1. Introduction
The distributions of net annual mass balance on the surfaces of the Greenland and Antarctic ice sheets are important components of the present climate system. Over the last 104–106 yr they have dictated the existence and evolution of the ice sheets, and during the next ∼100 yr anthropogenic changes in these distributions may induce relatively rapid changes in ice-sheet volume and sea level. The present network of station data issparse over Greenland aud Antarctica, and observational estimates of precipitation and net mass balance are quite uncertain (e.g., van der Veen 1991). Global climate models (GCMs) have the capability of simulating precipitation, ablation, and mass balance over ice sheets, and moreover can predict changes in both thermal regime and circulation that may be important for ice sheets in climates other than the present (Bromwich 1995; Kapsner et al. 1995).
The main difficulty in simulating ice-sheet mass balances in GCMs is that of scale. The coarse horizontal resolutions in early GCMs (∼500 km) were inadequate to resolve the steep topographic slopes around the edges of Antarctica and Greenland, and the relatively narrow east–west extent of Greenland causes spurious lowering of its maximum elevation by ∼500 m in low resolution spectral GCMs. As a result, the simulated mass balances over Greenland and paleo-ice sheets were often much too negative (Manabe and Broccoli 1985; Rind 1987; Broccoli and Manabe 1993). Also the precipitation over Antarctica in such models was often too high due to problems with coarse spectral representation of the water vapor field (e.g., Tzeng et al. 1993; Bromwich et al. 1994). Another problem unrelated to resolution is that the snow and ice physics in most GCMs does not include important ice-sheet-specific processes such as the refreezing of meltwater that does not contribute to the mass balance.
Most studies of the Antarctic and Greenland response to future global warming have used fine-grid (≤50 km) ice-sheet models, forced by relatively simple surface mass balance parameterizations with assigned meteorological variations (e.g., Huybrechts and Oerlemans 1990; Letreguilly et al. 1991; Huybrechts 1994). Also considerable effort has been devoted to empirical parameterizations of accumulation and ablation based on degree days, altitude, slope, distance from the coast, and other variables (e.g., Giovinetto et al. 1990; Braithwaite et al. 1992; Oerlemans 1993; Bøggild et al. 1994). Many of these studies have included simple parameterizations to account for refreezing of meltwater. While this approach basically solves both the horizontal-scale and the refreezing problems, it cannot adequately capture changes in ice-sheet climate due to large-scale circulation shifts, for which GCMs are necessary.
Several recent studies have used higher-resolution GCMs to examine Greenland and Antarctic mass balances and obtained better results for Antarctica (Connolley and Cattle 1994; Genthon 1994; Tzeng et al. 1994) and both ice sheets (Gregory 1995; J. Gregory 1995, personal communication; Ohmura et al. 1996). The highest resolution reported to date is spectral T106 (1.1° lat/long) in 5.5-yr runs by Ohmura et al. (1996). Although such high resolutions can in principle yield very realistic mass balances, they generally limit the length of runs to only a few years, which introduces uncertainty due to interannual variability over the ice sheets and limits the scope of possible applications. The main purpose of this paper is to describe an alternate approach, whereby a medium-resolution atmospheric GCM (3.75° lat/long) is used and the meteorological fields are spatially interpolated at each time step to a higherhorizontal resolution (2° lat/long) for the surface models including ice sheets. In addition, a correction is made over ice sheets for the error between the GCM and the true topography, and a post-processing correction is made for the refreezing of meltwater following Pfeffer et al. (1991). With these techniques, the simulated present-day mass balances over Greenland and Antarctica are realistic, and the model is cheap enough to run multiple decade-long simulations for other climate scenarios. Such a model opens the door to a number of important applications:
Predicting rates of change of sea level due to anthropogenically induced changes in Greenland and Antarctic mass balance in the next century;
Performing “snapshot” simulations of the last glacial maximum at 21 kyr BP to examine the net budgets of the Laurentide and Eurasian ice sheets;
Performing ice-sheet initiation experiments at the end of the last interglacial at ∼115 kyr BP, allowing for small-scale high-latitude topography;
Driving two- or three-dimensional ice-sheet models through ∼104 yr, employing asynchronous coupling techniques (Pollard et al. 1990).
Verbitsky and Oglesby (1992, 1995), Schlesinger and Verbitsky (1996), and Verbitsky and Saltzman (1995) have taken pioneering steps in coupling a GCM with a dynamical ice-sheet model, mainly addressing the first topic above. In the first three papers, the GCM and ice-sheet models both used coarse horizontal resolutions (4.5° lat × 7.5° long or 4° lat × 5° long). In Verbitsky and Saltzman (1995), a high-resolution (40 km) ice-sheet model was used; however, the surface mass-balance forcing signal due to increased CO2 was scaled drastically to compensate for the GCM’s several-fold overprediction of present-day Antarctic precipitation, computed on the coarse GCM grid, and used without accounting for the scale or refreezing problems mentioned above. In a sense, the present paper supports their work by suggesting techniques to improve the coupling between GCMs and ice sheets.
In this paper we focus on the first topic above, and use the ice-sheet-specific techniques and a new version of the Global and Environmental and Ecological Simulation of Interactive Systems (GENESIS) global climate model to simulate Greenland and Antarctic mass balances in a pair of experiments for the present-day and for doubled atmospheric CO2. The difference in the net surface mass balance between the two experiments is an estimate of the influence on sea level toward the end of the next century, before the ice sheets’ dynamic response becomes significant (Budd 1988; Huybrechts and Oerlemans 1990; Huybrechts et al. 1991). Work is in progress to apply these techniques within the GENESIS global model to the other three topics listed above.
The global climate model used here is version 2.0.a of GENESIS, henceforth abbreviated to Version 2. It has been developed at the National Center for Atmospheric Research (NCAR) with emphasis on terrestrial physical, biophysical, and cryospheric processes for the purpose of performing greenhouse and paleoclimatic experiments. The model resolutions are spectral T31 (∼3.75° lat × 3.75° long) and 18 levels for the atmospheric GCM, and 2° lat × 2° long for the surface models. An earlier version 1.02 of the model with coarser AGCM resolution has been described in Thompson and Pollard (1995a,b) and Pollard and Thompson(1994, 1995), and has been used for both paleoclimate and future climate applications (e.g., Bonan et al. 1992; Barron et al. 1993; Crowley et al. 1993; Otto-Bliesner 1993; Sloan and Rea 1995; Pollard and Schulz 1994; Wilson et al. 1994). Version 2 of GENESIS includes several significant physical changes over earlier model versions, and the basic structure of the model and the recent changes are outlined in the appendix.
There are three basic tests that are available to gauge the quality of a GCM’s present-day mass balances over Greenland and Antarctica: (i) by comparing the global climate results with observed climatology, especially in high latitudes; (ii) by comparing seasonal fields such as temperature, precipitation, etc., over the ice sheets themselves with available observations; and (iii) by comparing the single values for the net annual precipitation and surface mass balance with observed estimates. This paper will address each of these tests in turn. Some aspects of the present-day climate for Version 2 are improved over earlier results, especially for high-latitude fields that strongly influence the mass balances of Greenland and Antarctica, and so the global distributions of some basic present-day fields for Version 2 are presented in section 2. Also, section 2 briefly describes the sensitivity of the global climate to doubled CO2 to provide some perspective for the local changes over the ice sheets. The remaining sections address the second and third tests. First, the seasonal cycles of precipitation over Greenland and Antarctica are compared to available observations (section 3). Then, the ice-sheet-specific techniques mentioned above are described in detail (section 4), followed by annual mean mass-balance results over present-day Greenland and Antarctica and their sensitivity to doubled CO2 (section 5). Uncertainties in the current results are discussed in section 6, and conclusions and directions for future work are summarized in section 7.
2. Global climate results
a. Present day
The present-day climate in GENESIS Version 2 is significantly improved in some regions compared to the earlier results for Version 1.02 described in Thompson and Pollard (1995a). Most of the improvements stem from the greater horizontal and vertical resolutions, the new prognostic cloud scheme, and adjustments to the convective plume scheme (see appendix). In particular precipitation, high-latitude surface temperatures, and snow and sea-ice distributions are all more realistic. Since these fields strongly influence the mass balances over Greenland and Antarctica and are not yet reported elsewhere, they are described in the following section and compared to observations where available. In addition, we will present the basic global sensitivity of the model to doubled CO2 since the CO2-induced changes over the ice sheets do not occur in isolation but are strongly influenced by the model’s regional and global sensitivity. All model results shown below are averages over the last 10 yr of fully equilibrated GCM runs.
The surface-air temperatures for January and July are compared with observations in Figs. 1a–d. The general magnitude of the temperature errors is considerably smaller than previously obtained (Thompson and Pollard 1995a, their Fig. 15), especially over Northern Hemispheric land masses in January. There is still a large cold bias in both hemispheres over the Himalayan plateau and China, which we suspect is due to erroneous wintertime dynamical effects of the Himalayan plateau. Also, much of the Antarctic interior is ∼10°C too warm in winter. Fortunately, this error does not affect the Antarctic snow and ice mass budgets since temperatures are still much too cold for any melting, and the interior precipitation is realistic (see below). Over central Greenland in both seasons temperatures are about 3°C too warm, even after the elevation correction described below in section 4. However, the error is much less than what it would be without the correction (∼7°–12°C over much of Greenland, see below).
The global and annual mean precipitation in Version 2 is 3.2 mm day−1, within the range of some climatologies (e.g., Legates and Willmott 1990) and much less than the earlier value of 4.5 mm day−1 in Version 1.02. Most of the reduction is due to correcting an erroneously high value of the ocean roughness length, and also to using wind-dependent drag coefficients over the ocean and adjusting properties of the convective plumes as noted in the appendix. The zonal means of precipitation now agree well with the observations (Fig. 2). The global patterns in Fig. 3 are quite similar to the earlier results except that the east Asian summer monsoon is somewhat more realistic (Fig. 3c). We found that the quality of the Indian monsoon is quite sensitive to small changes in nearby sea surface temperatures (SSTs); for instance, it is improved considerably by using prescribed monthly SSTs instead of a slab ocean, as shown in Fig. 4. We discuss precipitation over Greenland and Antarctica in more detail below.
The snow distribution for January is shown in Fig. 5. Correct snowfall amounts are of course vital to good simulations of ice-sheet mass balance, and over the Northern Hemispheric continents the predicted fractional cover is quite realistic and the snow depths are within observational uncertainty (e.g., Robock 1980; Foster and Chang 1993). There is some improvement over the snow cover in Version 1.02 (Foster et al. 1996), due mostly to a more realistic parameterization of fractional cover versus depth (see appendix). By July (not shown) all Northern Hemispheric snow melts except on Greenland, and trace amounts (too little to accumulate annually) on Arctic sea ice and the central Himalayas.
Sea-ice distributions are particularly important to the regional climate and mass balances of Greenland and coastal Antarctica. The Arctic distributions shown in Fig. 6 are improved over Version 1.02 (cf. Pollard and Thompson 1994) and are surprisingly close to observed, especially considering that the winds driving the ice dynamics in Version 2 are the instantaneous AGCM winds and are not prescribed from climatology as in Version 1.02. The earlier spurious “spiky” behavior near the North Pole is no longer as evident, probably due to a new smoother prescription of the surface ocean currents as noted in the appendix. The publication of the Beaufort gyral with observed sea-ice distribution is anticipated (J. Maslanick, personal communication 1996). circulation in the Arctic creates larger ice thicknesses on the western (Greenland–Canadian) side of the basin as observed (Bourke and Garrett 1987), and iceis advected down the eastern coast of Greenland realistically (Gloersen et al. 1992). The Norwegian Sea region is kept ice-free in winter as observed via a crude modification of the prescribed ocean heat flux as in Thompson and Pollard (1995a). A more detailed assessment of the Arctic climate in GENESIS, including comparisons with observed sea-ice distribution, will be given in J. Maslanik et al. (1996, in preparation).
Around Antarctica (Fig. 7) the sea-ice fractions and thicknesses are also reasonable. By March most of the ice has melted except in the western Weddell Sea, as observed. In September high (>80%) fractional cover extends out to ∼60° in all sectors, which is realistic for the Atlantic and Indian sectors but not for the Pacific hemisphere (∼135°–270°E) where the observed ice extends only to ∼65°S (Gloersen et al. 1992). The seasonal ice thicknesses in September range from ∼0.5 to 2.0 m, within the bounds of observations (Budd 1986).
