Abstract

A data set of 43 confirmed earthquakes and 42 explosions (1.3 ≤ ML ≤ 3.8) is analyzed by multivariate statistical analysis. The explosion events include quarry blasts and explosions detonated for seismic experiments as well as an explosion of an ammunition storage site. The events of this training data set are mainly located in the Alpine mountains and Central Switzerland, with the seismic sources having excellent distance (R < 300 km) and azimuthal coverage for the recording stations. Multivariate statistical analysis is used to derive general discriminant functions based on regional waveform data for the training set. The discriminant functions are then tested for consistency on the events of the training data set, with the result that 97% of these events are correctly identified by the multivariate discriminant functions. Of the total training data set, 17% is classified as ambiguous cases, based on marginal discriminant functions. The method is subsequently applied to an extended data set including nighttime events from 1992-1996, which are hence regarded as earthquakes, and a set of presumed explosions from the years 1984-1997. For Central Switzerland, the events of the extended data set can clearly be identified with the discriminant functions derived from the training data set. The identification rate is lower for other regions in Switzerland, which can be related to tectonic units. For these regions, considered to have a laterally homogeneous crust, a linear correction term in the discriminant function providing for station corrections can be found that significantly decreases the misidentification rate.

Online material: table of earthquake and explosion events

You do not currently have access to this article.