← Back to Sessions

Improving the Detection of Microearthquakes Without Prior Events: Application to Large-N Arrays

Description: 

In recent years, the development of high-resolution earthquake catalogs has enabled new insights into earthquake source dynamics. These catalogs, which lower the magnitude-of-completeness, have been made possible by new data-dense seismic experiments (e.g., Large-N arrays) and advances in algorithms that include Template Matching (TM) and Machine Learning (ML). However, most TM and ML must be trained using prior events, which may not always be available. In this study, we explore a new approach for the detection of microearthquakes that exploits dense seismic network data to backproject waveform correlations or Local Similarity (LS) waveforms. The Local Similarity-Back Projection method does not require templates or training. Furthermore, it enhances signal-to-noise ratio (SNR) by stacking coherent signals thus enabling us to detect even smaller events. We apply the method to Large-N data including the IRIS Wavefield Experiment in Oklahoma. Continuous ground motion velocity data were processed to compute LS, which were then backprojected through grid-search over a volume of event hypotheses to find the best origin times and hypocenters. We demonstrate that this effort resulted in a 30-fold increase in detection as compared to analyst detection.

Session: Network Seismology: Recent Developments, Challenges and Lessons Learned

Type: Oral

Date: 4/20/2023

Presentation Time: 10:45 AM (local time)

Presenting Author: Ketan Singha Roy

Student Presenter: Yes

Invited Presentation: 

Authors