Explicit filtering and reconstruction turbulence modeling for large-eddy simulation of neutral boundary layer flow

Standard turbulence closures for large-eddy simulations of atmospheric flow based on finite-difference or finite-volume codes use eddy-viscosity models and hence ignore the contribution of the resolved subfilter-scale stresses. These eddy-viscosity closures are unable to produce the expected logarithmic region near the surface in neutral boundary layer flows. Here, explicit filtering and reconstruction are used to improve the representation of the resolvable subfilter-scale (RSFS) stresses, and a dynamic eddy-viscosity model is used for the subgrid-scale (SGS) stresses. Combining reconstruction and eddy-viscosity models yields a sophisticated (and higher order) version of the well-known mixed model of Bardina et al.; the explicit filtering and reconstruction procedures clearly delineate the contribution of the RSFS and SGS motions. A near-wall stress model is implemented to supplement the turbulence models and account for the stress induced by filtering near a solid boundary as well as the effect of the large grid aspect ratio. Results for neutral boundary layer flow over a rough wall using the combined dynamic reconstruction model and the near-wall stress model show excellent agreement with similarity theory logarithmic velocity profiles, a significant improvement over standard eddy-viscosity closures. Stress profiles also exhibit the expected pattern with increased reconstruction level.