Parameterization of two-dimensional turbulence using an anisotropic maximum entropy production principle

We consider the modeling of the effect of unresolved scales, for two-dimensional and geophysical flows. We first show that the effect of small scales on a coarse-grained field, can be approximated at leading order, by the effect of the strain tensor on the gradient of the vorticity, which exactly conserves the energy. We show that this approximation would lead to unstable numerical code. In order to propose a stable parameterization, while taking into account of these dynamical properties, we apply a maximum entropy production principle. The parameterization acts as a selective diffusion proportional to the mean strain, in the contraction direction, while conserving the energy. We show on numerical computation that the obtained \foreignlanguage{french}{anisotropic relaxation equations} give an important predictability improvement, with respect to Navier-Stokes, Smagorinsky or hyperviscous parameterizations.