Adaptive Bayesian Algorithms for the Estimation of Source Term in a Complex Atmospheric Release

In this paper, we present an adaptive algorithm for the estimation of source parameters when a release of pollutant in the atmosphere is observed by a sensor network in complex flow field. Due to the error-based observations, inverse statistical methods have to be used to perform an estimation of the parameters (position of the source, time and mass of the release) of interest. However, given the complexity of the dispersion model, even with a Gaussian assumption on the sensor-based errors, direct inversion cannot be done. In order to have quick results, classical MCMC, while accurate, is too slow. We then demonstrate the accuracy of using adaptive techniques such as the AMIS (Population Monte-Carlo based). We finally compare the results with the classical MCMC estimation in term of accuracy and velocity of implementation.