Large-eddy simulation (LES) has emerged as a powerful tool to study geophysical turbulence over the past few decades. In principle, LES resolves explicitly the large eddies which account for most of the energy and models the smaller, unresolved eddies analytically through a 'subgrid model.' When the grid resolution is much finer than the length scales of the large eddies, LES yields reliable results with negligible sensitivity to the subgrid model, a desirable feature. When the grid resolution is comparable to the length scales of the large eddies, however, the results can be sensitive to the subgrid model as the subgrid scales account for a significant fraction of the total fluxes.

In this talk, I will discuss the subgrid conservation equations and their utility in developing subgrid models when the grid resolution is of the order the integral length scale of turbulence. The subgrid conservation equations describe the evolution of subgrid-scale variables from first principles and hence, have the potential to offer useful insights into the balance between different mechanisms that generate and destroy subgrid fluxes. Using LES, I will also evaluate the performance of a subgrid model that solves for the subgrid fluxes prognostically using a truncated version of their conservation equations.