Meso-scale meteorological models, both diagnostic and prognostic, are widely used to provide meteorological inputs to atmospheric dispersion models. Traditionally, a deterministic approach is applied with a best estimation of the meteorological fields provided. This is not sufficient for uncertainty study and sensitivity study of the dispersion models, which are important for decision makers to take meaningful actions using the results of the dispersion models.

Uncertainty study and sensitivity analysis, have been popular in the atmospheric dispersion modeling and other environmental modeling studies, and they are crucial for the application of the results of these modeling studies. However, uncertainty study and sensitively analysis of meso-scale numerical weather prediction models were not performed systematically in a manner similar to those carried out in environment modeling. This gap renders the advancement in uncertainty study and sensitivity study for atmospheric dispersion modeling.

Fundamental concepts in uncertainty study and sensitivity analysis are reviewed and applied to works on uncertainty study of the meteorological models, especially meso-scale numerical prediction models. In particular, the ensemble forecast approach used in the weather forecast community and uncertainty study in environmental modeling and policy analysis are compared. Using the Advanced Regional Prediction System (ARPS) and other widely used meso-scale numerical weather prediction models, uncertainty in the model output is estimated using Monte-Carlo simulation and ensemble forecasting techniques, and the preliminary results with horizontal resolution of 3km or less will be reported in this presentation.