A global ocean general circulation model has been coupled with a dynamicthermodynamic sea-ice model. This model has been spun-up in a 1000 yearintegration using daily atmosphere model data. Main water masses andcurrents are reproduced as well as the seasonal characteristics of the icecover of the Northern and Southern Hemispheres. Model results for theSouthern Ocean, however, show the ice cover as too thin, and there are largepermanent polynyas in the Weddell and Ross Seas. These polynyas are due to alarge upward oceanic heat flux caused by haline rejection during thefreezing of sea ice. Sensitivity studies were performed to test several waysof treating the sea-surface salinity and the rejected brine. The impact onthe ice cover, water-mass characteristics, and ocean circulation aredescribed.

A global coupled atmosphere-ocean-ice model is used to examine theinterdependence among several components of the hydrologic cycle in theArctic Ocean, including river discharge, sea-ice cover, and the flow of seaice through Fram Strait. Since the ocean model has a free surface,fresh-water inflow from rivers is added directly to the ocean. The timing ofthe peak spring river flow depends on snowmelt runoff and its subsequentrouting through the river system. Thermodynamic sea ice is included, and anew sea-iee advection scheme is described. The model’s river dischargeaffects salinity at the mouth of large rivers. The effect of the riverdischarge on sea-ice cover is not clear, either locally or at the basinscale. There is significant inter-annual variability of ice flow throughFram Strait, but the model’s flow is about half of that observed. Theanomalous ice flow through Fram Strait is most highly correlated with themeridional wind stress. Potential implications for the “great salinity”anomaly are discussed.

Different vegetation models impact the atmospheric response of a regionalclimate model in different ways, and hence have an impact upon the abilityof that model to match an observed climatology. Using a multivariateprincipal-component analysis, we investigate the relationships betweenseveral land-surface models (BATS, LSM) coupled to a regional climate model,and observed climate parameters over the North Slope of Alaska. In thisapplication, annual cycle simulations at 20 km spatial resolution arecompared with European Centre for Medium-Range Weather Forecasts (ECMWF)climatology. Initial results demonstrate broad agreement between all models;however, small-scale regional variations between land-surface modelsindicate the strengths and weaknesses of the land-surface treatments in aclimate system model. Specifically, we found that the greatersurface-moisture availability and temperature-dependent albedo formulationof the LSM model allow for a higher proportion of low-level cloud, and alater, more rapid transition from the winter to the summer regime. Crucialto this transition is the seasonal cycle of incoming solar radiation. Thesepreliminary results indicate the importance of the land-surface hydrologiccycle in modelling the seasonal transitions.

During snowmelt over a continuous snow cover, the vertical turbulentexchanges of sensible and latent energy are influenced by regional air-masscharacteristics, which exert a strong control on air temperature. Inhigh-latitude sites, the melting surface rapidly becomes heterogeneous, withpatches of snow and snow-free areas. Local advection occurs whennear-surface air layers are warmed due to sensible heal flux from thesnow-free areas, with the resulting heat transferred horizontally toadjacent snowpatches. This advection greatly increases the rate of snowmeltalong the leading edges of the snowpatches. In order to estimate correctlythe average melt rates of the snowpatches and the bulk energy balance of theentire landscape, it is necessary to estimate the local advection component.To date, few studies have dealt with this problem. This paper reportsresults from an Arctic tundra site located approximately 55 km northeast ofInuvik, Northwest Territories, Canada. The importance of local advection isestimated by comparing the sensible heat flux of the snowpatches toestimates of sensible heat without local advection. This latter term isderived from a relationship between upper air temperature and sensible heatflux over a continuous snow cover. This work has important implications fordeveloping models that correctly represent the cryosphere of tundra regions,and in developing appropriate scaling techniques for heterogeneouslandscapes.

The response of the Antarctic ice sheet to climate change over the next 500years is calculated using the output of a transient-coupled ocean-atmospheresimulation assuming the atmospheric CO2 value increases up tothree times present levels. The main effects on the ice sheet on thistime-scale include increasing rates of accumulation, minimal surfacemelting, and basal melting of ice shelves. A semi-Lagrangian transportscheme for moisture was used to improve the model’s ability to representrealistic rates of accumulation under present-day conditions, and therebyincrease confidence in the anomalies calculated under a warmer climate. Theresponse of the Antarctic ice sheet to the warming is increased accumulationinland, offset by loss from basal melting from the floating ice, andincreased ice flow near the grounding line. The preliminary results of thisstudy show that the change to the ice-sheet balance for thetransient-coupled model forcing amounted to a minimal sea-level contributionin the next century, but a net positive sea-level rise of 0.21 m by 500years. This new result supercedes earlier results that showed the Antarcticice sheet made a net negative contribution to sea-level rise over the nextcentury. However, the amplitude of the sea-level rise is still dominated Inthe much larger contributions expected from thermal expansion of the oceanof 0.25 m for 100 years and 1.00 m for 500 years.

