Journal of Geophysical Research Solid Earth Supporting Information for Fault-Slip Source Models for the 2014 Mw 6.9 Samothraki-Gökçeada Earthquake (North Aegean Trough) Combining Geodetic and Seismological Observations Vasso Saltogianni (1), Michail Gianniou (2), Tuncay Taymaz (3), Seda Yolsal-Çevikbilen (3), and Stathis Stiros (1)* (1) Department of Civil Engineering, University of Patras, Patras, Greece, (2) National Cadastre and Mapping Agency SA, Athens, Greece, (3) Department of Geophysical Engineering, The Faculty of Mines, İstanbul Technical University, İTÜ Maslak Campus, 34469, Sarıyer, İstanbul, Turkey *corresponding author: Laboratory of Geodesy and Geodetic Applications, Dept. of Civil Engineering, Patras University, Rio 26500, Greece, tel: +30 2610 996511, fax: +30 2610 997877, email: stiros@upatras.gr Contents of this file Text S1 to S2 Figures S1 to S7 Tables S1 to S3 Additional Supporting Information (Files uploaded separately) Captions for Movie S1 Introduction This file contains additional Text, Figures and Tables supporting the main article. In particular Text S1: Detailed description of the analysis of the seismological data. Text S2: Details of geodetic data and validation of the PPP solutions. Figure S1: Location map of the stations used in seismological analysis. Figure S2: Distributions of P- wave first motion polarities of the May 24, 2014 earthquake. Figure S3: Comparison of the source parameters of the May 24, 2014 earthquake. Figure S4: Finite-fault slip model of the May 24, 2014 earthquake. Figure S5: Timeseries of the daily vertical coordinates. 1 Figure S6: Predicted vertical displacements for the geodetic models. Figure S7: Coulomb stress changes due to the double fault of the 2014 earthquake. Table S1: Epicenter list of the main events in NAT the last 50 years. Table S2: Epicenter list of the main aftershocks of the May 24, 2014 earthquake. Table S3: Displacement GPS vectors of the May 24, 2014 earthquake. Movie S1: Movie of the finite-fault slip evolution. 2 Text S1. Details of the methodology of the Waveform Inversion for Source Parameters and for Finite-Fault Slip Modeling Waveform Inversion for Source Parameters In order to compute the strike, dip and rake angles, the focal depth, the seismic moment and source time functions of the source teleseismic long-period waves were inverted using the Moment Tensor 5 (MT5) body-waveform algorithm [Nàbělek 1984; McCaffrey et al., 1991; Zwick et al., 1994]. The epicenter location was assumed fixed to that of USGS-NEIC values. For long-period P- and SH- waveforms the attenuation parameter t* (t*=t/Q, t: travel time and Q: average attenuation along the ray path) was selected to be 1.0s and 4.0s, respectively [Futterman, 1962]. The effect of the uncertainties in t* was predominantly assigned on source duration and on seismic moment, rather than source orientation or focal depth [McCaffrey and Aber’s, 1988; Taymaz et al., 1990; 1991; 2007a; 2007b; Taymaz and Price, 1992; Tan and Taymaz, 2006; Yolsal-Çevikbilen and Taymaz, 2012; Fielding et al., 2013; Yolsal-Çevikbilen et al., 2014; Çubuk et al., 2014]. Velocity responses were deconvolved from the records and then were re-convolved using the response of the 15–100s long-period instruments of the old World Wide Standard Seismographic Network (WWSSN). Synthetic waveforms were formed by the combination of direct (P or SH) and reflected (pP and sP, or sS) phases from a point source embedded in a given velocity structure. We used a half-space source velocity model, including a water layer with a depth appropriate to USGS-NEIC epicenter, consisting of P-wave velocity Vp =6.5 km/s, S-wave velocity Vs =3.7 km/s, and density ρ = 2.9 g/cm3 given