slip. NS: Normal with strike-slip component. N: Normal. RS: Reverse with strike-slip component. R: Reverse. SN: Strike-slip with normal component. SR: Strike-slip with reverse component. Earthquake locations are represented by grey circles and scaled by magnitude. Faults are shown by black lines (Emre et al. 2013). Dashed red lines indicate segment boundaries.

Figure S13. Classification of the focal mechanism catalogue (Alvarez-Gomez 2014). Triangular representation based on the fault slip. Circles represent the source mechanisms scaled by magnitude and coloured by longitude. (b) Percentages of different type of faulting obtained from the focal mechanism catalogue.

Figure S14. Slip vectors and relative plate motion on the EAF. White slip vectors are calculated for the earthquakes which have fault planes parallel to the mapped faults. Relative plate motion (blue vectors) on the EAF is calculated for the Euler Pole from Bletery et al. 2020 using GPS velocity field of Reilinger et al. (2006). KMTJ: Kahramanmaraş Triple Junction. LH: Lake Hazar. BZFTB: Bitlis-Zagros Fold and Thrust Belt. SF: Sürgü Fault. AF: Amanos Fault. YsF: Yesemek Fault. YF: Yumurtalık Fault. TF: Toprakkale Fault. NFZ: Narlı Fault Zone.

Figure S15. Upper panel: comparison of strain amplitudes and cumulative moment release distribution. Background colours represent the second invariant of the strain rate tensor derived from a joint inversion of GNSS and InSAR (Weiss et al. 2020). Black contours show the logarithm of the cumulative moment release isolines. Lower panel: the cumulative seismic moment release from the long-term seismicity catalogue between 1900 and 2020 including the largest ever reported earthquakes.

Table S1. The 1-D velocity model obtained in this study with the station distribution in Fig. S1 and event distribution in Fig. S2 using VELEST inversion code (Kissling 1994).

Table S2. Focal mechanism solutions obtained in this study. The earthquakes with available focal mechanism solutions from GMCT are shown in bold letters. (Comparison is in Fig. S7).

Table S3. Stress inversion results from focal mechanisms. A and P refer to azimuth and plunge angles for principle stress directions (σ₁, σ₂, σ₃), respectively.

Please note: Oxford University Press is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the paper.

ACKNOWLEDGEMENTS

This work was supported by TÜBİTAK Project numbers 114Y250 and 118Y435. Local and regional seismic data are downloaded from Boğaziçi University Kandilli Observatory and Earthquake Research Institute (https://doi.org/10.7914/SN/KO) and AFAD (https://tdvms.afad.gov.tr/) web sites. The relocated seismicity and aftershock location catalogue is provided in the Zenodo repository https://doi.org/10.5281/zenodo.5220633. We would like to thank to E.E. Papadimitriou and the anonymous reviewer for their valuable comments that helped to improve this paper. We also acknowledge the editor Huajiane Yao for detailed comments.

DATA AVAILABILITY

The data that support the findings of this study are available in http://www.koeri.boun.edu.tr/sismo/2/en/and https://tdvms.afad.gov.tr/.