| Description |
The present work focuses on developing a holistic understanding of flow and dispersion in urban environments. Toward this end, ideas are drawn from the fields of physical modeling, inverse modeling, and optimization in urban fluid dynamics. The physical modeling part of the dissertation investigates flow in the vicinity of tall buildings using wind tunnel two-dimensional particle image velocimetry (PIV) measurements. The data obtained have been used to evaluate and improve urban wind and dispersion models. In the inverse modeling part of the dissertation, an event reconstruction tool is developed to quickly and accurately characterize the source parameters of chemical / biological / radiological (CBR) agents released into the atmosphere in an urban domain. Event reconstruction is performed using concentration measurements obtained from a distributed sensor network in the city, where the spatial coordinates of the sensors are known a priori. Source characterization comprises retrieving several source parameters including the spatial coordinates of the source, the source strength, the wind speed, and wind direction at the source, etc. The Gaussian plume model is adopted as the forward model, and derivative-based optimization is chosen to take advantage of its simple analytical nature. The solution technique developed is independent of the forward model used and is comprised of stochastic search with regularized gradient optimization. The final part of the dissertation is comprised of urban form optimization. The problem of identification of urban forms that result in the best environmental conditions is referred to as the urban form optimization problem (UFOP). The decision variables optimized include the spatial locations and the physical dimensions of the buildings and the wind speed and wind direction over the domain of interest. For the UFOP, the quick urban and industrial complex (QUIC) dispersion model is used as the forward model. The UFOP is cast as a single optimization problem, and simulated annealing and genetic algorithms are used in the solution procedure. |
