LAUSR

Abstract Many CFD studies have investigated the influence of computational parameters on the predicted concentration distribution of pollutants around isolated buildings, but such studies for building arrays are still lacking. This study systematically evaluated the influence of four computational and two physical parameters on pollutant dispersion in building arrays, including turbulence models, grid resolution, discretization of time step size Δ t , length of sampling period, aspect ratio of the arrays, and release rate of tracer gas. Throughout these evaluations, a set of published wind tunnel experimental data was used to validate the CFD models. For concentration simulations, the Large Eddy Simulation (LES) model gave the most accurate results but still had limitations in areas near the source, whereas the Detached Eddy Simulation (DES) and the Reynolds Averaged Navier-Strokes (RANS) RNG k − e models underperformed in some areas. The results of the LES and DES simulations varied with changes in Δ t ∗ and sampling length until Δ t ∗ was less than 0.24 and the sampling length was higher than 2400 Δ t ∗ for LES and 1200 Δ t ∗ for DES. A larger aspect ratio did not necessarily result in a higher concentration field than a smaller ratio. An increase in the tracer gas release rate did not change the general dispersion characteristics, but it still affected the concentration distribution in the areas near the source and resulted in a larger polluted area. The findings of this study are intended to contribute to improvements in the quality of CFD simulations of pollutant dispersion in building arrays.