Real-time realizations of the Bayesian Infrasonic Source Localization Method

Friday, 18 December 2015
Poster Hall (Moscone South)
Vladimir Pinsky1, Stephen Arrowsmith2, Abraham Hofstetter1 and Alexandra Nippress3,4, (1)Geophysical Institute of Israel, Lod, Israel, (2)Sandia National Laboratories, Albuquerque, NM, United States, (3)AWE, Blacknest, United Kingdom, (4)Los Alamos National Laboratory, Los Alamos, NM, United States
The Bayesian Infrasonic Source Localization method (BISL), introduced by Mordak et al. (2010) and upgraded by Marcillo et al. (2014) is destined for the accurate estimation of the atmospheric event origin at local, regional and global scales by the seismic and infrasonic networks and arrays. The BISL is based on probabilistic models of the source-station infrasonic signal propagation time, picking time and azimuth estimate merged with a prior knowledge about celerity distribution. It requires at each hypothetical source location, integration of the product of the corresponding source-station likelihood functions multiplied by a prior probability density function of celerity over the multivariate parameter space. The present BISL realization is generally time-consuming procedure based on numerical integration. The computational scheme proposed simplifies the target function so that integrals are taken exactly and are represented via standard functions. This makes the procedure much faster and realizable in real-time without practical loss of accuracy. The procedure executed as PYTHON-FORTRAN code demonstrates high performance on a set of the model and real data.