Earthquake relocation of Varzaghan- Ahar 6.5 Mw, 6.3 Mwand their aftershocks using probabilistic nonlinear algorithm

Earthquake relocation has an important role in the investigation of tectonic settings of a region. An increase in accuracy of a relocation problem can enhance the calculation of the local velocity model as well as associated studies of a risk analysis. There are some important procedures to make a reliable catalog of earthquakes. Verifying the seismic waveform to correct the picked phases, utilizing other seismic networks and using an appropriate local or regional velocity model, we can improve the final results. In a case with a well-conditioned network geometry, using an accurate local velocity model has a significant effect on increasing the depth accuracy. In this study, we have tried to use all available information from seismic networks in NW Iran and south Azerbayjan. We relocated the double main-shock of Varzaghan-Ahar Mw 6.5, Mw 6.3 and more than 1800 aftershocks with Ml > 2.0 over a 10-month period after the main shocks. To increase the accuracy of the earthquake location, we merged the recorded information of eight short periods of the Iranian Seismological Center (IRSC), one broad band station of the International Institute of Engineering and Earthquake Seismology (IIEES) and five broad band stations of the Azerbaijan National Seismic Network (ANSN). We also calculated a local velocity model using a P arrival time inversion scheme (using VELEST code) not only to obtain a more reliable earthquake location especially in depth, but also to decrease the hypocentral error. We have used all the data between 2006 and 2013 after applying a band-pass filter to get a uniform dataset of the earthquake within 250 km around the main shocks. The final velocity model indicated two velocity layers in the upper-crust with p-velocities of 5.87, 6.01 km/s and 6, 18 km thicknesses, respectively. These layers lay on a half-space with a p-velocity of 6.40 km/s. To minimize the effect of the initial velocity model on the final result, we implemented 50 random depth-increasing velocity models. Furthermore, making use of a non-linear probabilistic approach for relocating the earthquake leads to more accurate results compared to linear location programs. A comparison of the results of hypocenters between the IRSC catalog and those calculated by this study shows better line-alignment in the direction of the infer fault. Epicenter and depth error reduction due to making use of an accurate local velocity model were clearly obvious. Plotting the five depth cross sections along and perpendicular to after-shock sequence, shows a more clear geometry compared to the fixed-depth results from the IRSC catalog. According to some statistical parameters such as the hypocentral error, RMS and also preliminary location conditions such as the azimuthal gap and the number of stations, we defined two classes of events and plotted them for both IRSC and this study in map view and cross sections. This was a better way to show accuracy of our dataset (relocated events) than what we obtained by IRSC. Using the focal mechanism of the two main shocks obtained by the IRSC, along with the event distribution especially in depth, showed that there was more consistency between the results of this study and the fault orientation. This showed that the infer fault had a near vertical plane with an east-west direction and this suggested that it would be a strike-slip fault.