In the framework of the Paleoclimate Modelling Intercomparison Project(PMIP), simulations of the Last Glacial Maximum (LGM) have- been performed.More than 10 different atmospheric general circulation models (AGCMs) havebeen used with the same boundary conditions: sea-surface temperaturesprescribed by CLIMAP (1981), ice-sheetreconstruction provided by Peltier (1994), change in insolation, and reduced CO2 content.One of the major questions is to investigate whether the simulations of theLGM are in equilibrium with the prescribed ice-sheet reconstruction. To answer this question, we have used two differentapproaches. First, we analyze the results of a sel of LGM simulationsperformed with different versions of the Laboratoire de Meteorolo-gieDynamique (LMD) AGCM and study the hydrologic and snow- budgets over theLaurcntide and Fennoscandian ice sheets. Second, we use the AGCM outputs toforce an ice-sheet model in order to investigate its ability to maintain theice sheets as reconstructed by CLIMAP (1981) or Peltier(1994).

Mean oxygen-isotope ratio values relative to standard mean ocean water (δ18 O, in ‰) reported for 406 sites in Antarctica arecompiled together with data on mean annual surface temperature, latitude,surface elevation, and mean annual shortest distance to open ocean denotedby the 20% sea-ice concentration boundary. Stepwise regression analyses with δ18O as the dependent variable are used as a model-buildingprocedure based on statistical, rather than physical, criteria. Multivariatemodels sensitive to covariation between independent variables are definedusing the whole dataset (N406 where Ndenotes the number of sites), as well as sub-sets for areas of conterminousgrounded ice (N206) and ice shell (N110).The models show improvement over bivariate regression models. Distance tothe open ocean enters all models at the second step. Inversions of the setand sub-set models applied to a database for 1351 gridpoint locations 100 kmapart (it excludes the regions of Graham Land and eastern Palmer Land) areused to produce contoured distributions of δ18O. These may beused to assess the effects of atmospheric advection, as well as deriveice-flow adjustments for δ18O series obtained from deep-core orablation-zone samples. Suggestions are made to improve modelreliability.

A major component of the climate system on the 10 000-100 000 yeartime-scales is continental ice sheets, yet many of the mechanisms involvedin the land-sea-ice processes that affect the ice sheets are poorlyunderstood. In order to examine these processes in more detail, we havedeveloped a coupled energy balance climate-thermodynamicsea-ice—continental-ice-sheet model (CCSLI model). This model includes ahydrologic cycle, a detailed surface energy and mass balance, athermodynamic sea-ice model, and a zonally averaged dynamic ice-flow modelwith bedrock depression.

Because of the variety of space and time-scales inherent in such a model, wehave asynchronously coupled the land—ice model to the other components ofthe model. In this paper the asynchronous coupling is described andsensitivity studies are presented that determine the values of theasynchronous coupling parameters. Model simulations using these values allowthe model to run nearly ten times faster with minimal changes in the finalstate of the ice sheet.

Both observational studies and numerical experiments demonstrate thesensitivity of the atmosphere to variations in the extent and mass of snowcover. There is therefore a need for simple but realistic snowparameterizations in forecast and climate models. A new snow hydrology hasrecently been developed at Météo-France. This parameterization has beenvalidated off-line against field measurements, as well as in a fullyinteractive numerical climate experiment. In this study, further tests wereperformed in stand-alone mode. The results indicate that the scheme is ableto reproduce snow-cover evolution, both on local and global scales, and somedirections to improve the parameterization further are suggested.

