Keywords: Bodrum; Kos; Earthquake; Normal Fault; Inversion; Extension Palabras clave: Bodrum; Kos; Terremoto; Falla normal; Inversión; Extensión; How to cite item Over, S., Ozden, S., Kalkan, E. E., Turhan, F., Coskun, Z., & Pinar, A. The 20 July 2017 Bodrum-Kos Earthquake (Mw 6.6) in southwestern Anatolia, Turkey. Earth Sciences Research Journal, 25(3), 309-321. DOI: https:// doi.org/10.15446/esrj.v25n3.87080 A 6.6 (Mw) earthquake struck the western part of Gökova Gulf in the eastern Aegean Sea on July 20, 2017. The fault plane solution for the mainshock shows an E-W striking normal fault with approximately N-S (N4°E) tensional axis (T-axis). Fault plane solutions for 33 aftershocks offer two groups of normal faulting with E-W and NE-SW to ENE- WSW orientations. The inversion of the focal mechanisms of the aftershocks yields two different extensional stress regimes. The stress regime obtained from 12 focal mechanisms of aftershocks and the mainshock is characterized by an approximately N-S (N5°E) σ3 axis, while the other regime calculated from 21 focal mechanisms of aftershocks exhibits σ3 axis in an NW-SE (N330°E) direction. The latter extension significantly affects the basin’s growth in the area where the earthquake occurred. Twenty-four focal mechanisms of earthquakes in and around Gökova Basin before the 2017 earthquake (1933-2017) were included in the inversion to determine the stress state effective in a larger area. The inversion yielded an extensional stress regime characterized by approximately N-S (N356°E) σ3 axis. E-W trending faults inferred in the central part of Gökova Fault Zone, bordering Gökova Gulf in the north, also indicate N-S extension. The NW-SE extension obtained from NE-SW aftershocks appears to be locally effective in the northwest of Gökova Gulf. N-S extension, which appears to act on a regional scale, may be attributed to geodynamic effects related to the roll-back of the African subduction beneath the Aegean. ABSTRACT The 20 July 2017 Bodrum-Kos Earthquake (Mw 6.6) in southwestern Anatolia, Turkey Terremoto de magnitud 6.6 en Bodrum-Kos, el 20 de julio de 2017, en el suroeste de Anatolia, Turquía ISSN 1794-6190 e-ISSN 2339-3459 https://doi.org/10.15446/esrj.v25n3.87080 Un terremoto de magnitud 6.6 golpeó la parte oeste del golfo de Gökova en el este del mar Egeo el 20 de julio de 2017. El mecanismo focal del evento principal muestra un movimiento E-O en la falla normal con un eje tensional (T-axis, en inglés) en dirección N-S (N4°E). El mecanismo focal de 33 réplicas muestra dos grupos de fallamiento normal con orientaciones E-O y NE-SO hacia ENE-OSO. La inversión del mecanismo focal de las réplicas produjo dos regímenes de fuerza extensiva diferentes. El régimen de fuerza obtenido a través de 12 mecanismos focales de réplicas y del movimiento principal se caracteriza por un desplazamiento aproximado N-S (N5°E) σ3 del eje, mientras que el otro régimen se calculó a partir de 21 mecanismos focales en réplicas y muestra un desplazamiento del eje en dirección NO-SE (N330°E). Esta última extensión afecta significativamente el crecimiento de la cuenca en el área donde ocurrió el terremoto. Veinticuatro mecanismos focales de terremotos en la zona de la cuenca Gökova y que ocurrieron entre 1933 y 2017 fueron incuidos en la inversión para determinar el estado de la fuerza efectiva en una área mayor. La inversión dio como resultado un régimen de fuerza extensiva caracterizada por un desplazamiento del eje aproximado N-S (N356°E) σ3. De las fallas con tendencia E-O se infiere que en la parte central de la Zona de Fallas de Gökova, que limita al norte con el golfo de Gökova, también tiene una extensión N-S. La extensión NO-SE obtenida de las replicas NE-SO aparece como localmente efectiva en el noroeste del golfo de Gökova. La extensión N-S, que según los resultados actúa a escala regional, puede ser atribuido a los efectos geodinámicos relacionados con el retroceso de la subducción africana bajo el Egeo. RESUMEN Record Manuscript received: 07/05/2020 Accepted for publication: 16/07/2021 EARTH SCIENCES RESEARCH JOURNAL Earth Sci. Res. J. Vol. 25, No. 3 (Septiembre, 2021): 309-321 Semir Över1, Süha Özden2, Esra Kalkan Ertan3, Fatih Turhan3, Zeynep Coşkun3, Ali Pınar3 1Iskenderun Technical University, Department of Civil Engineering, 31200, Iskenderun, Turkey. 2Çanakkale Onsekiz Mart University, Department of Geological Engineering, 17100, Çanakkale, Turkey. 