The model sea level pressure fields (Fig. 8) show some improvement over those for Version 1.02 (Thompson and Pollard 1995a; their Fig. 12), as might be expected due to the higher horizontal resolution of the AGCM dynamics. The core of the Icelandic low is now correctly situated in January, which is important for the wintertime precipitation over southern Greenland (see below); however, its spatial extent is greater than observed. There is a suggestion of a wintertime Baffin Bay low to the west of Greenland, with Greenland situated in a weak saddle between these lows as observed (Ohmura and Reeh 1991). The depth of the circum-Antarctic trough at ∼60°S is now more realistic although still not as great as observed by about 10 mb in January. The greatest model error occurs over central Antarctica, where the sea level pressure (∼1020 mb) is much greater than observed (∼995 mb). Although some of this error could be due to different algorithms used to extrapolate surface pressure to sea level, most must be due to model error. It is common to many coarse and medium-resolution GCMs and may be due to insufficient resolution to handle the dynamical effects of the steep Antarctic topography (Bromwich et al. 1995). In our model this does not seem to have any serious effects on the simulated precipitation or summer temperatures in Antarctica, which are reasonably realistic (see below), and so we remain confident in our simulations of Antarctic mass balance. A similar error of excessive sea level pressure over the Arctic and Greenland in July is evident in Fig. 8c. However, the spatial pattern of Greenland summer precipitation does not seem to be seriously affected (see section 3), perhaps because the same caveat as above about extrapolation to sea level applies over Greenland. Moreover, the model sea level pressure and low-level winds over the Arctic in all other seasons are realistic (J. Maslanik 1996, personal communication).
b. Doubled CO2
One of the purposes of this paper is to predict the change in net annual surface mass balance of Greenland and Antarctica after the climate equilibrates to a doubling of atmospheric CO2. This should correspond roughly to the response toward the end of the next century, neglecting effects due to ocean thermal lag (Schneider and Thompson 1981; Gregory 1995) and ice-sheet dynamics (Budd 1988; Huybrechts and Oerlemans 1990; Huybrechts et al. 1991). Since there is significant scatter in the global and regional sensitivities to doubled CO2 of different GCMs, there would presumably be similar scatter in their results for ice-sheet mass balance if the techniques used here were applied in each model. Consequently, it is necessary to bear in mind the global CO2 sensitivity of our GCM in interpreting our predicted changes in ice-sheet mass balance.
As a result of doubling atmospheric CO2 from 345 to 690 ppmv, the annual and global mean surface-air temperature in GENESIS Version 2 warms by 2.5°C and the global precipitation increases by 4.4%. These changes are within the envelope of recent GCM results (IPCC 1990, 1996; Mitchell 1991). As expected, the surface warming is amplified in the high latitudes of each winter hemisphere (Figs. 9a,b), due to (i) albedo feedback of snow and sea ice, (ii) weakening of shallow polar-winter surface inversion layers, and (iii) reductions in winter sea ice exposing the atmosphere to the drastically warmer ocean surface (Schlesinger and Mitchell 1987). The patch of relatively small warming in the Norwegian Sea in winter is due to our prescription of a large oceanic heat flux keeping the region ice-free for both 1 × and 2 × CO2 (as observed for the present day and presumably for warmer worlds, but see section 6c). The changes in winter and summer surface-air temperatures are significant at the 95% confidence level nearly everywhere, as shown in Figs. 9a,b (i.e., there is less than 5% chance that they could be due to interannual variability and the 10-yr averaging period alone).
Changes in annual mean precipitation (Fig. 10a) are largest in the Tropics due to intensification and latitudinal shifts of the ITCZ (Fig. 10a). In most of the subtropics and midlatitudes, the changes are smaller and not significant at the 95% confidence level [cf. Boer et al. (1992), who also found widespread areas of statistically insignificant precipitation changes]. One exception is the summer monsoonal rainfall in the Sahel and central Sahara, which decreases markedly. Of more relevance to the present application, over most high-latitude regions, there are quite small but statistically significant increases in annual precipitation on the order of 0.1–0.3 mm day−1, due mostly to increases in winter snowfall. Much the same small but significant high-latitude increases in zonal mean precipitation were found by Boer et al. (1992). Precipitation and ablation over the ice sheets will be examined further below; we will find that the CO2-induced changes in the mean ice-sheet budget terms are highly significant and much bigger than their interannual variations, as might be expected from the areas of local significance shown in Figs. 9 and 10a.
The change in annual mean upper-soil moisture is shown in Fig. 10b and closely follows the patterns of precipitation change. There is considerable drying in northern and central Africa, and wetting in the Northern Hemispheric continental interiors. Elsewhere, the changes are smaller and statistically insignificant. Future changes in soil moisture may have large impacts on human societies, but, unfortunately, there has been little agreement to date in the regional patterns predicted by various GCMs.
3. Seasonal precipitation over ice sheets
Inthis section we begin to focus on the model performance over present-day Greenland and Antarctica, starting with the seasonal cycles of precipitation. As discussed by Barry and Kiladis (1982) and Ohmura and Reeh (1991), and as evident in station data (Putnins 1970), different areas of Greenland experience quite different seasonal cycles of precipitation and wind regimes. According to Ohmura and Reeh, in the wintertime the northern flank of the Icelandic low steers moist air from the Atlantic into the southeast slope of Greenland, producing a region with wintertime precipitation maxima and a coastal belt of large annual precipitation (shown later in section 5). In the summer, onshore westerly winds flow into the western and northwestern flanks of the ice sheet, causing pronounced summertime orographic precipitation there and a band of large annual values running roughly north–south along the western slope. Figure 11 shows that the model is simulating this general picture well, with strong fall–winter (SON–DJF) maxima around the southern coasts and summer–fall (JJA–SON) maxima on the western flanks of the ice sheet in its midlatitudes. The simulated cycles also agree qualitatively with much of the station data in Putnins (1970) but with a few notable exceptions such as Godthaab (on the west coast at 64°N), which has an observed maximum in late summer and fall and a minimum in winter. We feel that a good simulation of the seasonal cycles of precipitation is particularly important for the net Greenland mass balance since there is significant summertime snow and ice melt at lower elevations and the local mass balances are affected by the temporal correlations of melting and precipitation.
The large-scale seasonal character of precipitation for Antarctica, shown for the model in Fig. 12, is not as well known as for Greenland. The general patterns inferred by Bromwich (1988) from a few station data are (i) maxima in March over the Ross and Weddell Seas, (ii) June–August maxima around the East Antarctic coast and probably also over the Eastern plateau, and (iii) January and July maxima at Byrd (center of West Antarctica). The model cycles in Fig. 12 only partially agree with this picture. We obtain general maxima over the Ross and Weddell Seas and/or ice shelves in austral fall (MAM), and a maximum at Byrd in fall–winter (MAM–JJA), more or less as observed. However, around much of the East Antarctic coast, the model has pronounced maxima in fall (MAM), not in June–August as observed (cf. Genthon and Braun 1995).
4. Ice-sheet-specific techniques
One obvious prerequisite for reliable predictions of future mass-balance changes is that the model should yield reasonably realistic values for present-day ice sheets. Initially, our values for present-day Greenland using GENESIS Version 2 without any of the ice-sheet-specific techniques described below were very unrealistic, with net annual surface ablation and mass-balance values of −80 and −37 cm yr−1, respectively, compared to observed estimates of about −15 and +14 cm yr−1.
a. Correction for AGCM versus true topography
There are two obvious candidates for the causes of the ablation error. First, even though the surface ice budgets are computed using a 2° lat × 2° long grid, the lowest-level temperature, downward infrared radiation flux, and other meteorological variables are still obtained from the AGCM with its 3.75° lat × 3.75° long resolution. The latter resolution is inadequate to resolve the steep topography on the flanks of Greenland, withonly 2–5 grid boxes spanning the ice sheet in the east–west direction equatorward of 70°N. Moreover, our AGCM topography and dynamics uses T31 spectral truncation, which spuriously lowers most Greenland elevations by ∼300–600 m. This leads to surface-air temperatures that are about 4°C too warm, producing summer melt rates that are too large by about ∼60 cm month−1, and an annual ablation bias of ∼30 cm yr−1 (Pollard 1980; van de Wal and Oerlemans 1994).
Since the difference between the true topography and the AGCM topography is known, it is possible to apply a simple correction to the AGCM lowest-level (∼50 m) meteorological fields as they are passed to the surface models at each time step, by assuming a globally uniform and constant lapse rate. For instance, the lowest-level air temperature is simply lowered by a constant lapse rate multiplied by the difference between the true elevation minus the AGCM elevation. This is done at each 2° lat × 2° long grid point after the meteorological variables have been interpolated from the AGCM grid, using the true 2° lat × 2° long topography and the truncated AGCM topography interpolated to 2° lat × 2° long. We emphasize that these corrections are applied only to the meteorological values passed to the surface models; the corresponding AGCM prognostic variables are not changed since that would constitute a nonconservation of energy in the AGCM and destroy the near-geostrophic dynamic balance between the temperature and wind fields. Note also that we do not use the AGCM vertical profiles to interpolate vertically since the free atmospheric profiles predicted by the AGCM have no obvious relation to changes in the lowest-level quantities resulting from a vertical shift in the surface itself.
Lowest-level air temperature, changed by γΔz where γ = −7.0 °C km−1 is the uniform lapse rate mentioned above, and Δz is the true elevation minus the AGCM elevation. The value of −7.0°C km−1 is chosen to represent observed values around Greenland in summer (Ohmura 1987, his Table 2).
Lowest-level specific humidity, adjusted to keep relative humidity constant at the value given by the AGCM.x
Surface pressure, obtained by starting at the AGCM’s value and integrating the hydrostatic equation through a vertical distance Δz with constant lapse rate.
Downward infrared (IR) radiative flux, obtained from the AGCM’s value multiplied by the ratio of two empirical estimates: one for conditions at the corrected surface and the other for conditions at the uncorrected surface. The empirical expression is from a clear-sky parameterization by Idso (1981) and a correction for cloudiness (Sellers 1965, 58):
As mentioned above, these corrections are applied to the meteorological variables after they are horizontally interpolated from the AGCM gridto the surface-model grid (3.75° lat × 3.75° long to 2° lat × 2° long). Hence the “AGCM” topography is actually the AGCM’s coarse-grid spectrally truncated topography interpolated to the surface grid, and the “true” topography is obtained from the 10-min FNOC dataset (Cuming and Hawkins 1981; Kineman 1985) aggregated to the 2° surface grid. The effect of the correction on summer surface-air temperatures over Greenland is shown in Fig. 13. The uncorrected AGCM temperatures (Fig. 13a) are much warmer than the observed temperatures by ∼7°–12°C over much of the interior, and they reflect the “truncated-blob” nature of the AGCM’s Greenland topography. The corrected temperatures (Fig. 13b) are still about 3°C warmer than observed, but the pattern is obviously much closer to reality.
We do not attempt to adjust the AGCM cloudiness for the elevation correction. Averaged over large space and timescales above the lowest few kilometers of the atmosphere, cloudiness does generally decrease with height, but we do not attempt this for several reasons. One is that in regions of steep topography, cloudiness depends strongly on wind direction relative to the slope, unlike surface-air temperatures, which are dominated by the lapse rate. Also, the effect of cloudiness on incident solar radiation is dramatically curtailed over high-albedo surfaces due to multiple reflections between the surface and the cloud base (Schneider and Dickinson 1976).
We also do not attempt to adjust precipitation rates for the elevation correction. Like cloudiness, orographically forced precipitation depends strongly on the wind direction relative to the local topographic slope, especially on the flanks of Greenland and Antarctica. Leung and Ghan (1995) have introduced a scheme that does compute subgrid precipitation from large-scale grid variables and subgrid topography, but still without taking wind direction into account, which could possibly be used in the future. However, as shown above seasonally and below for annual means, our AGCM precipitation fields over Greenland and Antarctica are adequately realistic and capture the basic regions of orographically forced precipitation on the flanks of the ice sheets and the “desert” regions of low precipitation on the plateaus. (This improvement compared to earlier coarse-grid GCMs is probably due to the use of semi-Lagrangian advection for water vapor, which avoids problems caused by spectral truncation of the water vapor field). We feel that as far as ice-sheet mass balance is concerned, the dominant effect of the imperfect AGCM topography is the error in air temperature, which is amenable to the simple corrections described above.
