Probabilistic characterization of atmospheric transport and dispersion

Dispersion models are used to assess the possible extent and severity of accidental and terrorist's releases of toxic materials. Most of the current operational dispersion models provide only a characterization of what is observed on average (1st-moment) given the stated conditions. For downwind transport distances near the release location (within 3 to 5 km) the size of the turbulent eddies in the atmosphere are as large or larger than the dispersing plume, and thus make the characterization of the size and location of the dispersing plume highly uncertain. In such situations, the prediction of possible outcomes involves the characterization of the effects of stochastic processes on the transport and diffusion, and it is no longer possible to adequately characterize the impact with a single deterministic prediction. What is needed is a prediction of the distribution of possible outcomes and their respective probabilities of occurrence. The uncertainties in the prediction can be characterized as coming from two primary sources, 1) wind field (trajectory) uncertainties, and 2) model parameterization uncertainties. For this study model parameterization uncertainties included horizontal and vertical dispersion and plume rise. An analytical scheme was developed to characterize the variability of dispersion and plume rise. The algorithms were incorporated in a Lagrangian puff model, INPUFF. The effects of variability in the dispersion and plume rise were simulated using Monte-Carlo methods. The variability in the plume trajectory will be addressed using ensemble Mesoscale meteorological models. The meteorological ensembles are generated from MM5 runs and two versions of the WRF (Weather Research and Forecasting) model. Additional members to the ensemble are generated from the two WRF models with perturbations superimposed on the initial and lateral boundary conditions.