• Alkama, M. R., , M. Kageyama, , G. Ramstein, , O. Marti, , P. Ribstein, , and D. Swingedouw, 2008: Impact of a realistic river routing in coupled ocean-atmosphere simulations of the Last Glacial Maximum climate. Climate Dyn., 30 , 855869.

    • Search Google Scholar
    • Export Citation
  • Braconnot, P., and Coauthors, 2007a: Results of PMIP2 coupled simulations of the Mid-Holocene and Last Glacial Maximum—Part 1: Experiments and large-scale features. Climate Past, 3 , 261277.

    • Search Google Scholar
    • Export Citation
  • Braconnot, P., and Coauthors, 2007b: Results of PMIP2 coupled simulations of the Mid-Holocene and Last Glacial Maximum—Part 2: Feedbacks with emphasis on the location of the ITCZ and mid- and high latitudes heat budget. Climate Past, 3 , 261277.

    • Search Google Scholar
    • Export Citation
  • Cai, M., 2005: Dynamical amplification of polar warming. Geophys. Res. Lett., 32 , L22710. doi:10.1029/2005GL024481.

  • Crucifix, M., 2006: Does the Last Glacial Maximum constrain climate sensitivity? Geophys. Res. Lett., 33 , L18701. doi:10.1029/2006GL027137.

    • Search Google Scholar
    • Export Citation
  • Hall, A., 2004: The role of surface albedo feedback in climate. J. Climate, 17 , 15501568.

  • Held, I. M., , and B. J. Soden, 2006: Robust responses of the hydrological cycle to global warming. J. Climate, 19 , 56865699.

  • Holland, M. M., , and C. M. Bitz, 2003: Polar amplification of climate change in coupled models. Climate Dyn., 21 , 221232.

  • Hourdin, F., and Coauthors, 2006: The LMDZ4 general circulation model: Climate performance and sensitivity to parameterized physics with emphasis on tropical convection. Climate Dyn., 27 , 787813.

    • Search Google Scholar
    • Export Citation
  • Kim, S., 2004: The effect of atmospheric CO2 and ice sheet topography on LGM climate. Climate Dyn., 22 , 639651.

  • Marti, O., and Coauthors, 2006: The new IPSL climate system model: IPSL-CM4. IPSL Note du Pôle de Modélisation 26, 332 pp. [Available online at http://dods.ipsl.jussieu.fr/omamce/IPSLCM4/DocIPSLCM4/HTML/].

    • Search Google Scholar
    • Export Citation
  • Masson-Delmotte, V., and Coauthors, 2005: Past and future polar amplification of climate change: Climate model intercomparisons and ice-core constraints. Climate Dyn., 26 , 513529. doi:10.1007/s00382-005-0081-9.

    • Search Google Scholar
    • Export Citation
  • Otto-Bliesner, B. L., , E. C. Brady, , G. Clauzet, , R. Tomas, , S. Levis, , and Z. Kothavala, 2006: Last Glacial Maximum and Holocene climate in CCSM3. J. Climate, 19 , 25262544.

    • Search Google Scholar
    • Export Citation
  • Peixoto, J., , and A. H. Oort, 1992: Physics of Climate. American Institute of Physics, 520 pp.

  • Peltier, W. R., 2004: Global glacial isostasy and the surface of the ice-age Earth: The ICE-5G (VM2) model and GRACE. Annu. Rev. Earth Planet. Sci., 32 , 111149. doi:10.1146/annurev.earth.32.082503.144359.

    • Search Google Scholar
    • Export Citation
  • Ramstein, G., , Y. Serafini-Le Treut, , H. Le Treut, , M. Forichon, , and S. Joussaume, 1998: Cloud processes associated with past and future climate changes. Climate Dyn., 14 , 233247.

    • Search Google Scholar
    • Export Citation
  • Shin, S., , Z. Liu, , B. Otto-Bliesner, , E. Brady, , J. Kutzbach, , and S. Harrison, 2003: A simulation of the Last Glacial Maximum climate using the NCAR-CCSM. Climate Dyn., 20 , 127151.

