GFDL - Geophysical Fluid Dynamics Laboratory

Topic: Eddy Kinetic Energy

Info associated with a Journal of Climate paper (Delworth et al., 2011, in press) and a poster presented at the October 2011 WCRP Open Science Conference (Dixon et al, 2011).

[icon version: maps of observed and modeled ocean eddy kinetic image]

For bookkeeping purposes, one can divide the ocean’s kinetic energy into two parts: one associated with the time mean and the other being the eddy component. Observations tell us that eddy kinetic energy (EKE) dominates across most of the global ocean. In other words, the ocean circulation is dynamic and turbulent. However, to date, most global coupled climate models used to study multi-decade to century time scale climate questions have had horizontal grid resolutions too coarse to explicitly resolve meso-scale eddies, and thus have relied on parameterizations (e.g., Gent & McWilliams, 1990; Gent, 2010) to estimate the net, bulk effect of sub-grid scale eddy processes on the transport of heat, salt, and other tracers. GFDL’s CM2.1 is one such model. Though sub-grid scale parameterizations have improved over the years, they have limitations, which is one of the reasons we are curious to see how GFDL’s newer, higher resolution CM2.5 and CM2.6 model simulations compare to the previous generation of models and with observations.

The figure below display maps  of near-surface Eddy Kinetic Energy calculated from satellite observations and from three GFDL global climate models.

[maps of ocean eddy kinetic energy in GFDL models and observations]

We note that the map of the CM2.6 model’s near-surface EKE is in excellent agreement with the observational estimates, both in terms of pattern and magnitude. EKE is largest in areas of strong currents (e.g, the Gulf Stream, Kuroshio, Agulhas, Antarctic Circumpolar Current, and Argentine Basin regions).  CM2.5 does a credible job of reproducing the general pattern of observed EKE. That CM2.5 tends to underestimate the EKE magnitude somewhat is not surprising, because CM2.5 grid resolution is considered eddy-permitting, not eddy-resolving. Recall that, though they are coupled to the same atmospheric model, the CM2.6 model has finer resolution in the ocean than does CM2.5 (horizontal grid spacing in CM2.5 is approximately 2.5 times greater than in CM2.6). In turn, over much of the ocean, the coarser resolution CM2.1 model’s ocean grid point spacing is about 4 times greater than in CM2.5. CM2.1 can resolve ocean meso-scale eddy activity, except in part of the tropics, and thus produces and EKE map that differs greatly from observations and from the higher resolution CM2.5 and CM2.6 model results.

[4 panel near-surface ocean log(EKE) maps for the Intra_American Seas - GFDL models and obs]

caption: log(EKE), same as figure above but focused on the Gulf Stream and Intra-American Sea region

  • Gent, P.R., and J.C. McWilliams, 1990: Isopycnal mixing in ocean circulation models. J. Phys. Oceanogr.,  20 , pp. 150?155. [LINK]
  • Gent, P.R., 2011: The Gent-McWilliams parameterization: 20/20 hindsight. Ocean Modelling, 39, pp 2-9. [LINK]