Using a genetic algorithm to assimilate transport and dispersion variables

There is a significant difference between the ensemble average estimates of contaminant transport and dispersion calculated by a model and the turbulent details of any particular realization of the phenomenon. In some cases, however, it is important to be able to compute a best estimate of the likely realization, such as when it is necessary to warn residents of a potentially hazardous situation. If there are concentration sensors in the field, then one can use that data to estimate the current conditions and assimilate those estimates into a model. This approach can enhance the accuracy of the forecast.

This work reports on a genetic algorithm used to optimize the variational problem (GA-Var). Given contaminant sensor measurements and a transport and dispersion model, one can back-calculate unknown source and meteorological parameters. In this case, we demonstrate the dynamic recovery of unknown meteorological variables, including the transport variables that comprise the “outer variability” (wind speed and wind direction) and the dispersion variables that comprise the “inner variability” (contaminant spread). The optimization problem is set up in an Eulerian grid space, where the comparison of the concentration field variable between the predictions and the observations forms the cost function. The dispersion parameters apply to Lagrangian puff space, where one estimates the transport and dispersion parameters.

This calculation is applied to an instantaneous release in a meandering wind field such as that observed during field experiments of a smoke plume. GA-Var proves to be successful at recovering the unknown transport and dispersion parameters.