    • Search Google Scholar
    • Export Citation
  • Solomon, S., , D. Qin, , M. Manning, , Z. Chen, , M. Marquis, , K. B. Averyt, , M. Tignor, , and H. L. Miller, Eds. 2007: Climate Change 2007: The Physical Science Basis. Cambridge University Press, 996 pp.

    • Search Google Scholar
    • Export Citation
  • Sutton, R. T., , B. Dong, , and J. M. Gregory, 2007: Land/sea warming ratio in response to climate change: IPCC AR4 model results and comparison with observations. Geophys. Res. Lett., 34 , L02701. doi:10.1029/2006GL028164.

    • Search Google Scholar
    • Export Citation
  • Swingedouw, D., , P. Braconnot, , P. Delecluse, , E. Guilyardi, , and O. Marti, 2007: Quantifying the AMOC feedbacks during a 2 × CO2 stabilization experiment with land-ice melting. Climate Dyn., 29 , 521534.

    • Search Google Scholar
    • Export Citation
  • Weaver, A. J., , M. Eby, , A. F. Fanning, , and E. C. Wiebe, 1998: Simulated influence of carbon dioxide, orbital forcing and ice sheets on the climate of the Last Glacial Maximum. Nature, 394 , 847853.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Annual-mean surface temperature change for an increase in GHG concentration from LGM to preindustrial values, for (a) LGM (“glacial”) and (b) modern (“interglacial”) boundary conditions, and (c) the difference between the temperature change under interglacial and glacial boundary conditions. Blue contours represent the ice sheets imposed in the model. Units: °C. Blank areas are regions for which the null hypothesis cannot be rejected at 95%.

  • View in gallery

    Zonal- and annual-mean (a) surface temperature change (°C), (b) surface temperature change divided by the global-mean temperature change, and (c) the ratio between the temperature change over land and ocean (for grid cells representing 100% of the surface type) for modern (black) and LGM (gray) boundary conditions. Horizontal lines represent global averages.

  • View in gallery

    Annual-mean surface energy flux [see Eq. (4)] differences (GHGpreind − GHGlgm) (W m−2) for modern boundary conditions: (a) longwave surface emission changes, (b) solar absorption changes due to GHG concentration changes, (c) solar absorption changes due to cloud changes, (d) solar absorption changes due to albedo changes, (e) direct longwave feedback differences, (f) effect of water vapor concentration changes on longwave incoming radiation, (g) effect of fixed GHG concentration changes on longwave incoming radiation, (h) effect of cloud changes on longwave incident radiation, (i) changes in latent heat fluxes, (j) changes in sensible heat flux, (k) heat advection changes, (l) heat storage differences, and (m) errors resulting from the decomposition. Positive fluxes point downward, that is, correspond to a warming of the surface. Blank areas are regions for which the null hypothesis of the results cannot be rejected at the 95% level.

  • View in gallery

    Mean contributions of the different surface fluxes from Eq. (4) to the changes in longwave surface emission for modern boundary conditions, for all types of surfaces (gray bars), continental surfaces only (grid cells with 100% land, black bars), and for oceanic surfaces (100% oceanic grid cells, white bars). Fluxes are normalized by the changes in longwave emission at each grid cell (a) before being averaged globally and (b) for high latitudes (superior to 60° in both hemispheres).

  • View in gallery

    Annual-mean difference for modern boundary conditions in (a) cloud cover change (in percentage of the sky covered by clouds), (b) total effect of cloud changes on solar and longwave radiation for modern boundary conditions (W m−2), and (c) precipitation changes (mm day−1). Blank areas are regions for which the null hypothesis cannot be rejected at 95%.

  • View in gallery

    Global-mean contributions of the different surface fluxes from Eq. (4) to the land–sea longwave emission change ratio [see Eq. (5)] for interglacial boundary conditions (white bars) and glacial boundary conditions (black bars). Ratios are calculated over 100% land and oceanic grid cells. Contributions are calculated for each latitude before being averaged.

