| Summary: | International audience Bayesian inference approach to source localization from infrasound waves typically involves computing marginals of the posterior probability density for source parameters. This is typically carried out with Markov-chain Monte Carlo methods, with various approaches applied to improve efficiency. In these methodologies, infrasound propagation models can be constructed by numerically propagating signals through a set of plausible atmospheric specifications so as to obtain distributions for arrival characteristics (arrival times, durations, amplitudes, etc.). Such an approach, however, drastically increases the number of model runs and for this reason, long-range monitoring of geophysical events is often based either on simplified stochastic propagation models or generative models. Such models, however, do not include current atmospheric specifications knowledge and additional analysis is often necessary to better refine the source location estimate. In this work, we combine the Bayesian framework and non-intrusive generalized Polynomial Chaos (gPC) to update the posterior probability density function describing the source localization. The main difference with the standard Monte Carlo method lies in the fact that the sampling is carried out over the gPC metamodel, which is built from an experimental design of limited size. Using this framework, the maximum posterior solution can be computed using full-wave modeling. This makes such propagation models more efficient than their stochastic counterparts and better suited for real-time monitoring. The performance of the method is demonstrated through reanalysis of a bolide that caused a huge explosion over the Bering sea, near Russia's Kamchatka Peninsula, on December 18th, 2018. It is shown how the prior distribution provided by the International Data Center can be updated, using as many metamodels as there are IMS stations that have presumably recorded the bolide and atmospheric specifications provided by the European centre for medium-range ...
|
|---|