Seismic wave field prediction using encoder-decoder networks: from learning transfer functions to Virtual Seismic Arrays

Abstract

The prediction of seismic wave fields between stations using machine learning offers great potential for geophysical monitoring, particularly in remote areas or in regions with sparse sensor coverage. This thesis introduces a novel encoder-decoder deep learning architecture that successfully learns the transfer function between seismic stations. By learning the complex signal transformations, this method enables accurate predictions of how seismic signals alter as they travel from one station to another. Notably, high quality predictions are achieved using only two days of data consisting solely of ambient seismic noise. The method’s robustness in a range of scenarios is demonstrated via validation at a seismic exploration site with a variety of noise sources. The network shows particular strength in capturing phase-related features, which is crucial to its performance in seismic wave prediction. A systematic parameter study reveals important insight about the variables affecting model performance and points out areas for future development. Virtual Seismic Arrays are introduced as a powerful proof of concept, extending the approach from individual station pairs to entire seismic arrays. By training the algorithm on all station pairs within an array, a set of predictive models is obtained that collectively form the Virtual Seismic Array. This enables the reconstruction of full-array recordings from a single reference station, even after physical sensors are no longer present. In the secondary microseism frequency band, beamforming analysis validates the effectiveness of Virtual Seismic Arrays by showing a high degree of agreement between the original and predicted waveforms. This novel application of encoder-decoder networks for modelling transfer functions has the potential to enhance seismic monitoring, while reducing the need for continuous sensor coverage. By reconstructing signals at multiple stations from a single reference station, the approach enables ongoing array functionality in remote regions while reducing costs and maintaining array capabilities. These improvements are beneficial in industries like advanced seismic instrumentation and ultra-precision manufacturing where even small vibrations have significant impact on results. This is particularly beneficial in projects like the Einstein telescope, where the sensitivity of gravitational wave detections depends on reducing seismic disturbances.