A series of simulations were carried out to test the accuracy of a CFD (Computational Fluid Dynamics) model for flow and dispersion problems around buildings. The basic equations involved are Reynolds-averaged Navier-Stokes equations. Two different cases were selected to estimate the accuracy of a CFD model. Case 1 adopted Euler equations, which are obtained by neglecting the viscous fluxes, which can be closed by the $textsc{k}$-$varepsilon$model for a turbulent close problem. The results of both cases were compared with wind tunnel data. The results for Case 2 were closer to the wind both cases were compared with wind tunnel data. The results for Case 2 were closer to the wind tunnel data than Case 1. Accordingly, this indicates that the inclusion of viscous fluxes in a CFD model is required for the simulation of flow and 야spersion around buildings.