Effect of Turbulence Models and Spatial Resolution on Resolved Velocity Structure and Momentum Fluxes in Large-Eddy Simulations of Neutral Boundary Layer Flow

This paper demonstrates the importance of high-quality subfilter-scale turbulence models in large-eddy simulations by evaluating the resolved-scale flow features that result from various closure models. The Advanced Regional Prediction System (ARPS) model was used to simulate neutral flow over a 1.2-km square, flat, rough surface with seven subfilter turbulence models [Smagorinsky, turbulent kinetic energy (TKE)-1.5, and five dynamic reconstruction combinations]. These turbulence models were previously compared with similarity theory. Here, the differences are evaluated using mean velocity statistics and the spatial structure of the flow field. Streamwise velocity averages generally differ among models by less than 0.5 m s(-1), but those differences are often significant at a 95% confidence level. Flow features vary considerably among models. As measured by spatial correlation, resolved flow features grow larger and less elongated with height for a given model and resolution. The largest differences are between dynamic models that allow energy backscatter from small to large scales and the simple eddy-viscosity closures. At low altitudes, the linear extent of Smagorinsky and TKE-1.5 structures exceeds those of dynamic models, but the relationship reverses at higher altitudes. Ejection, sweep, and upward momentum flux features differ among models and from observed neutral atmospheric flows, especially for Smagorinsky and TKE-1.5 coarse-grid simulations. Near-surface isopleths separating upward fluxes from downward are shortest for the Smagorinsky and TKE-1.5 coarse-grid simulations, indicating less convoluted turbulent interfaces; at higher altitudes they are longest. Large-eddy simulation (LES) is a powerful simulation tool, but choices of grid resolution and subfilter model can affect results significantly. Physically realistic dynamic mixed models, such as those presented here, are essential when using LES to study atmospheric processes such as transport and dispersion-in particular at coarse resolutions.