Fast Earthquake Relocation Using ANFIS Neuro-Fuzzy Network Trained Based on the Double Difference Method

One of the most important issues in seismology is to accurately determine the epicenter of earthquake, mainly because the accuracy of dependent studies such as seismic risk analysis, fault mechanism, depends on the accuracy of mentioned parameter. Double difference method is one of the conventional relative earthquake relocation methods which is used vastly by seismological centers and it is used as the basis of an algorithm that relocates earthquake events about 0.004 second. In the proposed method, Adaptive Neuro-Fuzzy Inference System (ANFIS) is trained based on the former relocated events and generalized for new earthquakes. In the training procedure, the hypocentral information, for the main event and the nearest one (determined by K nearest neighbor search), and the velocity model of the region are considered as inputs and hypocentral information are set as target for the network.  Here, with reference to the information of about 8000 earthquakes, as the training events, 490 earthquakes were relocated in about 0.1 (s) where RMS for predicating latitude and longitude of the events were 0.002 and for the depth parameter, RMS was 0.45. It is concluded that the proposed method could be considered as a powerful algorithm for near real time relocation of earthquakes.