1-Slide Summary Explicit Southern Ocean eddies respond to forcing differently than parameterizations. We need eddy resolving ocean climate models. Spurious diapycnal mixing is a serious problem with traditional ocean climate modeling techniques, particularly when the models are eddy resolving. We need to develop better models, not to just increase the resolution of existing models.
Most of the dense-to-intermediate water conversion occurs in the Southern Ocean. The Southern Ocean dynamics influence the deep stratification worldwide. Eddies are of leading-order importance in the dynamics of the Southern Ocean and essential for obtaining the observed overturning circulation. Do eddyless ocean models respond to forcing changes very differently from the real ocean? (Hallberg & Gnanadesikan, JPO 2006) Southern Ocean eddies and climate
Layers in a Sample Pacific Cross-Section
MESO Time Mean Sea Surface Height and Surface Speed
Modeled and Observed Sea Surface Height Variability
Eddy energy at 4 Resolutions
Overturning Streamfunction NADW CDW AABW AAIW AAMW
Overturning Streamfunction with Parameterized Eddies NADW CDW AABW AAIW AAMW
Drake Passage Transport at 4 Resolutions
Eddy Response to 20% Increase in Winds
Overturning Response to 20% Increase in Winds after 20 years Explicit EddiesParameterized Eddies
Change in Annual Mean Overturning at 40 o S at 3 Resolutions
Summary of MESO Findings The Southern Ocean eddy dynamics are of leading order importance in regulating the watermasses that fill the deep ocean. The fine and coarse resolution models are qualitatively similar in many regards, but the models differ in two key ways: 1.The explicit eddies contribute a poleward extension of the subtropical overturning cell and heat fluxes that may be difficult to parameterize diffusively. 2.The overturning response to wind forcing changes is substantially weaker and has a shallower compensating flow than with parameterized eddies. “Eddy-saturation” theories seem appropriate. Eddy-permitting ocean models should be used to evaluate key climate predictions. Ref: Hallberg & Gnanadesikan, JPO 2006.
Numerical Diapycnal Diffusion in Ocean Models Non-rotated horizontal diffusion operator leads to “Veronis effect”, The same issue arises with a nonrotated biharmonic diffusion (Roberts & Marshall 1998) Avoided in Z-coordinate models by using Gent/McWilliams closure (Gent et al. 1995) and rotating the diffusion tensor to align with isopycnals (Griffies et al. 1998) Advective truncation errors lead to spurious watermass modification in Z- coordinate and -coordinate models (Griffies et al. 2000). In coarse resolution models, diffusion can be acceptably small, provided that western boundary currents are well resolved. The cascade of variance to small scales in eddy rich models leads to diffusivities of order 1 cm 2 s -1. Avoiding advective spurious mixing in Z- and - coordinate models is an area of active research, but there is no known solution without suppressing eddy variability. Isopycnal coordinate models avoid numerical diapycnal diffusion at all resolutions (apart from effects of the nonlinear equation of state).
Numerical Diapycnal Diffusion in Low Resolution Gyres from Griffies, Pacanowski, and Hallberg, Mon. Wea. Rev., 2000 Well Resolved Western Boundary Currents Marginally Resolved Western Boundary Currents Veronis EffectPhysical Diffusion m 2 s -1 Diagnosed effective diffusivity from advective truncation errors.
Numerical Diapycnal Diffusion in Eddy-Rich Z-models from Griffies, Pacanowski, and Hallberg, Mon. Wea. Rev., 2000 Diagnosed spurious diffusion Diagnosed Diapycnal Diffusivity (cm 2 s -1 ) K d =10 -5 m 2 s Depth (m) Diffusivity from Quicker at 1/6° Res. In a representative channel flow, diagnosed numerical diapycnal diffusion is unacceptably large. Eliminating this diffusion requires: (A) unexploited higher resolution (B) suppressing the cascade of variance, or (C ) much better approaches to tracer advection? Lee, Coward, & Nurser (JPO 2002) find similarly large numerical diapycnal diffusion in an analysis of a ¼ o global ocean-only model. Marchesiello et al. (Ocean Modelling 2009) report that the problem is even worse in -coordinate models, peaking when the eddies start to be well resolved, only becoming “acceptable at resolutions finer than 1 km”.
Spurious diapycnal mixing in coupled climate models?
Spurious diapycnal mixing increases with resolution?
Isopycnal coordinates avoid spurious diapycnal mixing.
1/8° Global Isopycnal-Coordinate Model Based on CM2G May 1, Year 5
Discussion? Do we need to explicitly resolve eddies in ocean climate models? Is spurious diapycnal mixing a serious problem in long-term simulations with eddy- rich ocean climate models? Are there good solutions to avoiding spurious diapycnal mixing (apart from using isopycnal coordinates)?