Time-Domain Techniques to Automatically Detect Local Earthquakes in the Wavetrain of Large Remote Teleseseismic Events Using Data within the Continental United States

Technology advances in combination with the onslaught of data availability now allow for large seismic data streams to automatically and systematically be processed. This processing allows for the detection of unique seismic events, including events triggered by the passage of seismic waves created by large, distant earthquakes. We develop an automated approach to identify small, locally recorded earthquakes on a single station within continuous seismic data. We apply a time domain short-term average (STA) to long-term-average (LTA) ratio algorithm to create a catalog of "detections" (a signal above the noise level, which may, or may not be an earthquake) at each station recorded on three components, and then remove any spurious detections by requiring that a detection is real only if recorded on a minimum of two channels. To calibrate the algorithm, we use a set of ~900 small earthquakes in the December 2008 Yellowstone Swarm. Of the four STA/LTA algorithms we tested (e.g., 1 s/10 s; 4 s/40 s; 8 s/80 s; 16 s/160s), the 4/40s method is the most effective at identifying the majority of events in the swarm. We apply this preferred method to data from 165 M≥7 earthquakes (±5-hours of data centered on the mainshock origin times) recorded at >400 seismic stations in EarthScope's USArray Transportable Array (TA) and regional seismic networks within the continental United States. Our algorithm nets, on average, hundreds of detections for each mainshock, and we find we can detect small events within the network. The 4/40s method is successful at identifying local earthquakes within the TA and regional networks. We find STA/LTA algorithms can successfully identify small local earthquakes within large datasets.