  • View in gallery

    Same as in Fig. 4, but for the differences between the changes under interglacial and glacial boundary conditions.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 82 82 20
PDF Downloads 52 52 17

Impact of Greenhouse Gas Concentration Changes on Surface Energetics in IPSL-CM4: Regional Warming Patterns, Land–Sea Warming Ratios, and Glacial–Interglacial Differences

View More View Less
  • 1 LSCE/IPSL, UMR CEA-CNRS-UVSQ 1572, Saclay, France
© Get Permissions
Full access

Abstract

The temperature response to a greenhouse gas (GHG) concentration change is studied in an ocean–atmosphere coupled model—L’Institut Pierre-Simon Laplace Coupled Model, version 4 (IPSL-CM4)—for both a glacial and an interglacial context. The response to a GHG concentration changing from Last Glacial Maximum (LGM) to preindustrial values is similar for both climatic contexts in terms of temperature pattern, but the magnitude is greater under modern ones. The model simulates the classical amplification of the temperature response in the northern high latitudes compared to lower latitudes and over the land surfaces compared to the ocean.

The physical reasons for the differential warming according to the latitude and to the surface type are studied through an analysis of the energy flux changes, which are decomposed to consider and quantify many different physical processes. The results highlight the role of many different factors in the thermal response to a GHG forcing for different regions, and stress, for instance, the large effect of increased water vapor concentration in the atmosphere. Concerning the land–sea warming ratio, several fluxes contribute to the final value of the ratio, with latent flux having the greatest influence. The different contributions are quantified. The comparison of the flux changes between the interglacial and glacial contexts shows that the differences are more than a simple effect of different surface emissions of the base state. It suggests that the climatic context is particularly important for the cloud and oceanic advection responses to the forcing, along with albedo effects.

Corresponding author address: Alexandre Laîné, LSCE/IPSL, UMR CEA-CNRS-UVSQ 1572, CE Saclay, L’Orme des Merisiers, Bâtiment 701, 91191 Gif-sur-Yvette CEDEX, France. Email: alexandre.laine@lsce.ipsl.fr

Abstract

The temperature response to a greenhouse gas (GHG) concentration change is studied in an ocean–atmosphere coupled model—L’Institut Pierre-Simon Laplace Coupled Model, version 4 (IPSL-CM4)—for both a glacial and an interglacial context. The response to a GHG concentration changing from Last Glacial Maximum (LGM) to preindustrial values is similar for both climatic contexts in terms of temperature pattern, but the magnitude is greater under modern ones. The model simulates the classical amplification of the temperature response in the northern high latitudes compared to lower latitudes and over the land surfaces compared to the ocean.

The physical reasons for the differential warming according to the latitude and to the surface type are studied through an analysis of the energy flux changes, which are decomposed to consider and quantify many different physical processes. The results highlight the role of many different factors in the thermal response to a GHG forcing for different regions, and stress, for instance, the large effect of increased water vapor concentration in the atmosphere. Concerning the land–sea warming ratio, several fluxes contribute to the final value of the ratio, with latent flux having the greatest influence. The different contributions are quantified. The comparison of the flux changes between the interglacial and glacial contexts shows that the differences are more than a simple effect of different surface emissions of the base state. It suggests that the climatic context is particularly important for the cloud and oceanic advection responses to the forcing, along with albedo effects.

Corresponding author address: Alexandre Laîné, LSCE/IPSL, UMR CEA-CNRS-UVSQ 1572, CE Saclay, L’Orme des Merisiers, Bâtiment 701, 91191 Gif-sur-Yvette CEDEX, France. Email: alexandre.laine@lsce.ipsl.fr

1. Introduction

The surface energy fluxes can be decomposed into several components, representing different physical processes that interact to eventually equilibrate on the long term. The equilibrium state of the surface heat budget depends on geographical and climatic factors. For example, at high latitudes, the incoming solar radiation is weak because of the strong inclination of the sun rays with respect to the earth’s surface. Therefore, the temperature is cold, and the surface can be covered either by snow or ice. This further reduces the heat absorption from solar radiation resulting from a strong albedo. In these regions, the net upward longwave radiation is usually stronger than the absorbed solar radiation; therefore, the other surface fluxes release energy to the surface, which is usually done mainly via the sensible heat flux and oceanic advection. At low latitudes, the strongly absorbed and almost vertically oriented incoming solar radiation is usually greater than the net emission of energy through infrared radiation plus latent heat flux. The surface releases energy to the air through sensible heat fluxes and advects it in the case of the ocean. Connecting these remote regions, heat is transferred by the atmosphere and the oceans from the low to high latitudes.

