• Boer, G. J., and M. Lazare, 1988: Some results concerning the effects of horizontal resolution and gravity-wave drag on simulated climate. J. Climate,1, 789–806.

  • Boyle, J. S., 1993: Sensitivity of dynamical quantities to horizontal resolution for a climate simulation using the ECMWF (Cycle 33) model. J. Climate,6, 796–815.

  • Bromwich, D. H., 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.

  • CLIMAP Project Members, 1981: Seasonal reconstruction of the earth’s surface at the last glacial maximum. Geol. Soc. Amer. Map. Chart. Ser.,MC-36, 18 pp.

  • Deque, M., C. Dreveton, A. Braun, and D. Cariolle, 1994: The ARPEGE/IFS atmosphere model: A contribution to the French community climate modelling. Climate Dyn.,10, 249–266.

  • Dong, B.-W., and P. J. Valdes, 1995: Sensitivity studies of Northern Hemisphere glaciation using an atmospheric general circulation model. J. Climate,8, 2471–2496.

  • ——, and ——, 1998: Simulations of the last glacial maximum climates using a general circulation model: Prescribed versus computed sea surface temperatures. Climate Dyn.,14, 571–591.

  • Glover, R. W., 1999: Influence of spatial resolution and treatment of orography on GCM estimates of the surface mass balance of the Greenland ice sheet. J. Climate,12, 551–563.

  • Hall, M. N. J., P. J. Valdes, and B.-W. Dong, 1996: The maintenance of the last great ice sheets: A UGAMP GCM study. J. Climate,9, 1004–1019.

  • Harrison, E. F., P. Minnis, B. R. Barkstrom, V. Ramanathan, R. D. Cess, and G. G. Gibson, 1990: Seasonal variation of cloud radiative forcing derived from the Earth Radiation Budget Experiment. J. Geophys. Res.,95, 18 687–18 703.

  • Hoskins, B. J., H. H. Hsu, I. N. James, M. Masutani, P. D. Sardeshmukh, and G. H. White, 1989: Diagnostics of the global atmospheric circulation. Tech. Document WMO/TD-326, WCRP-27, 217 pp.

  • Joussaume, S., and K. E. Taylor, 1995: Status of the Paleoclimate Modeling Intercomparison Project (PMIP). Proc. First Int. AMIP Scientific Conf., Monterey, CA, AMIP, 425–430. [WMO/TD-732, WCRP-92.].

  • Kageyama, M., P. J. Valdes, G. Ramstein, C. D. Hewitt, and U. Wyputta, 1999: Northern Hemisphere storm tracks in present day and last glacial maximum climate simulations: A comparison of the European PMIP models. J. Climate,12, 742–760.

  • 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 the past 18 000 years. Nature,373, 52–54.

  • Kiehl, J. T., and D. L. Williamson, 1991: Dependence of cloud amount on horizontal resolution in the National Center for Atmospheric Research Community Climate Model. J. Geophys. Res.,96, 10 955–10 980.

  • Peltier, W. R., 1994: Ice age paleotopography. Science,265, 195–201.

  • Phillips, T. J., L. C. Corsetti, and S. L. Grotch, 1995: The impact of horizontal resolution on moist processes in the ECMWF model. Climate Dyn.,11, 85–102.

  • Rind, D., 1988: Dependence of warm and cold climate depiction on climate model resolution. J. Climate,1, 965–997.

  • Senior, C. A., 1995: The dependence of climate sensitivity on the horizontal resolution of a GCM. J. Climate,8, 2860–2880.

  • Thompson, S. L., and D. Pollard, 1997: Greenland and Antarctic mass balances for present and doubled atmospheric CO2 from the GENESIS version-2 global climate model. J. Climate,10, 871–900.

  • Warrick, R. A., and J. Oerlemans, 1990: Sea level rise. Climate Change, the IPCC Assessment, J. T. Houghton, G. J. Jenkins, and J. J. Ephraums, Eds., Cambridge University Press, 257–281.

  • Williamson, D. L, J. T. Kiehl, and J. J. Hack, 1995: Climate sensitivity of the NCAR Community Climate Model (CCM2) to horizontal resolution. Climate Dyn.,11, 377–397.

  • View in gallery

    Changes in ice sheet height and extent of the updated ice sheets (Peltier 1994) at LGM at three truncations: (a) T63, (b) T42, and (c) T21. Contours are at 250 and 500 m, and then increasing by 500 m. Shading indicates ice height greater than 1000 m.

  • View in gallery

    Seasonal evolution of the surface air temperature at three horizontal resolutions for the PD and LGM simulations: (a) over globe, (b) over land, and (c) over ocean.

  • View in gallery

    Similar to Fig. 2, but total precipitation.

  • View in gallery

    Annual mean surface air temperature anomalies between LGM and PD simulations and the anomaly difference due to change in model horizontal resolution: (a) at T21, (b) T42 − T21, and (c) T63 − T42. Thick line indicates the extent of ice sheets at LGM. Shading indicates that the anomalies are statistically significant at 99% confidence level using the Student’s t test.

  • View in gallery

    850-hPa air temperature anomalies between the LGM and PD simulations in DJF, JJA, and the anomaly difference due to change in model horizontal resolution: (a) and (b) at T21, (c) and (d) T42 − T21, and (e) and (f) T63 − T42. (a), (c), and (e) show DJF and (b), (d), and (f) show JJA.

  • View in gallery

    Zonally asymmetric streamfunction and zonal wind at 500 hPa in DJF for the (left) PD simulations and (right) the corresponding anomalies between the LGM and PD simulations: (a) and (d) T63, (b) and (e) T42, and (c) and (f) T21. Contour intervals are 2.0 × 106 m s−2 with negative values dashed. Zonal wind (m s−1) is plotted in shading scale.

  • View in gallery

    Similar to Fig. 6, but for JJA.

  • View in gallery

    The high-pass transient eddy kinetic energy at 250 hPa in DJF for the PD simulations: (a) T63, (b) T42, (c) T21, (d) T63 − T42, (e) T42 − T21, and (f) ECMWF analysis 1983–89. Regions with values greater than 100 m2 s−2 are shaded.

  • View in gallery

    Similar to Fig. 8, but for DJF for the LGM simulations.

  • View in gallery

    Similar to Fig. 8, but for JJA for the PD simulations.

  • View in gallery

    Similar to Fig. 8, but for JJA for the LGM simulations.

  • View in gallery

    (left) Large-scale and (right) total precipitation anomalies in DJF over the Northern Hemisphere between the LGM and PD simulations: (a) and (d) T63, (b) and (e) T42, and (c) and (f) T21. Positive anomalies are within solid line and negative anomalies within dotted line. The thick line outlines the extent of ice sheets at LGM. Shading indicates the anomalies are statistically significant at 99% confidence level using the Student’s t test.

