Simulating Seismic Wavefields Using Generative Artificial Intelligence
Description:
Simulating realistic seismic wavefields is crucial for a range of seismic tasks, including acquisition designing, imaging, and inversion. Conventional numerical seismic wave simulators are computationally expensive for large 3D models;, and discrepancies between simulated and observed waveforms arise from wave equation selection and input physical parameters such as the subsurface elastic models and the source parameters. To address these challenges, we adopt a data-driven artificial intelligence (AI) approach, and propose a Conditional Generative Modeling (CGM) framework for seismic wave simulation. The novel CGM framework learns complex 3D wave physics and subsurface heterogeneities from the observed data without relying on explicit physics constraints. As a result, trained CGM-based models act as stochastic wave-propagation operators encoded with a local subsurface model defined by training datasets. Given these models, we can simulate multicomponent seismic data for arbitrary acquisition settings within the area of the observation, using source and receiver geometries and source parameters as input conditional variables. In this study, we develop four models within the CGM framework (CGM-GM-1D/3D, CGM-Wave and CGM-FAS), and demonstrate their performance using two seismic data sets: one small low-density data of natural earthquake waveforms from the San Francisco Bay Area, a region with high seismic risks; and one large high-density data from induced seismicity records of the Geysers geothermal field. The CGM framework reproduces the waveforms, the spectra, and the kinematic features of the real observations, demonstrating the ability to generate waveforms for arbitrary source locations, receiver locations, and source parameters. We address key challenges, including data density, acquisition geometry, scaling and generation variability, and outline future directions for advancing the CGM framework in seismic applications and beyond.
Session: Scientific Machine Learning for Forward and Inverse Wave Equation Problems - I
Type: Oral
Date: 4/15/2025
Presentation Time: 04:45 PM (local time)
Presenting Author: Nori
Student Presenter: No
Invited Presentation:
Poster Number:
Authors
Nori Nakata Presenting Author Corresponding Author nnakata@lbl.gov Lawrence Berkeley National Laboratory |
Rie Nakata rnakata@lbl.gov Lawrence Berkeley National Laboratory |
Pu Ren PRen@lbl.gov Lawrence Berkeley National Laboratory |
Zhengfa Bi zfbi@lbl.gov Lawrence Berkeley National Laboratory |
Maxime Lacour maxlacour@berkeley.edu University of California, Berkeley |
Benjamin Erichson erichson@icsi.berkeley.edu International Computer Science Institute |
Michael Mahoney mmahoney@stat.berkeley.edu University of California, Berkeley |
|
|
Simulating Seismic Wavefields Using Generative Artificial Intelligence
Category
Scientific Machine Learning for Forward and Inverse Wave Equation Problems