Data-driven Genetic algorithm in Bayesian estimation of the abrupt atmospheric contamination source

We have applied the methodology combining Bayesian inference with Genetic algorithm (GA) to the problem of the atmospheric contaminant source localization. The algorithms input data are the on-line arriving information about concentration of given substance registered by sensors' network. To achieve rapid-response event reconstructions the fast-running Gaussian plume dispersion model is adopted as the forward model. The proposed GA scan 5-dimensional parameters' space searching for the contaminant source coordinates (x,y), release strength (Q) and atmospheric transport dispersion coefficients. Based on the synthetic experiment data the GA parameters, best suitable for the contamination source localization algorithm performance were identified. We demonstrate that proposed GA configuration can successfully point out the parameters of abrupt contamination source. Results indicate the probability of a source to occur at a particular location with a particular release strength. We propose the termination criteria based on the probabilistic requirements regarding the parameters' value.