  • View in gallery

    Total precipitation anomalies in JJA over the tropical regions in the Eastern Hemisphere between the LGM and PD simulations: (a) T63, (b) T42, and (c) T21. Positive anomalies are within solid line and negative anomalies within dotted line. Shading indicates the anomalies are statistically significant at 99% confidence level using the Student’s t test.

  • View in gallery

    Annual mean precipitation minus evaporation anomalies between the LGM and PD simulations:(a) T21, (b) T42, and (c) T63. Positive anomalies are within solid line and negative anomalies within dotted line. The thick line outlines the extent of ice sheets at LGM. Shading indicates the anomalies are statistically significant at 99% confidence level using the Student’s t test.

  • View in gallery

    Net snow accumulation rate in the LGM simulations. (a) T63, (b) T42, and (c) T21. Contours are at 50, 100, 200, and 400 kg m−2 yr−1, and then increasing by 400 kg m−2 yr−1 with negative values dashed. Areas with net accumulation rate greater than 200 kg m−2 yr−1 are shaded.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 135 133 3
PDF Downloads 13 13 1

Climates at the Last Glacial Maximum: Influence of Model Horizontal Resolution

View More View Less
  • 1 Department of Meteorology, University of Reading, Reading, United Kingdom
© Get Permissions
Full access

Abstract

The climate at the last glacial maximum (LGM) has been simulated using the U.K. Universities Global Atmospheric Modeling Programme (UGAMP) general circulation model (GCM) truncated at total wavenumbers 21, 42, and 63 (T21, T42, and T63) with prescribed SSTs based on the Climate: Long-Range Investigation Mapping and Prediction Study data. Consistent with the Paleoclimate Modeling Intercomparison Project, the other boundary conditions include the changes in ice sheet topography and geography, a lower sea level, a lower concentration of CO2 in the atmosphere, and a slightly different insolation pattern at the top of the atmosphere. The influence of the model horizontal resolution on the simulated global climate and regional circulation changes has been analyzed.

The simulations indicate that both the global and regional climate changes due to the imposed ice age boundary conditions are affected by the model horizontal resolution. The largest climate change differences occur when the model horizontal resolution increases from T21 to T42. Further increase in resolution to T63 only results in quantitative differences. However, for some variables, such as measures of storm track activity, the LGM simulation shows that there are clear systematic differences between T42 and T63. This result was not seen in the equivalent fields at present day. The regional climate changes simulated in a lower-resolution model (T21) differ significantly from those in higher-resolution models, because of the poor simulations of planetary waves and storm tracks. Results also imply that the insensitivity of simulated storm tracks to model horizontal resolution change from T42 to T63 at present day may not be applicable to a different climate regime. The lower-resolution model cannot reproduce the present-day climate as accurately as the high-resolution model can. In addition, the simulated climate changes in the lower-resolution model due to ice age boundary conditions in some regions contradict those simulated with higher-resolution models with the latter giving better agreement with geological evidence. The authors conclude that a higher-resolution model, at least T42, is needed for the UGAMP GCM to simulate climate changes.

Corresponding author address: Dr. Buwen Dong, Department of Meteorology, University of Reading, Reading, RG6 6BB, United Kingdom.

Email: swsdong@met.reading.ac.uk

Abstract

The climate at the last glacial maximum (LGM) has been simulated using the U.K. Universities Global Atmospheric Modeling Programme (UGAMP) general circulation model (GCM) truncated at total wavenumbers 21, 42, and 63 (T21, T42, and T63) with prescribed SSTs based on the Climate: Long-Range Investigation Mapping and Prediction Study data. Consistent with the Paleoclimate Modeling Intercomparison Project, the other boundary conditions include the changes in ice sheet topography and geography, a lower sea level, a lower concentration of CO2 in the atmosphere, and a slightly different insolation pattern at the top of the atmosphere. The influence of the model horizontal resolution on the simulated global climate and regional circulation changes has been analyzed.

The simulations indicate that both the global and regional climate changes due to the imposed ice age boundary conditions are affected by the model horizontal resolution. The largest climate change differences occur when the model horizontal resolution increases from T21 to T42. Further increase in resolution to T63 only results in quantitative differences. However, for some variables, such as measures of storm track activity, the LGM simulation shows that there are clear systematic differences between T42 and T63. This result was not seen in the equivalent fields at present day. The regional climate changes simulated in a lower-resolution model (T21) differ significantly from those in higher-resolution models, because of the poor simulations of planetary waves and storm tracks. Results also imply that the insensitivity of simulated storm tracks to model horizontal resolution change from T42 to T63 at present day may not be applicable to a different climate regime. The lower-resolution model cannot reproduce the present-day climate as accurately as the high-resolution model can. In addition, the simulated climate changes in the lower-resolution model due to ice age boundary conditions in some regions contradict those simulated with higher-resolution models with the latter giving better agreement with geological evidence. The authors conclude that a higher-resolution model, at least T42, is needed for the UGAMP GCM to simulate climate changes.

Corresponding author address: Dr. Buwen Dong, Department of Meteorology, University of Reading, Reading, RG6 6BB, United Kingdom.

Email: swsdong@met.reading.ac.uk

1. Introduction

Past climate change presents a challenging test of our understanding of climate and our ability to simulate climatic change using computer models. These models have been developed in the context of present-day climate and are used to predict future climate change. Model skill is usually measured in terms of accuracy of the present-day climatic simulation. However, accurate simulation of the current climate does not guarantee a model’s ability to simulate climate change. It is, therefore, vital that these models are tested on climatic regimes that are very different from today. The last glacial maximum (LGM), which corresponds to a period with maximum southward extent of the Fennoscandian and Laurentide ice sheets, represents an example of an extreme cold climate. Studying this period is important to understand how the ice age boundary conditions can influence climate change.

The Paleoclimate Model Intercomparison Project (PMIP; Joussaume and Taylor 1995) designed two experiments to investigate the LGM. The first used prescribed SSTs based on Climate: Long-Range Investigation Mapping and Prediction (CLIMAP; CLIMAP Project Members 1981) Study data; the other computed SSTs using a slab ocean model. A total of eight modeling groups used prescribed SSTs, and seven used computed SSTs. A wide range of horizontal resolutions were used (from R15 to T42). Therefore, a key question that arises is, how do the results depend on the model horizontal resolution? From the perspective of climate research, one would be interested in the coarsest resolution that adequately simulates the features of interest and climate sensitivity.

From the analyses of the ice age climates with the 4° × 5° and 8° × 10° version of the Goddard Institute for Space Studies (GISS) model, Rind (1988) found a greater reduction in hydrological cycle and atmospheric temperature, a greater eddy energy, and an energy transport increase in the ice age finer-grid simulations. However, currently many climate models are at even higher resolution than 4° × 5° and so it is of interest to revisit the resolution studies but using grids nearer to contemporary models.