Crocus, a one-dimensional model of snow-cover stratigraphy and evolution,was developed by the Cenire d’Etudes de la Neige (CEN, Météo-France) andextensively validated in temperate Alpine conditions. We present here astudy of Crocus’s ability to reproduce the characteristics of polar snow atthe surface of ice sheets. Crocus simulates the evolution of the thermal andstructural features of snow cover as a function of meteorological parametersat the snow-atmosphere interface. Only models can provide the necessarymeteorologic at information with full ice-sheet spatial coverage, and withthe temporal resolution needed by Crocus. Meteorological data have beenextracted from the European Centre for Medium-Range Weather Forecasts(ECMWF) archives (analyses and short-term predictions), over the entiresurface of Antarctica with a spatial resolution of 1.5°. Here, the ECMWFdata from the South Pole are first compared with observations to check theirquality. Then, 20 year simulations of snow covet are computed to test thesensitivity of Crocus to inaccuracies in the meteorological input. Thesimulated snow characteristics exhibit a strong sensitivity to airtemperature, accumulation rate and the initial density of depositing snow.However, even with no major model adaptation to polar conditions, Crocusdoes reproduce a number of thermal and structural features of polarsnow.

In situ measurements of ice stress were made on a multi-year floe in theAlaskan Beaufort Sea over a 6 month period, beginning in October 1993. Thedata suggest that, in this region of the Arctic during this experiment,there were two main sources of stress: a thermally induced stress caused bychanges in air temperature, and a stress generated by ice motion. Due to thenatural damping of the snow and ice above the sensor, the thermally-inducedstresses are low frequency (order of days). Stresses associated with periodsof ice motion have both a high-frequency (order of hours), andlow-frequency, content. The relative significance of these sources of stressis seasonal, reflecting the changes in the strength and continuity of thepack.

Experiments with dynamic thermodynamic sea-ice models indicate a strongdependence of the net freezing rate, sea-ice transport and variability ondynamic model parameters. Although current dynamic—thermodynamic sea-icemodels show relatively good agreement with observations, an optimizationseems to be necessary, especially for the parameterizations of dynamicprocesses.

Presently, only a few coupled climate models use dynamic-thermodynamicsea-ice models. In order to promote, by means of coordinated numericalexperiments, the development of an optimal sea-ice model for climateresearch, the Sea Ice Ocean Modelling Panel of the Arctic Climate SystemStudy (ACSYS, a project of the World Climate Research Programme hasinitiated the Sea Ice Model Intercomparison Project (SIMIP). The firstresults from this model hierarchy approach are presented.

The emphasis of this paper is on the partitioning of solar energy in an openplane-parallel melt pond with a Lambertian bottom. Spectral radiative-energyfluxes into the meltwater and underlying ice ocean layers are calculatedanalytically as a function of pond-bottom spectral albedo, pond depth andillumination condition Albedo of the pond bottom is reconstructed from dataon pond albedo and depth. Results of calculations are presented for meltponds of comparatively high and comparatively low reflectivity for a broadrange of pond depths and for various illumination conditions. In the 350–700nm spectral band, spectrally averaged pond albedo and solar-energy inputsare a function of pond-bottom albedo, pond depth and illumination condition.In the 700–2400 nm spectral band, the partitioning of solar energy in meltponds depends on pond depth and illumination condition only. The effect ofuncertainty in specifying pond-bottom albedo on total energy input into thewater layer is relatively small compared to that on spectrally averaged pondalbedo and total energy input into the ice-ocean layer.

There are currently a variety of one- and two-dimensional sea-ice modelsbeing used for climate simulations and sensitivity studies. Though all themodels can be timed to simulate current-day conditions to some degree ofaccuracy, the responses of each model to perturbations in forcing from theatmosphere or ocean are different. Thus, climate-change prediction dependson the choice of sea-ice model. In this study, the sensitivities of varioussea-ice models to external heat-flux perturbations are examined in asystematic manner. Starting from similar baseline annual thicknesses, eachmodel is subjected to an applied heat-flux perturbation to assess icemelt.Separate experiments are conducted to compare the response of each model toheat fluxes applied at the atmospheric and the oceanic interfaces. It isfound that the magnitude of the heat-flux perturbation required to melt icevaries greatly among different models, with the largest difference arisingbetween models that include ice dynamics vs those that do not. Most modelsshow an asymmetry in the response to heat-flux perturbations applied at thetop and bottom surfaces of the ice. This study has implications for thechoice of sea-ice models used for climate-change simulations. It also givesinsight to the accuracy required for observations and model simulations ofthe surface heat fluxes.

Many questions remain concerning whether or not an ice cap existed duringthe mid-Cretaceous period (120–90 Ma). Other than data and observations fromice-rafted materials, atmospheric general circulation models (AGCMs) may beappropriate tools to investigate whether changes in atmospheric composition,land-sea distribution, or oceanic circulation (used as boundary conditionsto constrain the model), provide the climatic conditions that enable icecaps to be formed.

