Forecasting Seismic Waveforms: A Deep Learning Approach for Einstein Telescope

2025-09-25

AI Summary

To address the real-time Newtonian noise suppression requirement for next-generation gravitational-wave detectors such as the Einstein Telescope, this work proposes SeismoGPT—the first end-to-end autoregressive Transformer model for joint short-term prediction of multi-station, three-component seismic waveforms. The model autonomously captures spatiotemporal dependencies in seismic signals and supports joint modeling of triaxial time-series inputs from both single stations and seismic arrays. Evaluated on real earthquake data, SeismoGPT achieves high-fidelity reconstruction of ground motion features within short prediction horizons (several seconds), with prediction error increasing gradually and consistently with forecast step size—aligning with fundamental predictability limits of dynamical systems. This study establishes the first data-driven, large-model-based framework for modeling multi-component seismic motion patterns, introducing a novel paradigm for real-time noise control in gravitational-wave observations.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

We introduce extit{SeismoGPT}, a transformer-based model for forecasting three-component seismic waveforms in the context of future gravitational wave detectors like the Einstein Telescope. The model is trained in an autoregressive setting and can operate on both single-station and array-based inputs. By learning temporal and spatial dependencies directly from waveform data, SeismoGPT captures realistic ground motion patterns and provides accurate short-term forecasts. Our results show that the model performs well within the immediate prediction window and gradually degrades further ahead, as expected in autoregressive systems. This approach lays the groundwork for data-driven seismic forecasting that could support Newtonian noise mitigation and real-time observatory control.

Problem

• Forecasting seismic waveforms using deep learning
• Modeling temporal and spatial dependencies in data
• Supporting noise mitigation for gravitational wave detectors

Innovation

• Transformer model forecasts seismic waveforms
• Autoregressive training captures spatiotemporal dependencies
• Supports single-station and array-based input configurations