The cryosphere is represented in some hydrological models by the areal extent of snow cover, a variable that has been operationally available in recent years through remote sensing. In particular, the snowmelt–runoff model (SRM) requires the remotely sensed snow-cover extent as a major input variable. The SRM is well-suited for simulating the hydrological response of a basin to hypothetical climate change because it is a non-calibrated model. In order to run the SRM in a climate-change mode, the response of the areal snow cover to a change in climate is critical, and must be calculated as a function of elevation, precipitation, temperature, and snow-water equivalent. For the snowmelt-runoff season, the effect of climate change on conditions in the winter months has a major influence. In a warmer climate, winter may experience more rain vs snow events, and more periods of winter snowmelt that reduce the snow water equivalent present in the basin at the beginning of spring snowmelt. As a result, the spring snowmelt runoff under conditions of climate warming will be affected not only by different temperatures and precipitation, but also by a different snow cover with a changed depletion rate. A new radiation-based version of the SRM is under development that will also take changes in cloudiness and humidity into account, making climate-change studies of the cryosphere even more physically based.

In polar regions, the exchange of heat, fresh water and salt water, and momentum between ocean and atmosphere is strongly affected by the presence of sea-ice cover. As a growing number of climate models include a dynamic–thermodynamic sea-ice component to take these effects into account, it might be asked whether sea ice is adequately represented in these simulations, and how far these simulations fit with physical observations.

Sea ice in the classical models (Hibler, 1979; Parkinson and Washington, 1979) that have been available for two decades, is regarded as a two-dimensional (2-D) continuum covering the ocean surface. The prognostic variables describing the ice pack are horizontal ice velocity, areal coverage (ice concentration), and ice thickness. In numerical models, these variables and their evolution in space and time are solved on an Eulerian grid.

A number of observational data are available to verify the model results. Sea-ice drift is observed from drifting buoys deployed on ice floes. Areal sea-ice coverage can be observed with satellite-borne passive-microwave sensors (SMMR, SSM/I). For ice thickness, which cannot be observed with remote-sensing techniques, rather few, scattered observations from upward-looking sonars on submarines and moorings are available.

This article gives an overview of three additional variables representing sea ice in large-scale climate models. These are (1) roughness, (2) age of the ice, introduced as two prognostic variables, and (3) simulated trajectories of ice motion, which are diagnosed from the Eulerian velocity grid. The new variables enable a more detailed look at sea ice in models, helping to understand better the coupled dynamic–thermodynamic processes determining the polar ice cover. Further, the new variables offer important, additional possibilities for comparing the simulated sea-ice properties with available observations.

Accurate information concerning snow cover, and associated impacts of snow on regional surface albedo, needs to be available for empirical studies and for the validation of climate models. Here, a new integrated dataset for Northern Hemisphere lands is discussed, including files of visible and microwave satellite-derived snow estimates and in situ station data. These files will be used to examine snow extent, snow depth and surface albedo over five-day intervals, and have been generated using geographic-information system techniques. Visible and station observations extend from 1972 to present, and microwave estimates from 1979 to present, The 1×1° gridded files permit the strengths and weaknesses of the individual data sources to be identified and quantified. Also included is a hemispheric time series of snow extern derived from the visible satellite file. Of note are the two pronounced regimes of Northern Hemisphere extent during the past several decades. Between 1972 and 1985, 12 month running means of snow extent fluctuated around a mean of 25.9 × 106 km2. An abrupt transition occurred in 1986 and 1987, and since then mean annual extern has been 24.2 × 106 km2. Recent decreases are found from late winter to early summer.

Results are presented testing the sensitivity of a two-layer model to changes in the Laurentide ice sheet’s geometry following the collapse of the Hudson Bay ice dome. Since the ice sheet is thought to induce cooling over the North Atlantic through its mechanical effect on atmospheric circulation, the model shows a surprising result in that the removal of the Hudson Bay ice dome led to a further cooling of 4°C over the North Atlantic. This finding suggests that fluctuations in ice-sheet topography could have contributed to the climate variability witnessed in the geologic record. Further study is needed to understand the mechanism behind these results.

