Studies of Low-frequency Earthquakes in Northern Cascadia Using a Cross-station Method

Abstract:

Low-frequency earthquake templates obtained for southern Vancouver Island reveal waveforms consistent with moment-tensor point sources and shallow thrusting in the direction of plate motion (Royer and Bostock 2013). For near-vertical source-receiver geometries, pre-critical S-wave arrivals are dominantly excited on horizontal component seismograms and likely to exhibit waveform similarity. As previously proposed by Rubin et al. (2013) and Armbruster et al. (2014), such similarity across stations can be exploited in cross-station detection approaches. In order to identify raw LFE waveforms that can be used to construct high SNR templates, we developed a multi-channel cross-correlation-based cross-station algorithm that we have applied to 4 composite arrays in northern Cascadia, employing data from EarthScope (Transportable and Flexible Arrays, Plate Boundary Observatory), POLARIS and permanent network (Canadian National Seismograph Network, Pacific Northwest Seismic Network ) sources. We obtained thousands of epicenters with horizontal errors under 3 km per ETS episode that reveal high spatio-temporal clustering, suggesting repeated failing at localized asperities on the plate interface. These detections allow for spatial targeting of initial templates that may be incorporated within network correlation approaches to extend LFE template catalogues. In addition, we examine the scales of clustering and the migration velocities and directions, some of which are consistent with the rapid tremor reversals (RTR) (Houston et al. 2011). During inter-ETS periods, our epicenters are clustered near the 40 km depth contour of plate interface models, downdip of asperities that dominate activity during ETS. We identify P-wave arrivals for a few hundreds of detections and obtain hypocenters concentrated along a surface consistent with the plate interface. The method performs best for the high SNR records of southern Vancouver Island, but we also present results for Washington state and northern Vancouver Island, where similar clustering behaviour is observed.