Comparison of Pick-Based and Waveform-Based Event Detectors for Local to Near-Regional Distance Data From Utah

Date: 4/26/2019

Time: 11:00 AM

Room: Elliott Bay

In this presentation, we compare a pick-based event detector, the Probabilistic Event Detection Association and Location (PEDAL) algorithm, to a waveform-based detector, the Waveform Correlation Event Detection System (WCEDS). Both algorithms were tested on 3-component data from stations in the University of Utah Seismic Station (UUSS) network. We chose to focus on Utah because the region is tectonically complex and includes both regularly occurring earthquakes from the north-south trending Intermountain Seismicity Belt (ISB) as well as various types of anthropogenic events (quarry blasts and mining induced seismicity), hence it presents a variety of challenges for event detection. The interval of time processed (January 1-14, 2011) was chosen because of the occurrence of a significant aftershock sequence (several hundred events) near the town of Circleville in southern Utah. The data set also includes a huge number of mining induced events from a coal mining region in central Utah. The events built by our two methods are scored against a master catalog, carefully built by an expert analyst who attempted to find all events with observable signals at 3 or more stations (resulting in a total of 7883 events for the 14 days). Comparison of the 3 catalogs is done using a Venn diagram, to investigate areas of overlap and isolation. Our results suggest that when tuned to achieve a comparable level of recall, the waveform-based method has better precision (i.e. fewer false events). We also found the waveform-based method to be more stable and easy to configure due to the lack of an additional processing step to generate signal detections.

Presenting Author: Stephen Heck

Authors