Dong and Valdes (1998, hereinafter referred to as DV) studied the LGM climate using the U.K. Universities Global Atmospheric Modelling Programme (UGAMP) general circulation model (GCM) by comparing the two simulations using a relatively high horizontal resolution model (T42) in the framework of PMIP. One simulation prescribed SSTs deduced from the CLIMAP dataset. The other computed SSTs using a thermodynamic slab ocean model with prescribed ocean heat transport. Emphasis was given to the changes in midlatitude transient eddy activities and in planetary waves and their role in regional climate changes at LGM. Many previous studies suggest the importance of horizontal resolution on the simulation of present-day climate (Boer and Lazare 1988; Boyle 1993; Deque et al. 1994; Phillips et al. 1995; Williamson et al. 1995) and on the different climate sensitivity to future climate (Senior 1995). More recently, Kageyama et al. (1999) have shown that the storm tracks simulated by a number of PMIP models appeared to be related to a model’s horizontal resolution, with the lower-resolution models generally having weaker storm tracks. However, the models were different in many ways, not just resolution. To examine the effect of horizontal resolution on the climate sensitivity to the imposed ice age boundary conditions, both present-day (PD) and LGM simulations have been run with the UGAMP GCM at resolutions of T21, T42, and T63 with prescribed SSTs. The resolution dependence of climate change and changes in atmospheric dynamics are investigated in this paper.

2. The model and boundary conditions

The UGAMP GCM used in this study was based on the forecast model of the European Centre for Medium-Range Weather Forecasts (ECMWF). The brief model description was given in Dong and Valdes (1995). The model is modified for the LGM simulations by changing SSTs, ice sheet topography and orography, CO2 concentration, and the orbital parameters. All the changes are consistent with the PMIP experiments. The CO2 concentration is set to 200 ppm at the LGM simulations, and the orbital parameters are set to those appropriate for 21 kyr ago. The land ice sheet extent and elevation are based on Peltier’s (1994) reconstructions, which are given in Fig. 1 for the model truncated at T63, T42, and T21. It is seen that the ice sheets are smoother and their peak height is a few hundred meters lower at T21 resolution than those at T42 and T63 (the original reconstructions are 1° resolution). Also important for regional climates is that the higher-resolution model is able to partially represent some of the isolated ice sheets, such as the Patagonian ice sheet over the southern tip of south America and the separation of the British ice sheet from the Scandinavian ice sheet. We will show that this improved representation of the smaller ice sheets is unimportant for global climate but is important on regional scales.

Both the PD and LGM simulations are 11 yr at T21 and T42, while they are 6 yr at T63. The analyses are based on the last 10-yr mean at T21 and T42 and the last 5-yr mean at T63. Because the changes of the climate in the LGM simulations relative to the present-day simulations are far above the level of the model climate natural variability, the statistical significance level will be discussed only for some selected quantities.

3. Global mean changes

Table 1 gives the globally averaged values of some climatic variables, over globe, land, and ocean, simulated by the model at three different horizontal resolutions. The decreases in the global averaged surface air temperature and whole atmosphere temperature due to the ice age boundary condition depend on the horizontal resolution. The annually or seasonally averaged surface air cooling and whole atmosphere cooling in the LGM simulation decreases with the increase in the horizontal resolution. Over land, it is seen that the largest difference in the simulated cooling occurs when horizontal resolution increases from T21 to T42. Further increase of resolution from T42 to T63 results in a only modest difference. These characteristics are true for the annual mean, DJF (December–February), and JJA (June–August) seasons (the latter two are not shown). It is also interesting to note that the resolution sensitivity of the (LGM − PD) change in land air temperature is 0.9°C (−7.6°C at T21 vs −6.7° at T42) and is bigger than the differences between the two models at present-day resolution sensitivity of 0.6°C (9.8°C at T21 vs 9.2°C at T42). The large decrease in temperature in the LGM simulation in the lower-resolution model is associated with the fact that the decrease in total precipitable water in the atmosphere is greater and the increase in total cloud radiative forcing (CRFT) is smaller than those in the simulations at T42 and T63.

For both PD and LGM simulations, the simulated total cloud cover decreases with increasing horizontal resolution of the model from T21 to T42. These features are similar to that found in the National Center for Atmospheric Research (NCAR) CCM1 by Kiehl and Williamson (1991). The decrease in cloud cover with increasing horizontal resolution is particularly evident in the medium level and convective cloud cover (not shown). This is because in the lower-resolution models the convection in the tropical convergence zone is less intense, but more broad compared to that in the higher-resolution models. This decrease in upward motion in the convergence zone, through continuity, leads to a weaker subsidence outside of the Tropics. Weaker subsidence, in turn, results in the increase of large-scale condensation cloud outside the regions of convergence in the lower-resolution models. The broad scale of the convergence zone in the lower-resolution models leads to an increase in convective cloud cover. However, details of resolution-dependent changes in the simulated clouds appear to be highly model dependent. In the GISS model, for example, mainly cirrus clouds were affected by resolution changes (Rind 1988), while in the NCAR CCM1 low subtropical cloud was most sensitive (Kiehl and Williamson 1991). In the UGAMP GCM, it is the convective and medium cloud cover that is affected by the model horizontal resolution.

Generally, globally mean total cloud cover in the LGM simulations at the three resolutions decreases relative to the present-day simulations. Over ocean, the magnitude of the decrease in cloud cover is similar in the LGM simulations at all three resolutions. However, over land, it is very sensitive to the horizontal resolution. The simulated changes in cloud cover at T42 are similar, in both DJF and JJA, to those simulated at T63. They indicate total cloud cover decrease in DJF and increase in JJA, while the simulation at T21 suggests little change in DJF and a large increase in JJA (not shown). As a result, the annual mean cloud cover over land in the LGM simulation at T21 increases while it decreases at T42 and T63 relative to the PD simulations.

Shortwave cloud radiative forcing (CRFS) is overestimated in the PD simulations at three resolutions in comparison with the observation (−48.4 W m−2) given in Harrison et al. (1990). The overestimation is larger in the simulation at T21 than in the simulations at T42 and T63. This is associated with the fact that the cloud is too reflective and that the total cloud cover in the simulation at T21 is higher than in the simulations at T42 and T63. The LGM simulations indicate similar characteristics.

Longwave cloud radiative forcing (CRFL) is also overestimated in the PD simulation at T21 by about 20% compared to observation (31.1 W m−2), while the simulations at T42 and T63 give values very close to the observed one. This leads to the overestimation of the CRFT being slightly smaller in the lower-resolution model than in the high-resolution models.

Changes of shortwave cloud radiative forcing in the LGM simulations relative to the PD simulations are positive. These positive differences (warming effect) mean a weakening of the negative shortwave cloud radiative forcing in the LGM simulations. This feedback is weaker in the LGM simulation at T21 than that in the LGM simulations at T42 and T63, implying a larger decrease in temperature in the LGM simulation in the lower-resolution model.

For longwave radiative forcing, the main effect of changes in clouds in the LGM simulations is the reduction of their greenhouse effect. This effect has a similar magnitude in the simulations at three resolutions. As a result, the sensitivity of shortwave cloud feedback to the resolution dominates, which implies a larger temperature sensitivity to the ice age boundary condition in the simulation at T21. Thus the processes and parameterizations of cloud radiative effects appear to be the main factors inducing the global climate sensitivity to be resolution dependent for the imposed ice age boundary condition in the UGAMP GCM.

