Dynamics of the Popocatépetl Volcano, Mexico, revealed by Machine Learning-Based Seismic Catalogs

Popocatépetl, one of Mexico's most active volcanoes, poses significant risks to the dense populations in its vicinity. Effective monitoring of its seismic activity is crucial for understanding and mitigating these hazards. This study employs data collected with a network of 19 seismic stations surrounding the volcano, combined with machine learning techniques and spatial coherence methods, to generate comprehensive seismic catalogs spanning from 2019 to the present. Our automated workflow includes the identification and localization of long period (LP) events, tremors, and volcano-tectonic (VT) earthquakes.

For this purpose, an improved classification model based on Support Vector Machines was developed to distinguish LP events and tremors within continuous recordings. Their locations were determined using a cross-correlation-based method. Additionally, the VT earthquake catalog was compiled using deep learning-based models for phase picking, followed by standard location methods. Our findings not only corroborate trends observed in manual analyses at the volcano's observatory but also uncover additional events, highlighting trends in the volcano’s dynamics not observed before.

To showcase the use of these catalogs, we will present a multiparametric analysis integrating seismic data with thermal anomalies, SO2 emissions, and GPS measurements. This research not only deepens our comprehension of volcanic processes but also underscores the transformative role of technology in geophysical research.

 

Research supported by the program UNAM-DGAPA-PAPIIT: IN103823.