This study uses an AGCM developed by the Laboratoire de MétéorologieDynamique (LMD) to perform a set of numerical sensitivity experiments toinvestigate plate tectonics (land–sea distribution and orography), CO2 partial pressure in the atmosphere and changes inprescribed sea-surface temperatures. The main goals are to quantify theeffect of each forcing factor on the increase of Northern Hemispherehigh-latitude temperatures, and to investigate whether combining thesefactors produce temperatures that would allow the ice-sheet formation.

Global climate models have pointed to the polar regions as very sensitiveareas in response to climate change. However, these models often do notcontain representations of processes peculiar to the polar regions such asdynamic sea ice, permafrost, and Arctic stratus clouds. Further, globalmodels do not have the resolution necessary to model accurately many of theimportant processes and feedbacks. Thus, there is a need for regionalclimate models of higher resolution. Our such model (ARCSy M) has beendeveloped by A. Lynch and W. Chapman. This model incorporates the NCARRegional Climate Model (RegCM2) with the addition of Flato–Hibler cavitatingfluid sea-ice dynamics and Parkinson–Washington ice thermodynamicformulation. Recently work has been conducted to couple a mixed-layer oceanto the atmosphere–ice model, and a three-dimensional (3-D) dynamical oceanmodel, in this case the S-Coordinate Primitive Equation Model (SPEM), to theice model. Simulations including oceanic circulation will allowinvestigations of the feedbacks involved in fresh-water runoff from sea-icemelt and sea-ice transport. Further, it is shown that the definition of themixed-layer depth has significant impact on ice thermodynamics.

Mean values of the oxygen-isotope ratio relative to standard mean oceanwater (δ18O, in ‰) reported for 46 sites on the Greenland ice sheet arecompiled together with data on mean annual surface temperature, latitude, δ18O elevation, and mean annual shortest distance to the openocean denoted by the 10% sea-ice concentration boundary. Stepwise regressionanalyses, with δ18O as the dependent variable, define two robust models. In theforward mode at the 99.9% confidence level, only temperature enters themodel. In the backward mode at the 95% confidence level, only temperature,latitude and distance to the open ocean remain in the model. Inversions ofthe models on the basis of 160 gridpoint locations 100 km apart in the areadelimited by the surface equilibrium line produce four contoureddistributions of δ18O. Two distributions are based on the bivarrate model and twoon the multivariate model. The second distribution for each model isobtained substituting mean annual surface-temperature values obtained fromthe Nimbus-7 Temperature Humidity Infrared Radiometer (THIR) database. Allfour distributions are considered valid, and differences between them areevaluated using contoured anomaly maps. It is suggested that the inversionof the multivariate model using THIR data provides the more reliable patternfor studies of atmospheric advection or for the derivation of ice-flowadjustments for δ18O series obtained from deep-core or ablation-zone sites.

A compilation of mean values of the oxygen-isotope ratio relative tostandard mean ocean water (δ18O, in ‰) for 22 siIesrepresentative of conditions in northeastern Canada is complemented withdata on mean annual surface temperature, latitude, surface elevation, andmean annual shortest distance to open ocean denoted by the 10% sea-iceconcentration boundary. Stepwise regression analysis is used to develop amultivariate model suitable to inter the distribution of δ18O inan area of complex topography and possibly mixed sourcing of advected watervapor. The best model is produced by a run in the backward mode at the 95%confidence level in which only temperature, latitude and distance to theopen ocean remain in the model (the correlation coefficient is 0.915, theadjusted coefficient of determination is 0.809, the root mean squareresidual is 1.62). This model is similar to the best δ18Opredictive model derived elsewhere for Greenland, suggesting a commonprincipal source of advected moisture.

Seasonal variations of water-balance components at an experimental watershedwere measured for three hydrologic years. Snowfall accounts for 47% of theannual precipitation, and snowmelt runoff comprises 45% of the total runoff.Evaporation changed markedly during the season; it increased during May toAugust, decreased in September and October, and was negligibly small duringthe period of snow cover in winter. Losses due to evaporation reached halfof the precipitation during summer, and total loss is about 26 % of theannual precipitation.

A conceptual precipitation–runoff model (HYCYMODEL) is used to simulate thedaily runoff from the 67% glacierized Lirung basin (16% of which is coveredby debris). The average thermal resistance of the debris cover of the LrungGlacier has been estimated from the surface temperature of the debris, whichwas obtained from Landsat 5 (TM Band 6) images. This enabled calculation ofthe melt rate of the debris-covered area of the glacier using meteorologicaldata collected at a nearby weather station. The calculated discharge resultsare roughly compatible with the observed daily discharges.