Speaker
Description
Localization and characterization of nuclear detonations can be accomplished using various observation techniques, including seismic and infrasound data. In both cases, accurate knowledge of the propagation medium - specifically, Earth's subsurface and atmospheric properties - is essential for reliable inversion results. This study introduces a joint inference framework designed to localize and characterize events while simultaneously estimating the properties of the wave propagation medium and quantifying associated uncertainties. We apply this methodology to the 2018 meteorite explosion over the Bering Sea, a widely studied event. Our reanalysis leverages a full waveform linear model to address the limitations of empirical models identified in previous studies. To estimate the source and medium parameters, we use a polynomial chaos-based surrogate model to efficiently explore the parameter space. Our results demonstrate that this approach effectively reduces overconfidence in the quantities of interest by providing a robust estimate of uncertainties. In this work, we combine Bayesian inference and recent developments in metamodeling to update the posterior PDF describing the event localization and the associated yield. The difference with the Monte Carlo method lies in the fact that the sampling is carried out over the metamodel, which is built from an experimental design of limited size.
| [email protected] |
