A Diagnostic Suite of Models for the Evaluation of Oceanic Mesoscale Eddy Parameterizations.

摘要

The practice of modeling geophysical fluid flows has grown tremendously in concert with recent advances in computing power. To study the climate models must simulate centuries of real time, a difficulty made worse by the need to capture fine-scale (eddy) activity. Turbulence at scales ranging from 10 km to 250 km, whose coherent structures are colloquially referred to as mesoscale eddies, is of particular interest because of its ability to transport and mix water masses, and because it dominates the oceanic kinetic energy budget. As of the writing of this dissertation, it also happens to represent the cutting edge in OGCM resolution, hence the need for skillful parameterizations. Calibration and evaluation of such parameterizations is the focus of this work.;An "eddy parameterization challenge suite" is being developed, consisting of a set of high- resolution tracer experiments designed to assist in parameterizing subgridscale processes in ocean models. In each experiment, multiple tracers are initialized in a frontal spindown simulation designed to mimic the stirring effect of mesoscale eddies. Diagnosis of an eddy transport tensor is performed by inverting a matrix of passive tracer gradients, each of which is assumed to satisfy an identical linear flux-gradient relationship. Aspects of the matrix inversion are explored, including the implications of overdetermining the linear relationship using a large number of tracers.;Two sets of simulations, featuring Eady-like and exponential stratification, allow us to investigate scaling laws and vertical structures of the eddy transport tensor. The diagnosed tensor reproduces the horizontal transport of an active tracer (buoyancy) to within ±7% and the vertical transport to within ±12%. The derived scalings are shown to be close in form to the Gent and McWilliams (1990) and Redi (1982) diffusivity tensors, with a magnitude that varies in space and time. The parameterization suite is also used to evaluate an extant scheme (Ferrari et al., 2010) and to recommend improvements. We also attempt a local scaling for the along-isopycnal diffusivity and argue that it is unlikely that any such scaling can be written as a simple function of the velocity, stratification, or eddy variances.