Using Waveform Cross-correlation to Detect and Identify Regional Seismic Phases Based on a Reference Event Set

It is widely observed that earthquakes and mining blasts occurring very close to each other can share very similar waveforms. An adaptive waveform cross-correlation automated phase detection method based on a reference event set of waveforms and picks has been developed and applied to process daily seismicity in New Mexico. This method can produce robust initial seismic phase estimates while greatly improving the handling of false triggers due to telemetry or other transient noise. 72 well-picked earthquakes and mining explosions in New Mexico during 1997-2003 were collected as an initial reference event set covering most historically active source regions. Each waveform in the reference event set has a high signal to noise ratio and accurately-picked P and S phases. All waveforms are converted to analytic time series envelopes before processing in order to increase the signal to noise ratio. If a new earthquake or mining blast occurs, it has a high probability of occurring close to a reference event. New event waveforms are compared to reference waveforms across all available stations using envelope cross-correlation. If the cross-correlation coefficient is sufficiently large, the phases of the most similar waveform in the reference set are assigned to the lag-aligned unknown waveform as initial picks for subsequent refinement. If cross-correlation coefficients are low, the new waveforms are either from a noise event or possibly from a seismic event occurring in a new source region. New source region events can readily be added to the reference event set following identification and picking. The NMT-SC network real-time data acquisition system, using the EarthWorm STA/LTA pick_ew/binde_rew modules, identifies hundreds of events per month, and most of them are noise transients. We have recently integrated this waveform cross-correlation phase detection and classification method into the Earthworm data flow and into Matseis (a Matlab-based analysis package developed at Sandia National Laboratories). The method makes adaptive use of evolving earthquake catalogues of waveforms and phases, improves the handling of false triggers, and increases the accuracy of initial P and S phases picking and automatic location and event classification.