Wall-resolved large-eddy simulation (LES) is assessed as a computationally efficient alternative to direct numerical simulation (DNS) for turbulent katabatic flows. Simulations using two variants of the Smagorinsky model are performed at varying horizontal resolutions and compared with DNS for the canonical Prandtl slope-flow configuration over a \(60^\circ \) slope at \(Gr = 10^{14}\) . A systematic sweep over the horizontal grid spacings reveals that the wall shear stress is robustly captured, while the errors in the surface buoyancy flux and the low-level jet velocity are controlled primarily by the horizontal grid aspect ratio, \(\textrm{AR}\) , rather than by the overall horizontal coarseness. These errors are traced to the \(\textrm{AR}\) -dependence of the resolved buoyancy flux, which modifies near-surface buoyancy mixing and consequently the wall buoyancy gradient and jet-driving buoyancy forcing. Applying a mean-shear correction to the static Smagorinsky model reduces the magnitude of these errors but does not eliminate the underlying \(\textrm{AR}\) trend. Horizontal grid coarsening is also found to alter turbulent kinetic energy production near the surface and above the jet, weaken pressure–strain redistribution among velocity components, and damp marginally resolved scales. When these scales are energetically significant, the spatial coherence of the flow is reduced and the energy cascade is weakened. Based on these findings, we recommend limiting the horizontal grid aspect ratio to \(\textrm{AR}\le 2\) in wall-resolved LES of katabatic flows to control errors in surface buoyancy fluxes and low-level jet characteristics. Overall, this study provides mechanistic insight into the link between horizontal grid resolution, subgrid terms, and turbulence organization, and demonstrates that wall-resolved LES is a viable, low-cost alternative to DNS for fundamental katabatic flows, even when using simple subgrid-scale closures.
