J4.1 Data Requirements for Assimilating Concentration Data with a Genetic Algorithm

In the event of a contaminant release, atmospheric transport and dispersion (AT&D) models would be used to predict the path of the contaminant plume. If monitored contaminant concentration data are available, various assimilation techniques can be applied to incorporate the data into the transport and dispersion model, and thus, more accurately predict the plume path. We refer to this as the forward assimilation problem. The AT&D models can also be combined with other techniques to estimate unknown source characteristics or to retrieve meteorological data, the backward assimilation problem. Recovering such data is equally important for AT&D prediction. Both the back-calculation techniques and the forward assimilation methods rely on obtaining sufficient concentration data monitored by either a stationary or a mobile sensor network. In addition, how much data is required is an open question. To be useful, the sensor network must be sited strategically or should be evolvable to follow the plume of contaminant. A second critical issue for AT&D is determining accurate local meteorological data. A relatively small error in wind direction can produce a large error the concentration field since the transport could be in the wrong direction (Peltier et al. 2008). Even when the wind direction is known, local effects can lead to large errors (Krysta et al. 2006). Therefore, our assimilation methods emphasize using field monitored concentration data to infer the correct wind data. Note that there is only one-way coupling between the AT&D ________________________________________