At the Last Glacial Maximum (LGM) about 21000 years ago (21 ka BP), the overall mass balance of the Laurentide and Eurasian ice sheets should have been close to zero, since their rate of change of total ice volume was approximately zero at that time. The surface mass balance should have been zero or positive to balance any iceberg/iceshelf discharge and basal melting, but could not have been strongly negative. In principle this can be tested by global climate model (GCM) simulations with prescribed ice-sheet extents and topography.

We describe results from a suite of 21 ka BP simulations using a new GCM (GENESIS version 2.0.a), with sea-surface temperatures (SSTs) prescribed from GLIMAP (1981) and predicted by a mixed-layer ocean model, and with ice sheets prescribed from both the ICE-4G (Peltier, 1994) and CLIMAP (1981) reconstructions. This GCM is well suited for ice-sheet mass-balance studies because (i) the surface can be represented at a finer resolution than the atmospheric GCM, (ii) an elevation correction accounts for spectral distortions of the atmospheric GCM topography, (iii) a simple post-processing correction for the refreezing of meltwater is applied, and (iv) the model's precipitation and mass balances for present-day Greenland and Antarctica are realistic. However, for all reasonable combinations of SSTs and ice-sheet configurations, the predicted annual surface mass balances of the LGM Laurentide and Eurasian ice sheets are implausibly negative. Possible reasons for this discrepancy are discussed, including increased ice-age aerosols, higher CLIMAP-like ice-sheet profiles in the few thousand years preceding the LGM, and a surface of the southern Laurentide just before the LGM to produce fleetingly the ICE-4G profile at 21 ka BP.

A three-dimensional (3-D), high-resolution, non-linearly viscous, non-isothermal ice-sheet model is employed to calculate the “present-day” equilibrium regime of the Antarctic ice sheet and its evolution during the last glacial cycle. The model is augmented by an approximate formula for ice-sheet basal temperature, based on a scaling of the thermodynamic equation for the ice flow. Steady-state solutions for both the shape and extent of the areas of basal melting (or freezing) are shown to be in good agreement with those obtained from the solution of the full 3-D thermodynamic equation. The solution for the basal temperature field of the West Antarctic Siple Coast produces areas at the pressure-melting point separated by strips of frozen-to-bed ice, the structure of which is reminiscent of Ice Streams A–E. This configuration appears to be robust, preserving its features in spite of climatic changes during the last glacial cycle. Ice Stream C seems to be more vulnerable to stagnation, switching to a passive mode at least once during the penultimate interglacial. We conjecture that the peculiarities of local topography determine the unique behavior of Ice Stream C: reduced basal stress and, consequently, relatively weak warming due to internal friction and basal sliding is not able to counteract the advective cooling during the periods of increased snowfall rate.

Using a dynamic-thermodynamic numerical sea-ice model, external oceanic and atmospheric forcings on sea ice in the Weddell Sea are examined to identify physical processes associated with the seasonal cycle of pack ice, and to identify further the parameters that coupled models need to consider in predicting the response of the pack ice to climate and ocean-circulation changes. In agreement with earlier studies, the primary influence on the winter ice-edge maximum extent is air temperature. Ocean heat flux has more impact on the minimum ice-edge extent and in reducing pack-ice thickness, especially in the eastern Weddell Sea. Low relative humidity enhances ice growth in thin ice and open-water regions, producing a more realistic ice edge along the coastal areas of the western Weddell Sea where dry continental air has an impact. The modeled extent of the Weddell summer pack is equally sensitive to ocean heat flux and atmospheric relative humidity variations with the more dynamic responses being from the atmosphere. Since the atmospheric regime in the eastern Weddell is dominated by marine intrusions from lower latitudes, with high humidity already, it is unlikely that either the moisture transport could be further raised or that it could be significantly lowered because of its distance from the continent (the lower humidity source). Ocean heat-transport variability is shown to lead to overall ice thinning in the model response and is a known feature of the actual system, as evidenced by the occurrence of the Weddell Polynya in the mid 1970s.

