Bayesian Optimal Experimental Design as a Tool to Quantify the Value of Seismic Waveform Features
Year: 2026
Type: article
Abstract: Higher-frequency waveform simulation and processing could improve seismic monitoring of low-magnitude events at local to regional distances, but it is unclear when the investment is worthwhile. Earth model uncertainty (EU) and background noise create information-theoretic limits on how informative a waveform can be. We introduce a Bayesian experimental design framework to rigorously predict the benefit of incorporating waveform data or extracted waveform features into seismic monitoring for source-parameter inference. Using a synthetic, model-based study, we leverage this framework to answer questions about the value of high frequencies for constraining event location and source parameters. We identify the minimum frequency requirements for inference and a maximum frequency at which there are diminishing returns under different EU and background noise assumptions. Ultimately, this informs how we should invest research and development efforts across Earth model refinement, higher frequency computational simulation, and reducing background noise.
Source: The Seismic Record
DOI: 10.1785/0320250041
Authors: Tommie A. Catanach, Erik L. Ellis
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7151739652