The hydrological cycle is more active in the high-resolution model with greater precipitation and greater evaporation. This is true for both the PD simulations and the LGM simulations. The greater precipitation in the higher-resolution model is due to the greater convective precipitation over ocean. This indicates that the models’s subgrid-scale convective processes are decoupled to some extent from the large-scale atmospheric moisture field since the total precipitable water over ocean in both the PD and LGM simulations at T21 is greater than in the higher-resolution models.

The decreases in precipitation and evaporation in the LGM simulations at three horizontal resolutions are similar in the annual and global averaged value. However, there are significant differences in changes over land and over oceans. In the LGM simulation at T21, precipitation decreases significantly over the ocean while over land there is a small change in comparison with the PD simulations. These are in contradiction to the changes simulated at T42 and T63, which indicate a large decrease in precipitation over land and little change over ocean.

Seasonal cycles of global mean surface air temperature and its contribution from land and ocean areas for both the PD and LGM simulations at three horizontal resolutions are shown in Fig. 2. They indicate the seasonal cycle of surface air temperature in both the PD and LGM simulations at T21 differs from that simulated at T42 and T63. The surface air temperature in the PD simulation at T21 is about 0.8°C warmer in the Northern Hemisphere summer than those simulated at T42 and T63, while it is about 0.8°C colder in the Northern Hemisphere winter. These differences are mainly from land areas because the sea surface temperature is prescribed. The warmer late northern spring and summer surface air temperature in the lower-resolution model at the PD simulation may be partly due to the snow–albedo and snow–hydrological effects. In the PD simulation at T21, the seasonal snow mass over the Northern Hemisphere melts too quickly in spring, in comparison with the higher-resolution models (not shown). In addition, orographic effect may also play a role. The surface air temperature differences over ocean in the LGM simulations at three horizontal resolutions mainly come from temperature differences over sea ice, with the lower-resolution model giving colder sea ice around Antarctic in the Southern Hemisphere winter.

Seasonal cycles of global, land, and ocean mean precipitation are shown in Fig. 3 for both the PD and LGM simulations at three horizontal resolutions. The hydrological cycle is more active in the high-resolution model for both PD and LGM simulations. As shown by Fig. 3, the largest differences occur when horizontal resolution changes from T21 to T42. Both the PD simulations and LGM simulations at T42 and T63 indicate similar seasonal evolution, respectively. Although the seasonal evolution of global mean precipitation anomalies between the LGM and PD simulations at T21 is similar to that at T42 and T63, it is significantly different over land and ocean areas. Precipitation over land decreases all year-round in the LGM simulation at T42 and T63 relative to the PD simulations, while T21 simulations indicate very little change. Over ocean, higher-resolution models simulate very little changes in precipitation while the lower-resolution model gives a decrease in precipitation all year-round.

The above results suggest that the global mean climate and climate change due to ice age boundary conditions over land and over ocean are dependent on the model horizontal resolution. The largest climate change differences to ice age boundary conditions occurs when the model horizontal resolution changes from T21 to T42. Further increase in horizontal resolution results in only small differences.

4. Dependence on model resolution of regional surface and lower-troposphere air temperature changes

The mean annual surface air temperature anomalies between the LGM and PD simulation at T21 and the anomaly differences due to the increase in model horizontal resolution are shown in Fig. 4. Area weighted mean method was used to regrid the results at higher-resolution grid to a lower-resolution grid in order to get anomalies due to the changes in horizontal resolution. We found that the results were relatively insensitive to the interpolation method, with only the much smaller scales being affected. The cooling is stronger at high latitudes and it decreases toward the Tropics. The largest cooling occurs over the two land ice sheets and the North Atlantic. Over the ocean, the surface air temperature changes follow the changes of SSTs (which are prescribed). These large-scale features are similar in the simulations with three resolutions.

However, there are some significant regional differences when resolution increases from T21 to T42 (Fig. 4b). Further increases of horizontal resolution to T63 (Fig. 4c) result in only modest change. It should be noted that coastline and orography are resolution dependent, which also has an impact on surface air temperature. The cooling over the two Northern Hemisphere land ice sheets is about 5°C stronger at T21. This is primarily due to the cloud feedback (mainly in JJA) and water vapor feedback. Over the Laurentide ice sheet, the increase in total cloud cover in the LGM simulations relative to the PD simulations is smaller, while decrease in total water vapor is greater at T21 than that at T42 and T63 (not shown). Over the land ice sheets in the LGM simulations, the presence of cloud has a warming effect due to large surface albedo. Therefore, the smaller increase in total cloud cover and the larger decrease in total water vapor result in the larger cooling over the land ice sheets in the LGM simulation at T21. The magnitude of cooling over North Africa, the near east, and southwest Asia is also horizontal-resolution-dependent with the simulation at T21 giving cooling 1°–2.5°C stronger than the simulations at T42 and T63. In large parts of Australia in the Southern Hemisphere, the LGM simulation at T21 indicates cooling of 1°–2.5°C, while simulations at T42 and T63 indicate little change. These differences mainly result from the local hydrological cycle and cloud feedback in DJF season. The Australian summer monsoon response to the ice age boundary conditions differs in the lower-resolution model from that in the higher-resolution models. The T21 simulation gives an enhanced monsoon precipitation over Australia at LGM relative to PD while the T42 and T63 simulations suggest a weakened monsoon precipitation (Fig. 14). The decrease of precipitation, which is associated with a decrease of cloud cover in the LGM simulations at T42 and T63, results in a decrease of soil moisture and evaporation. Decrease in cloud cover allows more solar radiation to reach the ground, and a decrease in evaporation reduces latent heat release. These two effects lead to the surface temperature increase over Australia in DJF in the LGM simulation at T42 and T63, which offsets the cooling in JJA season.

The lower-tropospheric air temperature anomalies between the LGM and PD simulations in DJF and JJA and the anomaly differences due to changes in horizontal resolution are illustrated in Fig. 5. Generally, the cooling patterns are similar to the surface air temperature changes. In DJF, the largest cooling occurs over the North Atlantic, which results in sharp temperature gradients along the sea ice edge there. This in turn has a significant effect on planetary waves and storm track activity, which will be discussed in the next section. One important feature is that the sharpness of temperature gradient along the sea ice edge over the Atlantic in the LGM simulations increases as resolution increases. The largest changes occur when horizontal resolution increases from T21 to T42. Similar to the surface air temperature changes, the lower-tropospheric cooling over North Africa, the near east, and southwest Asia is stronger at T21 simulation. It is also worth noting that the lower-tropospheric temperature gradient along 70°S latitude for PD and 60°S for LGM simulations sharpens as horizontal resolution increases (not shown).

