Sea Ice Modeling: A Mini-Workshop
4. Rapporteur Summaries
Session 3. Sea Ice in Global Models
(Rapporteur: Gavin Schmidt)
Greg Flato gave an overview of the differences found in the Sea Ice Model Intercomparison Project (SIMIP) for a hierarchy of different sea ice dynamical schemes. The viscous-plastic model (using the Hibler rheology) had the best overall performance. Other simpler schemes, including free-drift + stoppage, cavitating fluid, did not perform as well. However, other viscous-plastic models have not yet been fully tested. He also gave an overview of the sensitivity of global models to the treatment of sea ice processes using the IPCC realizations. The inclusion of dynamics (regardless of the complexity of the dynamical scheme) did not necessarily improve the simulation. There was no correlation between flux-corrected/un-flux-corrected models and dynamics/no-dynamics and results for sea ice extent (though flux corrected models did slightly better). He concluded that errors in the sea ice were mainly due to extraneous factors (such as heat transport convergence in atm/ocean, errors in wind stress, and SLP patterns). Sea ice models are not the limiting factor.
Bruno Tremblay discussed the various dynamical models. They can be classified by 1) how they deal with small deformations (not important for the overall dynamics - but covers the differences between elastic and viscous rheologies), and 2) how they deal with large plastic deformations. Different plasticity models depend on the yield curve (which defines the transition to plastic behavior) and flow rule (what happens when critical yield is reached) used. As long as dynamics includes both failure under compression and shear (which does not include the cavitating fluid approach) results are similar. The plastic rheology based on granular material theory is a good candidate for global models and includes important dilation effects.
Gary Russell presented the latest runs from the GISS coupled model. Northern Hemisphere sea ice extent is stable with about the right level of interannual variability and amplitude of the seasonal cycle. Southern Ocean sea ice has much greater variability than observed (which is possibly related to poor simulation of the SO temperature/salinity structure). Southern Hemisphere variability is sensitive to increased CO2 forcing but does not appear to be correlated amongst different members of the ensemble. In contrast, NH changes are highly coherent among different runs. Problems still remain with the coarse resolution spherical grid at the pole and the larger than observed vertical mixing in the Southern Ocean.
Jinlun Zhang presented his work on the POLar Exchange at the Sea surface (POLES) model. This included a new ice dynamics routine for the National Center for Atmospheric Research Climate System Model (NCAR CSM) model solved using an Alternate Direction Implicit (ADI) method on a spherical grid (using 5 pseudo-time-steps). He stated that this was more efficient and more stable than the previous Hibler scheme. Twelve thickness categories were used in the model.
Elizabeth Hunke presented the Los Alamos National Laboratory (LANL) model results. The LANL model is a very high resolution (1/6 deg) version of the Parallel Ocean Program (POP) model on a distorted latitude-longitude grid (pole in North America). The sea ice dynamic component uses the Elastic-Viscous-Plastic scheme which improves scalability as the number of parallel processors increases. Recent improvements increase fidelity to the yield curve physics. She also noted that the extra metric terms in the ice dynamic equations that arise on a spherical grid are small for these very high resolution runs and therefore they are ignored.
Bill Lipscomb presented a three-category ice thermodynamics model (first-year ice, multiyear ice and ridged ice). All first-year ice becomes multiyear ice at the end of the summer and all ridged ice is in the ridged ice category. His model also includes the effect of salinity in brine pockets on heat capacity and on the energy needed to melt ice.
In discussion the number of sea-ice categories necessary was heavily debated: seven categories seemed to be the minimum that could achieve acceptance by all participants. The point was made that since ice thickness is not a well observed quantity, how could the various ice thickness distribution schemes be properly verified? Judy Curry pointed out that a statistical distribution approach was not promising since most of the sensitivity depends on the higher moments, i.e., on the fraction of open water and the thickest ice, and so parameterization is very difficult. Prasad Gogineni stated that the satellite derived data currently can probably only deliver three categories: open-water, first-year ice and multiyear ice categories.