by Zwick et al. [1994], Makris and Stobbe, [1984] and Taymaz et al. [1990; 1991], a simplified crustal model for teleseismic waveform modeling with the MT5 algorithm. Based on bathymetry data, we defined a water layer with a velocity of Vp = 1.5 km/s and thicknesses of 750 m [Smith and Sandwell, 1997a; 1997b]. More complex velocity models may appear more tempting, but it is questionable whether they may lead to better estimates of the centroid depth. Also, that receiver structures and the medium below them were assumed to correspond to homogeneous half-spaces. The seismograms were weighted according to the azimuthal distribution of stations, and smaller weights were given to clusters of stations and larger weights to isolated stations [McCaffrey et al., 1991]. Jeffreys and Bullen [1940, 1958] travel-time tables were used to calculate the arrival times. It was found that the estimates of the above inversion were weakly only correlated. Hence, if one parameter is fixed to an a priori value not differing more than a few degrees or kilometers from the value offering minimum misfit between observed and synthetic seismograms, the inversion procedure was leading to estimates of the other parameters also corresponding to optimal (minimum misfit) solutions. The obvious complexities of P- and of SH- waves are indicative of strike-slip faulting at shallow focal depth, as well as directivity effect of the rupture front. Furthermore, we have done uncertainty test to find the degree of this misfitting between observed and synthetic waveforms. The inversion returns the minimum misfit solution for the strike, dip, rake angles and focal depth and seismic moment values between the synthetic and observed seismograms in a weighted least square’s sense. It is widely reported that [see Taymaz et al., 1990; 1991; 2004; 2007a; Kiratzi and Louvari, 2001; Tan and Taymaz, 2006; YolsalÇevikbilen et al., 2012, 2014; Fielding et al., 2013] the covariance matrix associated with the minimum misfit solution usually underestimates the true uncertainties. Hence, we followed the procedure proposed by McCaffery and Nabelek [1987] and Molnar and Lyon-Caen [1989] to find more realistic uncertainties. 3 Details of the Finite-Fault Slip Modeling Finite-fault slip distributions were investigated using an inversion scheme developed by Yagi and Kikuchi [2000] based on Jeffreys-Bullen [1940, 1958] velocity-depth models. Teleseismic broad-band P-waveforms are windowed for 60 s, starting 10 s before the origin time. Velocity seismograms were first band-filtered between 0.01 and 0.8Hz and then were converted into displacements adopting a sampling interval of 0.25s. The earthquake source model is constructed using, a standard waveform inversion scheme given by Hartzell and Heaton [1983]. The basic assumption is that the rupture can be approximated by a single fault plane, and that the slip (rake) angle remained constant during the strong motion. The fault plane was assumed to consist of a grid of sub-faults, as analyzed in the main text. The source-time (sliprate) function of each sub-fault was expanded in a series of overlapping triangle functions each with a rise time of 1.25 s. Green's functions are calculated using the method of Kikuchi and Kanamori [1991]. We also applied smoothing on the slip distribution in space and time in order to prevent an instability that might occur due to local minima through high number of model parameters. Further details of finite-fault slip inversion can be found at relevant studies [Hartzell and Heaton, 1983; Yagi and Kikuchi, 2000; Yagi et al., 2003]. The stress drop (Δσ) was estimated using the equation below based on the assumption of a circular crack [Aki, 