Parameter-Efficient Adaptation of Pre-Trained Vision Foundation Models for Active and Passive Seismic Data Denoising
Abstract
The demand for high-resolution subsurface imaging and continuous Earth monitoring has driven rapid growth in active and passive seismic data from dense geophone deployments, distributed acoustic sensing (DAS) arrays, and large-scale 2D and 3D surveys. This expansion makes complex noise suppression increasingly challenging, especially when signal fidelity must be preserved. Conventional supervised deep learning methods are often task-specific, require large paired datasets, and can suffer from domain shift under new acquisition conditions. Foundation models offer a promising alternative, but pre-training seismic foundation models from scratch requires massive domain-specific data and substantial computation. We propose an efficient framework that repurposes general-purpose Vision Foundation Models (VFMs) for geophysical tasks through Parameter-Efficient Fine-Tuning. The architecture uses a pre-trained VFM, a DINOv3 encoder, adapted with Low-Rank Adaptation (LoRA) to enable effective feature adaptation with few additional parameters. To improve robustness under unseen field conditions without ground truth, we introduce a kurtosis-guided unsupervised test-time adaptation module that updates only LoRA parameters during inference. This module self-calibrates the model to site-specific noise by identifying information-rich regions via kurtosis and performing self-training without labeled data. Experiments on public exploration seismic images and DAS vertical seismic profiling data from the Utah FORGE site show that the framework matches or outperforms domain-specific models. Tests on unseen cross-site data from a land survey in China and the Groß Schönebeck geothermal site in Germany further demonstrate strong generalization and effective signal-noise separation. These results highlight the potential of adapting pre-trained VFMs to data-intensive problems in exploration seismology.
Related
- Xaurora: Generative Weather Forecasting with Denoising Stochastic Interpolants from a Foundation Model Prior
- A Review of Deep-learning-based Seismic Data Denoising and Its Promising Paradigm Shift to Foundation Models
- Agentic Large Language Model Workflows for Earth Observation: Tool Orchestration, Foundation-Model Adaptation, and the Evaluation Gap
- ConLoRA-Sat: Parameter-Efficient Synthetic Satellite Imagery Generation for Data-Scarce Humanitarian Contexts
- Deep Learning Denoising of Real SWOT Sea Surface Height Observations
BibTeX
@misc{zhao2026parameterefficient,
title = {{Parameter-Efficient Adaptation of Pre-Trained Vision Foundation Models for Active and Passive Seismic Data Denoising}},
author = {Jiahua Zhao and Umair bin Waheed and Jing Sun and Yang Cui and Nikos Savva and Eric Verschuur},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2605.10953},
eprint = {2605.10953},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2605.10953},
}
RIS
TY - UNPB
TI - Parameter-Efficient Adaptation of Pre-Trained Vision Foundation Models for Active and Passive Seismic Data Denoising
AU - Jiahua Zhao
AU - Umair bin Waheed
AU - Jing Sun
AU - Yang Cui
AU - Nikos Savva
AU - Eric Verschuur
PY - 2026
DA - 2026/09/22
PB - arXiv
DO - 10.48550/arxiv.2605.10953
UR - https://arxiv.org/abs/2605.10953
AB - The demand for high-resolution subsurface imaging and continuous Earth monitoring has driven rapid growth in active and passive seismic data from dense geophone deployments, distributed acoustic sensing (DAS) arrays, and large-scale 2D and 3D surveys. This expansion makes complex noise suppression increasingly challenging, especially when signal fidelity must be preserved. Conventional supervised deep learning methods are often task-specific, require large paired datasets, and can suffer from domain shift under new acquisition conditions. Foundation models offer a promising alternative, but pre-training seismic foundation models from scratch requires massive domain-specific data and substantial computation. We propose an efficient framework that repurposes general-purpose Vision Foundation Models (VFMs) for geophysical tasks through Parameter-Efficient Fine-Tuning. The architecture uses a pre-trained VFM, a DINOv3 encoder, adapted with Low-Rank Adaptation (LoRA) to enable effective feature adaptation with few additional parameters. To improve robustness under unseen field conditions without ground truth, we introduce a kurtosis-guided unsupervised test-time adaptation module that updates only LoRA parameters during inference. This module self-calibrates the model to site-specific noise by identifying information-rich regions via kurtosis and performing self-training without labeled data. Experiments on public exploration seismic images and DAS vertical seismic profiling data from the Utah FORGE site show that the framework matches or outperforms domain-specific models. Tests on unseen cross-site data from a land survey in China and the Groß Schönebeck geothermal site in Germany further demonstrate strong generalization and effective signal-noise separation. These results highlight the potential of adapting pre-trained VFMs to data-intensive problems in exploration seismology.
ER -