Large-Eddy Simulation of Stably Stratified Winds Over Arbitrary Complex Terrain

Accurate prediction of wind flow over complex terrain at meter-scale resolution is increasingly important for energy, environmental, and atmospheric applications. However, current computational tools often struggle to represent realistic boundary conditions, including arbitrary wind directions, thermal stratification, and heterogeneous surface properties. This dissertation develops a GPU-accelerated, wall-modeled large-eddy simulation (LES) framework tailored for stably stratified flows over complex terrain with arbitrary inflow directions. Key contributions include a turbulent inflow generation method consistently coupled to lateral boundary conditions, and an immersed boundary formulation for enforcing surface heat and momentum fluxes on both scalar and momentum fields. In addition, direct numerical simulations (DNS) of stably stratified open-channel flows and flows over periodic hills are conducted using a high-order spectral element solver to provide detailed reference data for model development and validation. Validation of the immersed boundary method reveals a critical limitation under stably stratified conditions. The cubic relationship governing the friction velocity introduces an ambiguity in selecting a single representative value, and the proposed approach systematically underpredicts friction velocity. This bias leads to discrepancies between the simulated and target friction Reynolds numbers, which in turn affects the agreement of turbulence statistics. Collectively, this dissertation establishes a computational foundation for terrain-resolving atmospheric modeling under stably stratified conditions, and provides high-fidelity datasets and methodological insights that are directly relevant for future efforts aimed at improving predictive capabilities for atmospheric boundary-layer flows over complex terrain.