Simulations of the atmospheric boundary layer using subfilter-scale reconstruction and the dynamic Wong-Lilly SGS model

Large-eddy simulations (LES) of the atmospheric boundary layer (ABL) can be particularly sensitive to near-surface parameterization of turbulent motions. The success of dynamic turbulence closure models in capturing flow behavior in the near-wall region in smooth-walled turbulent flows motivated the application of such models to the ABL. In the ABL, however, the added difficulty of surface roughness has stymied many previous attempts to use the popular dynamic Smagorinsky model (DSM) for atmospheric flows. ABL simulations using the DSM typically do not give good near-wall profile results. This is partially due to the sensitivity of the model to calculations of shear stresses in the under-resolved regions at the wall, resulting in large kinks in the stress profiles. We therefore adopted the dynamic Wong-Lilly model (DWL) (Wong & Lilly 1994) which limits the need to calculate the strain rate to determine the dynamic eddy viscosity and thus exhibits smoother profiles in the near-wall region.

The DWL was used in the context of our scale-partitioning turbulence modeling approach which separates resolved subfilter-scale (RSFS) and subgrid-scale (SGS) motions in the velocity field. The former can be calculated using reconstruction models (series expansions) for the unfiltered velocity, while the latter are represented using an eddy viscosity model, in this case the DWL. A near-wall stress model is applied to represent the effects of surface roughness. Results are presented showing the combined effects of these pieces of the dynamic reconstruction model (DRM) for LES of neutral boundary layer flow over flat and complex terrain. The DRM is able to capture the near-wall logarithmic velocity profile expected from similarity theory for neutral boundary layer flow over flat terrain. Similarly, simulations of flow over Askervein Hill show that use of the DRM gives the expected wind speed in the lee of the hill, a significant improvement over results using standard eddy viscosity models for this flow.