It is generally accepted that fresh-water fluxes due to ice accretion or melting profoundly influence the formation of Antarctic bottom water (AABW). This is investigated by means of a global, three-dimensional ice–ocean model. Two model runs were conducted. At the high southern latitudes, the control experiment exhibits positive (i.e. towards the ocean) fresh-water fluxes over the deep ocean, and large negative fluxes over the Antarctic continental shelf, because of the intense ice-production taking place in this region. The salinity of shelf water can increase in such a way that deep-water formation is facilitated. The simulated net fresh-water flux over the shelf has an annual mean value of −1 m a−1. This flux induces a transport of salt to bottom waters, which corresponds to an increase of their salinity estimated to be around 0.05 psu. In the second model run, the fresh-water fluxes due to ice melting or freezing are neglected, leading to a rearrangement of the water masses. In particular, the AABW-formation rate decreases, which allows the influence of North Atlantic deep water (NADW) to increase. As NADW is warmer and saltier than AABW, the bottom-water salinity and temperature become higher.

The performance of an explicit cloud physics parameterization is examined with simulations of high southern latitude winter climate using a version of the Pennsylvania State University/National Center for Atmospheric Research Mesoscale Model, version 4. The results reveal that there are three moist physics regimes in the vertical over the elevated interior of Antarctica: the very cold upper troposphere, the relatively warm middle troposphere and the cold boundary layer. Deficiencies for these layers include excessive cloud ice in the upper troposphere, excessive cloud ice in the inversion layer near the ice surface, overly warm temperatures in the lower troposphere, overly cold temperatures in the upper troposphere and excessive downward longwave radiation at the Earth’s surface. Three sensitivity experiments were performed to investigate possible improvements in the cloud parameterization. The results indicate that a reduction of the numerous cloud condensation nuclei, while reducing some errors, appears to be insufficient to improve the simulation. A reduction in the excessive cloud ice in the upper troposphere significantly improves the simulation of upper-tropospheric temperature.

Snow falling uniformly on a distribution of ice thicknesses results in a distribution of snow-cover thicknesses. These snow depths depend on the amount of snow-fall, the time of year at which it falls, and the thickness of the underlying ice. The effect of snowfall on snow-ice-thickness distribution is examined using a single-column ice ocean model. The time at which snow begins to accumulate, and melt ponds and leads freeze, affects the surface albedo. The rate of snowfall affects ice-growth rates and, as a result, the ice-thickness distribution. During the period of rapid ice growth between the autumn freeze and mid-winter, snow falling on newly formed ice is rapidly depleted due to sublimation. Snow falling on thicker ice remains throughout the winter to create a source of melt-water for ponds and runoff into the ocean.

A major concern of the cryospheric community is the availability of reliable data for use with various types of models. The National Science Foundation’s Arctic System Science program (ARCSS) integrates specific Arctic investigations ranging from the Greenland Ice Sheet Project, the Land-Almosphere-Ice Interactions, the Ocean-Atmosphere-Ice Interactions and its associated Surface Heat Energy Budget of the Arctic program to the recently initiated Human Dimensions of the Arctic System project. All data collected from these investigations are archived at the ARCSS Data Coordination Center and the National Snow and Ice Data Center at the University of Colorado, U.S.A. A primary goal of the ARCSS Data Coordination Center is to facilitate better use of observed data within cryospheric modeling efforts. Working with the ARCSS Data Management Group and the Modeling Working Group, we are developing data appropriate for the cryospheric modeling community. This paper summarizes the data currently available at the ARCSS Data Coordination Center, and details the mechanisms by which Arctic researchers can access these data.

Most dynamic ice-sheet studies currently use either empirically based parameterizations or simple energy-balance climate models for the surface mass-balance forcing. If three-dimensional global climate models (GCMs) could be used instead, they would greatly improve the potential realism of coupled climate ice-sheet simulations. However, there are two serious problems in simulating realistic mass balances on ice sheets from GCM simulations: (i) dynamic ice-sheet models and the underlying bedrock topography need horizontal resolution of 50–100 km or less, but the finest practical resolution of atmospheric GCMs is currently ˜250 km, and (ii) GCM surface physics usually neglects the local refreezing of meltwater on ice sheets.

