Forecasting Surface Velocity Fields Associated With Laboratory Seismic Cycles Using Deep Learning

Abstract

Over the last few years, Machine Learning (ML) has been used for predicting laboratory earthquakes in diverse experimental settings (Bergen et al., 2019;Ren et al., 2020). The majority of these studies have used ML to decrypt the acoustic signals emitted by a laboratory fault analog in double direct shear experiments (e.g.

Recent advances in earthquake seismology using machine learning

“…Inspired by the double direct shear experiment, ML predictions have also been attempted in an analog experiment simulating a subduction zone (Rosenau et al 2017). In these studies, issues of predicting the time to macroscopic slip (Corbi et al 2019), determining whether slip will occur immediately after the analyzed data window (Corbi et al 2020), and predicting future surface velocity fields (Mastella et al 2022), were addressed based on surface deformation data, by assuming the future application to the actual GNSS data.…”