Evaluation of Distributed Acoustic Sensing Phase Pick Quality and Performance for Operational Earthquake Monitoring

Abstract Over the past several years, distributed acoustic sensing (DAS) has been gaining use in seismological research and even seismic monitoring. The Southern California Seismic Network recently started incorporating DAS data from Ridgecrest, California, into routine regional seismic monitoring operations. We present an analysis of the expected and observed effects of adding DAS data to monitoring. Before adding DAS data to real-time (RT) operations, we conducted a test on 228 events from 28 May to 1 August 2023 to evaluate the number and quality of phase picks from DAS data and the potential effects of those additional picks on earthquake locations. We also compared automatic picks from the machine learning picker PhaseNet-DAS to analyst picks to assess its accuracy, although it is not yet being used in operations. We repeated this analysis for 331 events from 6 May to 8 July 2025, after seven DAS channels were added to the RT systems, to assess whether the performance was as expected. PhaseNet-DAS generated about twice as many picks as the analysts. When the arrival times were compared, PhaseNet-DAS was mostly within ∼0.5 s of the analysts, with P picks more similar than S picks. The short-term average/long-term average picker in the RT system generated about half as many P picks as the analysts, though they were typically accurate. The detection threshold of PhaseNet-DAS was a little better than that of the analysts. For the event locations and depths, we found little difference in the accuracy compared to a reference hypocenter with or without DAS picks included. Overall, we found that adding DAS to our monitoring systems can provide more picks, especially in areas with poor seismic station coverage, with potential benefit to earthquake parameter quality. PhaseNet-DAS may be able to provide accurate automatic picks to reduce additional analyst work from the extra stations.