In JJA, the largest cooling occurs over two land ice sheets, which is the result of large ice sheet albedo and ice sheet elevation, with the simulation at T21 giving the strongest cooling. This is primarily due to the cloud feedback and water vapor feedback. Due to the largest cooling being over the two land ice sheets, the temperature gradient along the south edge of the Laurentide and Fennoscandian ice sheets is enhanced, which in turn results in vigorous baroclinic activity. Once again, there are large-scale differences in lower-troposphere air temperature anomalies between T21 and T42 simulations. The difference in the anomalies between T63 and T42 is relatively smaller. The cooling over North Africa and south Asia in JJA is 1° to 2°C larger in the LGM simulation at T21 than that in the simulations at T42 and T63. This is also associated with the local hydrological and cloud feedback. The stronger cooling over these two regions in the LGM simulation at T21 are accompanied with slightly enhanced monsoon precipitation (see Fig. 13).

5. Response of planetary waves

The planetary waves are of substantial importance to ensure the fidelity of the simulation. The physical processes that contribute to the forcing of the zonal asymmetries include land–ocean contrasts, SST pattern, orographic features, diabatic heating, and synoptic transient eddy activities. At LGM, large orographic features of the continental ice sheets result in changes of the planetary scale waves. The changes in the transient eddy transport of heat and momentum and in the diabatic heating are also responsible for changes of planetary waves at LGM.

Shown in Fig. 6 are the asymmetric streamfunction and zonal wind at 500 hPa in DJF for the PD simulations and the corresponding anomalies between the LGM and PD simulations at three horizontal resolutions. The simulations at T42 and T63 present a credible simulation of the features of the planetary waves. They are in good agreement with the ECMWF analyses (Hoskins et al. 1989). However, the ridge–trough system over North America is too strong in the T21 PD simulation. The ridge–trough also shifts westward in comparison with either of the simulations at higher resolution or the analysis. Associated with this, the jet is located over the central part of North America in the T21 simulation rather over eastern North America–North Atlantic. The ridge over the North Atlantic is also stronger and extends meridionally over Greenland in the T21 PD simulation. The enhanced southwesterly flow ahead of the North Atlantic ridge advects more warm, moist air over Greenland and is favorable for more large-scale precipitation there. The structure of the wave pattern in the northwest Pacific sector and in the Southern Hemisphere in DJF is similar in the three PD simulations with different horizontal resolutions. However, the Southern Hemisphere jet, at T21, is centered at about 45°S, while it is centered at about 55°S at T42 and T63.

In DJF, the LGM simulations show an enhanced wave train across Canada and into the Atlantic (Fig. 6). The ridge over northwestern North America and the trough in northeastern North America, tilting from southwest to northeast, are enhanced; that is, the Laurentide ice sheet enhances the upstream longwave ridge and the downstream longwave trough. The North Atlantic ridge in the LGM simulations shrinks southward due to cold sea ice temperatures and extends downstream to central Siberia with a secondary high center there. The upper-level jet over the North Atlantic is significantly altered. The jet shifts downstream and northward, and its strength is enhanced at the LGM with the simulation at T21, giving a weaker increase. The splitting of the jet by the Laurentide ice sheet is not pronounced. This is especially true in the LGM simulation at the higher-resolution models. Consistent with changes in lower-level temperature gradients, the upper-level jet over the North Pacific in the LGM simulations is also enhanced slightly. The planetary waves in the Southern Hemisphere are weak and hardly change in the LGM simulations.

In JJA, the planetary wave patterns in both hemispheres are similar in the PD simulation at three horizontal resolutions (Fig. 7). However, the south Asian anticyclone is too strong in the T21 simulation. This is associated with latent heat release due to unrealistically large precipitation over the Bay of Bengal. In addition, the trough over the eastern Pacific in the Southern Hemisphere is stronger and extends into Australia. Associated with this is that the westerly jet curves toward southern Australia.

The effect of ice age boundary conditions has a relatively minor effect on planetary waves in JJA (Fig. 7). The zonal wind along the southern edges of the two land ice sheets increases slightly in the LGM simulations. This enhancement is consistent with the changes in low-level baroclinicity, which is enhanced over these two regions. In the Southern Hemisphere, the westerlies around 60°S are enhanced in the LGM simulations with the enhancement being the strongest in the simulation at T21. The relatively large changes in both planetary waves and westerlies over the Southern Hemisphere in the LGM simulation at T21 relative to the PD result from the fact that the present-day patterns are distorted there in the lower-resolution model.

6. Sensitivity of simulated storm tracks and their changes due to ice age boundary conditions to model horizontal resolution

Land–sea temperature contrasts are of primary importance for transient developing systems at PD, whereas the temperature contrasts over the sea ice are the most important feature as regards transient eddy activity at LGM. This is particularly true over the North Atlantic.

Shown in Fig. 8 are high-passed transient eddy kinetic energy at 250 hPa in DJF for the PD simulations at three resolutions, the anomalies due to changes in horizontal resolution, and that based on the ECMWF analysis (1983–89). The transient eddy has been temporally filtered to include only the relatively fast moving midlatitude systems with timescales for growth and decay of the order of 6 days or less, using the filter described in Hoskins et al. (1989). The high-pass transient eddy activity shows the regions of maximum high-frequency variability, or “storm tracks.”

In DJF, the positions and strengths of the two storm tracks in the North Atlantic and North Pacific in the simulations at T42 and T63 are in reasonable agreement with the analysis. The simulations at T42 and T63 also capture the slight northward tilt of the North Atlantic storm track. The Southern Hemisphere storm track and its zonal variation in intensity are well simulated at T42 and T63. However, the intensity is overestimated and latitude position is about 5° more poleward than the analysis suggests. This is associated with the cold bias (5°–10°C) of simulated surface and lower-troposphere temperatures over Antarctica.

The shapes of the North Atlantic, North Pacific, and Southern Ocean storm tracks are distorted in the simulation at T21. The North Atlantic storm track curves northeast toward eastern Greenland, and the North Pacific storm track curves toward Alaska. In addition, the intensity of the three storm tracks is significantly weaker in the simulation at T21. As the model horizontal resolution increases from T21 to T42, significant enhancement of the three major storm tracks occur. The Southern Hemisphere storm track also shifts poleward by about 5°–7° as resolution increases from T21 to T42, consistent with a poleward shift of Southern Hemisphere westerlies with resolution (Fig. 6). Further increases in model horizontal resolution from T42 to T63 result in hardly any changes in the North Atlantic and North Pacific storm tracks. However, the intensity of Southern Ocean storm track still increases.

The distortion in storm tracks in the lower-resolution model has a significant effect on the precipitation pattern over high-latitude regions and will be discussed later. Bromwich et al. (1994) also found in the lower-resolution version of the NCAR GCM1 that the topography of Greenland was distorted. This resulted in a major dislocation of the simulated North Atlantic storm track. They pointed out that this bias was substantially alleviated when the horizontal resolution increased to T42.

