Poster #181, Seismology

Initial Tests of a Bayesian Framework for Directivity and Source Spectral Analysis: Application to the July 2019 Ridgecrest Earthquake Sequence

Daniel T. Trugman
Poster Image: 

Poster Presentation

2021 SCEC Annual Meeting, Poster #181, SCEC Contribution #11109 VIEW PDF
Earthquake source spectra can provide crucial observational constraints for understanding the physics of earthquake rupture. While there exist a number of approaches to extract key earthquake source parameters like corner frequency and stress drop from waveform spectra, the most widely-used workflows are based on simple optimization procedures that do not accurately characterize the inherent data and modeling uncertainties. Perhaps because of this, source parameter estimates exhibit a puzzling degree of scatter, and results from different methods often disagree even when using the same underlying datasets. Here we develop a Bayesian framework for source spectral analysis that allows for the ...encoding of prior physical knowledge to guide inversions while solving for the full posterior probability distribution that fully quantifies parameter tradeoffs and uncertainties. In these inversions, input data include spectral ratios between a target event and one or more nearby empirical Green’s function (EGF) events, and the Bayesian inversion solves for corner frequencies and spectral falloff rates for all events, as well as a directivity metrics for the target event. Initial application of this computationally intensive technique to large (M5 and greater events) during the July 2019 Ridgecrest earthquake sequence appears promising, with physically viable priors enabling stable corner frequency estimates for EGFs and target events alike. We anticipate that the work performed here will form the basis of a future, larger study applying the method at scale across California and beyond, likely with the help of high-performance computing resources.