A High-Order Multi-Moment Finite Volume Model for Fully Compressible Large-Eddy Simulation of Atmospheric Boundary Layer Turbulence

Accurately representing atmospheric boundary layer turbulence is essential for advancing numerical weather prediction. This study presents the development and validation of a novel, fully compressible large-eddy simulation (LES) model—the multi-moment constrained finite volume LES (MCV_LES). Built on a high-order multi-moment constrained finite volume (MCV) dynamical core, the model excels in accurately simulating nonhydrostatic flows. It utilizes a Favre-filter and a three-dimensional (3D) Smagorinsky–Lilly closure for subgrid-scale (SGS) turbulence. A key innovation is its hybrid Riemann solver strategy, meticulously designed to minimize numerical dissipation: the low-dissipation Roe solver preserves turbulent structures in the horizontal, while the AUSM+-up (Advection Upstream Splitting Method, u representing velocity and p being pressure) scheme ensures stability for atmospheric flows in the vertical. This configuration allows the modeled turbulence physics to dominate. The model’s performance is not only rigorously validated against established references (UCLA_LES, ICON_LES, PALM) in an idealized dry convective atmospheric boundary layer simulation, but also tested in an idealized dry neutral ABL simulation with reliable results obtained. The MCV_LES successfully reproduces key turbulent statistics—including mean potential temperature profiles, heat flux distributions, vertical velocity variance, and energy spectra—with comparable fidelity. Sensitivity analysis identifies the Smagorinsky coefficient ( C_s ) as critical; C_s=0.1 delivers optimal performance by maintaining the inertial sub-range and realistic turbulence intensity, whereas C_s=0.2 introduces excessive dissipation. This work lays a foundation for future extensions, such as incorporating moist physics and real-terrain applications.