Source identification for unsteady atmospheric dispersion of hazardous materials using Markov Chain Monte Carlo method

Abstract

This paper is to study source inversion and identification of hazardous gas dispersion in a three-dimensional urban area. An unsteady adjoint transportation model was adopted, and an advanced numerical scheme based on adaptive mesh refinement was used. A time-dependent concentration database over the entire parameter space was generated. Markov Chain Monte Carlo sampling based on the Bayesian inference was used to invert the parameters such as source location and its strength obtained from the database at different sampling time of sensor readings and simulation results. The probability distributions of source parameters were calculated and predicted source location and strength agree well with actual values which indicates the feasibility of the proposed method and procedure. Numerical studies also show that the computational scheme is efficient when using unsteady adjoint transportation equation with MCMC methods. The unsteady inversion method proposed here can also improve the accuracy of source location in the wind direction compared with the steady inversion method for the case of atmospheric release of hazardous materials.