1972; Kanamori, 1994] 7 3/2 M 0 16 A3/2 Text S2. Details of geodetic data and validation of the PPP solutions As was mentioned in the main text data from 11 permanent GPS stations were analyzed three weeks before and after the earthquake on 2014.05.24 (data period : from 2014.05.03 to 2014.06.14). Daily 30-s records were obtained for all the stations, except for station IPSA for the days 2014.05.05, 2014.05.20-21 and 2014.06.06-14 and for station CANA for the days 2014.05.08, 2014.05.18-19, 2014.05.31 and 2014.06.01-02. In addition, the available records of CANA on 2014.05.07, 2014.05.09, 2014.05.30, 2014.06.03, 2014.06.11 and on 2014.05.13 were less than 24 hours long. The GPS data for each station have been processed using the Precise Point Positioning [PPP; Zumberge et al., 1997; Héroux and Kouba, 2001] method and the GrafNav software ver. 8.40. Elevation-dependent weighted solutions have been obtained using an elevation mask of 7.5o and final precise orbits (IGb08) and clock data (sp3 and clk files) computed at the CODE IGS Analysis Centre, providing higher frequency information compared to conventional final IGS products. The ionosphere-free linear combination was used for the elimination of the ionospheric influence. The tropospheric zenith delay was modeled as a state in a Kalman filter. For the validation of the PPP results in the vertical component (see section 3.2 in main text) part of the data have been processed in an alternative way. The distant 020A station (about 130 km away from the epicentral area; Figure 3) was regarded as a reference station (fixed). The GPS observations of stations 036A, 022A, 019A, 018B and 089A were then adjusted using the baseline technique and the Trimble Business Center ver. 1.12 software. The results from this alternative analysis show that no significant vertical displacements can be documented in any of the above stations, including near-field stations 089A and 018B. This confirms previous, PPPbased results and indicates that within the limits of accuracy of elevation changes, a few millimeters, GPS stations captured only horizontal crustal deformation. 4 Figure S1. Location of the (a) near field (0°<Δ<30°) and teleseismic (30°<Δ<90°) stations used in P- wave first motion polarity distribution (b) teleseismic long-period P- and SH- stations used in point-source inversion (c) Teleseismic broad-band P- stations used in finite-fault source inversion. Figure S2. Distributions of P- wave first motion polarities of the, 2014 earthquake recorded by regional and teleseismic stations. Black and white circles show up and down P-wave polarities, respectively. Lower hemisphere equal area projections are used. The station positions have been plotted with the same velocity model beneath the source used in our minimum misfit solution. Nodal planes are those of the minimum misfit solution. Station codes and epicentral distances are given above and below waveforms, respectively. An arrow marks the onset of second-event 13 s after the first event (see also Figure 2a-b in main text). ANTO station is equipped with Geotech KS-36000-I Borehole seismometer, and the others are Strekeisen (STS) Very Broad Band (VBB) seismometers. 5 Figure S3. Comparison of our minimum misfit solution of the 2014 earthquake with the source parameters reported by USGS-NEIC and Harvard-CMT catalogs. The top row shows selected waveforms from our preferred long-period minimum misfit solution. The stations are identified at the top of each column, with the type of waveform marked by P- and SH- and followed by the instrument type. At the start of each row are the P- and SH- focal spheres for the focal parameters represented by the five numbers (strike, dip, rake, depth and seismic moment), showing the positions on the focal spheres of the stations chosen. X and show matches of observed to synthetic waveforms that are worse and better than in the minimum misfit solution, respectively. An arrow marks the onset of second-event 13 s after the first event (see also Figure 2a-b in main text). 