Prof. Francois Septier
(Universite Bretagne Sud, France)
Title
Bayesian inference for pollutant source reconstruction in built-up environments
Abstract
In this talk, a Bayesian inference procedure that estimates the complete distribution of pollutant source term parameters in case of an adverse (accidental or malevolent) atmospheric release will be presented. After introducing the source term estimation (STE) problem in a probabilistic manner, we will describe the proposed Bayesian inference algorithm which uses a Lagrangian Particle Dispersion Model (LPDM) in backward mode to optimize its computational cost as well as its convergence speed. Performances of the proposed approach will be illustrated with a synthetic example, using Retro-SPRAY, the backward implementation of SPRAY, the LPDM of the PMSS suite developed by the the French Alternative Energies and Atomic Energy Commission (CEA).
