Nearest Neighbor Method for Discriminating Aftershocks and Duplicates When Merging Earthquake Catalogs

Abstract

Early aftershocks contain important information about the physics of earthquake occurrence and postseismic relaxation processes. However, the standard catalogs of early aftershocks are usually incomplete. Many events can be missed in the main shock coda, some of which are strong enough due to the extremely high noise level. Under these conditions, the process of event identification becomes largely stochastic. Due to different network configurations and record processing methods, different agencies may register/miss different events, thus merging catalogs can improve the completeness of the aftershock sequence. When merging catalogs, the problem of identifying duplicates (records related to the same seismic event) arises. The main difficulty is discriminating aftershocks and duplicates, since both are events close in space and time. The problem is analogous to the problem of discriminating aftershocks and independent events. The solution methods are usually similar too. In this paper, we apply the nearest neighbor method modified for our problem. This method has become widespread in recent years in the problem of identifying aftershocks, and a probabilistic metric in the space of network errors in determining the epicenters and times of seismic events. It is applied for automatic identification of duplicates when merging catalogs of aftershocks for the Tohoku earthquake. An analysis of the space-time structure of duplicates and aftershocks shows their significant difference, which makes it possible to successfully solve the problem. In a sample from the global Advanced National Seismic System (ANSS) catalog ((Formula presented.)), were found more than 700 events missed by the Japan Meteorological Agency (JMA) seismic network, which is one of the best in the world. Among the misses, there are several events with (Formula presented.) in the first hours after the main shock. Duplicate identification reliability is >97%. The method can be used to improve the completeness of aftershock sequences. The reliable identification of duplicates allows, in addition, to study the correspondence of the magnitudes determined by different agencies. Therefore the present method is an effective tool for creating merged catalogs of earthquakes with a uniform magnitude.