Significant changes occur for the two Northern Hemisphere storm tracks in the LGM simulations (Fig. 9). In the Pacific, the largest changes occur at the end of the storm track. The peak intensity is unaltered, but there is a downstream and equatorward shift of about 5°. This is consistent with a similar equatorward shift in the ice edge and hence in the maximum surface temperature gradient. The changes in the North Atlantic storm track at LGM are more complicated. The region of maximum transient eddy activity is considerably more confined meridionally but extends much farther into Europe. These changes are generally consistent with the changes in surface temperature gradient. The temperature gradients at the edge of the sea ice result in the midlatitude depressions closely following the edge of the sea ice. These features are true in the LGM simulations at three horizontal resolutions.

The changes in storm tracks have a significant effect on the high-latitude ice sheet mass budget. By comparing accumulation rates and temperature derived from oxygen isotopic measurements of ice in the deep core over Greenland, Kapsner et al. (1995) found that atmospheric circulation, not temperature, seems to have been the primary control on snow accumulation in central Greenland over the past 18 000 yr. Bromwich et al. (1993) demonstrated that changes in snow accumulation in central Greenland over the period 1963–88 were controlled primarily by changes in the strength and position of the dominant storm track in the region, with little relation to changes in temperature. The downstream development of the North Atlantic storm track in the LGM simulations steer the storms toward southern Europe and away from Greenland. This, in turn, has a contribution to reduction of the snow accumulation over Greenland at LGM. The mass budget over various ice sheets will be discussed in section 7c.

Similar to the PD simulations, the intensity of the three major storm tracks are significantly weaker in the LGM simulation at T21. As model horizontal resolution increases, the intensity of the storm tracks increases and they develop farther downstream. There is a significant increase of the intensity of both the North Atlantic and North Pacific storm tracks as model horizontal resolution increases from T42 to T63. This is associated with sharpened surface and lower-tropospheric temperature gradients along the sea ice edge as model horizontal resolution increases. It also implies that the eddies at LGM shift to high-frequency variability and smaller-scale eddies have a significant contribution to the north Atlantic and north Pacific storm tracks (Hall et al. 1996). This suggests that the insensitivity of model simulated storm tracks to horizontal resolution change from T42 to T63 at PD may be not applicable to different climate regimes.

In JJA, the two Northern Hemisphere storm tracks are weaker and they shift poleward by about 10° (Fig. 10), in comparison with the intensity and position in DJF. The Southern Ocean storm track extends over a broader range of latitudes, in part reflecting the presence of the double-jet structure of the troposphere. The main storm track is associated with the polar jet stream. Its position and intensity are similar to those in DJF, indicating that there is less seasonal variability in the Southern Hemisphere storm track. The model captures these features at T42 and T63. Similar to that in DJF, the intensity of the storm tracks is significantly underestimated in the simulation at T21.

In JJA at LGM (Fig. 11), the transient eddy activity along the southern edge of the land ice sheets increases, due to the enhanced temperature contrast between land and land ice in JJA. The Southern Hemisphere storm track intensifies at LGM and shifts poleward by about 3°–5°, consistent with an enhanced equator–Pole temperature gradient and poleward displacement of the maximum temperature gradient and westerlies (Fig. 7), in which the cold Antarctic is accentuated by the reduced moisture content at LGM. In JJA at LGM, significant changes in three storm tracks occur when model resolution changes from T21 to T42. In contrast to that in DJF, further increase in model horizontal resolution results in hardly any changes.

7. Changes in surface hydrology

a. Precipitation

Considering the tremendous increase in the transient eddy activity in the eastern North Atlantic along the sea ice edge in DJF, one would expect an increase in the associated precipitation in the region, because of condensation as warm, moist air is moved upward and poleward. Precipitation arising from this large-scale condensation process does indeed increase, as shown in Fig. 12, which gives both large-scale precipitation and total precipitation anomalies in DJF between the LGM and PD simulations at three resolutions. The meridional confinement of the storm track is also evident in the large-scale precipitation anomalies with the dipole structure. The large-scale precipitation also slightly increases at LGM over southern Europe at T42 and T63 simulations. There is also an increase in large-scale condensation on the northern west coast of North America, consistent with the eastward migration of the Pacific storm track.

The changes in the midlatitude large-scale condensation is to some extent compensated for by changes in the precipitation due to deep convection. Over sea ice in the Northern Hemisphere, the air is dry, cold, and stable at the LGM. As a result, there is a corresponding reduction in deep convection and the associated convective precipitation (not shown). In addition, the cooler air is holding less water vapor thus resulting in less precipitation globally. As a result, total precipitation in DJF at LGM over the North Atlantic and Europe decreases.

The changes in total precipitation over the Northern Hemisphere monsoon regions in JJA are shown in Fig. 13. Simulated regional changes are resolution dependent. The T21 simulation indicate that the monsoon precipitation over northwestern India and Africa is enhanced at LGM while the T42 and T63 simulations indicate slightly decreased precipitation over the same regions. The precipitation changes along the east coast of the Asian continent in the lower-resolution simulation are also in contradiction with those in the higher-resolution simulations. This indicates that the changes in deep convection are in some extent dependent on the model horizontal resolution. The largest difference occurs when horizontal resolution changes from T21 to T42. Further increase in horizontal resolution only results in quantitative changes.

b. Precipitation minus evaporation

The introduction of the ice age boundary conditions induces changes in precipitation minus evaporation (PE). It is of interest since it can provide some insight into changes in geographical distribution of vegetation type. In addition, changes in PE can be inferred from botanical and lake-level geological records of past climates. This allows a comparison of the model results with paleodata. The anomalies of the annual mean PE associated with the introduction of ice age boundary conditions at three horizontal resolutions are shown in Fig. 14. The main signal is of drier conditions over the Eurasian continent, and large parts of North and South America. Over the tropical continent, the changes in annual mean PE follow the changes in summer precipitation. The wetter conditions over Australia, central Africa, and north-western India in the LGM simulation at T21 are contradicted by the drier conditions in the LGM simulations at T42 and T63. The above-mentioned changes clearly indicate that there is a significant difference in the regional climate response to the ice age boundary conditions even with the same model but different horizontal resolutions. The surface hydrological changes are more horizontal-resolution dependent. In some regions, the lower-resolution model gives a signal opposite to that from the higher-resolution models.

c. Ice sheet mass balance

The mass balance over the Greenland and Antarctic ice sheets in the PD and LGM simulations at the three resolutions is given in Table 2, and the geographical distributions of snow accumulation rate in the LGM simulations are shown in Fig. 15. The area averaged net snow accumulation rates over the Greenland ice sheet in the PD and LGM simulations at T21 are about 80% larger than those in the simulations at T42 and T63. This is largely associated with the overestimation of large-scale condensation resulting from distortion of simulated storm tracks and atmospheric moisture availability in the lower-resolution model over this region.