In the case of an increased greenhouse gas (GHG) concentration in the atmosphere, the first-order global effect to be expected is a warming of the air resulting from stronger longwave absorption, and a warming of the surface through stronger backward infrared downward radiation. Another general aspect in the temperature response to a given GHG increase is its “polar amplification” (Solomon et al. 2007), that is, the fact that the warming is increasing with latitude. To first order, this can be explained by the fact that at high latitudes, the temperature change modifies the sea ice and land snow and ice limits, especially in the Northern Hemisphere. The albedo change considerably affects the solar radiation budget toward a stronger absorption for snow/ice retreat, which makes the temperature response in these areas particularly strong. This albedo feedback has been quantified by Hall (2004) to be responsible for about half of the polar temperature response in the case of a doubling of CO2 in a coarse-resolution coupled general circulation model (GCM). Braconnot et al. (2007b) have estimated that the role played by the feedback from snow and sea ice at mid- and high latitudes contributed to about half the cooling of the Northern Hemisphere at the Last Glacial Maximum (LGM; 21 kyr BP). Holland and Bitz (2003) have studied the response simulated by coupled models to CO2 doubling, and they show that the models’ polar amplification is significantly correlated to changes in sea ice extent [and to sea ice thickness in the control (present) climate, influencing the latter], to wintertime polar cloud cover changes, and to poleward ocean heat transport anomalies. They do not find correlations either with control climate snow cover or with summertime polar cloud cover changes. Cai (2005) shows that an increased GHG forcing can increase the atmospheric poleward heat transport and lead to a polar warming amplification. Finally, Masson-Delmotte et al. (2005) have highlighted the polar amplification for both future climate (increased GHG) and paleoclimate simulations (especially for the LGM), and they show that this phenomenon is attested by temperature reconstructions from proxy data for the glacial climate.

Another contrast in the temperature response to GHG concentration change is the land–sea warming ratio. Sutton et al. (2007) explain that for the equilibrium case of an increased GHG concentration, the warming of the ocean surface is weaker than for land, not because of the stronger heat capacity of the oceans (which would only affect the transient response), but because of a greater evaporation response. The wetter a surface is, the more the increased surface energy input resulting from increased backward longwave radiation is balanced by evaporation, effectively reducing the surface warming. According to their interpretation, the obvious contrast in surface humidity between continental and oceanic surfaces explains the stronger increase in temperature for the former surfaces than for the latter. The land–sea temperature change difference is also suggested by proxy data and GCMs for paleoclimates like the LGM (Braconnot et al. 2007a).

In this study, we consider the surface temperature response to changes in GHG concentration for a glacial (LGM) and an interglacial (preindustrial) context (which are different mostly in terms of ice sheet boundary conditions) using the ocean–atmosphere coupled model, L’Institut Pierre-Simon Laplace Coupled Model, version 4 (IPSL-CM4). We use a decomposition of the surface energy fluxes for understanding the changes in the surface heat budget and quantifying the relative roles of different processes involved in the surface temperature change. The main questions we are addressing are as follows:

  • What is the relative role of the different energy fluxes associated with the surface temperature changes in different regions (oceans and continents, high and lower latitudes)?
  • What is the relative role of the different fluxes for determining the land–sea warming contrast?
  • Which energy fluxes are involved in the difference between the responses to the GHG forcing under glacial and interglacial contexts?

In the next section, we present the model and the simulations used in this study. The following sections provide the results of the experiments, in terms of surface temperature changes (section 3) and surface energy flux changes (section 4). The global-mean temperature changes for the different sets of experiments are presented in section 3a. A quantification of the relative roles of GHG concentration change, of changes in other boundary conditions, and their interaction in the temperature change between the LGM and the preindustrial climates in IPSL-CM4 is performed and compared to previously published results of other models. Section 3b presents the regional patterns of warming for the increase in GHG concentration. The decomposition of the surface energy fluxes is detailed in section 4a. The changes in these fluxes in response to the GHG forcing are discussed in section 4b in the modern context. Their relative roles are quantified in section 4c for the land–sea warming contrast, and in section 4d for the glacial–interglacial differential response. Finally, section 5 summarizes the main results and concludes the article.

