The previous models required a "hero calculation" to simulate a decade of model time, now similar models run for centuries. However, mesoscale eddy parameterizations are still needed: increasing system complexity, simulation length, ensemble size, and running ocean models on desktop computers require a continuous use of coarse models for decades to come. Ideally, eddy parameterizations should be validated directly against observations, but the density of observations required renders this option impracticable. Tests against higher resolution calculations are an effective alternative, but these comparisons are rarely or incompletely done.
This project will provide such a comparison globally in a direct and thorough manner. Three products will emerge: 1) a global diagnosis of the complete eddy stirring tensor in a fine-resolution simulation, 2) a comparison to the stirring tensor predicted using extant parameterizations, and 3) a parameterization challenge suite based on eddying regions in the fine-resolution model. Most extant eddy parameterizations may be written completely in terms of the eddy stirring tensor. Diagnosis of this tensor represents a significant advance over presently state-of-the-art studies diagnosing only a scalar eddy diffusivity or otherwise limited tensor structure. Agreement with the full tensor globally is a stringent test that will lead to immediate improvement or wholesale rejection of parameterizations. In addition to a dataset for comparison, it is useful to have test cases for parameterization development. Traditionally, eddy parameterizations have been tested in artifact-prone idealized settings. The third product of this study will be a suite of perhaps 6 challenges constructed from prototypes in the global high-resolution simulation. A challenge will provide enough information to be run as a stand-alone forward model. The full suite of challenges will be designed so that a successful parameterization in all challenges will predict likely success in a global model. Much progress has been made in parameterizing mesoscale eddies in coarse ocean models, but the next level of accuracy requires better defined metrics in eddy-rich datasets and test suites against which extant and forthcoming parameterizations can be validated. This study will provide direct products accessible with moderate computer resources, and the global nature of the diagnosis in a realistic circulation will prevent common artifacts.
The project will increase understanding of mesoscale eddy dynamics on a global scale and the assumptions underlying eddy parameterizations, improve or reject extant eddy parameterization schemes, and test a novel diagnostic methodology. The project will support two early-career researchers and a graduate student, publicly provide manageably-sized versions of a huge simulation, lead to improvements in ocean and climate models, lead to performance improvement of the POP model, and support and encourage collaboration among research institutions and modelers with the larger community.
