Numerical predictions on the wake interference flow in two-dimensional street canyon based on various RANS turbulence closure models

The precise numerical prediction of urban flow patterns is crucial for evaluating ventilation performance, pollution dispersion, and pedestrian comfort in densely built environments. Among these types of patterns, the wake interference flow poses a distinct modeling difficulty due to its complex vortex dynamics. This study performed a series of steady Reynolds-Averaged Navier-Stokes (RANS) simulations to evaluate the predictive efficiency of five turbulence closure models: standard k-ε (STD), renormalisation group k-ε (RNG), realizable k-ε (RLZ), shear-stress transport k-ω (SST), and Reynolds stress model (RSM) in a two-dimensional (2D) idealized street canyon with an aspect ratio of 3 within the wake interference flow regime. The predicted results were compared with wind tunnel experimental data using velocity profiles, statistical validation metrics, and streamlines visualization. The results demonstrate that the quantitative assessment utilizing the factor of two observations (FAC2) distinctly revealed a satisfactory predicted of streamwise velocity within the street canyon, topped by RNG (0.92) and followed by STD (0.91), RSM (0.90), SST (0.89), and RLZ (0.88). Nevertheless, all models inadequately predicted the vertical velocity, as the FAC2 values fell below the threshold of 0.5. The qualitative assessment of the velocity streamlines indicates that the RNG and STD predictions closely resembled the flow pattern obtained from the experimental results that determine the main characteristics of the wake interference flow regime. Other models exhibited inadequate performance due to the observation of completely different flow patterns. Consequently, it can be concluded that while all models can estimate the streamwise velocity in the wake interference regime with good accuracy, substantial constraints persist in predicting the in-canyon vertical velocity. The observed limitations, together with the apparent variation between models in replicating secondary vortex formations, suggest several avenues for future investigations.