6 Figure S4. Focal mechanism, co-seismic slip distribution, and total moment rate function of the main earthquake and comparison of the observed (black) and synthetic (red) broadband P waveforms used in body-wave finite-fault slip-distribution inversion. Station code and maximum amplitude are shown above the waveforms, station azimuth, and distance below. 7 Figure S5. Variations of daily coordinates for four representative stations in vertical axis for an interval of 21 days before and after the day of the earthquake (dashed line). 2-σ uncertainty intervals are shown. Figure S6. Contour pattern of the modeled vertical displacements for (a) the single fault and (b) the double fault (see main text, Section 3.3). The location of GPS stations is marked by triangles with the station code. Results were produced using the Coulomb 3.3 software. 8 Figure S7. Coulomb stress changes due to the double fault of the 2014 earthquake. The scale of the changes is indicated by the color bar to the right (in bars) The calculations were made at a depth of 10km (approximate middle of the geometric fault model, compatible to focal depth), for the average values of strike, dip and rake, 72o, 90o and 179o, respectively. Aftershocks are marked by gray circles. Results produced using the Coulomb 3.3 software. 9 Table S1. Source parameters of North Aegean Trough earthquakes (see Figure 3). List reported by TT91: Taymaz et al. [1991] and TY15: Taymaz and Yolsal-Çevikbilen [2015]. Lat: Latitute (°N), Lon: Longitude (°E), : strike angle, : dip angle, : rake angle, Mo: Seismic Moment (Newton – meter), h: focal depth (km). For earthquake magnitude, w and s represent moment and surface wave magnitudes, respectively. Epicenter locations are taken from ISC earthquake catalogue except for 2013-2014 earthquakes (USGS-NEIC). Origin time (to) (h: m :s) 09.03.1965 17:57:54 27.03.1975 05:15:07 Date (d. m. y) Ms Mw Lat. (°N) Lon. (°E) 39.34 40.45 23.82 26.12 6.3 6.6 18.01.1982 19:27:25 39.96 24.39 6.9 06.08.1983 06.07.2003 24.05.2014 Source-2 40.14 40.45 40.29 40.44 24.74 26.04 25.45 25.87 6.9 5.7 6.7 6.5 15:43:51 19:10:27 09:25:03 09:25:16 * Strike (, °) Dip (, °) 135±5 68±8 85±5 55±8 Rake (, °) h (km) Mo (101 6 Nm) 147 200 15+3/-7 7±2 -145±8 15±3 187233±5 62±5 7±1 732 10/+7 47-5/+8 83±6 180±10 7±1 1940 168 86 10 8 40.6 75±5 85±5 -178±5 11±2 1550 242 80 -176 11 677 Slip vector (°) 045 046 Ref. TT91 TT91 050 TT91 047 078 075 061 TT91 TY15 This study Table S2. Important (Mw>4.0) events of the 2014 seismic sequence. Aftershocks correspond to rather distinct clusters (see Figure 1 for location and epicenter distribution). Data retrieved from ECP14: Evangelidis [2014] (relocated) and National Observatory of Athens (NOA) database (preliminary). a/a Date-UTC time Longitude (o) Latitude (o) Depth (km) Magnitude Ref 1 24/5/2014-09:25 25.3982 40.2970 15.86 6.8 Mw ECP14 2 24/5/2014-09:31 26.1335 40.4675 18.98 4.9 ML ECP14 3 24/5/2014-09:34 24.3791 40.0510 22.11 4.6 ML ECP14 4 22/8/2014-04:27 23.4610 39.9228 28.7 5.1 Mw NOA 5 04/9/2014-17:43 24.9012 40.1515 15.08 4.9 Mw ECP14 10 Table S3. GPS-derived displacements used in the analysis. For location see Figure 1 in main text. Longitude Latitude dE dN σE σN (o) (o) (mm) (mm) (mm) (mm) 018B 25.524 40.474 96.9 32.0 0.7 1.0 089A 25.201 39.908 -23.0 -52.0 0.6 0.8 019A 25.745 40.855 14.3 13.1 0.6 0.8 036A 24.612 40.777 9.4 -5.4 0.7 0.8 076A 