High-Rate Phase Association with Travel Time Neural Fields

14 Jul 2023  ·  Cheng Shi, Maarten V. de Hoop, Ivan Dokmanić ·

Our understanding of regional seismicity from multi-station seismograms relies on the ability to associate arrival phases with their originating earthquakes. Deep-learning-based phase detection now detects small, high-rate arrivals from seismicity clouds, even at negative magnitudes. This new data could give important insight into earthquake dynamics, but it is presents a challenging association task. Existing techniques relying on coarsely approximated, fixed wave speed models fail in this unexplored dense regime where the complexity of unknown wave speed cannot be ignored. We introduce Harpa, a high-rate association framework built on deep generative modeling and neural fields. Harpa incorporates wave physics by using optimal transport to compare arrival sequences. It is thus robust to unknown wave speeds and estimates the wave speed model as a by-product of association. Experiments with realistic, complex synthetic models show that Harpa is the first seismic phase association framework which is accurate in the high-rate regime, paving the way for new avenues in exploratory Earth science and improved understanding of seismicity.

PDF Abstract

Datasets


  Add Datasets introduced or used in this paper

Results from the Paper


  Submit results from this paper to get state-of-the-art GitHub badges and help the community compare results to other papers.

Methods