The Effect of Turbulence Models on Numerical Prediction of Air Flow within Street Canyons

This study assesses the sensitivity of numerical results of air flow modelling within an urban environment to the six different turbulence models applied in a specific case of a studied staggered cross-road. The principal parameters investigated in this comparison are position of centres of vortices within a street canyon, two components of the spatial velocity vector, turbulent kinetic energy, and dimensionless CPU time effectiveness of turbulence models. Comparison with the experimental results has been made. The results on turbulence modelling suggest the existence of strong three-dimensional coherent structures within the street canyon which is perpendicular to the approaching flow. The standard version of the k-ε model produces too small separation region with unrealistic reattachment on the roof. The RNG model, and the Two-Scale model show much more realistic qualitative results and, despite the dramatic change of velocity gradient, the quantitative results are much less discrepant then the ones obtained using the other modifications of the standard k-ε model. Using the RNG and the Two-scale model the separation locations on the internal windward sides of the square shaped buildings and the separation region on the slanted roofs are predicted correctly. As expected the two-scale performance improvements are paid for in computing time which is doubled compared to the other turbulence models used in this study.