Source Depths Utilizing Broad Band Data

This report covers the first year's effort of a two year research project to improve our ability to discriminate between nuclear explosions and earthquakes based on the depth of an event determined from local and regional phases. The approach taken has been to improve phase identification and picking abilities for depth phases pP and sP throughy adaptive filtering to reduce effects of mode conversion and scattering. Three-component polarization methods are applied to broad-band digital data from a set of earthquakes located in the northeastern United States and adjacent Canada. The cleaned up data are compared with suites of synthetic seismograms for different depths and focal mechanisms obtained using two velocity/attenuation models appropriate for northeastern United States and Canada. This comparison provides a basis for identification of the depth phases and discrimination of other reflected/refracted arrivals in the P wavetrain. Polarization state filtering (following procedures proposed by Samson) is used to emphasize those frequencies exhibiting the desired particle motion within a time window that slides through the P wavetrain. Adaptive orientation of the state filtered data into radial and transverse horizontal components and determination of azimuths and apparent angles of incidence further facilities phase identification and comparison with synthetics. Eight local and regional earthquakes and one quarry blast were selected for study. Polarization filtering reduces much of the P coda and emphasizes discrete portions of the wavetrain as anticipated.