Abstract
Wind and dispersion in a complex urban environment have been numerically simulated with large-eddy simulation (LES) and Reynolds-averaged Navier-Stokes (RANS) and compared to wind-tunnel measurements. This paper presents a systematic comparison of LES and RANS across mesh resolutions and turbulent Schmidt numbers, linking dispersion predictions to momentum and scalar flux modelling. Model performance was quantitatively assessed using several statistical metrics, including ‘factor-of-two’ (FAC2).
The velocity magnitude showed excellent agreement (FAC2 0.94) between numerical and experimental data for both RANS and LES, regardless of the adopted mesh. For the turbulence kinetic energy (TKE), excellent agreement was found for the LES (FAC2 0.99), whereas RANS systematically underpredicted TKE (FAC2 0.90).
Scalar concentration predictions were satisfactory for the LES (FAC2 0.85) but significantly poorer for RANS (FAC2 0.54). Notably, RANS required substantially higher mesh refinements to obtain mesh convergence, likely due to large scalar gradients near the point-like release source. Moreover, the RANS agreement with experimental data deteriorated with increasing resolution.
Turbulent momentum and scalar (FAC2 0.53) fluxes computed from LES agreed well with the measurements. In contrast, the Boussinesq eddy-viscosity model and the gradient-diffusion hypothesis used in the RANS framework estimated less accurate momentum fluxes and very poor scalar fluxes (FAC2 0.30), respectively. In many cases, RANS predicted turbulent scalar flux in the opposing direction to the experimental and LES results. The results indicate that inaccuracies in RANS concentration predictions are dominated by limitations in scalar-flux closure assumptions, which highlights a RANS modelling limitation.
•Mesh-convergence can require significantly higher resolutions for RANS than LES.•LES dispersion predictions are better than RANS for all (from 0.2 to 1.3).•Poor RANS representation of turbulent scalar fluxes is the likely cause.• is recommended for complex urban environments.