Characterizing Contaminant Source and Meteorological Forcing Using Data Assimilation with a Genetic Algorithm

The release of harmful contaminants is a potentially devastating threat to homeland security. Accurate identification of the source strength and location is essential to minimize the impact. Insufficient spatial and temporal resolution as well as inherent uncertainty in wind field data makes characterizing the source and predicting subsequent transport and dispersion extremely difficult. The solution requires a robust technique such as a genetic algorithm (GA) in order to precisely characterize the source and obtain the required wind information. The method uses a GA to find the combination of source location, source strength, and surface wind direction that best matches the monitored receptor data with the forecast pollutant dispersion model output. The approach is validated with an identical twin experiment that generates the observation data using the same model embedded in the solution method (Daley 1991). Such an experiment allows the validation of the solution methodology in a controlled environment. Researchers have characterized the source of a release using a Bayesian probability density algorithm to determine the mass of material released to within an order of magnitude as well as the location, type, and time of release (Robins, et al. 2005a). Robins, et al. (2005b) use a probabilistic dispersion model to better simulate the small scale effects of a meandering plume. In the current study, we use the Gaussian puff equation as the dispersion model and add wind direction to the list of parameters sought.