SciPost Submission Page
Towards a Seismology Foundation Model
by Waleed Esmail, Alexander Kappes, Stuart Russell, Christine Thomas
Submission summary
| Authors (as registered SciPost users): | Waleed Esmail |
| Submission information | |
|---|---|
| Preprint Link: | scipost_202509_00067v1 (pdf) |
| Date submitted: | Sept. 30, 2025, 9:34 p.m. |
| Submitted by: | Waleed Esmail |
| Submitted to: | SciPost Physics Proceedings |
| Proceedings issue: | The 2nd European AI for Fundamental Physics Conference (EuCAIFCon2025) |
| Ontological classification | |
|---|---|
| Academic field: | Physics |
| Specialties: |
|
| Approach: | Computational |
The author(s) disclose that the following generative AI tools have been used in the preparation of this submission:
OpenAI ChatGPT was used only for paraphrasing, sentence reformulation, and improving readability. No scientific results or data were generated with AI tools.
Abstract
In this article we introduce SeismoGPT, a step toward a foundation model for seismology, developed to forecast three-component seismic waveforms with direct application to future gravitational wave detectors such as the Einstein Telescope. SeismoGPT learns to predict the next token in an autoregressive framework. It can be applied to both single-station and multi-station seismic array, learning both temporal and spatial dependencies directly from raw waveform. Beyond immediate forecasting, SeismoGPT represents a step toward a general-purpose framework for seismology, one that could support Newtonian-noise mitigation, real-time observatory control, and ultimately broader seismic monitoring and prediction tasks.