Source Reconstruction of Unknown Model Parameters in Atmospheric Dispersion Using Dynamic Bayesian Inference

A source reconstruction capability driven by time-evolving data is tried out by coupling the observed data and predictive model through dynamic Bayesian inference to obtain solutions to inverse problem. Solutions are determined by posterior probability distributions describing unknown model parameters and the Markov Chain Monte Carlo method with the Metropolis-Hastings sampling algorithm is employed to obtain the solutions. The posterior distributions of model parameters are obtained by performing stochastic sampling by the likelihood function test indicating the agreements between the measurements and the predictions from Gaussian plume model. The Yonggwang atmospheric tracer experiment in Korea is selected to testify the source reconstruction algorithm. The simulation has shown that the posterior modes of the release point and the released source rate for this experiment obviously converged to their true values.