Using a statistical model for the description of uncertainties associated with dispersal models

A quick and accurate prediction of the dispersion of the contaminated material is crucial in case of environmental disasters (Nuclear or chemical accidents). Conventional atmospheric dispersion models (physical models) are widely used for forecasting toxic contamination and obtaining results in real-time with varying degrees of accuracy. These models are deterministic, and one of the most signi cant problems associated with their use in prediction is the large degree of uncertainty inherent in their predictions. The objective of this work is to present a Bayesian model Smith and French [1993] which embeds a dispersal model in a description of the uncertainties associated with the dispersal model. This both allows the assimilation of data to update current forecasts and also expresses an appropriate degree of uncertainty associated with any forecasts or estimates.