On the Fidelity of Large-Eddy Simulation of Shallow Precipitating Cumulus Convection

The present study considers the impact of various choices pertaining to the numerical solution of the governing equations on large-eddy simulation (LES) prediction and the association of these choices with flow physics. These include the effect of dissipative versus nondissipative advection discretizations, different implementations of the constant-coefficient Smagorinsky subgrid-scale model, and grid resolution. Simulations corresponding to the trade wind precipitating shallow cumulus composite case of the Rain in Cumulus over the Ocean (RICO) field experiment were carried out. Global boundary layer quantities such as cloud cover, liquid water path, surface precipitation rate, power spectra, and the overall convection structure were used to compare the effects of different discretization implementations. The different discretization implementations were found to exert a significant impact on the LES prediction even for the cases where the process of precipitation was not included. Increasing numerical dissipation decreases cloud cover and surface precipitation rates. For nonprecipitating cases, grid convergence is achieved for grid spacings of 20 m. Cloud cover was found to be particularly sensitive, exhibiting variations between different resolution runs even when the mean liquid water profile had converged.