3Boğaziçi University, Kandilli Observatory, and Earthquake Research Institute, 36684, İstanbul, Turkey. Corresponding Author: semir.over@iste.edu.tr SE IS M O LO G Y https://doi.org/10.15446/esrj.v25n3.87080 https://doi.org/10.15446/esrj.v25n3.87080 https://doi.org/10.15446/esrj.v25n3.87080 mailto:semir.over@iste.edu.tr 310 Semir Över1, Süha Özden, Esra Kalkan Ertan, Fatih Turhan, Zeynep Coşkun, Ali Pınar Introduction Plate boundary forces caused by relative movements between Africa, Arabia, and Eurasia resulted in complex tectonic structures in the western Anatolia-Aegean region (Fig. 1). Plate movements involving the convergence of Arabia and Africa with Eurasia in the north; 1) give rise to the westward extrusion of Anatolia along the strike-slip North Anatolian Fault (NAF) and East Anatolian Fault (EAF), 2) cause crustal extension in western Anatolia-Aegean, and 3) are responsible for complex deformation in the Eastern Mediterranean region (McKenzie, 1972; Dewey and Şengör, 1979; Le Pichon and Angelier, 1979; Angelier et al., 1981; Jackson and McKenzie, 1984, 1988; Taymaz et al., 1990; 1991; Över et al., 2010; Özden et al., 2018). The complex tectonics of western Anatolia-Aegean involve different extensional regimes: N-S, NE-SW, and NW-SE (McKenzie, 1972; Jolivet and Brun, 2010; Alçiçek et al., 2006; Shah, 2015; Över et al., 2016; Ocakoğlu et al., 2018). Gökova Gulf, which runs east-west, is an asymmetrical basin that is growing towards the sea. The Gökova Graben formed on the Lycian Nappes and filled with Plio-Quaternary units known as the Gökova Formation (Görür et al., 1995; Gürer and Yılmaz, 2002; Gürer et al., 2013; Tur et al., 2015). Gürer et al. (2013) suggested that the northern edge of the gulf is bounded by the south- dipping E-W Gökova Fault, one of the most active structures in the southern part of west Anatolia-Aegean, based on geomorphological and geological features. The southern border of the basin is not as clear and uninterrupted as the northern border. The active fault map also shows that the Datça Fault Zone, which represents the western part of the southern boundary of the basin, is more active (Fig. 2). The Datça Fault Zone is divided into two parts: north and south. The northern segment extends to the east of Kos (Karasözen et al., 2018), while the southern segment is listric (Kurt et al., 1999). Analyzing both bathymetric and seismic reflection data, İşcan et al. (2013) revealed the existence of active strike-slip faults with various directions in the Gulf of Gökova. From analysis of the focal mechanisms, Shah (2015) asserted that the area is dominated by normal events with few strike-slip events at the western margin of Gökova Gulf. The interpretation of new multichannel seismic profiles by Ocakoğlu et al. (2018) also shows the existence of active strike-slip and normal faults in different directions in the Gulf of Gökova. They pointed out southern-dipping Figure 1. Map of simplified tectonic framework of the Aegean-Mediterranean Sea region (modified from Barka, 1992; Över et al., 2010). Figure 2. Location map and active faults in Gökova Gulf and surroundings (Fault data are taken from Duman, et al., 2011; Emre, et al., 2013; İşcan et al., 2013; Karasözen 311The 20 July 2017 Bodrum-Kos Earthquake (Mw 6.6) in southwestern Anatolia, Turkey normal faults on the north border (i.e. Gökova Fault Zone) and northern- dipping normal faults on the southern border of the basin (i.e. Datça Fault Zone). Normal faults were also implied in the Gökova Basin by several studies (e.g., Kurt et al., 1999; Uluğ et al., 2005; İşcan et al., 2013; Tur et al., 2015; Ocakoğlu et al., 2018). Following the earthquake on July 20, 2017, researchers obtained contrary findings about the source fault and the fault’s dip direction: (1) from geophysical and geodetic data the Bodrum-Kos earthquake ruptured a south-dipping E-W normal fault (Saltogianni et al., 2017; Tiryakioğlu et al., 2018; Ocakoğlu et al., 2018) or (2) a north-dipping E-W normal fault (Karasözen et al., 2018; Ganas et al., 2019; Konca et al., 2019). Karasözen et al. (2018) asserted that the source fault was the northern Datça Fault. A section of Gökova Bay extending from the Bodrum Peninsula to the Datça Peninsula shows the presence of E-W striking faults dipping both south and north (Fig. 3). In the last 30 years, GPS studies about the tectonics of the Eastern Mediterranean have yielded important results, including: 1) the velocity and direction of movement of Africa and Arabia relative to Eurasia (McClusky et al., 2000; Reilinger et al., 2006), 