Additional flux arising from unresolved scales in eddying ocean models

Abstract

An analysis is presented of snapshot data (eastward and northward velocity components: u and v; tracer such as potential temperature: τ) from an eddy-resolving (Rgrid: 1/12°) ocean model experiment, in order to explore a method for improving eddy-permitting model performance. Horizontal 3 × 3 R-grid averages give the eddy-permitting grid (P-grid: 1/4°) variables: 〈u〉, 〈v〉, and 〈τ〉, where 〈〉 denotes the spatial P-grid scale average. The difference between the horizontal tracer flux across the boundary face of a P-grid and that across the corresponding faces of R-grids is estimated as F2E. It is found that the correlations among the gradients of u, v, and τ give a good approximation F2C to the estimated flux F2E. The approximated flux is a function of these gradients and the grid size. A method is presented for implementing the F2C for density to an eddying ocean model as an additional advection. Practical experiments were conducted with a realistic configuration. It is shown that the zonal mean isotherms in the Kuroshio extension region are more flattened in the run using the proposed method than in another run using the conventional horizontal biharmonic operator, suggesting that the additional flux correction leads to an enhancement of sub-basin scale mixing.