2. Model and numerical experiments

The coupled model used in this study is IPSL-CM4 (Marti et al. 2006), developed at IPSL. The coupling involves the atmosphere, the land surface, the ocean, and the sea ice. The Laboratoire de Météorologie Dynamique atmospheric general circulation model with zooming capability, version 3.3 (LMDZ3.3) has a regular horizontal grid of 96 × 72 points, and 19 levels in the vertical. The oceanic model, Océan Parallélisé (OPA) in its global version ORCA2, has an irregular grid of 182 × 149 points horizontally and 31 levels in depth. The sea ice model is the Louvain Ice Model (LIM), Organizing Carbon and Hydrology in Dynamic Ecosystems (ORCHIDEE) is the land surface model, which also includes the river routing scheme, and the coupling between all components is performed using Oklahoma Mesonet–Oklahoma Atmospheric Surface Layer Instrumentation System (OASIS; version 3). More information on IPSL-CM4 and its different components can be found in Marti et al. (2006). The fully coupled model and its various components have successfully been used for present, future, and paleoclimate studies (e.g., Braconnot et al. 2007a; Swingedouw et al. 2007; Hourdin et al. 2006).

Two sets of simulations have been run: one with GHG concentrations corresponding to the LGM values, and one with GHG concentrations corresponding to preindustrial conditions (1860). The low levels of GHG concentrations are 185 ppm for CO2, 350 ppb for CH4, and 200 ppb for N2O; the preindustrial levels are 280 ppm for CO2, 760 ppb for CH4, and 270 ppb for N2O.

In each GHG concentration case, one simulation is run under modern conditions, and one is run under LGM conditions, in terms of sea level changes (i.e., bathymetry and land–sea distribution adapted to LGM), ice sheets topography and albedo, and orbital parameters. These parameters, excluding the GHG concentrations, are referred to as “boundary conditions” in the rest of the paper. In the case of modern geographical conditions with preindustrial GHG concentration, the simulation corresponds to the preindustrial run performed for the Fourth Assessment Report (AR4) of the Intergovernmental Panel on Climate Change (IPCC). The LGM boundary conditions are the ones defined in the second phase of the Paleoclimate Modeling Intercomparison Project (PMIP2; Braconnot et al. 2007a). The ICE-5G (Peltier 2004) reconstructions of the LGM ice sheets are used, and the corresponding coastline changes are prescribed. The Laurentide ice sheet covers present-day Canada and the Great Lakes, with elevations as high as 4000 m. The Fennoscandian ice sheet covers present-day Ireland, Scotland, Scandinavia, and the Barents Sea up to the Spitzberg Islands. Its elevation reaches about 2500 m. Antarctica is also raised by a few hundreds of meters at the LGM in the reconstruction. The coupled model does not have an interactive ice sheet module. The ice sheets are prescribed through their topography and their surface type and cannot disappear. Snow can accumulate on them to the limit of 3000 kg m−2 after a time filtering of 10 yr, a value over which the excess of snow is sent back to the ocean. It is performed in order to compensate for the lack of ice sheet dynamics, and hence calving, and to prevent continuous snow accumulation. In both the LGM and modern geographical conditions, vegetation is the modern one, except of course where ice sheets are present in the case of LGM simulations. The orbital parameters, although changed at LGM, do not imply significant insolation changes at any latitude.