The total precipitable water over the Antarctic is about 30% larger in the simulations at T21 than those at T42 and T63 in the southern summer due to slightly warmer air for both PD and LGM simulations at T21 (not shown). In the southern winter, moisture convergence due to baroclinic eddies is larger in the simulations at T21 than those at T42. As a result, precipitation over the Antarctic is overestimated in the lower-resolution model in both DJF and JJA seasons, resulting in overestimated snow accumulation rates.

In the LGM simulation, the mass budget characteristics over Greenland are quite different from the PD simulation. The ablation is negligible. The net accumulation rate over Greenland is consistent with the estimated snow accumulation rate derived from the oxygen isotopic composition of ice in the deep core by Kapsner et al. (1995). The reduction of the net snow accumulation rate is mainly due to the reduction in snowfall, which is the direct result of the colder climate and downstream shift of the North Atlantic storm track. Over the Antarctic, the features of the snow mass budget in the LGM simulations are similar to those of the PD simulations because changes in the storm track are smaller and the air temperature in this region is very low for both the PD and LGM. However, similar to the PD, the snowfall in the LGM simulation at T21 is higher than those at T42 and T63.

The regional snow accumulation rate over the Laurentide and Fennoscandian ice sheets in the LGM simulation at T21 is also different from those at T42 and T63. Generally, snow accumulates over the Laurentide and Fennoscandian ice sheets, particularly to the north, with a band of net ablation along the southern edges. Typical values of accumulation are 100–200 kg m−2 yr−1. However, the accumulation rate over central and northern parts of the ice sheets and the ablation along the southern edges at T21 simulation are significantly larger than those at T42 and T63. The net ablation zone between the Cordilleran ice sheet and the Laurentide ice sheet in the LGM simulations at T42 and T63 does not appear in the simulation at T21.

The ablation rates are much more sensitive to resolution. This is related to ablation occurring in a very narrow zone at the edge of the ice sheet (Glover 1999;Thompson and Pollard 1997). The higher-resolution models resolve this zone better.

8. Conclusions

This study of climate change due to the ice age boundary conditions suggests that both the global and regional climate responses to the ice age boundary conditions are horizontal-resolution dependent in the UGAMP GCM. There are significant regional differences in the response to the perturbed boundary conditions at different resolutions. The largest differences generally occur when horizontal resolution changes from T21 to T42. Further increase in horizontal resolution to T63 generally results in only quantitative differences. In some regions, the response to the ice age boundary conditions in the lower-resolution model is opposite to that in the higher-resolution models, with the simulations at higher-resolution models giving consistent signals with geological evidence. For example, the Indian and Australian summer monsoon precipitation at the LGM is enhanced at T21, which is contradicted by the weakened monsoon precipitation at T42 and T63. Given the fact that the lower-resolution version of the UGAMP GCM cannot reproduce the present-day climate as accurately as the higher-resolution one does, its simulation for the perturbed climate is questionable, especially over high latitudes where the climate is more sensitive to the external forcing. It should be borne in mind that the physical parameterizations in the model are probably not resolution independent. A parameterization scheme developed in a model at one resolution may not be correctly tuned when run at a different resolution. The present-day simulation using the model at T42 and T63 horizontal resolution gives a better agreement with observation in many aspects, such as storm track activity and the snow accumulation rates over the polar ice sheets, indicating that a T42 resolution is needed, at least for the UGAMP GCM, to simulate climate changes. Results also indicate that the insensitivity of simulated storm tracks to model horizontal-resolution change from T42 to T63 at PD may not be applicable to a different climate regime. The model horizontal resolution is an important factor that should be borne in mind in exploring the model–model and model–data differences in the PMIP. Some of these differences may arise from the difference in the horizontal resolution.

Acknowledgments

We would like to thank two anonymous reviewers for their constructive suggestions and comments on the paper. This work was funded by the EC through Grants EC5V-CT94-057 and ENV4-CT95-0122. The computing time was provided by the UGAMP, which is funded by the U.K. Natural Environment Research Council.

REFERENCES

  • Boer, G. J., and M. Lazare, 1988: Some results concerning the effects of horizontal resolution and gravity-wave drag on simulated climate. J. Climate,1, 789–806.

  • Boyle, J. S., 1993: Sensitivity of dynamical quantities to horizontal resolution for a climate simulation using the ECMWF (Cycle 33) model. J. Climate,6, 796–815.

  • Bromwich, D. H., 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.

  • CLIMAP Project Members, 1981: Seasonal reconstruction of the earth’s surface at the last glacial maximum. Geol. Soc. Amer. Map. Chart. Ser.,MC-36, 18 pp.

  • Deque, M., C. Dreveton, A. Braun, and D. Cariolle, 1994: The ARPEGE/IFS atmosphere model: A contribution to the French community climate modelling. Climate Dyn.,10, 249–266.

  • Dong, B.-W., and P. J. Valdes, 1995: Sensitivity studies of Northern Hemisphere glaciation using an atmospheric general circulation model. J. Climate,8, 2471–2496.

  • ——, and ——, 1998: Simulations of the last glacial maximum climates using a general circulation model: Prescribed versus computed sea surface temperatures. Climate Dyn.,14, 571–591.

  • Glover, R. W., 1999: Influence of spatial resolution and treatment of orography on GCM estimates of the surface mass balance of the Greenland ice sheet. J. Climate,12, 551–563.

  • Hall, M. N. J., P. J. Valdes, and B.-W. Dong, 1996: The maintenance of the last great ice sheets: A UGAMP GCM study. J. Climate,9, 1004–1019.

  • Harrison, E. F., P. Minnis, B. R. Barkstrom, V. Ramanathan, R. D. Cess, and G. G. Gibson, 1990: Seasonal variation of cloud radiative forcing derived from the Earth Radiation Budget Experiment. J. Geophys. Res.,95, 18 687–18 703.

  • Hoskins, B. J., H. H. Hsu, I. N. James, M. Masutani, P. D. Sardeshmukh, and G. H. White, 1989: Diagnostics of the global atmospheric circulation. Tech. Document WMO/TD-326, WCRP-27, 217 pp.

  • Joussaume, S., and K. E. Taylor, 1995: Status of the Paleoclimate Modeling Intercomparison Project (PMIP). Proc. First Int. AMIP Scientific Conf., Monterey, CA, AMIP, 425–430. [WMO/TD-732, WCRP-92.].

  • Kageyama, M., P. J. Valdes, G. Ramstein, C. D. Hewitt, and U. Wyputta, 1999: Northern Hemisphere storm tracks in present day and last glacial maximum climate simulations: A comparison of the European PMIP models. J. Climate,12, 742–760.

  • 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 the past 18 000 years. Nature,373, 52–54.

  • Kiehl, J. T., and D. L. Williamson, 1991: Dependence of cloud amount on horizontal resolution in the National Center for Atmospheric Research Community Climate Model. J. Geophys. Res.,96, 10 955–10 980.

  • Peltier, W. R., 1994: Ice age paleotopography. Science,265, 195–201.

  • Phillips, T. J., L. C. Corsetti, and S. L. Grotch, 1995: The impact of horizontal resolution on moist processes in the ECMWF model. Climate Dyn.,11, 85–102.