For the two GCM experiments reported in this paper the elevation corrections were applied just over Greenland and Antarctica. In principle, they could be applied at all surface points and could potentially affect snow accumulations over the Himalayas and Andes where the AGCM elevations are severely truncated. However, for other regions of the world well below the annual snowline, the corrections would not strongly affect the AGCM climate. That is because the solid-surface temperatures would adjust to maintain the same net energy flux with the atmosphere (whose annual mean over land must stay close to zero since the surface models conserve energy) and, in the absence of snow-albedo feedback, the AGCM’s surface fluxes and climate would remain essentially unchanged.
b. Correction to true topography near ice-sheet edge
The true topography used for the elevation correction at each surface grid point is computed by simply aggregating the 10-min FNOCelevation dataset (Cuming and Hawkins 1981; Kineman 1985) within each 2° lat × 2° long surface cell. However, for cells spanning the edge of an ice sheet some of the fine-grid elevations represent open ocean, so the aggregated value does not represent the mean elevation of the ice alone. As mentioned in the appendix, the land-surface tranfer (LSX) model can perform two independent calculations for each cell, one for land/ice-sheet and one for ocean/sea-ice. Since we are most interested in ice-sheet mass balance, we can improve the relevance of the elevation correction by aggregating only those 10-min points over ice sheet and ignoring those over ocean or land. The effect is to raise the true topography somewhat over the real-world values for the 2° lat × 2° long boxes spanning the ice-sheet edge, which has been done for the GCM experiments described in this paper. Of course the computation only needs to be performed once before the GCM is run to calculate and save a single two-dimensional field of Δz values needed for the elevation corrections at each time step.
c. Correction for refreezing of meltwater
5. Ice-sheet mass-balance results
a. Present-day Greenland and Antarctica
Observational networks on Greenland and Antarctica are sparse, and the harsh environment makes measurements of precipitation, accumulation, and ablation difficult. Measurements are further complicated by local refreezing, windblown snow, and other factors (van der Veen 1991). Nevertheless, recent maps of observed annual precipitation or accumulation for Greenland and Antarctica are available, and generally agree on the broadest scales but differ significantly in smaller regions (Bromwich 1988; Ohmura and Reeh 1991). Several estimates of annual precipitation and surface net mass balance integrated over the entire ice sheets are also available (e.g., Warrick and Oerlemans 1990; van der Veen 1991; Warrick et al. 1996), although these values have large uncertainties (Meier 1990, 1993). Also, it is not clear how close the present ice volumes are to equilibrium, that is, how closely the mean annual surface budget is balanced by other forms of mass loss (mainly basal runoff, iceberg/iceshelf discharge, and wind drift).
In presenting our results below, we will use precipitation (snowfall plus rainfall) rather than accumulation, which has various usages in the literature (sometimes including or excluding rainfall, evaporation, and/or windblown snow). We will use the term “ablation” to mean that part of the local snow melt, ice melt, and rainfall on bare ice that escapes according to Eqs. (2) and (3) plus evaporation, and “mass balance” to mean the net surface budget given by B in (2).
The predicted annual precipitation for Greenland is compared with Ohmura and Reeh’s (1991) map in Fig. 14. All large-scale features agreewell with the observations, including the minimum values of ∼10–20 cm yr−1 on the northern plateau, a value of 23 cm yr−1 at the Greenland Ice Sheet Project 2 site (Alley et al. 1993), and maximum values of ∼80–100 cm yr−1 near the southeast and southern coasts. The band of high precipitation trending north–south along the western flank is also captured to some extent, although not quite as pronounced as observed. The overall agreement is at least as good as other GCM results using comparable or higher-resolution models (Genthon et al. 1994; Ohmura et al. 1996).
The annual ablation and the mass balance are shown for Greenland in Fig. 15. As expected, large amounts of ablation (Fig. 15a) occur at low elevations near the southern perimeter, due mostly to summer ice melt. The small values of ablation (<5 cm yr−1) in the central region are due to the refreezing parameterization not allowing local runoff to escape; our values of local runoff (mostly snow melt and ice melt, not shown) are 10–20 cm yr−1 over most of the central plateau. In reality there is little or no local snow melt over much of the plateau (e.g., Abdalati and Steffen 1995), and the simulated melt is probably due to our summer temperatures still being ∼3° too warm even after the elevation correction (see Figs. 1d and 13b). The net mass balance (Fig. 15b) shows a pronounced maximum centered at ∼64°N that corresponds to a local topographic plateau near high bedrock, and negative values are limited to near the edges as expected. Incidentally, the ice caps on Ellesmere island (northwest of Greenland) have unrealistic negative mass balances in Fig. 15b because no elevation or refreezing correction was applied there. (They are not included in the area-average budgets for Greenland reported below.)
The corresponding annual fields for Antarctica are shown in Figs. 16 and 17. The overall patterns and magnitudes of precipitation agree well with observations (Fig. 16). Very low values (<∼5 cm yr−1) extend over much of the East Antarctic plateau, and most of the flanks nearer the coast have values on the order of 20–40 cm yr−1. Higher values are experienced in some regions very close to the coast, except on the Ross and Filchner–Ronne ice shelves where the precipitation is relatively low as observed. The model fails to capture the observed high values (up to 100 cm yr−1) on the coast west of the Antarctic peninsula at ∼100°W and incorrectly predicts a small region of very high values at the coast in Enderby Land (∼50°E); however, these anomalies are isolated and are too small in area to seriously bias the overall mass balance. High values of ablation (Fig. 17a) are limited to isolated coastal pockets, especially around the Antarctic peninsula. The small ablation values of less than 2 cm yr−1 over most of the interior are due entirely to sublimation. As will be shown below, the all-ice-sheet ablation is a very small component of the net annual mass balance but is quite significant in monthly averages for December and January.
In the remainder of this section, we discuss the net annual budgets averaged over the entire surface area of each ice sheet. These annual budget terms are compared with observed estimates in Table 1. There is somescatter in the various observed values (Warrick and Oerlemans 1990; van der Veen 1991; Meier 1993; Warrick et al. 1996), and we have tried to choose values roughly in the midrange of recent estimates. Our areal integrations for Antarctica include the floating ice shelves, which are often excluded in observed estimates since they do not directly contribute to sea level change; however, their exclusion makes only very small differences in our mean values in Table 1 (less than ∼0.5 cm yr−1). For both Greenland and Antarctica, the precipitation (top row of Table 1) is somewhat greater than observed, but not drastically so, considering the uncertainty in the data. The next few rows of Table 1 show the effects of not applying the various corrections described in section 4. With no corrections at all, the ablation and net mass balance on Greenland are wildly in error (−80 and −37 cm yr−1, respectively). The departures from reality are removed more than halfway by the elevation corrections alone (−43 and +1 cm yr−1, respectively). With the additional use of the refreezing correction, the ablation and net mass balance are in much better agreement with observations (−31 and +13 cm yr−1, compared with observed estimates of −15 and +14 cm yr−1).
For Antarctica, the corrections have relatively little effect (as expected since ablation is very small), but they do reduce ablation from −5 cm yr−1 to a possibly more realistic −3 cm yr−1. Using all corrections, the net surface ablation and mass balance are −3 and +18 cm yr−1, compared to observed estimates of ∼0 and +15 cm yr−1.
b. Doubled CO2
When atmospheric CO2 is doubled and the GCM is rerun to climatic equilibrium, the annual precipitation, ablation, and mass balance on Greenland change as shown in Figs. 18a–c. Both precipitation and ablation increase everywhere, but the latter dominates and the mass balance becomes more negative in nearly all locations. As shown in Table 2, the areal mean mass balance drops from +13 to −12 cm yr−1, equivalent to a global sea level rise of +1.2 mm yr−1. This rate is of course a prediction of the change in Greenland’s contribution to sea level due to doubled CO2, in addition to whatever its present-day contribution is and whatever other factors may influence future sea level. The same approximately two-fold increase in Greenland ablation has been found in a recent high-resolution (T106) GCM study by Ohmura et al. (1996), although their precipitation decreased very slightly. Our Greenland results are also quite consistent with the more parameterized mass balance sensitivities in Huybrechts et al. (1991) and van de Wal and Oerlemans (1994). [Boer et al. (1992) report a decrease in the area and amount of permanent snow cover over Greenland in their doubled CO2 experiment but do not report changes of total precipitation and ablation.]
As shown in parentheses in Table 2, the uncertainty in the 10-yr means of Greenland precipitation and ablation due to interannual variability is quite small, as expected from the areas of local significance in Figs. 9and 10a. The uncertainties in the net balances are relatively larger, but the resulting range in Greenland’s sea level influence, from +0.9 to +1.5 mm yr−1, still does not swamp the basic result. The same point will be evident in Fig. 22 (later in this section), which shows graphically that the mean budget changes due to doubled CO2 are considerably larger than the interannual variability.
The crudity of the elevation and refreezing corrections described in section 4 is another source of uncertainty for the mass-balance results. As an example amenable to postprocessing analysis, we have varied the “ramp” value in the refreezing parameterization [0.7 in Eq. (3)] over a large but plausible range (0.5–0.9) and recomputed the mass-balance results. The ramp value is quite uncertain and represents the effects of the density, winter temperature, and mobility of the snowpack, as well as the horizontal heterogeneity of meltwater flow networks. The resulting ranges in ablation and net balance are shown in Table 2 (second value in parentheses) and are still quite small compared to the mean values. Furthermore, we may hope that this type of systematic error stays nearly the same in each 1 × and 2 × CO2 experiment and so cancels in the final sea level result, which is why it does not appear in the last row of Table 2. This hope of course lies behind a great many GCM sensitivity experiments where the signal under investigation is comparable to the errors in the model’s present-day climate. Whether it is justified for our application, or whether the 2 × CO2 and the sea level results may have larger systematic errors than suggested in Table 2, will be discussed in section 6.
For Antarctica (Figs. 19a–c), precipitation increases by ∼5–15 cm yr−1 all around the flanks, whereas ablation increases only in a few limited areas (Ross Ice Shelf, western Ronne Ice Shelf, and the Antarctic peninsula). Unlike Greenland, the changes in mass balance are dominated by the increased precipitation. As shown in Table 2, the areal mean mass balance increases from +18 to +21 cm yr−1, equivalent to a global sea level drop of −1.3 mm yr−1. Again, this is the predicted future change in Antarctica’s contribution to sea level, in the same sense as mentioned above for Greenland. Similar CO2-induced increases in Antarctic precipitation that dominate changes in ablation have also been predicted by a variety of other models (Warrick and Oerlemans 1990; Huybrechts and Oerlemans 1990; Gregory 1995; J. Gregory 1995, personal communication; Ohmura et al. 1996; Warrick et al. 1996). In doubled-CO2 experiments with the GISS GCM, Rind et al. (1995) found a decrease in Antarctic surface mass balance due to large increases in their “runoff;” however, their coarse-grid model did not use elevation or refreezing corrections as here, and their present-day mean Antarctic precipitation and runoff are both considerably greater than observed.
The uncertainties due to interannual variability of the overall precipitation and net balance for Antarctica (Table 2) are smaller than for Greenland, probably because of the larger area of Antarcticacompared to individual storms. The uncertainty in the sea level influence, −1.6 to −1.0 mm yr−1, is as large as Greenland’s (due to the much greater area of Antarctica), but again it does not swamp the basic result. Again, Fig. 23 (later in this section) will demonstrate graphically that changes in the Antarctic means are well above the general level of interannual variability. Varying the ramp value in the refreezing parameterization between 0.5 and 0.9 has little effect on the net mass balance simply because surface ablation is a relatively minor component of the Antarctic budget for up to 2 × CO2 levels.