The simulations using modern boundary conditions are referred to as “MOD-GHGpreind” and “MOD-GHGlgm” for preindustrial and LGM GHG concentrations, respectively. These boundary conditions are also referred to as “interglacial” or “modern boundary” conditions herein. The simulations using LGM or “glacial” boundary conditions are referred to as “LGM-GHGpreind” and “LGM-GHGlgm.” We use 50 yr of each simulation to derive the climatic variables. The MOD-GHGpreind simulation has been run for 500 yr starting from the Levitus oceanic conditions prior to the selected years of analyses. It corresponds to years 2361–2410 of the IPCC AR4 simulation of the preindustrial IPSL-CM4. The simulation is stable and the drift in global surface temperature calculated from monthly data during the period considered is 0.001°C century−1. MOD-GHGlgm has been run for 400 yr from initial conditions of the MOD-GHGpreind simulation. The drift is equal to 0.021°C century−1. LGM-GHGlgm has been run for 300 yr starting from a glacial initial state corresponding to the output of a previous LGM simulation. It corresponds to the LGM simulation presented in Alkama et al. (2008). The drift is of −0.006° century−1. Finally, LGM-GHGpreind has been run for 200 yr starting from the same initial conditions as LGM-GHGlgm. Its drift equals −0.011°C century−1.

We use the Student’s t test of null hypothesis at 95% in order to isolate regions where the changes can be considered significant. The 50 yr of each dataset are considered independent and the number of degrees of freedom considered for the threshold value is therefore 49.

3. Surface temperature responses

a. Forcing contributions to the preindustrial LGM global surface temperature change

Table 1 indicates the global-mean surface temperatures for the different simulations, and the differences for the changes in GHG concentration and other boundary conditions (mostly ice sheet) from LGM to preindustrial levels. The global-mean temperature change from preindustrial to LGM conditions is −3.4°C in this model, which places it among the atmosphere–ocean coupled general circulation models (AOGCMs) with the lowest sensitivity (Braconnot et al. 2007a). It is at odds with the climate sensitivity of this model for 2 × CO2 concentrations, which is high (4.4°C; see Table 8.2 in Solomon et al. 2007) compared to others. A discussion on the nonlinear processes resulting in different climate sensitivities for LGM and 2 × CO2 climates can be found in Crucifix (2006).

The increased GHG concentration from LGM to preindustrial levels contributes to a warming of 2.4°C in the preindustrial context and 2.0°C in the LGM context. The imposition of the boundary conditions other than the GHG concentration from the LGM and the preindustrial context results in a global surface temperature increase of 1.4°C under preindustrial GHG concentration levels and 1.0°C under LGM preindustrial GHG concentration levels. The differences in the temperature change under both contexts indicate the interaction between the different LGM forcings.

The contribution of the reduced GHG concentration to the total decrease in temperature between the preindustrial climate and the LGM is about 71% if we consider the preindustrial climate as the reference one (about 59% if we consider the LGM as the base state) and 41% for the influence of the ice sheets (29% respectively), and the role of the interaction between both forcings is evaluated to be −12% (+12%, respectively) in IPSL-CM4. Locally, the role played by the ice sheets, GHG concentration reduction, and the interaction between both varies a lot. The temperature response to the ice sheet forcing is mostly localized to the LGM ice sheet regions, whereas the response to the GHG concentration change is globally distributed. These figures differ slightly from previously published results from other models, for which the influence of the GHG effect accounts for about 50%–60% to the total cooling from preindustrial conditions (Otto-Bliesner et al. 2006; Kim 2004; Shin et al. 2003), although Weaver et al. (1998) find values closer to ours (about 78% for the CO2 effect alone). Nevertheless, the types of models used in these studies and/or the boundary conditions were not precisely similar to the ones used in this paper.

b. Regional temperature differences to GHG concentration changes

Figures 1a–c show the response simulated by the model in terms of annual surface temperature change to an increase from LGM to preindustrial GHG concentration for glacial boundary conditions, modern boundary conditions, and the difference between modern and glacial boundary conditions, respectively. Both simulations indicate a global warming as a response to the increase in GHG concentration.