Two techniques are described that address these problems: an elevation correction applied to the atmospheric GCM fields interpolated to the ice-sheet grid, and a refreezing correction involving the annual totals of snowfall, rainfall and local melt at each grid-point. As an example of their use, we have used the GENESIS version 2 GCM at 3.75° × 3.75° resolution to simulate the climate at the end of the last interglaciation at ˜116 000 years ago. The atmospheric climate is then used to drive a standard two-dimensional dynamic ice-sheet model for 10 000 years on a 0.5° × 0.5° grid spanning northern North America. The model successfully predicts ice-sheet initiation over the Baffin Island highlands and the Canadian Archipelago, but at a slower rate than observed. A large ice sheet nucleates and grows rapidly over the northwestern Rockies, in conflict with geologic evidence. Possible reasons for these discrepancies are discussed.

This paper describes the application of an automated cross-correlation technique to pairs of 85.5 GHz Special Sensor Microwave Imager (SSM/I) images to obtain ice motion over the entire Arctic Basin for a contiguous two month period between December 1993 and January 1994. Although the surface ice information in the imagery is coarse and noisy, the area cross-correlation method is quite successful in picking up ice-motion information. The accuracy of 85.5 GHz SSM/I derived ice motions is evaluated by comparing results with Arctic buoy drift. Over 390 comparisons with buoy-drift estimates of ice displacement were made with an overall correlation of 0.75 and an average vector magnitude error in ice velocity of 3.5 km d−1. The main difficulty with the automated technique is the tendency to overestimate ice displacement compared to buoy data by about 14%. Two detailed examples of ice motion are presented. The first occurred in December 1993, when a major westward shift in the ice pack took place in the Canada Basin and opened up a very large lead off Banks and Prince Patrick Islands. The second example occurred in January 1994, when an intense anticyclone over the Canada Basin produced a strong Beaufort Gyre.

The areal distribution of snow cover and the variability of its characteristics were investigated at various locations in the eastern Swiss Alps. An areal energy-balance (AEB) model was used to calculate the predominant energy fluxes at the snow–atmosphere interface based on automatic meteorological measurements as input. By coupling the AEB model with a one-dimensional, physically based mass and energy-balance model of the snowpack, temperature distribution as well as energy and mass flow in the snowpack were simulated at three different locations in the topographically complex environment at Weissfluhjoch-Davos, 2540 m a.s.l. On a horizontal test site, calculated energy fluxes and characteristics of the snow cover are in good agreement with their measured counterparts. On inclined slopes, the temperature distribution is well represented by the coupled models, but the snow depth and density are not yet satisfactorily simulated. This discrepancy may be attributed to inhomogeneous accumulation and deposition of snow on the weather and lee sides.

A key measure of our understanding of polar glaciation is the ability to model the initiation, maintenance, and elimination of glaciation over Antarctica and high-latitude land masses in the Northern Hemisphere. Studies that address questions of Cenozoic Antarctic glaciation, as well as studies that address questions of Pleistocene glaciation in the Northern Hemisphere, are described in some detail. The intention is to emphasize and discuss issues that are important in modeling these types of glacial events as much as to present specific results.

Inter-annual sea-ice variability north of 55°N in eastern Canada was explored with a box model incorporating annually invariant, bathymetrically dependent ocean heat fluxes and near-surface currents. Using inputs of ice concentrations, regional surface temperatures and geostrophic winds at external model boundaries, ice composition was simulated in seven thickness categories at 10 day intervals during three annual ice seasons. Comparisons indicated good reproduction of observed inter-annual differences in regional ice volumes during critical January-March periods. Additional simulations of artificial cutoffs in southward ice fluxes showed that advective influence decreases with latitude and dominates the development of spring ice conditions in areas south of 60°N.