These results are for present ice-sheet configurations and for a climate that has fully equilibrated to a doubling of CO2. During the next century, the real climate will lag the transient CO2 increase by a few decades due to the thermal inertia of the upper ocean (Schneider and Thompson 1981), and the ice sheets will begin to respond dynamically and change their surface elevations, grounding line locations, iceberg/iceshelf discharge rates, and possibly the rate of basal melting and outflow. However, these dynamic changes will probably not become significant for at least 100 yr or more, at least for Antarctica (Budd 1988; Huybrechts and Oerlemans 1990; Huybrechts et al. 1991), so our results should be most relevant toward the end of the next century when CO2 levels are projected to reach twice the preindustrial value. According to our results, the surface mass balance changes in Greenland and Antarctica largely cancel out and will yield a sea level drop of only 1.2 minus 1.3 = −0.1 mm yr−1 compared to their present contribution. Similar cancellation between the ice sheets has been found recently by at least two other GCM studies (Gregory 1995; J. Gregory 1995, personal communication; Ohmura et al. 1996). This change in ice-sheet contribution is much smaller than the net sea level rise of ∼5 mm yr−1 expected in the next century due mostly to thermal expansion of the oceans and increased melting of smaller glaciers (Warrick and Oerlemans 1990; Warrick et al. 1996); however, that conclusion could be weakened by uncertainties in our results as discussed in section 6.
c. Seasonal cycles
Since the refreezing parameterization [Eqs. (2) and (3)] depends intrinsically on annual mean quantities and is nonlinear, there is no validity in applying it on a month-by-month basis, and so we cannot show seasonal cycles of ablation and mass balance. However, a general idea of the seasonality of the budget can be obtained from the areal mean precipitation and local runoff (i.e., snow melt, ice melt, and rainfall on bare ice regardless of whether it escapes the ice sheet), as shown in Figs. 20 and 21.
The seasonal cycles of areally averaged precipitation and local runoff over present-day Greenland (Fig. 20) are as expected. The precipitation has a relatively weak cycle (averaging out strong seasonal variations over different regions as mentioned above), and local runoff has a strong pulse in the summer and drops to nearly zero in the winter. Bromwich et al. (1993) have used analyzed National Meteorological Center (now National Centers for Environment Prediction) data to parameterize precipitation over Greenland for1963–88, and they find a seasonal cycle of precipitation (their Fig. 15) with about the same amplitudes as ours but with a sharp maximum in August and minimum in February, compared to our broad maximum in August–November and minimum in March–April. Perhaps some of the discrepancy is due to their underestimation of orographic precipitation on the northwest flanks (as they discuss), which in our model occurs mainly in summer and fall as seen in Fig. 11. The increases in precipitation due to doubled CO2 shown in Fig. 20 are relatively small and uniform throughout the year, whereas local runoff rates increase by ∼50% in the peak summer months.
For present-day Antarctica (Fig. 21), there is a small seasonal cycle in precipitation that peaks in March and April. As discussed earlier in connection with Fig. 12 above, different regions of Antarctica experience different seasonal precipitation cycles (Bromwich 1988), and the net cycle is not well constrained by the sparse station data. Observed estimates of circum-Antarctic water vapor convergence rates in Peixoto and Oort (1992, their Fig. 12.12) are larger in June–August than in December–February, but, as they state, “The corresponding (monthly) curve for the transport across 70°S is not shown because the present transport data are not reliable enough to determine the seasonal variation at those latitudes.” The local runoff in Fig. 21 is relatively small and occurs only in December and January due mostly to snow and ice melt in isolated coastal regions, as shown for ablation in Fig. 15a. When CO2 is doubled, the local runoff rates more than double in December and January, but this is dominated by the uniform increase in precipitation throughout the year.
d. Interannual variability
The interannual variability of the large-scale budgets over Greenland and Antarctica are of interest for several diverse reasons. Oerlemans (1979) has suggested that the “white-noise” forcing due to interannual variability can cause a significant long-term red-noise response in ice-sheet extent. In projects using Global Positioning System crustal motion measurements around Antarctica, the large-scale interannual variability of precipitation is a factor in their interpretation (James 1995). And as discussed above and shown in Table 2, interannual variability introduces some uncertainty in both observed and GCM results that are necessarily averaged over a finite number of years.
Figure 22 shows the annual budget quantities for Greenland over the 10 individual years of both 1 × and 2 × CO2 experiments. The interannual variability of the net mass balance for Greenland is surprisingly high, with standard deviations of 5.4 and 6.8 cm yr−1 for the two experiments. Precipitation variations contribute 24% and ablation variations contribute 76%, and there is negligible correlation between the two. In future work we will examine the spatial and seasonal characteristics of the variability and its cause. Bromwich et al. (1993) found a somewhat higher level of interannual variability than ours in the overall Greenland precipitation but with a general downward trend from 1963–88 and a 3–5 yr periodicity (their Fig. 9). These surprisingly large values suggest caution in using only ahandful of years to estimate the net mass balance over Greenland. However, as evident in Fig. 22 and shown quantitatively in Table 2, the change in the 10-yr mean mass balance due to doubled CO2 is considerably larger than the uncertainty caused by interannual variability.
The levels of interannual variability are much lower for Antarctica (Fig. 23), probably due to its greater size compared to individual storms. Bromwich (1988) discusses the interannual variability of Antarctic precipitation and suggests that for the continent as a whole it is less than ∼7% of the mean, consistent with Fig. 23. However, this does not preclude larger interannual variations on smaller spatial scales, especially around the coastal regions, which will be investigated in future work.
6. Uncertainties in mass-balance results
This study is one of the first to apply ice-sheet-specific downscaling techniques from GCM to ice-sheet scales, and there are some uncertainties in the results, especially for doubled CO2. Scatter in our 10-yr means due to interannual variability was shown in section 5 to be relatively minor, but the possibility of systematic biases in the model physics certainly exists. Two types of questions can be asked: (i) to what extent are our realistic present-day mass balances over Greenland and Antarctica due to canceling errors in the GCM or correction physics and (ii) do these errors remain the same in the 2 × CO2 mass-balance results and so cancel with the present day, or is their effect so different in the 2 × CO2 world that our sensitivity results are seriously in doubt?
a. Greenland
For present-day Greenland, our precipitation is overestimated by about 50% and is canceled by a similar overestimate of ablation (Table 1). Although the observed values are uncertain by at least ±20% (Warrick and Oerlemans 1990; van der Veen 1991; Warrick et al. 1996), our simulated values still lie somewhat outside the probable range. Our overestimate of precipitation could be due to several factors: excessive orographic precipitation on the steep western flanks of Greenland, spectral truncation of Greenland topography allowing too warm and moist air over the ice sheet, or a large-scale bias in the AGCM precipitation (perhaps associated with the summertime sea level pressure error noted for Fig. 8c). However, none of these would obviously vary systematically as CO2 is doubled. Moreover, our change in zonal-mean high-latitude GCM precipitation due to doubled CO2 agrees well with Boer et al. (1992); also the predicted slight increase over Greenland agrees with Huybrechts et al. (1991) and is consistent with Ohmura et al. (1996) in the sense that the change is much smaller than the change in ablation. Hence, we surmise that the uncertainty in our mean precipitation change for Greenland is relatively small.
Although our corrected summer surface-air temperatures over Greenland agree well with the overall observed pattern, especially the location of the freezing line (Fig. 13b and c), temperatures on the central plateau are still about 3°C too warm. Even though the simulated July temperatures there are −5° to −10°C, the diurnal cycle allows some melting, whereas in reality there is little or no snow melt over much of the plateau (e.g., Abdalati and Steffen 1995). The resulting annual ablation (after refreezing) in those regions is only 0–5 cm yr−1 (Fig. 15a), not enough to account for the areal mean overestimate of 16 cm yr−1 in Table 1. The mean error is probably due to excessive melting around the southern and western flanks, where the annual ablation locally reaches 0.5–3 m yr−1.
However, small errors in present-day temperature can potentially cause serious problems in the sensitivity of ablation to doubled CO2. The response of total ice-sheet ablation to climate warming depends not only on the magnitude of the warming in the existing ablation zone below the present-day snowline (roughly where summer temperatures are above freezing), but also on the areal expansion of the ablation zone upslope to the new snowline. If the elevation of our present-day snowline is too close to the relatively flat interior, then a climate warming might unrealistically expand the ablation zone over vast areas of the plateau instead of over a small strip on the much steeper flanks. Figure 18b shows that the model response is closer to the former, with significant increases of ablation over much of the interior (although most of the response comes from larger increases at lower elevations around the south). Therefore, our approximately two-fold increase in mean ablation (Table 2) may be overestimated, due to the warm error in present-day summer temperatures and the nonlinear response of ablation to temperature change.
This can be tested by comparing with other model studies that do not have our bias in present-day summer temperatures. The T106 GCM experiment of Ohmura et al. (1996) and the more parameterized mass-balance/ice-dynamics models of Huybrechts et al. (1991) and van de Wal and Oerlemans (1994) all find approximately two-fold increases in mean Greenland ablation due to doubled CO2, as here. (The latter two studies prescribe temperature increases representing transient CO2 scenarios and include results corresponding roughly to our doubled CO2 warming.) However, the present-day and doubled-CO2 ablation values in these studies are both somewhat less than ours (about −22 and −44 cm yr−1, respectively, compared to −31 and −69 cm yr−1 here). This rough agreement implies that our ablation sensitivity is not overwhelmingly biased but may be too great by about 50%. If we were to retune our present-day precipitation and ablation to agree with the other models, we estimate that our CO2-induced change in Greenland mass balance could be reduced by about 50%, and its contribution to sea level could shrink from +1.2 to about +0.6 mm yr−1.
b. Antarctica
For Antarctica, the situation is simpler. Our present-day distribution of precipitation is realistic (Fig. 16), and the areal mean value is only slightly overestimated and probably within the range of observational uncertainty (Table 1). When CO2 is doubled, mean precipitation increases by a modest amount, as predicted by many studies (Warrick and Oerlemans 1990; Huybrechts and Oerlemans 1990; Boer et al. 1992; Gregory 1995; J. Gregory 1995, personal communication; Ohmura et al. 1996; Warrick et al. 1996), although ourincrease of 6 cm yr−1 is about double that found in most of the other models.
Surface ablation is a minor component of the Antarctic budget since away from the low-lying coastal regions surface-air temperatures never get warm enough to allow significant snow melt. The same remains true for doubled CO2, and ablation increases only in narrow strips near the coast. The resulting small increase in areal mean ablation is dominated by the increase in mean precipitation, as found in all of the studies mentioned above. The large wintertime warm temperature errors over Antarctica (Fig. 1d) have no bearing on ablation since the model temperatures are still much too cold to allow any melting. The much smaller summertime temperature errors in the coastal ablation zones (Fig. 1a) are more of a concern but probably have little influence on ablation-zone expansion and the mean ablation change because of the steepness of the ice-sheet flanks. The sensitivity problem discussed above for Greenland (due to present-day summer temperature errors and the nonlinear response of ablation) does not arise for Antarctica since the narrow coastal ablation zone remains a very small fraction of the total area when CO2 is doubled (Figs. 17a and 19b).
Hence, comparison of our Antarctic results with those of other models suggests that our CO2-induced increase in precipitation might be a factor of 2 too large and ablation might actually remain close to zero instead of increasing slightly. If both of these possibilities were true, the predicted Antarctic influence on sea level would remain at about −1.3 mm yr−1. But if either one were true and not the other, that value could range between about −2.6 and 0 mm yr−1.
c. Model limitations
Apart from uncertainties suggested by the results-oriented discussion above, other types of uncertainty stem of course from known limitations in the model formulation itself. In particular, the refreezing parameterization used here is only a first step toward including ice-sheet-specific surface physics in a GCM. The snow and ice-sheet surface models used within the current version of GENESIS were developed without ice-sheet mass-balance applications foremost in mind. In the future it would be preferable to use a somewhat more sophisticated snow model to explicitly capture some of important processes on real ice sheets: percolation of surface melt, refreezing, and compaction of snow to firn and ice. Loth et al. (1993) have shown that at least some of these processes are feasible within GCM snow models. In addition, the assumption that rainfall on bare ice immediately becomes local runoff should be improved. These extensions could conceivably replace the current refreezing parameterization that is based on annual total quantities and so necessarily ignores seasonal correlations. Also, the distinction between snowfall and rainfall is currently made by the AGCM depending on the uncorrected AGCM lowest-level temperature; however, it would be preferable for the elevation correction procedure to convert the phase of incoming precipitation between snow and rain if necessary based on the corrected air temperature. This distinction would affect the ratio of snowfall to rainfall especially around the flanks of Greenland and so could change the mass balance via the refreezing parameterization (3). We have evaluated this effect crudely (ignoring diurnal cycles) by recomputing the ice-sheet budgets a posteriori using stored monthly mean fields of precipitation and either uncorrected or corrected surface-air temperatures; using the latter, we find that the mean annual ablation for Greenland decreases by 2 cm yr−1 (present-day)and 4 cm yr−1 (doubled CO2) from the values in Table 2, so that the net Greenland sea level influence decreases from +1.2 to +1.1 mm yr−1—a small change compared to the uncertainties discussed above. For Antarctica there is negligible change to the results in Table 2 since scarcely any precipitation falls as rain.
