Investigating the Relationship Between Waveform Correlation and Seismic Source Mechanisms

A common practice in nuclear monitoring is to use waveform correlation as a method of event detection. Earthquake waveforms, for example, are correlated with seismic data to detect other similar earthquakes. Likewise, explosion waveforms are used to identify other explosions from the same area. However, for monitoring purposes high confidence in the identification of source type is important. A high correlation coefficient does not necessarily guarantee that the template event and a new detected event are of the same source type. For example, an explosion waveform could correlate well with an earthquake template if much of the waveform's characteristics are due to propagation effects. Here, we investigate what we can confidently infer about source type from waveform correlations by correlating seismograms of earthquakes from the Geysers Geothermal Field. There are 53 earthquakes from this region with full moment tensor solutions, which range from 80% double couple to 66% isotropic to 75% CLVD (Boyd, et al. 2015). We cross-correlate the P and S waveforms using all 3 components for three filter bands (1-4Hz, 0.1-2Hz, 0.3-1Hz) and then cluster the waveforms based on their cross-correlation coefficients to see whether similar waveforms also have similar source mechanisms. We also decompose the waveforms into their subspaces using singular value decomposition to look for physical meaning in their basis functions.