Cold Diffusion Model for Seismic Denoising

• Published in: Journal of geophysical research. Machine learning and computation Vol. 1; no. 2
• Main Authors: Trappolini, Daniele, Laurenti, Laura, Poggiali, Giulio, Tinti, Elisa, Galasso, Fabio, Michelini, Alberto, Marone, Chris
• Format: Journal Article
• Language: English
• Published: 01.06.2024
• Subjects: cold diffusion; deep learning; denoising; diffusion model; seismology
• ISSN: 2993-5210; 2993-5210
• DOI: 10.1029/2024JH000179

Seismic waves contain information about the earthquake source, the geologic structure they traverse, and many forms of noise. Separating the noise from the earthquake is a difficult task because optimal parameters for filtering noise typically vary with time and, if chosen inappropriately, may strongly alter the original seismic waveform. Diffusion models based on Deep Learning have demonstrated remarkable capabilities in restoring images and audio signals. However, those models assume a Gaussian distribution of noise, which is not the case for typical seismic noise. Motivated by the effectiveness of “cold” diffusion models in speech enhancement, medical anomaly detection, and image restoration, we present a cold variant for seismic data restoration. We describe the first Cold Diffusion Model for Seismic Denoising (CDiffSD), including key design aspects, model architecture, and noise handling. Using metrics to quantify the performance of CDiffSD models compared to previous works, we demonstrate that it provides a new standard in performance. CDiffSD significantly improved the Signal to Noise Ratio by about 18% compared to previous models. It also enhanced Cross‐correlation by 6%, showing a better match between denoised and original signals. Moreover, testing revealed a 50% increase in the recall of P‐wave picks for seismic picking. Our work show that CDiffSD outperforms existing benchmarks, further underscoring its effectiveness in seismic data denoising and analysis. Additionally, the versatility of this model suggests its potential applicability across a range of tasks and domains, such as GNSS, Lab Acoustic Emission, and Distributed Acoustic Sensing data, offering promising avenues for further utilization. Plain Language Summary Seismic waves contain clues about earthquakes and what's beneath the Earth's surface, but any recording of these waves is often mixed with unwanted sounds or disturbances to varying degrees. It's important to filter out these disturbances from the earthquake recordings to improve their clarity and, as a result, make any further analysis more accurate. However, this can be tricky because the nature of these disturbances can change over time, including their amplitude, or by analogy to audio: how loud they are and their pitch of high and low notes. Our work removes noise and thus cleans up recordings to make them more understandable. Recently, advanced computer methods that are good for improving images and sounds have shown promising results. But, these methods usually look for disturbances that follow a certain pattern, which does not always work for more complex disturbances found in earthquake data. To address this, we introduce a strategy called the Cold Diffusion Model for Seismic Denoising. This strategy is tailor‐made to deal with the specific kinds of disturbances found in earthquake data, and it does a better job than previous methods at removing noise and making the earthquake recordings clear again, providing a new standard in this area of study. Key Points Our approach is the first to utilize the Cold Diffusion model with noise recorded from seismic stations Our technique is promising at making sense of earthquake data, even when the background noise is almost as loud as the earthquake signals Using Cold Diffusion Model for Seismic Denoising, we have demonstrated that we surpass benchmark results, such as those achieved by DeepDenoiser