In both cases (Figs. 1a,b), the maximum warming occurs in the high northern latitudes, especially over oceanic regions such as the Greenland and Norwegian Seas, and also the Hudson Bay under modern conditions. It corresponds to polar amplification that we hereafter refer to as the “Arctic amplification,” because it is more restricted to the Northern Hemisphere. The maximum warming reaches 13.0° and 18.5°C under glacial and interglacial conditions, respectively. Over the continents, the strongest warming in the high latitudes is found south of the ice sheets under LGM boundary conditions and around the oceanic regions of maximum warming for the modern context. This stronger warming in the high northern latitudes over the continents is also part of the Arctic amplification of the global warming under increased GHG concentration. The latter is also obvious when the zonal-mean temperature change is plotted as a function of latitude (Fig. 2a).

Localized patterns of temperature change are also found in different regions. Over the oceans, a local maximum of the warming is found in the low latitudes, especially in the central equatorial Pacific, in the northwestern part of the Northern Hemisphere subtropical gyres in the detachment zones of the Kuroshio and Gulf Stream, and off the southern tip of Africa. Over the continents, local maxima in the warming are found in the midlatitudes over Asia, Amazonia, and the northwestern, eastern, and southern parts of Africa.

The global and local patterns of warming are very similar for glacial and interglacial boundary conditions (Figs. 1a,b), indicating similar responses under both climatic contexts. Nevertheless, the response is usually enhanced under modern boundary conditions, with most regions experiencing a stronger warming in this context, except for some localized regions, especially in the Southern Ocean (Fig. 1c). The zonal-mean temperature changes, normalized by the global-mean temperature increase for both boundary conditions (Fig. 2b), indicate that the local warming patterns are very similar and can be considered as being amplified under modern boundary conditions. Indeed, the departure from the global mean is greater under modern boundary conditions for latitudes experiencing either a stronger or a weaker warming than that on average, north of 40°N and between 20° and 60°S, respectively.

Concerning the land–sea warming ratio at any given latitude, the model usually simulates a stronger warming under increased GHG concentration over the continents than over the oceans for both LGM and modern contexts, except for the higher latitudes where the warming is the strongest over the oceans. Figure 2c represents the ratio between the longitudinal-mean temperature change over the continents and over the oceans (land–sea warming ratio) as a function of latitude. The latitudinal variations of the warming ratio are very similar for the two sets of experiments in terms of pattern and amplitude, with a mean value around 1.5, the weakest ratios found in the high latitudes (with values of about 0.5), and the highest ratios found around 25°S and 45°N (about two). The significance of the ratio as a large-scale indicator is not very relevant between 40° and 50°S because grid cells fully covered by land in the model are only a few at a given latitude.

To understand these temperature responses, and which are the main physical processes involved, we perform an analysis of the surface energy fluxes. We then attempt to understand the global and local warming patterns, the land–sea warming ratio, and the differences in the responses under glacial and interglacial conditions.

4. Surface flux analysis

a. Surface heat budget

The instantaneous heat budget per unit area at the surface is
i1520-0442-22-17-4621-e1
where ρ refers to the density of the surface, cp its specific heat capacity at constant pressure, H the depth until which the energy fluxes are absorbed, SWdn and SWup the downward and upward shortwave radiation fluxes at the surface, respectively, LWdn and LWup the downward and upward fluxes for longwave radiation, lat and sens the latent and sensible heat fluxes, respectively, and adv the advection in the case of oceanic surfaces. Diffusion is neglected. The signs of the fluxes are all considered according to the same vertical direction. The fluxes are positive when pointing downward, that is, when corresponding to a warming of the surface.

The term on the left-hand side consists of heat storage or release. Over land, there is no advection and the surface heat variation is evaluated from the residue of other surface fluxes, and should include variations in snow amount, especially snow melting. Over ice-free oceans, we consider that the energy from the surface fluxes is spread within the mixed layer, and H therefore represents the mixed layer depth, and the advection term, which is calculated as the heat variation minus surface fluxes, represents the heat transport within this mixed layer. For surfaces covered by sea ice, the heat storage is evaluated using H of the snow plus sea ice thickness, and the energy flux at the base of sea ice is removed in order to consider the heat variations resulting from the surface fluxes and horizontal advection only. The term adv is then evaluated as the residue of this heat variation minus other surface fluxes.