23.944 40.039 3.0 -1.7 0.7 0.8 020A 26.389 41.270 6.7 5.9 0.6 0.8 022A 26.022 41.233 5.5 8.6 0.6 0.8 069A 25.289 41.090 8.5 3.4 0.4 0.6 091A 26.106 39.249 -0.7 -4.0 0.6 0.8 CANA 26.414 40.111 -27.0 14.0 0.7 0.9 IPSA 26.380 40.918 12.7 15.4 0.7 0.8 Site Movie S1. WMV movie of the finite-fault slip evolution for the model shown in Figure 3 of the main text is represented as an animation of the cumulative slip distribution for various times (1s – 40s) after rupture initiation at the hypocenter (white star). The animation is approximately in real time. The color scale shows the amount of slip in meters. References Aki, K. (1972), Earthquake mechanism, Tectonophysics, 13, 423-446. Çubuk, Y., S. Yolsal-Çevikbilen, and T. Taymaz (2014), Source parameters of the 2005–2008 Balâ–Sırapınar (central Turkey) earthquakes: Implications for the internal deformation of the Anatolian plate, Tectonophysics, 635, 125–153. doi:10.1016/j.tecto.2014.07.005. Evangelidis, C. P. (2014), Imaging supershear rupture for the 2014 Mw 6.9 Northern Aegean earthquake by backprojection of strong motion waveforms, Geophys. Res. Lett., 2014GL062513. doi:10.1002/2014GL062513. Fielding, E. J., P. R. Lundgren, T. Taymaz, S. Yolsal-Çevikbilen, and S. E. Owen (2013), Fault Slip Source Models for the 2011 M 7.1 Van Earthquake in Turkey from SAR Interferometry, Pixel Offset Tracking, GPS, and Seismic Waveform Analysis, Seismological Research Letters, 84(4), 579–593. doi:10.1785/0220120164. Futterman, W. I. (1962), Dispersive body waves, J. Geophys. Res., 67(13), 5279–5291. doi:10.1029/JZ067i013p05279. 11 Hartzell, S. H., and T. H. Heaton (1983), Inversion of strong ground motion and teleseismic waveform data for the fault rupture history of the 1979 Imperial Valley, California, earthquake, Bulletin of the Seismological Society of America, 73(6), 1553–1583. Héroux, P., and J. Kouba (2001), GPS precise point positioning using IGS orbit products, Physics and Chemistry of the Earth, Part A: Solid Earth and Geodesy, 26(6-8), 573–578. doi:10.1016/S1464-1895(01)00103-X. Jeffreys, H., K.E. Bullen (1940, 1958). Seismological Tables. Office of the British Association, Burlington House, London. Kanamori, H. (1994), Mechanics of Earthquakes, Annual Review of Earth and Planetary Sciences, 22(1), 207–237. doi:10.1146/annurev.ea.22.050194.001231. Kikuchi, M. and H. Kanamori (1991), Inversion of complex body waves – III, Bull. Seism. Soc. Am., 81, 2335-2350. Kiratzi, A., and E. Louvari (2001), Source parameters of the Izmit-Bolu 1999 (Turkey) earthquake sequences from teleseismic data, Annali di Geofisica, 44 (1), 33-47. Makris, J., and C. Stobbe (1984), Physical properties and state of the crust and upper mantle of the Eastern Mediterranean Sea deduced from geophysical data, Marine Geology, 55(3– 4), 347–363. doi:10.1016/0025-3227(84)90076-8. McCaffrey, R., G. Abers, (1988), SYN3: A program for inversion teleseismic body waveforms on microcomputers. Air force Geophysics Laboratory Technical Report, AFGI-TR-880099, Hanscomb Air Force Base, MA, 50 pp. McCaffrey, R., P. Zwick, and G. Abers, (1991). SYN4 Program. IASPEI Software Library 3, 81166. Molnar, P., and H. Lyon-Caent (1989), Fault plane solutions of earthquakes and active tectonics of the Tibetan Plateau and its margins, Geophysical Journal International, 99(1), 123– 154. doi:10.1111/j.1365-246X.1989.tb02020.x. Nàbělek, J. L. (1984), Determination of earthquake source parameters from inversion of body waves Thesis, Massachusetts Institute of Technology. Smıth W. H. F. and D.T. Sandwell (1997a), Measured and estimated seafloor topography (version 4.2) World Data Centre-A for Marine Geology and Geophysics Research Publication. Smıth W. H. F. and D.T. Sandwell (1997b), Global seafloor topography from satellite altimetry and ship depth soundings. Science 277, 1957–1962. Tan, O., and T. Taymaz (2006), Active tectonics of the Caucasus: Earthquake source mechanisms and rupture histories obtained from inversion of teleseismic body waveforms, Geological Society of America Special Papers, 409, 531–578. doi:10.1130/2006.2409(25). Taymaz, T. (1999). Seismotectonics of the Marmara region: Source characteristics of 1999 Gölcük-Sapanca-Düzce earthquakes, ITU-IAHS International Conference On Kocaeli 12 Earthquake - 17 August 1999: A Scientific Assessment and Recommendations for ReBuilding, pp 55-77, Istanbul Technical University, Accession Number: WOS:000180001400009, ISBN:975-561-166-5. Taymaz, T., and S. Price (1992), The 1971 May 12 Burdur earthquake sequence, SW Turkey: a synthesis of seismological and geological observations, Geophysical Journal International, 108(2), 589–603. doi:10.1111/j.1365-246X.1992.tb04638.x. Taymaz, T. and S. Yolsal-Çevikbilen (2015), Source Parameters of Major Earthquakes in the Aegean during 2013-2014: Implications on Recent Tectonics and Deformations, Geophysical Research Abstracts, Vol. 17, EGU2015-4744, 12-17 April 2015, Vienna, Austria. Taymaz, T., J. Jackson, and R. Westaway (1990), Earthquake mechanisms in the Hellenic Trench near Crete, Geophysical Journal International, 102(3), 695–731. doi:10.1111/j.1365246X.1990.tb04590.x. Taymaz, T., J. Jackson, and D. McKenzie (1991), Active tectonics of the north and central Aegean Sea, Geophys. J. Int., 106(2), 433–490. doi:10.1111/j.1365246X.1991.tb03906.x. Taymaz, T., R. Westaway, and R. Reilinger (2004), Active faulting and crustal deformation in the Eastern Mediterranean region, Tectonophysics, 391(1–4), 1–9. doi:10.1016/j.tecto.2004.07.005. Taymaz, T., T. J. Wright, S. Yolsal, O. Tan, E. Fielding, and G. Seyitoǧlu (2007a), Source characteristics of the 6 June 2000 Orta–Çankırı (central Turkey) earthquake: a synthesis of seismological, geological and geodetic (InSAR) observations, and internal deformation of the Anatolian plate, Geological Society, London, Special Publications, 291(1), 259–290. doi:10.1144/SP291.12. Taymaz, T., Y. Yilmaz, and Y. Dilek (2007b), The geodynamics of the Aegean and Anatolia: introduction, Geological Society, London, Special Publications, 291(1), 1–16. doi:10.1144/SP291.1. Yagi, Y., and M. Kikuchi (2000), Source rupture process of the Kocaeli, Turkey, earthquake of August 17, 1999, obtained by joint inversion of near-field data and teleseismic data, Geophys. Res. Lett., 27(13), 1969–1972. doi:10.1029/1999GL011208. Yagi, Y., M. Kikuchi, and T. Nishimura (2003), Co-seismic slip, post-seismic slip, and largest aftershock associated with the 1994 Sanriku-haruka-oki, Japan, earthquake, Geophys. Res. Lett., 30(22), 2177. doi:10.1029/2003GL018189. Yolsal-Çevikbilen, S., and T. Taymaz (2012), Earthquake source parameters along the Hellenic subduction zone and numerical simulations of historical tsunamis in the Eastern Mediterranean, Tectonophysics, 536–537, 61–100. doi:10.1016/j.tecto.2012.02.019. Yolsal-Çevikbilen, S., T. Taymaz, and C. Helvacı (2014), Earthquake mechanisms in the Gulfs of Gökova, Sığacık, Kuşadası, and the Simav Region (western Turkey): Neotectonics, seismotectonics and geodynamic implications, Tectonophysics, 635, 100–124. doi:10.1016/j.tecto.2014.05.001. 13 Zumberge, J. F., M. B. Heflin, D. C. Jefferson, M. M. Watkins, and F. H. Webb (1997), Precise point positioning for the efficient and robust analysis of GPS data from large networks, J. Geophys. Res., 102(B3), 5005–5017. doi:10.1029/96JB03860. Zwick P., R. McCaffrey and G. Abers (1994), MT5 Program, In IASPEI Software Library, number 4, IASPEI. 14