Localizing of the atmospheric contamination source based on the Kori field tracer experiment data

Accidental releases of hazardous material into the atmosphere pose high risks to human health and the environment. Thus it would be valuable to develop an emergency reaction system which can recognize the probable location of the source based only on concentrations of the released substance as reported by a network of sensors. We apply a methodology combining Bayesian inference with Sequential Monte Carlo (SMC) methods to the problem of locating the source of an atmospheric contaminant. The input data for this algorithm are the concentrations of a given substance gathered continuously in time. We employ this algorithm to locating a contamination source using data from a field tracer experiment covering the Kori nuclear site and conducted in May 2001. We use the second-order Closure Integrated PUFF Model (SCIPUFF) of atmospheric dispersion as the forward model to predict concentrations at the sensors’ locations. We demonstrate that the source of continuous contamination may be successfully located even in the very complicated, hilly terrain surrounding the Kori nuclear site.