Extended Abstract (676K)J1.1
Assimilating Concentration Data into Dispersion Models with a Genetic Algorithm
Sue Ellen Haupt, Penn State Univ., University Park, PA; and K. J. Long, A. Beyer-Lout, and G. S. Young
There is a need to assimilate concentration observations into atmospheric transport and dispersion systems in an accurate and efficient manner. Few current efforts focus on assimilating the concentration data into the wind forecasting system. The difficulty with this process is the one-way coupling inherent in the systems: the wind model determines the transport and dispersion of contaminant but the concentration model does not directly influence the wind model. Therefore, the goal is to use concentration observations to infer the wind in the absence of direct wind measurements. This study demonstrates assimilation in one such system by assimilating a field of concentration data into a coupled fluid/dispersion model: a multi-puff dispersion system embedded in a shallow water model. Our assimilation method uses a genetic algorithm to assimilate the entire field of concentration observations and optimize the agreement between predicted and observed concentration fields. This GA assimilation method is tested in the context of an identical twin experiment in which the same model used to construct the synthetic data is used in the iterative solution. Sensitivity studies define the robustness of the assimilation system.
Joint Session 1, Joint Session between the 7th Conference on Artificial Intelligence and the Meteorological Aspects of Air Pollution Committee—I
Monday, 12 January 2009, 10:45 AM-11:45 AM, Room 125A
Next paper
Recorded presentation