Declustering of Seismicity Flow: Statistical Analysis

Abstract

The known methods for identifying the clusters of seismic events that are mainly formed by the aftershocks frequently include debatable initial assumptions or a complicated system of successive approximations. One of the most recent and the most logically consistent methods for identifying clusters of aftershocks is the nearest neighbor distance (NND) method, which is however algorithmically most challenging. In this paper, we propose a new declustering method based on a generalized distance (GD) metric, which employs some assumptions of the NDD method but is as simple in practical implementation as the window methods previously proposed for this purpose. In analyzing and substantiating this new method, a procedure of random shuffling of seismic events with respect to time is used for generating a real catalog, which is however free of genetic relationships between different events. The efficiency of the existing window methods, the GD method, and the NND methods is compared by a number of the criteria for 17 regions. It is shown that the GD method is, on average, noticeably more efficient than the standard window methods and compares very favorably with the NND method. In the conclusions, a certain speculativeness of separating the events into the main and dependent shocks is discussed.