Another limitation of the model formulation is of course the use of a 50-m slab mixed layer instead of a 3D ocean general circulation model (OGCM). Transient experiments over the next century or more with coupled ocean–atmosphere GCMs have shown that the deep ocean can have a significant effect on the surface warming due to its thermal lag and to changes in the thermohaline circulation. When CO2 is held constant at two times its present value for several hundred more years, the thermohaline circulation does recover to nearly its present value (Manabe and Stouffer 1994). But since the real world will actually undergo a transient experiment in the next few centuries, coupled atmosphere–OGCM (A–OGCM) transient experiments obviously have more potential as a predictive tool than equilibrated 2 × CO2 simulations. However, a hierarchy of coupled models and types of experiments is still useful to explore basic sensitivities of the climate system and to explore new techniques and regional applications as here (Meehl 1992). The lack of a deep ocean introduces some error in our current predictions, and our ice-sheet-specific techniques will hopefully be included in some future transient A–OGCM simulations. But the cost of such experiments and the additional complexity of the deep-ocean dynamical response would have hindered the development of these techniques to date.
In our slab ocean model we have chosen not to use “flux corrections” that would guarantee exact replication of present-day SSTs at every point, but have attempted to parameterize the actual present-day oceanic heat convergence (see appendix). As part of this parameterization the region of enhanced heat flux in the Norwegian Sea maintains perennial ice-free conditions as observed for the present day. There is no guarantee of course that this Norwegian Sea parameterization will remain as realistic as CO2 increases; the perennial ice-free region may expand as found in several coupled A–OGCMs (e.g., Washington and Meehl 1996) or may shrink due to a more stable halocline caused by increasing Arctic runoff (Manabe and Stouffer 1994). However, within the limitations of the slab ocean model we feel that the Norwegian Sea parameterization for present and increased CO2 simulations is certainly better than the alternative of none at all, which in our model would allow excessive perennial ice across the entire North Atlantic down to ∼65°N for the present day and excessive warming feedback as the spurious ice retreats for doubled CO2.
7. Conclusions
There are two main problems in attempting to simulate realistic mass balances on ice sheets from GCM simulations.
The relative small scale of the ice-sheet topography, especially for Greenland which is severely truncated by AGCM resolutions of ∼250 km or coarser.
Ice-sheet-specific snow and ice processes, especially the refreezing of meltwater.
Although to some extent this agreement may be due to canceling errors in the GCM and the correction physics (section 6), without any corrections at all the mass balances produced by the GCM would be wildly unrealistic (Table 1). An important conclusion of this paper is that the elevation and refreezing corrections are large, approximate, but physically justifiable adjustments that are necessary to produce realistic ice-sheet budgets in low to medium resolution GCMs. We feel that Fig. 13 especially bears this out for the elevation correction. Section 6 discussed uncertainties in our CO2 sensitivity results due mainly to errors in the GCM climate, but such discussion would not even be possible without the correction techniques to get the model ice-sheet budgets in the right ball park.
In summary, when atmospheric CO2 is doubled our net annual surface mass balance decreases on Greenland from +13 to −12 cm yr−1 and increases on Antarctica from +18 to +21 cm yr−1. The effects on global sea level are +1.2 and −1.3 mm yr−1, yielding a residual change in sea level contribution of −0.1 mm yr−1. However, errors in the GCM climate and comparisons with other models (see section 6) suggest that there is significant uncertainty in these values: the sea level influence of Greenland could range between about +0.6 and +1.2 mm yr−1 and that of Antarctica could range between about −2.6 and 0 mm yr−1. The combined change in contribution could thus range from −2.0 to +1.2 mm yr−1. This is an approximate projection of the change in contribution toward the end of the next century, assuming ice-sheet dynamical effects (changing elevations, iceberg/iceshelf, discharge, and basal melt) remain unimportant for that long. Our estimated change in the ice sheets’ contribution to sea level would be a relatively small and possibly insignificant component of the total sea level rise of ∼5 mm yr−1 expected in the next century due predominantly to thermal expansion of the oceans and increased melting of smaller glaciers (Warrick and Oerlemans 1990; Warrick et al. 1996).
Acknowledgments
We thank Wei-Chyung Wang, Xin-Zhong Liang, and others (SUNY at Albany) for a new infrared radiation code that includes individual greenhouse gases; Jon Bergengren (NCAR) for providing the noninteractive part of his equilibrium vegetation ecology model; and Tad Pfeffer (University of Colorado, Boulder) for a helpful conversation on refreezing parameterizations. Thanks are also due to two anonymous reviewers, especially for one’s request for the extensive discussion of uncertainties in section 6. The development of the GENESIS earth system model at NCAR is supported in part by the U.S. Environmental Protection Agency Interagency Agreement DW49935658-01-0. The National Center for Atmospheric Research is sponsored by the National Science Foundation.
REFERENCES
Abdalati, W., and K. Steffen, 1995: Passive microwave-derived snow melt regions on the Greenland ice sheet. Geophys. Res. Lett.,22, 787–790.
Alley, R. B., and Coauthors, 1993: Abrupt increase in Greenland snow accumulation at the end of the Younger Dryas event. Nature,362, 527–529.
Ambach, W., 1989: Effects of climaticperturbations on the surface-ablation regime of the Greenland ice sheet, West Greenland. J. Glaciol.,35, 311–316.
Anthes, R. A., 1977: A cumulus parameterization scheme utilizing a one-dimensional cloud model. Mon. Wea. Rev.,105, 270–286.
Barron, E. J., W. W. Peterson, D. Pollard, and S. L. Thompson, 1993: Past climate and the role of ocean heat transport: Model simulations for the Cretaceous. Paleoceanography,8, 785–798.
Barry, R. G., and G. N. Kiladis, 1982: Climatic characteristics of Greenland. Climatic and Physical Characteristics of the Greenland Ice Sheet, U. Radok, R. G. Barry, D. Jenssen, R. A. Keen, G. N. Kiladis, and B. McInnes, Eds., CIRES, 7–33.
——, ——, and D. Pollard, 1996: Modeling the effects of vegetation change on climate sensitivity: Coupled GENESIS–EVE experiments with 1× and 2× CO2. Climate Change, in press.
Boer, G. J., N. A. McFarlane, and M. Lazare, 1992: Greenhouse gas-induced climate change simulated with the CCC second-generation general circulation model. J. Climate,5, 1045–1077.
Bøggild, C. E., N. Reeh, and H. Oerter, 1994: Modelling ablation and mass-balance sensitivity to climate change of Storstrommen, Northeast Greenland. Global Planet. Change,9, 79–90.
Bonan, G. B., D. Pollard, and S. L. Thompson, 1992: Effects of boreal forest vegetation on global climate. Nature,359, 716–718.
Bourke, R. H., and R. P. Garrett, 1987: Sea ice thickness distribution in the Arctic Ocean. Cold Reg. Sci. Technol.,13, 259–280.
Braithwaite, R. J., O. B. Olesen, and H. H. Thomsen, 1992: Calculated variations of annual ice ablation at the margin of the Greenland ice sheet, West Greenland, 1961–90. J. Glaciol.,38, 266–272.
Briegleb, B., and V. Ramanathan, 1982: Spectral and diurnal variations in clear sky planetary albedo. J. Appl. Meteor.,21, 1160–1171.
Broccoli, A. J., and S. Manabe, 1993. Climate model studies of interactions between ice sheets and the atmosphere–ocean system. Ice in the Climate System, W. R. Peltier, Ed., NATO ASI Series, Vol. I, Springer-Verlag, 271–290.
Bromwich, D. H., 1988: Snowfall in high southern latitudes. Rev. Geophys.,26, 149–168.
——, 1995: Ice sheets and sea level. Nature,373, 18–19.
——, F. M. Robasky, R. A. Keen, and J. F. Bolzan, 1993: Modeled variations of precipitation over the Greenland ice sheet. J. Climate,6, 1253–1268.
——, R.-Y. Tzeng, and T. R. Parish, 1994: Simulation of the modern Arctic climate by the NCAR CCM1. J. Climate,7, 1050–1069.
——, B. Chen, and X. Pan, 1995: Intercomparison of simulated polar climates by global climate models. Preprints, Sixth Symp. on Global Climate Change, Dallas, TX, Amer. Meteor. Soc., 14–19.
Budd, W. F., 1986: The Southern Hemisphere circulation of atmosphere, ocean, and sea ice. Preprints, Second Int. Conf. on Southern Hemisphere Meteorology, Wellington, New Zealand, Amer. Meteor. Soc., 101–106.
——, 1988: The expected sea level rise from climate warming in the Antarctic. Greenhouse: Planning for Climate Change, G. I. Pearman, Ed., E. J.Brill Publishing, 74–82.
Cess, R. D., and Coauthors, 1995: Absorption of solar radiation by clouds: Observations versus models. Science,267, 496–499.
Chervin, R. M., and S. H. Schneider, 1976: On determining the statistical significance of climate experiments with general circulation models. J. Atmos. Sci.,33, 405–412.
Clapp, R. B., and G. M. Hornberger, 1978: Empirical equations for some soil hydraulic properties. Water Resour. Res.,14, 601–604.
Cogley, J. G., 1991: GGHYDRO—Global hydrographic data release 2.0. Trent Climate Note 91-1, 10 pp. [Available from Dept. of Geography, Trent University, Peterborough, ON K97 7B8, Canada.].
Connolley, W. M., and H. Cattle, 1994: The Antarctic climate of the UKMO unified model. Antarc. Sci.,6, 115–122.
Cosby, B. J., G. M. Hornberger, R. B. Clapp, and T. R. Ginn, 1984: A statistical exploration of the relationships of soil moisture characteristics to the physical properties of soils. Water Resour. Res.,20, 682–690.
Crowley, T. J., S. K. Baum, and K. Y. Kim, 1993: General circulation model experiments with pole-centered supercontinents. J. Geophys. Res.,98, 8793–8800.
Crutcher, H. L., and J. M. Meserve, 1970: Selected level heights, temperatures and dew points for the Northern Hemisphere. NAVAIR Publ. 50-1C-52, 161 pp. [Available from Superintendent of Documents, U.S. Government Printing Office, Washington, DC 20402.].
Cuming, M. J., and B. A. Hawkins, 1981: TERDAT: The FNOC system for terrain data extraction and processing. Tech. Rep. M11 Project M254, 2d ed. [Available from U.S. Navy Fleet Numerical Oceanography Center, Code 42, Monterey, CA 93943.].
Dickinson, R. E., 1984: Modeling evapotranspiration for three-dimensional global climate models. Climate Processes and Climate Sensitivity, Geophys. Monogr., No. 29, Amer. Geophys. Union, 58–72.
——, A. Henderson-Sellers, P. J. Kennedy, and M. F. Wilson, 1986: Biosphere–Atmosphere Transfer Scheme (BATS) for the NCAR Community Climate Model. NCAR Tech. Note NCAR/TN-275+STR, 69 pp. [Available from National Center for Atmospheric Research, P.O. Box 3000, Boulder, CO 80307.].
Donald, J. R., E. D. Soulis, N. Kouwen, and A. Pietroniro, 1995: A land cover-based snow cover representation for distributed hydrologic models. Water Resour. Res.,31, 995–1009.
Eltahir, E. A. B., and R. L. Bras, 1993: Estimation of the fractional coverage of rainfall in climate models. J. Climate,6, 639–644.
Flato, G. M., and W. D. Hibler, 1990: On a simple sea-ice dynamics model for climate studies. Ann. Glaciol.,14, 72–77.
——, and ——, 1992: Modeling pack ice as a cavitating fluid. J. Phys. Oceanogr.,22, 626–651.
Foster, J. L., and A. T. C. Chang, 1993: Snow cover. Atlas of Satellite Observations Related to Global Change, R. J. Gurney, J. L. Foster, and C. L. Parkinson, Eds., Cambridge University Press, 361–370.
——, G. Liston, R. Koster, R. Essery, H. Behr, L. Dumenil, D. Verseghy, S. Thompson, D. Pollard, and J. Cohen, 1996: Snow cover and snow mass intercomparisons of general circulation models and remotely sensed datasets. J. Climate,9, 409–426.
