Advanced Monte Carlo Method for model uncertainty propagation in risk assessment

In this paper, an Advanced Monte Carlo Method based on interval analysis approach and Monte Carlo simulation is proposed in order to propagate uncertainties in an atmospheric dispersion model. The purpose is to compute with accuracy the geographical region in which the concentration of the considered toxic gas is less than the threshold of irreversible effects. The problem of uncertainty propagation is tackled in order to assess the risk at the event of an accident, which may have an important impact on population. The estimation of gas concentration is based on an effect model associated with the studied dangerous phenomenon where some model inputs are known with imprecision. The principle of the proposed method is to generate random interval supports of model inputs instead of random values in order to increase accuracy and reduce the sampling size. The Advanced Monte Carlo Method is applied and compared for estimating uncertainty on the computed region with the classical Monte Carlo simulation.