Can Large Eddy Simulation Techniques Improve Mesoscale Rich Ocean Models?

Large-eddy simulations (LES) differ from other fluid flow computations in that the largest eddies are explicitly resolved and the smaller eddies are modeled (as engineers call it) or parameterized (as oceanographers call it). Modeling the ocean and atmosphere inspired the first large-eddy simulations, but true large-eddy simulations of the ocean and atmosphere are surprisingly rare. Ocean models almost always have eddy parameterizations, be they just larger than natural values of constant diffusivities and viscosities, perhaps oriented along isopycnal surfaces [Redi, 1982], or more sophisticated models such as the eddy extraction of potential energy by along-isopycnal bolus flux model of Gent and McWilliams [1990], perhaps with nonlinear transfer coefficient scaling [Visbeck et al., 1997; Held and Larichev, 1996], the “Neptune” effect parameterizing eddy-topography interaction [Holloway, 1993], or parameterizations based on linear instability analysis [Stone, 1972; Branscome, 1983; Killworth, 2005]. All of these parameterizations are designed for situations where eddies are not present or weak, and, as a result, the model resolution does not appear explicitly. The Herculean—perhaps Sisyphean—task of parameterizing the important effects of eddies in ocean models is done by offline theoretical analysis in lieu of resolved eddies. It is no wonder that there is enormous model sensitivity to the parameterization choices made [e.g., Steiner et al., 2004]. Leaving all to theory seems very unlikely to succeed, as eddies are sensitive to difficult to quantify parameters, such as interaction with unresolved topography [Holloway, 1993], regional variations of hydrodynamic instability [Killworth, 2005], bottom drag [Arbic and Flierl, 2004a; Thompson and Young, 2006], and subtle sensitivity to the Coriolis parameter variation [Arbic and Flierl, 2004b; Thompson and Young, 2007]. Subgrid-scale parameterizations for LES are quite different in character because they attempt only to represent the effects Can Large Eddy Simulation Techniques Improve Mesoscale Rich Ocean Models?