Sea Ice Modeling: A Mini-Workshop
2. Science summary for Eos, Transactions of AGU
Large changes in Arctic sea ice cover have occurred in the last few years, though whether they are part of a long-term natural variability or are related to increased radiative forcing from anthropogenic greenhouse gases is unclear. Knowing how sea ice changes have affected, and will continue to affect, climate (through changes to the surface albedo and through alterations to the ocean-atmospheric exchange of freshwater and heat) may be crucial for making future climate predictions.
Processes in the Arctic that affect climate include melt-ponds, lead fractions, brine rejection, and ice-ice interactions. Quite a number more may exist. But, however many, they operate at relatively small scales that cannot be directly incorporated into current climate models. The high latitudes are expected to be the most sensitive regions to possible anthropogenically forced climate changes, yet observations of polar climate and understanding of polar processes have not proceeded as fast as our appreciation of their importance.
Recent developments such as the Surface HEat Budget of the Arctic (SHEBA) project, which included a ship that was allowed to freeze into the permanent ice pack for a year (October 1997 to October 1998), have provided a wealth of new data that are being used for small-scale process studies. New analyses of satellite-derived data (available from October 1978) are providing detailed information about interannual variability in sea ice extent, concentration, and motion. Homogeneous compilations of historical observations of sea ice extent are now starting to show the long-term history of sea ice cover. Finally, large-scale climate models are reaching a level of complexity that should allow significant improvements in their representation of sea ice processes.
A recent workshop at the NASA Goddard Institute for Space Studies brought together polar observers, remote sensing scientists, sea ice process modelers, and global climate modelers, who collectively assessed the adequacy of existing models and data sets and attempted to map out future strategies that would improve current understanding.
How large have recent changes been, and what kind of changes have occurred? Variability in the Arctic over the last decade seems to have been mainly regional in nature and mostly driven by altered atmospheric circulation patterns associated with changes in the Arctic Oscillation/North Atlantic Oscillation. Evidence exists that the amount of relatively warm North Atlantic water coming into the eastern Arctic has increased. In the Southern Hemisphere, there is no strong evidence for large overall changes over the last decade, although regionally sea ice changes are apparent.
Over the longer term, the difficulties in putting together very heterogeneous data sets (whaling records, old charts, satellite microwave imagery, ship reports, and so forth) make determining longer term trends very problematic. Previous data sets, such as Global Ice and Sea Surface Temperatures (GISST3.0), from the United Kingdom Meteorological Office, have had large spurious decreasing trends in sea ice extent (particularly in the Southern Hemisphere), at least in part as a result of these inhomogeneities. More recent data compilations (such as GISST4.0) are taking a more conservative approach that attempts to homogenize the data using satellite microwave measurements as a standard. Large Southern Hemisphere trends are no longer apparent, but significant trends are still found in the Arctic.
Remote sensing from satellites (using passive-microwave imagery, for example) provides reasonably accurate near-global coverage of sea ice extent since 1979. However, problems still exist in estimating sea ice concentration in the summer when melt ponds become extensive (since satellites confuse melt ponds with open water), and in combining data sets from different instruments. Some very promising analyses of sea ice motion from the passive microwave, the NASA scatterometer, advanced very high resolution radiometer, and synthetic aperture radar data are beginning to be produced. The error bars on the displacement fields can be as low as 300 m, comparable to those derived from the Arctic Buoy Program but with much greater coverage.
Combined analyses dating back to 1979 will soon be available. It may be possible to combine these motion fields (and hence shear fields and patterns of divergence/convergence) with simple ice growth models in order to estimate ice ages and thickness patterns — currently the most important unknown variables. However, more ground-truth validation is required for all the satellite-derived quantities.
