Dynamical Subgrid-Scale Parameterizations from Direct Numerical Simulations

Abstract Dynamical subgrid-scale parameterizations of stochastic backscatter, eddy drain viscosity, and net eddy viscosity have been formulated and calculated for two-dimensional turbulent flows on the sphere based on the statistics of direct numerical simulations (DNSs) with the barotropic vorticity equation. A relatively simple methodology based on a stochastic model representation of the subgrid-scale eddies, but which takes into account the memory effects of turbulent eddies, has been employed. The parameterizations have a cusp behavior at the cutoff wavenumber of the retained scales and have closely similar forms to those based on eddy damped quasi-normal Markovian (EDQNM) and direct interaction approximation (DIA) closure models. Large-eddy simulations (LESs) incorporating DNS-based subgrid-scale parameterizations are found to have kinetic energy spectra that compare closely with the results of higher-resolution DNS at the scales of LES for both isotropic turbulence and Rossby wave turbulence. The methodology presented is general and should be equally applicable to parameterizations of baroclinic processes and convective processes. Applications of the parameterizations to climate models and prediction models are discussed.