Source location using 3-D Gaussian beam migration imaging

Traditional earthquake location methods usually need to pick first arrivals of signals, which have a low precision when these arrivals are unremarkable or submerged by strong noise. In this paper, we employ the 3-D varying-grid Gaussian beam migration imaging method to determine seismic source locations. We apply this method to deal with the synthetic seismograms and real seismic events in the Capital Circle region around Beijing to verify the validity of this method. We migrate every record of the station backward and take the position where all the imaging results merge relatively as the source location that releases the dominant energy. We propose to use the numerical scheme of varying grid. Firstly we utilize the method of global search to obtain the quick result of the epicenter in the coarse grid condition, and then we determine the final source location just by the local search within the epicenter area with a fine grid. We also employ a 3-D spatial Gaussian filter to obtain the source location and suppress the background noise automatically when the imaging results are scattered or unfocused, getting rid of searching for the major spatial energy points in the conventional least square method. The results based on the synthetic data show that the method combining the varying-grid 3-D Gaussian beam and spatial Gaussian filter can not only ensure the high efficiency with less steps of ray tracing but also can be more adaptable to the low signal-to-noise ratio data and scattered or unfocused imaging energy. Besides, we also test the proposed method on the data of Zhuolu, Luanxian and Fangshan earthquakes in the Capital Circle region around Beijing, which can further confirm the efficiency and accuracy of the method. The earthquake location method based on 3-D varying-grid Gaussian beam migration imaging just needs the rough picking of first arrivals, which is more robust to the pickup error than conventional methods. The method implements the unification of imaging accuracy and computational efficiency, and is highly automatic and suitable for noisy data especially in low signal-to-noise ratio cases.