2) the rate and direction of movement of Anatolia towards the west relative to Eurasia along the dextral NAF (North Anatolian Fault) and sinistral EAF (East Anatolian Fault) (McClusky et al., 2000; Reilinger et al., 2006), 3) rate and direction of motion in west Anatolia- Aegean toward the Hellenic Trench (McClusky et al., 2000; Aktuğ et al., 2009; Howell et al., 2017), and 4) the crustal extension in the west Anatolia- Aegean (LePichon et al., 1995; McClusky et al., 2000; Reilinger et al., 2010). This area is currently undergoing a N-S extensional process at a rate of 30-40 mm/yr, which is well documented (Oral et al., 1995; LePichon et al., 1995), 6) the determination of both earthquake parameters and the source fault geometry (Saltogianni et al., 2017; Tiryakioğlu et al., 2018; Karasözen et al., 2018), and finally 7) revealed the coseismic deformation of the 20th July Bodrum-Kos earthquake (Tiryakioğlu et al., 2018). The GPS velocity vectors calculated relative to Anatolia (fixed) indicate an increase in the velocity of the southwestward extrusion of Anatolia in western Turkey and the Aegean region compared to other areas of Turkey accompanied by anticlockwise rotation (McClusky et al., 2000; Howell et al., 2017). To define the stress regimes and their tectonic implications, we used a method proposed by Carey-Gahillardis and Mercier (1987). We introduced the focal mechanisms of both the main shock and its aftershocks into the inversion algorithm. We also compiled the published focal mechanisms of earthquakes that occurred in Gökova and its environs between 1933 and 2017 in order to et al., 2018). Figure 3. A cross-section of Gökova Gulf from Bodrum Peninsula to Datça (A-B section line shown in Fig 2). Figure 4. Map of the distribution of earthquake (20 km depth) epicenters between 20.07.2017 and 20.02.2018 in Gökova Gulf. The earthquakes are compiled from KOERI. 312 Semir Över1, Süha Özden, Esra Kalkan Ertan, Fatih Turhan, Zeynep Coşkun, Ali Pınar determine the tectonic regime responsible for the great Bodrum-Kos Earthquake in terms of regional significance. Seismicity The historical catalogues show that Gökova Gulf and its surrounding settlements have been hit many times by a number of moderate to major destructive earthquakes, such as in 24 BC, 412 BC (Ambraseys and White, 1997), 227 BC, 199-198 BC, AD 141, 144, 174, 344, 1493, 1851, 1863, and 1869 (Ergin et al., 1967; Guidoboni et al., 1994; Yolsal et al., 2007). Luttrell (1999) reported that the city of Bodrum was totally demolished by the 1493 earthquake. The activity in the region is also confirmed by earthquakes occurring during the instrumental period, such as 23 April 1933 (Mw: 6.2), 23 May 1941 (Mw: 5.4), 13 December 1941 (Mw: 6.3), 25 April 1959 (Mw: 6.1), 19 February 1989 (Mw: 5.6), 5 October 1999 (Mw: 5.2), 4 August 2004 (Mw: 5.4), and 10 January 2005 (Mw: 5.2) (Eyidoğan and Barka, 1996; Şaroğlu et al., 1992; Uluğ et al., 2005; Yolsal and Taymaz, 2010). Kalafat and Horasan (2012) observed swarm type activity in the western Gökova Gulf during 2004 and 2005. Historical and instrumental era earthquakes show that the basin contains significant faults that enable its development. On July 20, 2017, one of the faults triggered a 6.6 (Mw) earthquake. Following the Bodrum-Kos Earthquake, more than 3000 aftershocks ranging in magnitude from 1.0 to 4.8 were recorded across Gökova Gulf according to KOERI (Kandilli Observatory and Earthquake Research Institute) data (Fig. 4). Aftershocks were scattered across a wide region, mainly in the west of the basin, and did not follow a linear distribution (Fig. 4). This event created a tsunami with a water wave reaching almost 2 meters height, which damaged the southern coast of Bodrum and the northern coast of Kos Island (Heiderzadeh et al., 2017; Yalçıner et al., 2017). Several studies reported tsunamis in relation to historical and instrumental period events (i.e. in 365, 554, 1303, 1481, 1822 and 1948) in the Eastern Mediterranean (Altınok and Ersoy, 2000; Yolsal et al., 2007; Çevikbilen and Taymaz, 2012, Çevikbilen et al., 2014). Methodology Moment Tensor Inversion In this study we used Moment Tensor (MT) inversion algorithms that model the waveforms recorded at one or more 3-component broadband seismic stations (e.g., Kuge, 2003; Sokos and Zahradnik, 2008). These techniques perform waveform inversions to find source parameters of small to moderate sized earthquakes. One of the most user-friendly MT inversion routines recently adopted by