SEISMIC EVENT CATALOG DEVELOPMENT FOR ROCK VALLEY, NEVADA, AND SURROUNDING AREAS USING NEURAL PHASE PICKERS: PRELIMINARY RESULTS AND FUTURE DIRECTIONS

Developing a comprehensive catalog of seismic events is a fundamental task in seismology. Several seismological data analyses, such as constraining the structure of the faults that are hosting the events and understanding the seismic source properties, depend on the quality of catalogs. The site for the Source Physics Experiment – an inter-organizational collaborative project for studying chemical explosion sources – located in Rock Valley, Nevada has recordings of continuous waveform data for over two decades. While there exists several catalogs available for the region, recent studies showed that reprocessing archived data with the state-of-the-art algorithms, particularly neural networks trained for detecting seismic phase arrivals (often referred to as neural phase pickers), can yield a better catalog with much increase in event count. We applied widely recognized pre-trained neural phase pickers to reprocess the archived data and to build a better event catalog. Although these pickers predict a large number of phase arrivals, preliminary results indicate a significant discrepancy. Specifically, they fail to reproduce many of the picks reported in existing catalogs, across all types of seismic events. Moreover, we observe high false positive rate, which yield many false events following phase association and event location. Consequently, ensuring catalog quality becomes a daunting task due to the vast number of events requiring analyst review. Our findings point to the necessity for developing custom neural phase pickers, designed either for the region or for individual stations. This approach is poised to lower false positive and false negative rates by accounting for local geology and station-specific noise characteristics. Here, we present a preliminary catalog and discuss our ongoing efforts towards developing custom neural phase pickers.