Genthon, C., 1994: Antarctic climatemodeling with general circulation models of the atmosphere. J. Geophys. Res.,99, 12 953–12 961.
——, and A. Braun, 1995: ECMWF analyses and prediction for the surface climate of Greenland and Antarctica. J. Climate,8, 2324–2332.
——, J. Jouzel, and M. Deque, 1994: Accumulation at the surface of polar ice sheets: Observation and modelling for global climate change. Global Precipitation and Climate Change, M. Desbois and F. Desalmand, Eds., NATO ASI Series, Vol. I, Springer-Verlag, 117–130.
Giovinetto, M., and C. R. Bentley, 1985: Surface balance in ice drainage systems of Antarctica. Antarc. J. U.S.,20, 6–13.
——, N. M. Waters, and C. R. Bentley, 1990: Dependence of Antarctic surface mass balance on temperature, elevation, and distance to open ocean. J. Geophys. Res.,95, 3517–3531.
Gloersen, P., W. J. Campbell, D. J. Cavalieri, J. C. Comiso, C. L. Parkinson, and H. J. Zwally, 1992: Arctic and Antarctic sea ice, 1978–1987: Satellite passive microwave observations and analysis. NASA Scientific and Technical Information Program SP-511, 290 pp. [Available from STIP, NASA, Washighton, DC 20402.].
Gregory, J. M., 1995: Prediction of sea-level changes using a coupled ocean–atmosphere GCM. Abstracts, Int. Union of Geodesy and Geophysics XXI General Assembly, Boulder, CO, IUGG, B317.
Harvey, L. D. D., 1988: Development of a sea ice model for use in zonally averaged energy balance climate models. J. Climate,1, 1221–1238.
Hibler, W. D., 1979: A dynamic thermodynamic sea ice model. J. Phys. Oceanogr.,9, 815–846.
——, and K. Bryan, 1987: A diagnostic ice-ocean model. J. Phys. Oceanogr.,17, 987–1015.
Huybrechts, P., 1994: Formation and disintegration of the Antarctic ice sheet. Ann. Glaciol.,20, 336–340.
——, and J. Oerlemans, 1990: Response of the Antarctic ice sheet to future greenhouse warming. Climate Dyn.,5, 93–102.
——, A. Letreguilly, and N. Reeh, 1991: The Greenland ice sheet and greenhouse warming. Palaeogeogr., Palaeoclimatol., Palaeoecol.,89, 399–412.
Idso, S. B., 1981: A set of equations for full spectrum and 8–14 μm and 10.5–12.5 μm thermal radiation from cloudless skies. Water Resour. Res.,17, 295–304.
Intergovernmental Panel on Climate Change, 1990: Equilibrium climate change and its implications for the future. Climate Change: The IPCC Scientific Assessment, J. T. Houghton, G. J. Jenkins, and J. J. Ephraums, Eds., Cambridge University Press, 131–172.
——, 1996: Climate models: Projections of future climate. Climate Change 1995: The Science of Climate Change, J. T. Houghton, L. G. Meira Filho, B. A. Callander, N. Harris, A. Kattenberg, and K. Maskell, Eds., Cambridge University Press, 285–358.
James, T. S., 1995: The effect of Antarctic ice mass changes on crustal motion and global geodetic observables. Abstracts, Int. Union of Geodesy and Geophysics XXI General Assembly, Boulder, CO, IUGG, B317.
Kapsner, W. R., R. B. Alley, C. A. Shuman, S. Anandakrishnan, and P. M. Grootes, 1995: Dominant influence of atmospheric circulation on snow accumulation in Greenland over thepast 18,000 years. Nature,373, 18–19.
Kineman, J., Ed., 1985: FNOC/NCAR Global Elevation, Terrain, and Surface Characteristics. Digital Dataset, 28 MB. NOAA National Geophysical Data Center.
Kreitzberg, C. W., and D. J. Perkey, 1976: Release of potential instability: Part I. A sequential plume model within a hydrostatic primitive equation model. J. Atmos. Sci.,33, 456–475.
Kuchler, A. W., 1983: World map of natural vegetation. Goode’s World Atlas, 16th ed., Rand McNally, 16–17.
Large, W. G., and S. Pond, 1981: Open ocean momentum flux measurements in moderate to strong winds. J. Phys. Oceanogr.,11, 324–336.
Leemans, R., and W. P. Cramer, 1991: The IIASA database for mean monthly values of temperature, precipitation and cloudiness on a global terrestrial grid. Research Rep. RR-91-18, WP-41, 60 pp. [Available from International Institute for Applied Systems Analysis, Laxenburg, Austria.].
Legates, D. R., and C. J. Willmott, 1990: Mean seasonal and spatial variability in gauge-corrected, global precipitation. Int. J. Climatol.,10, 111–127.
Letreguilly, A., N. Reeh, and P. Huybrechts, 1991: The Greenland ice sheet though the last glacial–interglacial cycle. Palaeogeogr., Palaeoclimatol., Palaeoecol.,90, 385–394.
Leung, L. R., and S. J. Ghan, 1995: A subgrid parameterization of orographic precipitation. Theor. Appl. Climatol.,52, 95–118.
Loth, B., H.-F. Graf, and J. M. Oberhuber, 1993: Snow cover model for global climate studies. J. Geophys. Res.,98, 10 451–10 464.
MacFarlane, N. A., 1987: The effect of orographically excited gravity wave drag on the general circulation of the lower stratosphere and troposphere. J. Atmos. Sci.,44, 1775–1800.
Manabe, S., and A. J. Broccoli, 1985: The influence of continental ice sheets on the climate of an ice age world. J. Geophys. Res.,90, 2167–2190.
——, and R. J. Stouffer, 1994: Multiple-century response of a coupled ocean–atmosphere model to an increase of atmospheric carbon dioxide. J. Climate,7, 5–23.
Matthews, E., 1983: Global vegetation and land use: New high-resolution data bases for climatic studies. J. Climate Appl. Meteor.,22, 474–487.
Meehl, G. A., 1992: Global coupled models: Atmosphere, ocean, sea ice. Climate System Modeling, K. E. Trenberth, Ed., Cambridge University Press, 555–581.
Meier, M. F., 1990: Role of land ice in present and future sea-level change. Sea-Level Change. Studies in Geophysics, National Academy Press, 171–184.
——, 1993: Ice, climate, and sea level; do we know what is happening? Ice in the Climate System, W. R. Peltier, Ed., NATO ASI Series, Vol. I, Springer-Verlag, 141–160.
Mitchell, J. F. B., 1991: The equilibrium response to doubling atmospheric CO2. Greenhouse-Gas-induced Climatic Change: A Critical Appraisal of Simulations and Observations, M. E. Schlesinger, Ed., Developments in Atmospheric Science, No. 19, Elsevier, 49–61.
Oerlemans, J., 1979: A model of astochastically driven ice sheet with planetary wave feedback. Tellus,31, 469–477.
——, 1993: Modelling of glacier mass balance. Ice in the Climate System, W. R. Peltier, Ed., NATO ASI Series, Vol. I 12, Springer-Verlag, 101–116.
Ohmura, A., 1987: New temperature distribution maps for Greenland. Z. Gletscherk. Glazialgeol.,23, 1–45.
——, and N. Reeh, 1991: New precipitation and accumulation maps for Greenland. J. Glaciol.,37, 140–148.
——, M. Wild, and L. Bengtsson, 1996: A possible change in mass balance of Greenland and Antarctic ice sheets in the coming century. J. Climate,9, 2124–2135.
Otto-Bliesner, B. L., 1993: Tropical mountains and coal formation: A climate model study of the Westphalian (306 Ma). Geophys. Res. Lett.,20, 1947–1950.
Paterson, W. S. B., 1969: The Physics of Glaciers. Pergamon Press, 250 pp.
Peixoto, J. P., and A. H. Oort, 1992: Physics of Climate. American Institute of Physics, 520 pp.
Pfeffer, W. T., M. F. Meier, and T. H. Illangasekare, 1991: Retention of Greenland runoff by refreezing: Implications for projected future sea-level change. J. Geophys. Res.,96, 22 117–22 124.
Pollard, D., 1980: A simple parameterization for ice-sheet ablation rate. Tellus,32, 384–388.
——, and M. Schulz, 1994: A model for the potential locations of Triassic evaporite basins driven by paleoclimatic GCM simulations. Global Planet. Change,9, 233–249.
——, and S. L. Thompson, 1994: Sea-ice dynamics and CO2 sensitivity in a global climate model. Atmos.–Ocean,32, 449–467.
,——, and ——, 1995: Use of a land-surface-transfer scheme (LSX) in a global climate model: The response to doubling stomatal resistance. Global Planet. Change,10, 129–161.
——, and ——, 1997a: Climate and ice-sheet mass balance at the last glacial maximum from the GENESIS Version 2 global climate model. Quat. Sci. Rev., in press.
——, and ——, 1997b: Driving a high-resolution dynamic ice-sheet model: Ice-sheet initiation at 116 Kyr BP. Ann. Glaciol.,25, in press.
——, I. Muszynski, S. H. Schneider, and S. L. Thompson, 1990: Asynchronous coupling of ice sheet and atmospheric forcing models. Ann. Glaciol.,14, 247–251.
Putnins, P., 1970: The climate of Greenland. World Survey of Climatology. Vol. 14, Climates of the Polar Regions, S. Orvig, Ed., Elsevier Publishing, 3–128.
Ramanathan, V., B. Subasilar, G. J. Zhang, W. Conant, R. D. Cess, J. T. Kiehl, H. Grassl, and L. Shi, 1995: Warm pool heat budget and shortwave cloud forcing: A missing physics? Science,267, 499–503.
Rind, D., 1987: Components of the ice age circulation. J. Geophys. Res.,92, 4241–4281.
——, R. Healy, C. Parkinson, and D. Martinson, 1995: The role of sea ice in 2×CO2 climate model sensitivity. Part I: The total influence of sea ice thickness and extent. J. Climate,8, 449–463.
Robock, A., 1980: The seasonal cycle of snow cover, sea ice, andsurface albedo. Mon. Wea. Rev.,108, 267–285.
Schlesinger, M. E., and J. F. B. Mitchell, 1987: Climate model simulations of the equilibrium climatic response to increased carbon dioxide. Rev. Geophys.,25, 760–798.
——, and M. Verbitsky, 1996: Simulation of glacial onset with a coupled atmosphere general circulation/mixed-layer ocean–Ice-sheet/asthenosphere model. Palaeoclimates,2, 179–201.
Schneider, S. H., and R. E. Dickinson, 1976: Parameterization of fractional cloud amounts in climate models: The importance of multiple reflections. J. Appl. Meteor.,15, 1050–1056.
——, and S. L. Thompson, 1981: Atmospheric CO2 and climate: Importance of the transient response. J. Geophys. Res.,86, 3135–3147.
Sellers, P. J., Y. Mintz, Y. C. Sud, and A. Dalcher, 1986: A simple biosphere model (SiB) for use within general circulation models. J. Atmos. Sci.,43, 505–531.
Sellers, W. D., 1965: Physical Climatology. University of Chicago Press, 272 pp.
Semtner, A. J., 1976: A model for the thermodynamic growth of sea ice in numerical investigations of climate. J. Phys. Oceanogr.,6, 379–389.
Senior, C. A., and J. F. B. Mitchell, 1993: Carbon dioxide and climate: The impact of cloud parameterization. J. Climate,6, 393–418.
Shea, D. J., 1986: Climatological atlas: 1950–1979. Surface-air temperature, precipitation, sea-level pressure, and sea-surface temperature. NCAR Tech. Note NCAR/TN-269+STR, 35 pp. plus 158 figs. [Available from National Center for Atmospheric Research, P.O. Box 3000, Boulder, CO 80307.].
——, K. E. Trenberth, and R. W. Reynolds, 1992: A global monthly sea surface temperature climatology. J. Climate,5, 987–1001.
Simmons, A. J., D. M. Burridge, M. Jarraud, C. Girard, and W. Wergen, 1989: The ECMWF medium-range prediction models development of the numerical formulations and the impact of increased resolution. Meteor. Atmos. Phys.,40, 28–60.
Sloan, L. C., and D. K. Rea, 1995: Atmospheric carbon dioxide and early Eocene climate: A general circulation modeling sensitivity study. Palaeogeogr., Palaeoclimatol., Palaeoecol.,93, 183–202.
Smith, R. N. B., 1990: A scheme for predicting layer clouds and their water content in a general circulation model. Quart. J. Roy. Meteor. Soc.,116, 435–460.