Due to large uncertainties in many of the parameters used to model sea ice, it is possible that models with significantly different physical processes can be tuned to obtain realistic present-day simulations. However, in studies of climate change, it is the response of the model it various perturbations that is important, in studies response can be significantly different in sea-ice models that include or exclude various physical feedback mechanisms. Because simplifications in sea-ice physics are necessary for general circulation model experiments, it is important to assess which physical processes are essential for the accurate determination of the sensitivity of the ice pack to climate perturbations. We have attempted to address these issues using a new coupled ice-thickness distribution ocean mixed-layer model. The sensitivity of the model to surface heat-flux perturbations is examined and the importance of the ice ocean and ice-albedo feedback mechanisms in determining this sensitivity is analyzed. We find that the ice ocean and ice-albedo feedback processes are not mutually exclusive, and that they both significantly alter the model response to surface heat flux perturbations.

A quasi-three-dimensional (3-D) climate model (Sellers, 1983) was used to simulate the climate of the Last Glacial Maximum (LGM) in order to provide climatic input for the modelling of the Northern Hemisphere ice sheets. The climate model is basically a coarse-gridded general circulation (GCM) with simplified dynamics, and was subject to appropriate boundary conditions for ice-sheet elevation, atmospheric CO2 concentration and orbital parameters. When compared with the present-daysimulation, the simulated climate at the Last Glacial Maximum is characterized by a global annual cooling of 3.5°C and a reduction in global annualprecipitation of 7.5%, which agrees well with results from other, more complex GCMs. Also the patterns of temperature change compare fairly with mostother GCM results, except for a smaller cooling over the North Atlantic and the larger cooling predicted for the summer rather than for the winter over Eurasia.

The climate model is able to simulate changes in Northern Hemisphere tropospheric circulation, yielding enhanced westerlies in the vicinity of the Laurentide and Eurasian ice sheets. However, the simulated precipitation patterns are less convincing, and show a distinct mean precipitation increase over the Laurentide ice sheet. Nevertheless, when using the mean-monthly fields of LGM minus present-day anomalies of temperature and precipitation rate to drive a three-dimensional thermomechanical ice-sheet model, it was demonstrated that within realistic bounds of the ice-flow and mass-balance parameters, veryreasonable reconstructions of the Last Glacial Maximum ice sheets could be obtained.

Observed and reconstructed snow-cover duration data from stations covering southern Canada, the Great Plains, the former Soviet Union and China were used to reconstruct spring snow-covered area over North America (NA) and Eurasia from 1915 to 1985. A combination of nine regions from NA and five from Eurasia were able to explain 81% and 67%, respectively, of the variance in satellite-derived sprint; snow-covered area (SCA) for each continent. The results suggested sprint; SCA had decreased significantly in Eurasia this century, but there was no evidence of a similar long-term decrease in NA spring SCA. Considerable caution should be used when interpreting these results because of the short period of calibration, and because of the less-than-optimal distribution of station data. Nonetheless, the reconstructed results are consistent with observed spring-temperature trends, which show a significant increase over Eurasia, but none Over NA.

The stochastic relationships between terrestrial snow water equivalent (SWE) and measures of the atmospheric circulation were investigated for the Canadian Prairies and the American Great Plains for the winter of 1988. Snow-cover extent, derived from EASE-grid SSM/I satellite data, and griddcd atmospheric data from the National Meteorological Center were averaged at five day intervals. Principal components analysis (PCA) were performed for the time series of SSM/I snow-cover imagery as well as for 700 mb geopotential height and temperature, 500 mb height and 700–500 mb thickness. Canonical correlation analysis of the derived principal component weights was used to identify relationships between atmospheric variables and SWE. Results of the PCA indicate that a high degree of variance in upper air variables (>75%) can be explained by the first three principal components, while the first three SWE components account for over 90% of the variance in the original data. Results of the canonical correlation analysis show positive relationships between snow-cover accumulation and a meridional pressure distribution pattern, while snow ablation is linked to a zonal atmospheric pressure pattern.