Gary A. Maykut
Department of Atmospheric Sciences
University of Washington, Seattle, WA 98195-1640
206-543-0164; 206-543-0308 (fax)
Some areas where the treatment of thermodynamics can potentially be upgraded in PIPS 3.0 are the following:
1. Ice Thickness Distribution Resolution. In general, the greater the number of ice thickness categories in the model, the better the prediction of quantities such as: ice production, salt fluxes, turbulent heat exchange and deformation. In practice, there will be a point of diminishing returns where additional resolution does not yield substantial improvement. Results from simpler, single column models may provide some guidance in choosing the appropriate number of categories, but such models contain different physics and do not include effects from surrounding grid cells. Sensitivity tests with the complete PIPS 3.0 model offer the best way to examine how the number and partitioning of categories affect predicted heat and mass fluxes for the particular physics used in this model. Of particular importance, both to the thermodynamics and dynamics, is adequate resolution of thinner, first-year ice categories which are the source of nearly all the net ice production, much of which ends up as deformed ice. The uniform thickness partitioning used in PIPS 2.0 is not the most efficient. Ice growth rates and heat fluxes over open leads rapidly drop by an order of magnitude or more by the time the ice reaches 50 cm in thickness; this is followed by continued, but much more gradual, decreases as the ice grows toward equilibrium. While high thickness resolution is needed to predict fluxes associated with the thinner ice, there seems little need for such resolution in thicker (H >100-150 cm) ice where these quantities are much less sensitive to thickness. Unequal partitioning of the thickness distribution, with highest resolution at the thin end and progressively decreasing resolution for thicker ice, would preserve the accuracy of heat and mass balance predictions while greatly decreasing the number of categories required. An objective way to select these categories might be to require that the total ice growth or heat exchange in each category be roughly comparable when integrated over the thickness range of that category. Alternatively, one might try to define categories that coincide with those reported by ice observers. It should also be noted that the ice thickness distribution model, itself, has serious uncertainties in characterizing the mechanical redistribution of ice thickness and the mass balance of deformed ice. Research being carried out in conjunction with SCICEX and SHEBA is likely to result in improved treatment of these processes. The new computer code for PIPS 3.0 should be designed to accommodate such improvements, as well as making it easy to change the number and partitioning of the ice thickness categories.
2. Vertical Transport and Storage of Heat in the Ice. While the assumption of linear temperature profiles in the ice and snow greatly reduces computational difficulties associated with the thermodynamics, it does a poor job of simulating the ice cover when there is rapid warming or cooling, i.e. during the spring and fall. In the fall, for example, when thin ice is growing rapidly in refreezing leads, the lower part of thicker ice is still warm and can continue to undergo bottom ablation well into November. We see this effect clearly in some SCICEX data we are analyzing. Two submarine-based thickness distribution surveys of the same area of ice were carried out about 35 days apart in September-October 1996. Even though the average thickness in the region increased substantially during this period due to ice production in the leads, the peak of the thickness distribution actually moved to a lower thickness in response to bottom melting. A linear temperature gradient model would have predicted that the thicker ice would also be growing. Similar difficulties also arise if one wants to predict diurnal or other short-term temperature variations in the upper part of the ice. The 3-level treatment devised by Semtner would do a much better job but, with low vertical resolution, such models have difficulty properly accounting for the storage and release of heat absorbed by the ice during the summer. With rapidly increasing computing capabilities, it is probably time to implement a more rigorous treatment of the heat conduction equation in PIPS 3.0.
3. Interaction of Shortwave Radiation with the Ice and Ocean. Numerous processes are involved in the interaction of shortwave radiation with the ice and upper ocean. Collectively, these processes are often referred to as the "ice-albedo feedback mechanism". Many of these processes appear to be extremely important, but are generally complex and difficult to treat. A major focus of the ongoing SHEBA program is to understand and develop models of these processes. To appreciate the impact of these processes on the heat and mass balance, consider the following example. We know that areally-averaged albedos over the arctic ice pack are on the order of .45-.50 in the middle of the summer. However, if albedos this low are used to force a 1-D thermodynamic model of the ice, the entire arctic ice pack disappears. The reasons for this are related to large horizontal variations in the composition of the summer pack which are not included in simple 1-D treatments. Melt ponds, first-year ice and leads all absorb and redistribute shortwave radiation differently than thick multiyear ice, and in a way that reduces the net loss of ice from the system over an annual cycle. Of particular importance is input of solar energy to the mixed layer through leads, thin ice and melt ponds. In the deep water portions of the Arctic Basin, this is the primary source of the oceanic heat flux at the underside of the ice, a major factor in determining the overall thickness of the ice cover. Shortwave energy absorbed in the water also affects the rate of decay and retreat of the pack over the shelves and Marginal Ice Zone, but treatment there is complicated by vertical exchanges that take place between the mixed layer and underlying water. In addition to the heat flux at the bottom of the ice, shortwave energy absorbed in the upper 2-3 m of leads produces lateral melting on floe edges, a positive feedback mechanism that causes a significant reduction in summer ice concentration. The input of shortwave radiation to the upper ocean and subsequent interactions with the ice are clearly of fundamental importance and should be included in PIPS 3.0.
Also important are melt ponds. During the summer, multiyear sea ice temporarily stores shortwave energy as latent heat in brine pockets. Most of this heat is returned to the atmosphere in the fall as the ice cools and the brine pockets freeze. Because the meltwater produced by internal absorption of shortwave radiation is retained within the ice and is not lost to the ocean, the net effect of this absorption on ice thickness is small. We have come to realize that melt ponds are, in effect, surface brine pockets that represent another portion of the absorbed shortwave energy that has only a small net effect on multiyear ice thickness. However, this is not the case with first-year ice where ignoring the effects of melt ponds may explain why the summer ice edge seems to retreat too slowly in PIPS 2.0. First-year ice is relatively flat and has a summer pond coverage that is typically between 50 and 90%. With an albedo of .25-.35, these ponds absorb large amounts of energy that produce rapid melting and decay of the first-year ice. Neglecting melt ponds would also result in summer albedos that are too high, making it difficult to compare model predictions with satellite-derived albedo observations. Analysis of SHEBA field data should provide a great deal of additional information on how ponds impact the heat and mass balance of the ice pack, as well as on ways to treat their effects. In designing PIPS 3.0, we should try to come up with a framework that makes it straightforward to include melt pond effects in the model.
4. Snow Cover. While snow has only a modest effect on the long-term equilibrium thickness of the ice, its depth and properties can have profound effects on short-term rates of ice growth and heat exchange, quantities which should be of concern in a forecasting model. Also, the date at which the snow cover vanishes strongly affects the total amount of ice ablated, the thickness of the ice that remains at the end of the summer, and possibly even whether the ice becomes seasonal or multiyear. At present, snow seems to be treated only crudely in PIPS 2.0. PIPS 3.0 should contain a more explicit treatment of the snow that includes spatially varying depths derived from climatological snowfall rates, along with spring albedos that more accurately reflect the structural changes that occur in the snow as temperatures warm. Other issues that may be important and need to be considered include the treatment of: (1) snow on thinner ice - 3 or 4 cm of snow on very thin ice can have a major impact on heat fluxes and growth rates, (2) melt metamorphism - recent work by Schramm et al. suggest that isolated spring rain events can cause a sharp reduction in snow albedo that hastens snow melt and increases the length of the ablation season, and (3) depth hoar - a ubiquitous, low density layer found near the ice-snow interface on level ice which reduces the rate of vertical heat transfer between the ice and snow.