Sugden, D. E., and B. S. John, 1976: Glaciers and Landscape. Wiley Publishing, 376 pp.
Taljaard, J. J., H. Van Loon, H. L. Crutcher, and R. L. Jenne, 1969: Climate of the Upper Air: Southern Hemisphere. Vol. 1, Temperatures, Dew Points and Heights at Selected Pressure Levels, NAVAIR Publ. 50-1C-55, 6 pp. plus 134 figs. [Available from Naval Weather Service Command, Washington Navy Yard Building 200, Washington, DC 20390.].
Thomas, G., and A. Henderson-Sellers, 1991: An evaluation of proposed representations of subgrid hydrologic processes in climate models. J. Climate,4, 898–910.
Thompson, S. L., and D. Pollard, 1995a: A global climate model (GENESIS) with a land-surface-transfer scheme (LSX). Part I: Present-day climate. J. Climate,8, 732–761.
——, and ——, 1995b: A global climate model(GENESIS) with a land-surface-transfer scheme (LSX). Part II: CO2 sensitivity. J. Climate,8, 1104–1121.
——, and ——, 1997: Ice-sheet mass balance at the last glacial maximum from the GENESIS Version 2 global climate model. Ann. Glaciol.,25, in press.
——, V. Ramaswamy, and C. Covey, 1987: Atmospheric effects of nuclear war aerosols in general circulation model simulations: Influence of smoke optical properties. J. Geophys. Res.,92, 10 942–10 960.
Trenberth, K. E., J. G. Olson, and W. G. Large, 1989: A global ocean wind stress climatology based on ECMWF analyses. NCAR Tech. Note NCAR/TN-338+STR, 93 pp. [Available from National Center for Atmospheric Research, P.O. Box 3000, Boulder, CO 80307.].
Tzeng, R.-Y., D. H. Bromwich, and T. R. Parish, 1993: Present-day Antarctic climatology of the NCAR Community Climate Model Version 1. J. Climate,6, 205–226.
,——, ——, ——, and B. Chen, 1994: NCAR CCM2 simulation of the modern Antarctic climate. J. Geophys. Res.,99, 23 131–23 148.
van der Veen, C. J., 1991: State of balance of the cryosphere. Rev. Geophys.,29, 433–455.
van de Wal, R. S. W., and J. Oerlemans, 1994: An energy balance model for the Greenland ice sheet. Global Planet. Change,9, 115–131.
Verbitsky, M. Ya., and R. J. Oglesby, 1992: The effect of atmospheric carbon dioxide concentration on continental glaciation of the Northern Hemisphere. J. Geophys. Res.,97, 5895–5909.
——, and ——, 1995: The CO2-induced thickening/thinning of the Greenland and Antarctic ice sheets as simulated by a GCM (CCM1) and an ice-sheet model. Climate Dyn.,11, 247–253.
——, and B. Saltzman, 1995: Behavior of the East Antarctic ice sheet as deduced from a coupled GCM/ice-sheet model. Geophys. Res. Lett.,22, 2913–2916.
Wang, W.-C., M. P. Dudek, X.-Z. Liang, and J. T. Kiehl, 1991: Inadequacy of effective CO2 as a proxy in simulating the greenhouse effect of other radiatively active gases. Nature,350, 573–577.
Warrick, R., and J. Oerlemans, 1990: Sea level rise. Climate Change: The IPCC Scientific Assessment, J. T. Houghton, G. J. Jenkins, and J. J. Ephraums, Eds., Cambridge University Press, 261–281.
——, C. Le Provost, M. F. Meier, J. Oerlemans, and P. L. Woodworth, 1996: Changes in sea level. Climate Change 1995. The Science of Climate Change. Contribution of Working Group 1 to the Second Assessment Report of the Intergovernmental Panel on Climate Change, J. T. Houghton, L. G. Meira Filho, B. A. Callander, N. Harris, A. Kattenberg, and K. Maskell, Eds., Cambridge University Press, 359–406.
Washington, W. M., and G. A. Meehl, 1996: High latitude climate change in a global coupled ocean–atmosphere sea ice model with increased atmospheric CO2. J. Geophys. Res.,101, 12 795–12 801.
Webb, R. S., C. E. Rosenzweig, and E. R. Levine, 1993: Specifying land surface characteristics in general circulation models: Soil profile data set and derived water-holding capacities. Global Biogeochem. Cycles,7, 97–108.
Williamson, D. L., and P. J. Rasch, 1989: Two-dimensional semi-Lagrangiantransport with shape-preserving interpolation. Mon. Wea. Rev.,117, 102–129.
——, J. T. Kiehl, V. Ramanathan, R. E. Dickinson, and J. J. Hack, 1987: Description of NCAR Community Climate Model (CCM1). NCAR Tech. Note NCAR/TN-285+STR, 112 pp. [Available from National Center for Atmospheric Research, P.O. Box 3000, Boulder, CO 80307.].
Wilson, K. M., D. Pollard, W. W. Hay, S. L. Thompson, and C. N. Wold, 1994: General circulation model simulations of Triassic climates: Preliminary results. Pangea: Paleoclimate, Tectonics and Sedimentation during Accretion, Zenith and Breakup of a Supercontinent, G. D. Klein, Ed., Geological Society of America, 91–116.
APPENDIX
Global Climate Model Description
The global climate model used here is Version 2.0.a of GENESIS. The version number is abbreviated below to Version 2. It has beendeveloped at NCAR since 1989 with emphasis on terrestrial physical, biophysical, and cryospheric processes, for the purpose of performing greenhouse and paleoclimatic experiments. An earlier Version 1.02 of the model has been described in Thompson and Pollard (1995a,b) and Pollard and Thompson (1994, 1995), and paleoclimatic applications of Version 2 to date are described in Pollard and Thompson (1997a, b) and Thompson and Pollard (1997). The model consists of an atmospheric general circulation model (AGCM) coupled to multilayer models of vegetation, soil or land ice, snow, sea ice, and a 50-m slab oceanic layer. The main features of the various submodels are outlined below, noting important differences between Version 1.02 and the new Version 2.
Atmospheric general circulation model
The AGCM uses spectral transform dynamics (Williamson et al. 1987) for mass, heat and momentum, and semi-Lagrangian transport in grid space for water vapor (Williamson and Rasch 1989). In Version 2 a hybrid coordinate system is used, which tends from sigma coordinates (pressure/surface pressure) near the ground to pressure coordinates at the top of the atmosphere (Simmons et al. 1989). There is a diurnal cycle, with solar-radiation calculations performed every 1.5 h. The solar radiative scheme of Thompson et al. (1987) is used, with two solar wave bands (0.25–0.9 μ and 0.9–4.0 μ); however, in Version 2 the single-effective-level cloud approximation is removed, and delta-Eddington calculations are now made for the clear and cloudy fractions of every layer. Convection in the atmosphere is treated using an explicit subgrid buoyant plume model along the lines of, but simpler than, Kreitzberg and Perkey (1976) and Anthes (1977, his section 4); the same model is also used to simulate planetary boundary layer mixing initiated by surface fluxes. In Version 2 the free parameters in the plume model have been adjusted for better results (notably the plume fractional areas and radii), and the planetary boundary-layer plumes can now condense and precipitate.
In Version 2 clouds are predicted using prognostic 3D water cloud amounts (e.g., Smith 1990; Senior and Mitchell 1993). Three separate prognostic cloud fields are kept for stratus, convective, and anvil cirrus clouds, which are advected by semi-Lagrangian transport and mixed vertically by convective plumes and background diffusion. Cloud evaporation, conversion to precipitation, aggregation by falling precipitation, reevaporation of falling precipitation, and turbulent deposition of lowest-layer cloud particles onto the surface are all included. Latent heat changes due to liquid versus ice clouds are neglected (as if all clouds are liquid), although the determination of cloud fraction, radiative properties, and microphysical parameters take the temperature (above/below freezing) into account. Cloud fractions for each type are proportional to the grid-box average cloud amounts.
Additional cloud absorption of solar radiation (Cess et al. 1995; Ramanathan et al. 1995) is included in Version 2 by a prescribed decrease in cloud single-scattering albedos except in high southern latitudes with a linear transition to higher albedos from 40° to 60°S. The global net solar absorption by the model atmosphere is now 86 W m−2 with about 30 W m−2 absorbed by clouds, and the ratio ofthe global mean cloud radiative forcing at the surface to that at the top of the atmosphere is 1.61.
Uniform mixing ratios of individual greenhouse gases CO2, CH4, N2O, CFC11, and CFC12 can be prescribed individually in Version 2, and their effects are treated explicitly in the infrared radiation module (Wang et al. 1991). Other new features in the Version 2 AGCM are gravity wave drag (MacFarlane 1987), Courant spectral truncation in the upper few layers to allow a time step of 0.5 h (Simmons et al. 1989), spectral filtering of the AGCM topography (Navarra et al. 1994), and the relatively small radiative effects of prescribed background tropospheric dust aerosols.
The standard AGCM resolution in Version 2 has been increased to spectral T31 (3.75° × 3.75°) and 18 vertical levels. As before, the AGCM horizontal grid is independent of the surface grid used for all surface models (nominally 2° × 2°), and fields are transferred between the AGCM and the surface by bilinear interpolation (AGCM fields to surface) or straightforward area-averaging (surface fluxes to AGCM) at each time step. These resolutions are used for all experiments in this paper.
Land-surface transfer model (LSX)
A land-surface transfer model (LSX) accounts for the physical effects of vegetation (Pollard and Thompson 1995). It is based on the earlier models of the Biosphere–Atmosphere Transfer Scheme (Dickinson et al. 1986) and simple biosphere model (SiB) (Sellers et al. 1986), with most of the physical components of SiB but less mathematical complexity. Up to two vegetation layers (“trees” and “grass”) can be specified at each grid point, and the radiative and turbulent fluxes through these layers to the soil or snow surface are calculated. Rain or snow can be intercepted by the vegetation and subsequently drips or blows off. Vegetation attributes such as leaf area indexes, fractional cover, leaf albedos, etc., are prescribed in Version 2 from present-day off-line results of a predictive natural vegetation model (Bergengren and Thompson 1997, manuscript submitted to Global Climate Change; Bergengren et al. 1997, manuscript submitted to Climate Change). These results are essentially global distributions of 110 life forms predicted using present-day observed climatology, and for each grid cell the net physical attributes of the vegetation community are determined and passed to LSX. The resulting global vegetation distributions compare well to datasets such as Kuchler (1983) and Matthews (1983).
A stochastic precipitation term is included in Version 2 to account for the observed statistics of precipitation time series at a point compared to large-scale averages. A random term is added to the precipitation at each surface cell (linearly interpolated from the AGCM grid) at each time step, with an exponential distribution based on observations, with the same mean as the current AGCM precipitation, and with implicit fractional areas of 0.3 and 0.7 containing nonzero precipitation for convective and stratiform precipitation, respectively (Thomas and Henderson-Sellers 1991; Eltahir and Bras 1993).
In Version 2 a fractional amount of open water representing lakes or coastal water is specified for each land surface grid cell, and separate calculations for the surface-AGCM fluxes are made for the open-water fraction and the land fraction. The two sets of fluxes are weighted together into the effective flux contribution for each surface cell. The open water is a well-mixed column with prognostictemperature, just as for the open-ocean slab model. Floating ice can form if the column reaches the freezing point and snow can accumulate on the ice, using exactly the same code as for thermodynamic sea ice. The water depth has one of two values representing “coastal ocean/deep lakes” or “shallow lakes” (50 or 5 m, respectively). New 2° × 2° maps for land-ocean-icesheet distribution, topography, and open-water fraction have been generated for Version 2 using the 10-min U.S. Navy FNOC global elevation dataset (Cuming and Hawkins 1981; Kineman 1985) and Cogley’s 1° × 1° surface datasets for ice-sheet distribution (Cogley 1991).
