← Back to Sessions

Quantifying the Impact of Modeling Uncertainty on the Performance of Waveform-Based Bayesian Inference for Seismic Monitoring

Description: 

To detect and characterize small magnitude seismic events at local distances, monitoring networks are increasing looking to algorithms that extract more information from observed waveforms. Regardless of the processing method, e.g., whether using machine learning, Bayesian inversion, or some other algorithm, different sources of uncertainty imply information theoretic limits on how useful waveform-based methods can be. This leads to the foundational question of how much information the waveform contains about the underlying seismic event given various assumptions such as observed frequency ranges, earth model uncertainties, source uncertainties, and background noise processes. By understanding the utility of processing full-waveform data, we can then better understand whether the added data gathering, modeling, and computational costs for processing this type of data is worth it for a given application.

In this research, we specifically investigate two sources of uncertainty. First, we explore the impact of Earth model uncertainty by adding different stochastic perturbations to the AK135 model, simulating the associated waveforms, and finally then quantifying the resulting uncertainty on the waveforms. Secondly, we explore the impact of the background noise process by adding noise models from the literature with different magnitudes and frequency characteristics. Taking these sources of uncertainty, we use tools from Bayesian inference and experimental design to quantify how tuning each of these uncertainty models impacts how informative processing the full-waveform would be for seismic monitoring.

This research was funded by the National Nuclear Security Administration, Defense Nuclear Nonproliferation Research and Development (NNSA DNN R&D). The authors acknowledge important interdisciplinary collaboration with scientists and engineers from LANL, LLNL, MSTS, PNNL, and SNL. SNL is managed and operated by NTESS under DOE NNSA contract DE-NA0003525.

Session: Exploiting Explosion Sources: Advancements in Seismic Source Physics [Poster]

Type: Poster

Date: 4/19/2023

Presentation Time: 08:00 AM (local time)

Presenting Author: Tommie A. Catanach

Student Presenter: No

Invited Presentation: 

Authors