Subfilter-scale turbulence modeling for large-eddy simulation of the atmospheric boundary layer over complex terrain

Our ability to accurately predict the evolution and dynamics of the atmospheric boundary layer over steep, mountainous terrain is limited by the sparsity of experimental observations and deficiencies in numerical models. The difficulties with the latter arise primarily from insufficient grid resolution, inadequate turbulence models, and poor representation of the rough lower boundary. In this dissertation, several new steps towards addressing these numerical difficulties are presented for large-eddy simulations (LES) of the atmospheric boundary layer over complex terrain.; First, a new series expansion model based on Taylor series is presented to reconstruct the resolved subfilter-scale (SFS) turbulent stresses. Variations of this series expansion are combined with dynamic eddy-viscosity models for the subgrid-scale stresses to create a dynamic reconstruction model (DRM). The effect of other numerical errors is also addressed in the context of explicit filtering for LES. The DRM yields significant improvements over standard eddy-viscosity closures in simulations of low Reynolds number turbulent channel flow. In particular, the SFS stress representation obtained with increasing reconstruction levels approaches the values predicted by direct numerical simulations.; The DRM is then evaluated for neutral atmospheric boundary layer flow (over flat terrain) at high Reynolds number and with bottom roughness. A near-wall stress model is used to account for the effects of subgrid roughness elements. The agreement of the mean velocity profiles with the expected logarithmic profile predicted by similarity theory shows great improvement with the DRM over traditional turbulence closure methods. The DRM is also validated for full-scale simulations of flow over Askervein Hill, Scotland. Improved comparisons to field data are obtained over other models in this complex flow with intermittent separation in the lee of the hill.; Finally, high-resolution simulations are performed for flow in the Riviera Valley in the Swiss Alps during the Mesoscale Alpine Programme. The steps necessary to achieve accurate simulations with such steep, complex terrain are described. Excellent agreement with field observations is obtained; valley wind transitions and the diurnal temperature variations are well reproduced. The sensitivity to soil moisture, land use data, topographic shading, and turbulence models is also determined.