Seiscmop3 is ISOLA, which is based on FORTRAN codes and offers a MATLAB graphic interface. ISOLA allows for both single and multiple point source iterative deconvolution (Kikuchi and Kanamori, 1991) and inversion of complete regional and local waveforms. In addition, we also used the technique developed by Kuge (2003). In this method, waveform fitting between the observed and synthetic displacement seismograms from one or more stations at local distances is achieved by searching for a MT point on a grid scheme for the best fit between the observed and the synthetic displacement seismograms. During the inversion process uniform weight is given to all seismograms. The broadband seismic station networks operated by KOERI, Disaster and Emergency Management Presidency (AFAD) and National Observatory of Athens (NOA) were utilized. For data selection to be used for the inversion process, firstly the signal quality check was completed for three-component broadband stations. Seismograms with gaps and signals with signal-to-noise ratio lower than 4.0 were not considered. Good azimuthal distribution of stations plays significant role in constraining the MT inversion results. The Green’s function (GF) calculation is performed using the frequency- wavenumber method (Bouchon, 1981). GFs were computed using the crustal structure from Akyol et al. (2006). When GFs are calculated, the positions of epicenters are fixed, and the depth is allowed to deviate from close to the surface to a depth of up to 30 km in steps of 2 km. The quality of fit between the observed and predicted seismograms is measured by variance reduction (VR); the larger the value of VR, the better the fit. The variance reduction is calculated for various depths for each time shift, and the faulting mechanism was selected with maximum VR for the analyzed event (Fig. 5). Figure 5. CMT solution result for the Mw 6.6 Bodrum-Kos earthquake (20 July 2017). 313The 20 July 2017 Bodrum-Kos Earthquake (Mw 6.6) in southwestern Anatolia, Turkey Figure 6. The correlation vs source number plot Figure 7. Observed versus synthetic waveform fits. Inversion of seismic slip-vector data sets to determine the present-day stress state To compute the state of stress responsible for present-day faulting from the population of focal mechanisms of earthquakes that occurred in the Gökova Gulf, we used an inversion method proposed by Carey-Gailhardis and Mercier (1987); this is one of several existing algorithms (Vasseur et al., 1983; Gephart and Forsyth, 1984). The inversion is a statistical method which allows calculation of the best mean fitting stress state from a population of focal mechanisms by selecting one of two nodal planes as the fault plane. According to this inversion process, the slip (s, defined by a slip vector corresponding to a striation for geological data or a rake for seismological focal mechanisms) on each fault plane occurs in the direction of the resolved shear stress (t), the fault plane being a pre-existing fracture. In the case of a seismic event, the inversion computes a mean best-fitting deviatoric stress tensor by minimizing the angular deviation between an expected slip vector (maximum shear, t) and the observed slip vector (s) deduced from the focal mechanism (Carey and Brunier 1974; Carey 1979). All inversion results include the orientation (azimuth and plunge) of the principal stress axes of a mean deviatoric 314 Semir Över1, Süha Özden, Esra Kalkan Ertan, Fatih Turhan, Zeynep Coşkun, Ali Pınar stress tensor, as well as a ‘stress ratio’ [R = (σ2 − σ1)/(σ3 − σ1)], a linear quantity describing relative stress magnitudes, where the principal stress axes, σ1, σ2, and σ3 correspond to compressional, intermediate, and extensional deviatoric stress axes, respectively. It is necessary to know the seismic slip vector, and consequently to select the preferred seismic fault plane for each pair of nodal planes to compute the stress state from earthquake focal mechanisms. For major earthquakes, the selection can be made if there is a co-seismic rupture, or from the spatial epicenter distribution of the aftershock sequence. There is another option for earthquake populations with low magnitude and no surface rupture- computation. It is possible to calculate the fault slip vector using Bott’s (1959) model since only one of the two slip vectors of a focal mechanism solution is in accordance with the principal stress axes. For this slip vector, the R ratio, defined [R = (σ2 − σ1)/(σ3 − σ1)], is such that 0