The SHEBA project has been highly regarded as a model for detailed study of sea ice processes. New information from this project includes the temporal/spatial and spectral variations of ice albedo and improved understanding of the effect of melt ponds on albedo and thermodynamics. Much of the data collected as part of this yearlong project is being used in a coherent four-step approach for evaluating sea ice process parameterizations. This is based on data from comparing in detail the parameterization and observations of the process; from seeing how it performs in single column models; from seeing how it affects feedbacks; and from reevaluating the results in three-dimensional and coupled models. Unfortunately, observations in the Marginal Ice Zones and around Antarctica have not been as comprehensive as those taken by SHEBA, and hence parameterizations relevant in these different conditions cannot be so thoroughly tested.
Much progress has been made in the modeling of sea ice dynamics. Small-scale deformations can be treated as elastic or viscous processes, but the differences between these approaches are mainly numerical and have little effect on the large-scale results. Hence dynamical modeling mainly involves specifying the large-scale deformation (plastic) behavior of ice-ice interactions between ice floes. This requires a model for when plastic (irreversible) deformations occur and for what happens subsequently.
Two essential components are that the ice has resistance to compression and resistance to shear. Models that do not contain both components do not perform as well. Many available models (the Hibler rheology or the granular material based approaches) contain these components. However, analyses of observations have not yet been able to distinguish clearly between the models. The new satellite-derived sea ice motion fields will be important in this regard.
Incorporation of the detailed sea ice processes in large-scale general circulation models (GCMs) is hampered by their relatively coarse resolution (typically hundreds of kilometers) and the limited amount of computing power available for the sea ice component. The challenge is to come up with simplified parameterizations for the net effect of the small-scale processes, valid at the large scales. The importance of sea ice for the overall sensitivity of the climate to increasing greenhouse gases is clear. Some recent modeling work indicates that it is the sea ice thickness in the Northern Hemisphere and sea ice extent in the Southern Hemisphere that exert the most significant control on the sensitivity.
At minimum, coupled GCMs should include many of the negative feedbacks on sea ice growth as well as the more obvious positive feedbacks (such as ice-albedo) if they are to produce a stable climate. Among the most critical processes are those that modify the albedo (melt-ponds, snow aging, possibly rainfall) and those that modify the ocean-atmosphere heat exchange (lead formation). Brine rejection and freshwater and salt transports are of crucial importance for the high-latitude oceans and should be included. Other issues such as the role of brine pockets on the heat capacity of the ice and the exact relationship between lateral and basal formation and melting do not seem as crucial, but analysis of very high resolution ice models should provide better guidance in the future.
Only a few of the couple models in the latest Intergovernmental Panel on Climate Change (IPCC) assessment contained sea ice dynamics, and of those that did, only a minority used realistic rheologies. Interestingly, no correlation was found between models that included sea ice dynamics and models with the best simulations of sea ice extent. If anything, models containing dynamics did slightly worse. This implies that the errors in sea ice in GCMs may be mostly caused by extraneous factors (such as the global atmosphere/ocean heat transports or the wind fields). However, scope clearly exists for including more sophisticated parameterizations in GCMs and, for the medium term, perhaps incorporating multiple sea ice thickness categories or nested higher resolution sea ice models.
Recent results provide some evidence that, in spite of imperfections, models can nonetheless provide reasonable simulations of high latitude climate. Very high resolution ocean-ice models have been able to generate realistic interannual and interdecadal variability of sea ice cover in the Arctic, and global coupled models seem to be able to reproduce the main features of Northern Hemisphere climate change over the last 50 years. In the Southern Hemisphere, problems simulating the thin halocline under the sea ice and deepwater formation on shelves lead to poorer model performance and less coherent results for climate change scenarios.
The workshop proved fertile ground for the different groups to understand and respect each other's priorities and expertise. Considerable scope exists for cross-disciplinary work to bridge the widely varying space scales and timescales at which the different groups work. In particular, there seems to be a need for the sea ice process community to work with the remote sensing teams to produce more sophisticated products from satellite data, and for the large- and small-scale modelers to work together to synthesize some very detailed modeling results to produce improved workable parameterizations for GCMs.
Workshop Homepage *
Exec. Summary *
Science Summary *
Agenda
Sessions:
1,
2,
3,
4,
5,
6 *
Participants
