Expanding Coherent Array Processing to Larger Apertures Using Empirical Matched Field Processing

摘要

This final report summarizes the results of a three-year collaborative project undertaken by NORSAR, Lawrence Livermore National Laboratory, and Deschutes Signal Processing to explore the applicability of empirical matched field processing (EMFP) to the detection and classification of seismic signals over increasing sensor apertures. EMFP uses the observations of historical events to calibrate the amplitude and phase structure of an incident wavefield over a given sensor configuration for particular repeating sources. We have demonstrated an excellent source classification both for mines on the Kola Peninsula, NW Russia and in Kazakhstan using calibrations constructed from multiple Ground Truth events. A multitaper procedure has been developed for evaluating covariance matrices over short and precisely defined time-series and, while there are advantages to characterizing source regions with ensemble covariance matrices, single-event calibrations appear to work well in many cases meaning that the method is applicable also to source regions with few observations. We have demonstrated single-phase EMFP to be a viable event detector. Given only single observations or rank-1 matched field statistics, it appears that an increase in the receiver aperture makes a calibration specific to a more limited source region. Higher rank matched field statistics may make broader regions of diffuse seismicity amenable to coherent processing over wider sensor apertures. Considering the aftershock sequence from the M=7.4 October 8, 2005, Kashmir event, we suggest a partially coherent procedure whereby 3-component matched field statistics from the individual stations of KNET can be stacked to form a robust detector of larger events from a broad source region.