Soil and ice-sheet surface model
A six-layer soil model extends from the surface to 4.25-m depth, with layer thicknesses increasing from 5 cm at the top to 2.5 m at the bottom. Physical processes in the vertical soil column include heat diffusion, liquid water transport (Clapp and Hornberger 1978; Dickinson 1984), surface runoff and bottom drainage, uptake of liquid water by plant roots for transpiration, and the freezing and thawing of soil ice. Version 2 also includes a surface ponding reservoir (with a maximum depth of 10 mm), which acts as a buffer between rainfall, infiltration, and runoff. Saturated soil layers are now possible by implicitly accounting for vertical hydrostatic pressure gradients. In Version 2, soil hydrologic properties (saturated matric potential and hydraulic conductivity, porosity, etc.) and wet surface albedo are determined from soil sand–silt–clay texture ratios, using empirical formulae in Cosby et al. (1984). The ratios are in turn prescribed from a new global soil-texture dataset (Webb et al. 1993), which includes variations with depth. The same six-layer model is used for ice sheets, with physical parameters appropriate for ice and with no internal liquid moisture.
Snow model
A three-layer snow model is used for snow cover on soil, ice-sheet, and sea-ice surfaces, including fractional areal cover when the snow is thin. Heat is diffused linearly through the snow, and the total thickness changes due to melting or snowfall on the upper layer. In Version 2 the fractional cover is diagnosed from the total snow mass, assuming a linear relationship between fraction and depth below 100% cover (Donald et al. 1995), and the depth at which the cover reaches 100% varies within the range 20–70 cm, depending on the height of the lower vegetation canopy. Of particular importance for the present study are the solar albedos of melting snow and ice-sheet surfaces. For the visible (<0.9 μ) and near-infrared (>0.9 μ) wave bands, these are 0.55 and 0.35 for snow and 0.70 and 0.50 for land ice, respectively. Albedos for colder snow and ice surfaces ramp linearly toward higher values for temperatures between 0° and −5°C. These albedos are used equally for direct and diffuse beams except for the direct beam on snow, which depends on solar zenith angle as in Briegleb and Ramanathan (1982). The effects of snow compactness, aging, and moisture content are currently ignored.
Sea-ice model
A three-layer, thermodynamic model predicts the local melting and freezing of sea ice essentially as in Semtner (1976). Fractional areal cover is included as in Hibler (1979) and Harvey (1988). Sea-ice advection is included using the “cavitating-fluid” model of Flato and Hibler (1990, 1992) in which the ice resists compressive stresses but offers no resistance to divergence or shear. In Version 2, the annual mean ocean currents prescribed for sea-ice dynamics are obtained from a 5-yr run of a 2° × 2° ocean GCM (E. Brady 1995, personal communication), and the surface winds used for sea-ice dynamics come from the AGCM itself, not from prescribed monthly climatological winds as in Version 1.02. The implicit statistical distribution of ice fraction versus thickness used to relate reductions in mean thickness to changes in mean ice fraction is now triangular between zero and two times the mean thickness, rather than uniform as in Version 1.02 (S. Vavrus 1995, personal communication).
Ocean slab model
The ocean is represented by a thermodynamic slab, which crudely captures the seasonal heat capacity of the surface mixed layer. The thickness of the slab is 50 m. Oceanic heat transport is treated differently in Version 2 to alleviate difficulties encountered with the earlier method of prescribing the heat convergence (W m−2) as a function of latitude. For Version 2 we first fitted the present-day observed zonal mean transport as a linear function of the latitudinal SST gradient, but with the diffusion coefficient depending on the zonal fraction of land versus ocean and on latitude itself. Those coefficients are used to calculate the two-dimensional linear diffusion of heat versus SST at each model time step. Convergence under sea ice is weighted toward 0 for 100% cover in the Northern Hemisphere and toward 6 W m−2 in the Southern Hemisphere. As in Version 1.02, to avoid unrealistic sea-ice formation in the Norwegian Sea region, we impose a crude local flux that warms the mixed layer whenever it drops below 1.04°C in a rectangular region between 66° and 78°N and −10° and 56°E. This flux would increase linearly to a maximum possible value of 500 W m−2 if the ocean were to cool to its freezing point (−1.96°C). This is meant to simulate the buffering effect of the deepening winter mixed layer and advection by warm ocean currents, and does produce wintertime heat convergences of about 200 W m−2 in agreement with Hibler and Bryan (1987). After making the sea-ice and Norwegian Sea adjustments at each time step, an additive global adjustment is made to ensure that the global integral of the convergence is zero. We emphasize that in both versions of the model, the ocean heat flux parameterization is based only on observed estimates and is not tuned to yield correct present-day SSTs, that is it is not a “flux-correction” method. This allows the model to be used for paleoclimatic applications with very different ocean configurations than the present, for which present-day flux corrections would be inappropriate.
The neutral drag coefficient used in the calculation of ocean surface momentum flux now depends on wind speed (Large and Pond 1981; Trenberth et al. 1989). The drag coefficient is increased slightly to allow for some scatter of wind speeds about the model mean. The roughness length for surface fluxes of heat and water vapor over ocean has been changed to 0.0001 m (which is the main factor in correcting the earlier overestimate of global precipitation, as mentioned in section 2).
Surface-air temperatures for model (2-m height) vs observed in °C. The observed data is from Crutcher and Meserve (1970) and Taljaard et al. (1969) over oceans and Antarctica, and from Leemans and Cramer (1991) over land except Antarctica. (a) Model, January. (b) Difference (model–observed), January. (c) Model, July. (d) Difference (model–observed), July.
Citation: Journal of Climate 10, 5; 10.1175/1520-0442(1997)010<0871:GAAMBF>2.0.CO;2
Zonal mean precipitation, for model (solid lines) and observed (dashed lines). The observed data are from Shea (1986): (a) January and (b) July.
Citation: Journal of Climate 10, 5; 10.1175/1520-0442(1997)010<0871:GAAMBF>2.0.CO;2
Precipitation in mm day−1. The observed data are from Shea (1986). (a) Model, January. (b) Observed, January. (c) Model, July. (d) Observed, July.
Citation: Journal of Climate 10, 5; 10.1175/1520-0442(1997)010<0871:GAAMBF>2.0.CO;2
Model and observed precipitation over Indian subcontinent, for July, in mm day−1. (a) Model with predicted SSTs. (b) Model with prescribed climatological SSTs (Shea et al. 1990). (c) Observed (Shea 1986).
Citation: Journal of Climate 10, 5; 10.1175/1520-0442(1997)010<0871:GAAMBF>2.0.CO;2
Model fractional snow cover for January, with a cutoff below 10%.
Citation: Journal of Climate 10, 5; 10.1175/1520-0442(1997)010<0871:GAAMBF>2.0.CO;2
Arctic monthly mean sea-ice concentrations and thicknesses for the model, with a cutoff below 10% fractional area. (a) Fractional area for March. (b) Thickness for March in m. (c) Fractional area for September. (d) Thickness for September in m.
Citation: Journal of Climate 10, 5; 10.1175/1520-0442(1997)010<0871:GAAMBF>2.0.CO;2
As in Fig. 6 except for the Antarctic.
Citation: Journal of Climate 10, 5; 10.1175/1520-0442(1997)010<0871:GAAMBF>2.0.CO;2
Sea level pressure in mb. (a) Model, January. (b) Observed (Shea 1986), January. (c) Model, July. (d) Observed (Shea 1986), July.
Citation: Journal of Climate 10, 5; 10.1175/1520-0442(1997)010<0871:GAAMBF>2.0.CO;2
Changes in seasonal mean surface-air (2-m) temperature due to doubling atmospheric CO2 in °C. Regions where there is 5% chance or greater that the change in the 10-yr mean could be due solely to interannual variability are hatched (Chervin and Schneider 1976): (a) December–January–February and (b) June–July–August.
Citation: Journal of Climate 10, 5; 10.1175/1520-0442(1997)010<0871:GAAMBF>2.0.CO;2
Changes in annual mean precipitation and upper-soil moisture due to doubling atmospheric CO2. Regions where there is 5% chance or greater that the change in the 10-yr mean could be due solely to interannual variability are hatched (Chervin and Schneider 1976). (a) Precipitation in mm day−1. (b) Fraction of soil pores filled by liquid or ice, averaged over the upper 30 cm of soil.
Citation: Journal of Climate 10, 5; 10.1175/1520-0442(1997)010<0871:GAAMBF>2.0.CO;2
Model seasonal precipitation over Greenland in mm day−1: (a) December–January–February, (b) March–April–May, (c) June–July–August, and (d) September–October–November.
Citation: Journal of Climate 10, 5; 10.1175/1520-0442(1997)010<0871:GAAMBF>2.0.CO;2
As in Fig. 11 except for the Antarctic.
Citation: Journal of Climate 10, 5; 10.1175/1520-0442(1997)010<0871:GAAMBF>2.0.CO;2
July surface-air temperatures over Greenland in °C. (a) Model, without elevation correction. (b) Model, with elevation correction. (c) Observed, redigitized and recontoured from Fig. 10 of Ohmura (1987).
Citation: Journal of Climate 10, 5; 10.1175/1520-0442(1997)010<0871:GAAMBF>2.0.CO;2
(a) Model annual precipitation over Greenland in cm yr−1. (b) Observed annual precipitation over Greenland, redigitized and recontoured from Fig. 2 of Ohmura and Reeh (1991).
Citation: Journal of Climate 10, 5; 10.1175/1520-0442(1997)010<0871:GAAMBF>2.0.CO;2
Model annual ablation and mass balance over Greenland in cm yr−1. The blank strips around the ice sheet edge in this and subsequent figures are plotting artifacts. (a) Ablation and (b) net surface mass balance.
Citation: Journal of Climate 10, 5; 10.1175/1520-0442(1997)010<0871:GAAMBF>2.0.CO;2
(a) Model annual precipitation over Antarctica in cm yr−1. (b) Observed annual accumulation over Antarctica, redigitized and recontoured from Fig. 2 of Bromwich (1988), adapted in turn from Giovinetto and Bentley (1985).
Citation: Journal of Climate 10, 5; 10.1175/1520-0442(1997)010<0871:GAAMBF>2.0.CO;2
Model annual ablation and mass balance over Antarctica in cm yr−1: (a) ablation and (b) net surface mass balance.
Citation: Journal of Climate 10, 5; 10.1175/1520-0442(1997)010<0871:GAAMBF>2.0.CO;2
Changes in annual budget terms over Greenland due to doubling atmospheric CO2 in cm yr−1 (double CO2 minus present): (a) precipitation, (b) ablation, and (c) net surface mass balance.
Citation: Journal of Climate 10, 5; 10.1175/1520-0442(1997)010<0871:GAAMBF>2.0.CO;2
As in Fig. 18 except for Antarctica.
Citation: Journal of Climate 10, 5; 10.1175/1520-0442(1997)010<0871:GAAMBF>2.0.CO;2
Seasonal cycles of model precipitation and local runoff, areally averaged over Greenland in cm yr−1: 1 × CO2 precipitation (thick solid line); 1 × CO2 local runoff (thick dashed line); 2 × CO2 precipitation (thin solid line); 2 × CO2 local runoff (thin dashed line).
Citation: Journal of Climate 10, 5; 10.1175/1520-0442(1997)010<0871:GAAMBF>2.0.CO;2
As in Fig. 20 except for Antarctica.
Citation: Journal of Climate 10, 5; 10.1175/1520-0442(1997)010<0871:GAAMBF>2.0.CO;2
Annual mean precipitation, ablation, and net surface mass balance for each of the 10 model years, areally averaged over Greenland in cm yr−1: 1 × CO2 precipitation (thick solid line); 1 × CO2 ablation (thick dashed line); 1 × CO2 surface mass balance (thick dotted line); 2 × CO2 precipitation (thin solid line); 2 × CO2 ablation (thin dashed line); 2 × CO2surface mass balance (thin dotted line).
Citation: Journal of Climate 10, 5; 10.1175/1520-0442(1997)010<0871:GAAMBF>2.0.CO;2
As in Fig. 22 except for Antarctica.
Citation: Journal of Climate 10, 5; 10.1175/1520-0442(1997)010<0871:GAAMBF>2.0.CO;2
Model and observed annual mean precipitation, ablation, and surface mass balance (cm yr−1) averaged over Greenland and Antarctica, showing the effects of the two correction procedures described in the text.
Model annual mean precipitation, ablation, and surface mass balance averaged over Greenland and Antarctica for present-day and doubled CO2. The last row shows the implied change in the ice sheet’s annual contribution to global sea level. The first value in parentheses is the uncertainty due to interannual variability at the 95% confidence level (i.e., ± the interannual standard deviation × 2.262/√(10 − 1)). The second value in parentheses indicates possible systematic error due to plausible variations in the refreezing parameterization. The possibility of other larger systematic errors is discussed in section 6.
