1. Detection and Characterization of Seismic and Acoustic Signals at Pavlof Volcano, Alaska, Using Deep Learning
- Author
-
Tan, Darren, Fee, David, Witsil, Alex, Girona, Társilo, Haney, Matthew, Wech, Aaron, Waythomas, Chris, and Lopez, Taryn
- Abstract
Volcanic tremor is a semi‐continuous seismic and/or acoustic signal that occurs at time scales ranging from seconds to years, with variable amplitudes and spectral features. Tremor sources have often been related to fluid movement and degassing processes, and are recognized as a potential geophysical precursor and co‐eruptive geophysical signal. Eruption forecasting and monitoring efforts need a fast, robust method to automatically detect, characterize, and catalog volcanic tremor. Here we develop VOlcano Infrasound and Seismic Spectrogram Network (VOISS‐Net), a pair of convolutional neural networks (one for seismic, one for acoustic) that can detect tremor in near real‐time and classify it according to its spectral signature. Specifically, we construct an extensive data set of labeled seismic and low‐frequency acoustic (infrasound) spectrograms from the 2021–2022 eruption of Pavlof Volcano, Alaska, and use it to train VOISS‐Net to differentiate between different tremor types, explosions, earthquakes and noise. We use VOISS‐Net to classify continuous data from past Pavlof Volcano eruptions (2007, 2013, 2014, 2016, and 2021–2022). VOISS‐Net achieves an 81.2% and 90.0% accuracy on the seismic and infrasound test sets respectively, and successfully characterizes tremor sequences for each eruption. By comparing the derived seismoacoustic timelines of each eruption with the corresponding eruption chronologies compiled by the Alaska Volcano Observatory, our model identifies changes in tremor regimes that coincide with observed volcanic activity. VOISS‐Net can aid tremor‐related monitoring and research by making consistent tremor catalogs more accessible. Volcanic tremor is a persistent vibration of the ground, atmosphere, or both that can occur before and during volcanic eruptions. Despite its importance in volcano monitoring and eruption forecasting, volcano observatories do not have a reliable way of automatically detecting and identifying tremor due to the variable intensities and frequencies at which it occurs. In order to accomplish this, we develop and test a pair of machine learning models that classify spectrograms (i.e., images representing a signal's frequency content over time) from seismic and low‐frequency acoustic data. The models are trained on manually labeled images derived from the recent 2021–2022 eruption of Pavlof Volcano, Alaska, which demonstrated substantial signal diversity (e.g., different tremor types, earthquakes, explosions and noise). Our models achieve 81.2% and 90.0% accuracy on the seismic and low‐frequency acoustic test sets respectively, and perform well when applied to data recorded from past Pavlof Volcano eruptions. In addition, transitions in tremor sequences identified from our analysis generally coincide with shifts in eruptive patterns from Pavlof Volcano. Our tools can help volcano observatories systematically monitor tremor, and advance tremor research by making catalogs of their occurrences more consistent and accessible. We develop a pair of convolutional neural networks that detect and classify volcano seismic and acoustic signalsWe apply our models to Pavlof Volcano eruptions (2007, 2013, 2014, 2016, and 2021–2022) and derive volcano seismoacoustic timelinesThe seismoacoustic timelines reveal shifts in unrest regimes linked to explosions and effusive activity We develop a pair of convolutional neural networks that detect and classify volcano seismic and acoustic signals We apply our models to Pavlof Volcano eruptions (2007, 2013, 2014, 2016, and 2021–2022) and derive volcano seismoacoustic timelines The seismoacoustic timelines reveal shifts in unrest regimes linked to explosions and effusive activity
- Published
- 2024
- Full Text
- View/download PDF