System for Automatic Recognition of Types of Sources of Regional Seismic Events

Abstract

An approach to recognizing the types of seismic event sources is proposed, based on the combination of several heterogeneous parameters of event records and information about the territory where the seismic event occurred. The following recording parameters are used: the ratio of the amplitudes of body waves, the ratio of parts of the spectra at high and low frequencies, the magnitude, and the spectral constancy parameter. Territorial information includes data on the presence of waterbodies, glaciers, mines, and simplified information on natural seismic activity. Their joint use is done with a special type of Bayesian belief network. The decisions made by the network are probabilistic in nature; the probability is understood in the Bayesian sense, i.e., as the degree of confidence in the truth of the judgment, which consists in attributing an event to one of the types (mine explosion, other explosion on land, underwater explosion, earthquake, icequake). The approach is implemented as a software system, which is included in the program for the interactive analysis of seismic event records LOS.