Bayesian-based approach to application of the genetic algorithm to localize the abrupt atmospheric contamination source

Abstract

We apply the Bayesian inference in combination with the Genetic algorithm (GA) to the problem of the atmospheric contaminant source localization. The algorithm input data are the on-line incoming concentrations of released substance registered by sensors network. The proposed reconstruction algorithm is firstly adjusted and tested based on the data from the synthetic experiment. The proposed GA scan 5-dimensional parameters space searching for the contaminant source coordinates (x,y), release strength (Q) and the atmospheric transport dispersion coefficients. Based on the performed tests the most efficient GA configuration is identified. To speed up the algorithm the dynamical termination criteria, founded on the probabilistic requirements regarding the searched parameters value, is proposed. Then, we apply developed algorithm to localize the release source utilizing data from the field tracer experiment conducted in May 2001 at the Kori nuclear site. We demonstrate successful localization of the continuous contamination source in very complicated hilly terrain surrounding the Kori nuclear site. Results indicate the probability of a source to occur at a particular location with a particular release strength.