arXiv.itsnick.co

an independent field guide to 23 million papers from arXiv & OpenAlex — not affiliated with arXiv.org

Computer Science · Artificial Intelligence

Detection and characterization of seismic and acoustic tremor at volcanoes using machine learning

Tan, Darren, Fee, David, Girona, Társilo, Haney, Matt, Witsil, Alex, Wech, Aaron

· Publication Database GFZ (GFZ German Research Centre for Geosciences)

See on the map

Abstract

<!--!introduction!--><b></b> Volcanic tremor is a semi-continuous seismic and/or acoustic signal that occurs at time scales ranging from seconds to years, with highly variable amplitudes and spectral features. Having a fast, robust, automated method to detect and characterize tremor in high temporal resolution would be beneficial to eruption forecasting and monitoring efforts, as well as to retrospectively building tremor catalogs for research purposes. As part of the NSF-funded PREEVENTS eruption forecasting project, we test and develop a pair of machine learning algorithms (one for seismic, one for acoustic) that can detect tremor in near-real time and classify it according to its spectral signature. First, we manually label tremor time windows on single station seismic and low-frequency acoustic (infrasound) spectrograms from the 2021-2022 eruption of Pavlof Volcano, Alaska, and train a convolutional neural network and random forest algorithm for both data types. We then use the trained models to classify seismic and infrasound data leading up to, during, and after the eruption, and compare temporal trends observed in our classifications alongside eruption chronologies compiled by the Alaska Volcano Observatory. We also implement a local network-wide weighting scheme to reduce false detections. Lastly, we assess the transferability of our models by applying them on unseen data recorded from past eruptions at Pavlof and other Alaska volcanoes. We envision that our machine learning models will aid tremor-related monitoring and research, and will advance our understanding of the relationship between tremor and subsurface volcanic processes by making consistent tremor catalogs more accessible.&nbsp;

More like this