Tracking aftershock sequences using empirical matched field processing

Extensive aftershock sequences present a significant problem to seismological data centres attempting to produce near real-time comprehensive seismic event bulletins. An elevated number of events to process and poorer performance of automatic phase association algorithms can lead to large delays in processing and a greatly increased human workload. Global monitoring is often performed using seismic array stations at considerable distances from the events involved. Empirical matched field processing (EMFP) is a narrow-frequency band array signal processing technique that recognizes the inter-sensor phase and amplitude relations associated with wavefronts approaching a sensor array from a given direction. We demonstrate that EMFP, using a template obtained from the first P arrival from the main shock alone, can efficiently detect and identify P arrivals on that array from subsequent events in the aftershock zone with exceptionally few false alarms (signals from other sources). The empirical wavefield template encodes all the narrow-band phase and amplitude relations observed for the main shock signal. These relations are also often robust and repeatable characteristics of signals from nearby events. The EMFP detection statistic compares the phase and amplitude relations at a given time in the incoming data stream with those for the template and is sensitive to very short-duration signals with the required characteristics. Significant deviations from the plane wavefront model that typically degrade the performance of standard beamforming techniques can enhance signal characterization using EMFP. Waveform correlation techniques typically perform poorly for aftershocks from large earthquakes due to the distances between hypocentres and the wide range of event magnitudes and source mechanisms. EMFP on remote seismic arrays mitigates these difficulties; the narrow-band nature of the procedure makes arrival identification less sensitive to the signals' temporal form and spectral content. The empirical steering vectors derived for the main shock P arrival can reduce the frequency dependency of the slowness vector estimates. This property helps us to automatically screen out arrivals from outside of the aftershock zone. Standard array processing pipelines could be enhanced by including both plane-wave and empirical matched field steering vectors. This would maintain present capability for the plane-wave steering vectors and provide increased sensitivity and resolution for those sources for which we have empirical calibrations.