  • Rind, D., 1988: Dependence of warm and cold climate depiction on climate model resolution. J. Climate,1, 965–997.

  • Senior, C. A., 1995: The dependence of climate sensitivity on the horizontal resolution of a GCM. J. Climate,8, 2860–2880.

  • Thompson, S. L., and D. Pollard, 1997: Greenland and Antarctic mass balances for present and doubled atmospheric CO2 from the GENESIS version-2 global climate model. J. Climate,10, 871–900.

  • Warrick, R. A., and J. Oerlemans, 1990: Sea level rise. Climate Change, the IPCC Assessment, J. T. Houghton, G. J. Jenkins, and J. J. Ephraums, Eds., Cambridge University Press, 257–281.

  • Williamson, D. L, J. T. Kiehl, and J. J. Hack, 1995: Climate sensitivity of the NCAR Community Climate Model (CCM2) to horizontal resolution. Climate Dyn.,11, 377–397.

Fig. 1.
Fig. 1.

Changes in ice sheet height and extent of the updated ice sheets (Peltier 1994) at LGM at three truncations: (a) T63, (b) T42, and (c) T21. Contours are at 250 and 500 m, and then increasing by 500 m. Shading indicates ice height greater than 1000 m.

Citation: Journal of Climate 13, 9; 10.1175/1520-0442(2000)013<1554:CATLGM>2.0.CO;2

Fig. 2.
Fig. 2.

Seasonal evolution of the surface air temperature at three horizontal resolutions for the PD and LGM simulations: (a) over globe, (b) over land, and (c) over ocean.

Citation: Journal of Climate 13, 9; 10.1175/1520-0442(2000)013<1554:CATLGM>2.0.CO;2

Fig. 3.
Fig. 3.

Similar to Fig. 2, but total precipitation.

Citation: Journal of Climate 13, 9; 10.1175/1520-0442(2000)013<1554:CATLGM>2.0.CO;2

Fig. 4.
Fig. 4.

Annual mean surface air temperature anomalies between LGM and PD simulations and the anomaly difference due to change in model horizontal resolution: (a) at T21, (b) T42 − T21, and (c) T63 − T42. Thick line indicates the extent of ice sheets at LGM. Shading indicates that the anomalies are statistically significant at 99% confidence level using the Student’s t test.

Citation: Journal of Climate 13, 9; 10.1175/1520-0442(2000)013<1554:CATLGM>2.0.CO;2

Fig. 5.
Fig. 5.

850-hPa air temperature anomalies between the LGM and PD simulations in DJF, JJA, and the anomaly difference due to change in model horizontal resolution: (a) and (b) at T21, (c) and (d) T42 − T21, and (e) and (f) T63 − T42. (a), (c), and (e) show DJF and (b), (d), and (f) show JJA.

Citation: Journal of Climate 13, 9; 10.1175/1520-0442(2000)013<1554:CATLGM>2.0.CO;2

Fig. 6.
Fig. 6.

Zonally asymmetric streamfunction and zonal wind at 500 hPa in DJF for the (left) PD simulations and (right) the corresponding anomalies between the LGM and PD simulations: (a) and (d) T63, (b) and (e) T42, and (c) and (f) T21. Contour intervals are 2.0 × 106 m s−2 with negative values dashed. Zonal wind (m s−1) is plotted in shading scale.

Citation: Journal of Climate 13, 9; 10.1175/1520-0442(2000)013<1554:CATLGM>2.0.CO;2

Fig. 7.
Fig. 8.
Fig. 8.

The high-pass transient eddy kinetic energy at 250 hPa in DJF for the PD simulations: (a) T63, (b) T42, (c) T21, (d) T63 − T42, (e) T42 − T21, and (f) ECMWF analysis 1983–89. Regions with values greater than 100 m2 s−2 are shaded.

Citation: Journal of Climate 13, 9; 10.1175/1520-0442(2000)013<1554:CATLGM>2.0.CO;2

Fig. 9.
Fig. 9.

Similar to Fig. 8, but for DJF for the LGM simulations.

Citation: Journal of Climate 13, 9; 10.1175/1520-0442(2000)013<1554:CATLGM>2.0.CO;2

Fig. 10.
Fig. 10.

Similar to Fig. 8, but for JJA for the PD simulations.

Citation: Journal of Climate 13, 9; 10.1175/1520-0442(2000)013<1554:CATLGM>2.0.CO;2

Fig. 11.
Fig. 11.

Similar to Fig. 8, but for JJA for the LGM simulations.

Citation: Journal of Climate 13, 9; 10.1175/1520-0442(2000)013<1554:CATLGM>2.0.CO;2

Fig. 12.
Fig. 12.

(left) Large-scale and (right) total precipitation anomalies in DJF over the Northern Hemisphere between the LGM and PD simulations: (a) and (d) T63, (b) and (e) T42, and (c) and (f) T21. Positive anomalies are within solid line and negative anomalies within dotted line. The thick line outlines the extent of ice sheets at LGM. Shading indicates the anomalies are statistically significant at 99% confidence level using the Student’s t test.

Citation: Journal of Climate 13, 9; 10.1175/1520-0442(2000)013<1554:CATLGM>2.0.CO;2

Fig. 13.
Fig. 13.

Total precipitation anomalies in JJA over the tropical regions in the Eastern Hemisphere between the LGM and PD simulations: (a) T63, (b) T42, and (c) T21. Positive anomalies are within solid line and negative anomalies within dotted line. Shading indicates the anomalies are statistically significant at 99% confidence level using the Student’s t test.

Citation: Journal of Climate 13, 9; 10.1175/1520-0442(2000)013<1554:CATLGM>2.0.CO;2

Fig. 14.
Fig. 14.

Annual mean precipitation minus evaporation anomalies between the LGM and PD simulations:(a) T21, (b) T42, and (c) T63. Positive anomalies are within solid line and negative anomalies within dotted line. The thick line outlines the extent of ice sheets at LGM. Shading indicates the anomalies are statistically significant at 99% confidence level using the Student’s t test.

Citation: Journal of Climate 13, 9; 10.1175/1520-0442(2000)013<1554:CATLGM>2.0.CO;2

Fig. 15.
Fig. 15.

Net snow accumulation rate in the LGM simulations. (a) T63, (b) T42, and (c) T21. Contours are at 50, 100, 200, and 400 kg m−2 yr−1, and then increasing by 400 kg m−2 yr−1 with negative values dashed. Areas with net accumulation rate greater than 200 kg m−2 yr−1 are shaded.

Citation: Journal of Climate 13, 9; 10.1175/1520-0442(2000)013<1554:CATLGM>2.0.CO;2

Table 1.

Annually global-averaged surface quantities for the PD and LGM simulations at different horizontal resolutions.

Table 1.
Table 2.

Mass balance over various ice sheets for the PD and LGM simulations at different horizontal resolutions. The PD observation is based on Warrick and Oerlemans (1990).

Table 2.
Save