Leveraging Machine Learning for Improved Earthquake Location: A Case Study from Mayotte

Léonard Seydoux

• Affiliations: Institut de physique du globe de Paris, Université Paris Cité, 75005 Paris, France

• Presentation type: Talk

• Presentation time: Monday 11:00 - 11:15, Room S150

• Programme No: 3.1.9

• Theme 3 > Session 1

---

Abstract

We introduce a machine learning-based method designed to refine hypocenter locations by leveraging data from dense, temporary seismic deployments alongside permanent seismic networks. The method learns to correct biases in hypocenter estimations from low-density permanent arrays using the detailed information provided by temporary stations. Under adequate assumptions, this approach enables the reanalysis of past and future seismic data, enhancing our understanding of seismic crises that occurred beyond the coverage of dense arrays. Applied to Mayotte Island following the 2018 eruption, the method combines data from ocean-bottom and land seismometers to significantly improve location accuracy, demonstrating its broad applicability in regions with sparse seismic station coverage.