Editors’ Highlights are summaries of recent papers by AGU’s journal editors.
Source: Journal of Geophysical Research: Solid Earth

Dozens of quasi-periodic caldera collapse events that generated a considerable amount of geophysical signals occurred during the three-month eruption of the Kilauea Volcano, Hawaii, in 2018. This unique dataset is ideal for testing the potential of deep learning to predict cyclical geophysical phenomena.

McBrearty and Segall [2024] train a deep learning model using seismicity, tilt, and GPS data recorded at Kilauea. They show that signals generated over only a half day at the beginning of each cycle are sufficient to predict the failure time of most caldera collapse events to within a few hours. Further analysis reveals accurate prediction also for long duration events that the models have never been trained on. These results demonstrate the potential of machine learning methods for predicting cyclical geophysical events under well-controlled conditions but with limited learning data.

Citation: McBrearty, I. W., & Segall, P. (2024). Deep learning forecasts caldera collapse events at Kı̄lauea Volcano. Journal of Geophysical Research: Solid Earth, 129, e2024JB029471. https://doi.org/10.1029/2024JB029471

—Olivier Roche, Associate Editor, JGR: Solid Earth

This paper is part of a special collection on Advanced Machine Learning in Solid Earth Geoscience.

Text © 2024. The authors. CC BY-NC-ND 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.