Automatic detection of low-frequency earthquakes (LFEs) based on a beamformed network response

Low-frequency earthquakes (LFEs), which frequently originate from multiplet-generating sources that are closely linked with tectonic tremor in subduction zones around the world, are difficult to observe and characterize due to their low signal-to-noise ratios. This obstacle can be sidestepped by detecting and then stacking all of the multiplets of a master LFE event, or template, using a matched-filter search; the difficulty however lies in finding an LFE event to use as a template. We implement here an automated beamforming algorithm to detect LFEs within the Mexican subduction zone that can then be used as templates in a matched-filter search. Seismograms recorded on a network of seismic stations are aligned to match the moveout of a potential source at depth and their energies are then summed; any spikes in the summed energy indicate an event originating from that potential source. We apply this method to a 1-d test case and we are able to detect 381 unique, potential LFE templates. We then compare our method to a previously introduced LFE detection scheme based on multiplet correlations for three test cases and find that the two methods are complementary.