Sea Ice Modeling: A Mini-Workshop
4. Rapporteur Summaries
Session 4. Sea Ice in Global Models Ð Continuation
(Rapporteur: Larissa Nazarenko)
Yuxia Zhang discussed modeling of the significant Arctic changes that have occurred in the past decade. European Center for Medium-Range Weather Forecasts (ECMWF) reanalysis for 10 m winds exhibit the recent tendency toward more cyclonic circulation. This change in atmospheric circulation has a profound impact on the underlying sea ice, because in the central Arctic Ocean the variance of the daily ice motion is largely explained by the local winds in both winter and summer, and cyclonic winds give rise to divergent ice motion in summer.
A high-resolution, coupled ice-ocean model, forced with 1983-1993 ECMWF reanalysis data, is used to explore the recent Arctic change. The sea ice model used in this study includes both dynamics and thermodynamics. The ice dynamics is characterized by an elastic-viscous-plastic ice rheology (Hunke and Dukowicz, 1997). The ice thermodynamics is determined from an energy budget at the ice surface following Parkinson and Washington (1979) and the zero-layer approximation of Semtner (1976) for heat conduction through the ice. The ocean model is based on the Semtner-Chervin free surface model (Semtner and Chervin, 1992), adapted to the Arctic grid by Parsons (1995) and Maslowski (1997).
The simulation demonstrates complex dynamic-thermodynamic interactions between the atmosphere and the underlying sea ice, which ultimately influence global climate. In response to the observed change of atmospheric circulation, stronger cyclonic circulation is present in the Arctic sea ice and the upper ocean in the late 1980's and early 1990's as compared to the early 1980's, which leads to a weakening of the Beaufort Gyre and a shifting of transpolar drift. The strengthening (weakening) of the summer atmospheric cyclonic vorticity from the earlier period to the later period over the Eurasian (Canadian) Basin leads to an increased (decreased) ice divergence over the Eurasian (Canadian) Basin, which then has significant impact on the ice production rate by increasing (reducing) the ice formation over the Eurasian (Canadian) Basin. The model simulation supports the recent observations of increased presence of Atlantic water in the Arctic Ocean. The warming and salinification of Atlantic layer may be linked to variations in the NAO, which had a high index in the later period.
Achim Stoessel discussed modeling of sea ice and the ocean around Antarctica. He noted that both atmospheric and oceanic circulation determine the air-sea heat exchange, which acts to moderate the equator-to-pole temperature difference. In general, a large heat flux from the ocean to the atmosphere is associated with strong vertical motion in the ocean. The most vigorous vertical motions of the world's oceans occur in high latitudes, where dense water is produced at the ocean's surface by cooling to the freezing point and by brine release through sea ice formation. Sea ice modifies the heat exchange and freshwater budget at the ocean surface. Specifically, sea ice may locally inhibit deep convection through melting and enhance it elsewhere through freezing. The latter may eventually modify the rates of deep- and bottom-water formation, which themselves determine the global deep ocean water-mass properties and influence the global thermohaline circulation.
Stoessel uses a global primitive-equation ocean general circulation model (OGCM) that was upgraded to allow for detailed surface heat-balance calculations over partially ice-covered grid cells in addition to the existing viscous-plastic ice dynamics. A coarse-resolution model (3.5x3.5x11-layers) is used to allow a series of near-equilibrium sensitivity experiments with climatological monthly mean forcing. Reasonable large-scale performance of the model is demonstrated by the global meridional overturning circulation, the distribution of the main water masses, the transport at various key points of the thermohaline conveyor-belt circulation, and the seasonal and spatial variations of convection. The model yields an acceptable seasonal representation of the integrated sea ice cover in both hemispheres; the spatial distributions, however, are less realistic. In the Northern Hemisphere, sea ice is generally too thick, especially in the Beaufort Sea, which is mainly due to the coarse grid resolution; in the South Hemisphere (SH), sea ice is generally too thin, especially in the central Weddell and Ross Seas, mostly due to vigorous convection associated with excessive tapping of warm water from deeper layers. Sensitivity experiments focus on various possible effects of modifications in the treatment of upper boundary conditions in the Southern Ocean (SO) by imposing changes to the description and forcing of sea ice. Stoessel investigated the effect of sea ice's low salinity, that is, brine release during sea ice formation and freshwater release during sea ice melt as suggested by the sea ice model. With respect to the impact on sea ice itself, brine release includes upward heat flux, which tends to melt sea ice, thus counteracting brine release. Vice versa, if brine release is neglected, ice growth is not associated with increasing density of the upper layer, stabilizing the water column and a significant increase in SO sea ice thickness, except for the central Weddell and Ross Seas where ice melt leads to enhanced convection. The inclusion of a snow cover increases the deep ocean's temperature consistent with an overall decrease in SH convection originating from a decrease in sea ice growth. Enhanced Antarctic Bottom Water formation, expressed by a significant increase of the strength of the SH overturning cell, occurs if the turbulent heat fluxes over the sea-ice-ocean admixture are calculated based on daily winds, or if they are just increased by an equivalent factor while using the monthly winds.
Marika Holland presented the results of a coupled atmosphere-sea ice-ocean model with present day forcing, as well as doubled CO2 forcing. She used the simple atmospheric model with no flux correction. The Modular Ocean Model (MOM2) was used. For the sea ice, Holland tested different models. One model included zero-layer approximation for heat conduction though the ice; and no minimum leads. Another sea ice model had a multi-category thickness distribution. One of her coupled model experiments had only thermodynamics for sea ice, another had dynamics and thermodynamics for sea ice, with elastic-viscous-plastic rheology used in the sea ice dynamics equations. The model with multi-category sea ice thicknesses gives, on average, thicker ice. The results with this model show a higher melting rate in summer because of different albedos for thin and thick ice. The coupled atmosphere-sea ice-ocean model shows more transport of sea ice than the stand-alone sea icece in quantities that are important for a proper coupled response (e.g., heat and fresh water exchange with ocean and atmosphere) for 5 and 15 categories. However, the parameterization of the compressive strength causes the ice to weaken as thin ice is better resolved, which allows more ridge building and makes the ice thicker as more categories are used. Bitz recommends using at least 5 categories for sea ice modeling.
In the discussions of sea ice modeling, Michael Steele presented the following table (Table 1) summarizing the history of ice thickness distribution modeling.
|1970s||Development of the sea ice thickness distribution g(h) theory (Thorndike et al., 1975) during the AIDJEX project; parameterization of individual terms such as internal stress, air drag, etc.|
|1979||Hibler simplifies the thickness distribution into the 2-level model which follows in each grid cell only two variables: mean ice thickness and the areal concentration of ice in a grid cell.|
|1980||Hibler publishes a study using a full thickness distribution with fixed thickness levels (the Eulerian approach). This model is not used again until it is revived by Flato and Hibler in 1995.|
|1981-1994||The "2-level era." All models including those used by Hibler use the 2-level approach.|
|1995||Flato and Hibler revive the g(h) model and test against submarine data.|
|1996||A number of single cell and full synoptic sea ice models are currently under development with to present Eulerian (fixed thickness bins) or Lagrangian (characteristics) thickness distributions. No currently published synoptic sea ice model resolves both the thickness distribution and the thermal profile through the snow and ice. An exception might be a conference paper by Curry or one of her co-workers. Such models are also under development by other groups (e.g., by Dr. Jinlun Zhang).|