Search of optimal crustal velocities using waveform modelling of local earthquakes

Abstract

One of the concerns of geophysicist during the last decade is to reduce the damage of earthquakes. As a result, seismic hazard studies have become an integral part of long term planning and mitigation. One approach is to calculate pe.ak ground acceleration (PGA) and use this scalar value in building design. However, this method has some disadvantages: it is very simple and it does not give spectral information related to the ground motion. Generally, the effects of faulting type, source mechanism, rupture directivity, asperities are ignored during the calculation of the PGA values. The study in this thesis contributes to the simulation of the ground motion by constructing optimal crustal velocity models based on 1- D synthetic seismogram modeling. For this purpose, different crustal models were generated using the discrete wave-number technique (Bouchon, 1981) and they have been tested by searching the best-fit between observed and synthetic seismograms. Waveforms from 5 earthquakes were analyzed in this study. The selected earthquakes have magnitudes larger than 3.5 and they are located along the main fault zone in the Sea of Marmara. The vertical, radial and transversal components were compared using the cross correlation coefficient between observed and synthetic seismograms. Crustal models having S-layers with fixed depths were used to calculate the synthetics for each selected event. First, the optimum P-wave velocities were searched within predefined velocity limits for each layer. Once the optimum P velocities were obtained, then the S wave velocities have been searched. In general a moderate level of fitting is obtained even for the optimal crustal models. Although numerically the correlation values are quite low, the shapes of the waveforms are roughly close to each other, at least for some selected parts of the total waveform. The degree of fitting is particularly low in the part of the waveform where the 3-dimentional effects in the crust start to dominate, such as the P-arrivals in the transversal component. The performance also degrades with the level of the local noise, which is known to be not negligeable at ISKB station. The use of a clever search algorithm that uses a feedback mechanism to guide the search in a selective parameter space and accelerates the convergence towards the optimum (such as steepest descent, etc) will allow the scanning of wider range of parameter (eg estimating the layer depths in parallel to velocities, etc). This will certainely improve the results.