SeisLM: a Foundation Model for Seismic Waveforms
This addresses the need for efficient seismic analysis in seismology, though it is incremental as it adapts existing language modeling techniques to a new domain.
The researchers tackled the problem of analyzing seismic waveforms by introducing SeisLM, a foundation model pretrained on large unlabeled datasets using self-supervised contrastive loss, which achieved strong performance in tasks like event detection and phase-picking when fine-tuned.
We introduce the Seismic Language Model (SeisLM), a foundational model designed to analyze seismic waveforms -- signals generated by Earth's vibrations such as the ones originating from earthquakes. SeisLM is pretrained on a large collection of open-source seismic datasets using a self-supervised contrastive loss, akin to BERT in language modeling. This approach allows the model to learn general seismic waveform patterns from unlabeled data without being tied to specific downstream tasks. When fine-tuned, SeisLM excels in seismological tasks like event detection, phase-picking, onset time regression, and foreshock-aftershock classification. The code has been made publicly available on https://github.com/liutianlin0121/seisLM.