Assessment of turbulence model effects on WRF-LES of separated turbulent flows past a 3D hill

Compared to traditional CFD models, weather research and forecasting model (WRF) can more realistically reproduce complex spatio-temporally varying wind fields under extreme weather disasters like typhoon. However, the large-eddy simulation mode of WRF (WRF-LES) to predict engineering-scale turbulence has yet to be clarified in terms of different turbulence models. This study selected a three-dimensional hill as the research object. We focus on separated flow past a 3D hill to systematically revisit the influence of four turbulence models (SMAG, TKE, NBA1, NBA2). The results show that four classical turbulence models under the default conditions can only reproduce the turbulent structure of the post-hill separation to a certain extent and that the nonlinear models (NBA1 and NBA2) simulate more hairpin vortices and small-scale vortex structures than the linear models (SMAG and TKE). Then, the parameter sensitivity is clarified by adjusting key parameters of four classical WRF-LES turbulence models. The results show that the ability of the linear models to simulate the separated flow and small-scale vortex structure is sensitive to the vortex viscosity coefficient. Once the nonlinear models are used, the simulation results are insensitive to the backscatter coefficient variation. •Effects of LES models are revisited on separated flow past a hill in WRF-LES.•Nonlinear models are found better resolving weak flow than linear models.•The separated flow is sensitive to key coefficients of linear models.•The backscatter coefficients are found insensitive to separated flow generation.