Given the definition of albedo, SWdn + SWup = (1 − α)SWdn. We consider the effect of clouds on solar radiation by simply decomposing SWdn in a clear-sky component (SWdn_clr) and a component depending on clouds, SWdn_cld, with SWdn = SWdn_clr + SWdn_cld. Given this decomposition,
i1520-0442-22-17-4621-e2

The first term on the right-hand side (rhs) of this equation corresponds to the part of the total change in solar radiation absorption resulting from changes in the incoming solar radiation at the surface, because of shortwave absorbing-gas (or aerosol) concentration changes. The aerosol concentration is fixed in the model, as is ozone. On the other hand, water vapor, whose concentration is expected to increase with temperature, because its saturation value depends nearly exponentially on temperature according to the Clausius–Clapeyron relationship, a priori changes between the two experiments. The second term corresponds to the part of total change related to changes in cloud cover and cloud radiative properties. Finally, the last term is the part of the total change in solar radiation absorption resulting from surface albedo changes.

If we consider that the emissivity of the surfaces is prescribed to one (blackbody in their emission range), all of the incoming longwave radiation is absorbed by the surfaces and their emissions correspond to the Stephan–Bolztmann law for black bodies: LWup = −σTsurf4 (negative because the fluxes positive downward), where σ refers to the Stephan’s constant and Tsurf to the temperature of the surface. Emissivities are close to one in reality, Peixoto and Oort (1992, p. 105) indicate values of 0.96 for oceanic surfaces, 0.92 for land, and 0.98 for vegetation. In IPSL-CM4, the emissivity is prescribed to one for all types of surfaces. Anomalies for the upward longwave emission are then approximately equal to .

We decompose the longwave incoming radiation into a clear-sky and a cloud-related component: LWdn = LWdn_clr + LWdn_cld. We further estimate LWdn_clr in the form of LWdn_clr = −LWup_clr(a + b × shum), where shum corresponds to the mean specific humidity in the atmospheric column; a and b are the parameters of a linear regression fit calculated at each grid cell for each experiment between the ratio of annual-mean values of LWdn_clr and LWup_clr, and of shum. This decomposition expresses the dependence of the backward longwave radiation to the magnitude of the upward longwave flux and to the amount of GHG, which absorbs part of it (the fixed ones like CO2, CH4, and N2O, and water vapor, which can fluctuate in the model). This decomposition is valid everywhere (95% significant test) except in localized regions that are ice-covered during some years and ice-free during others, situations for which the regression parameters change significantly (cf. masked areas in Figs. 3e–g). For differences we then get
i1520-0442-22-17-4621-e3

The first term on the rhs represents the direct feedback between the upward (hence surface temperature) and downward longwave radiations, with more emission by the surface implying more radiation being absorbed and reemitted back by the atmosphere, further contributing to a warming the surface. The second term quantifies the changes in LWdn_clr resulting from the changes in the humidity content of the atmosphere, with the other parameters kept constant. Finally, the last term represents the effect of different fixed GHG concentrations on the relationship between LWdn_clr and LWup_clr for the same values of shum and LWup_clr.

Given the above decompositions of the radiative fluxes, Eq. (1) yields, for differences Δ related to a given forcing, and expressing changes in longwave radiation to the changes of the other energy fluxes,
i1520-0442-22-17-4621-e4

In the following analyses, the constant term in a given expression [e.g., SWdn in (Δα)SWdn] is calculated as the mean values for two simulations considered.

b. Surface flux changes

Figure 3 shows the changes in surface energy fluxes for the modern boundary conditions. At first order, the patterns and relative magnitudes of the different terms are very similar under glacial and interglacial boundary conditions (not shown); therefore, we present the results for only one climatic context (the modern one; the differences for the two climatic contexts will be the focus of section 4d). The fluxes are presented in the same order as in Eq. (4). Their global-mean relative contributions to the total longwave emission change are shown in Fig. 4a and for the high latitudes (poleward of 60° for both hemispheres) in Fig. 4b. The changes for each flux are normalized by the change in longwave surface emission at each grid cell, and then averaged spatially for 100% land grid cells (black bars) and 100% oceanic grid cells (white bars), and for all surface types (gray bars).

The term , which represents the changes in the surface longwave emission, is directly related to surface temperature changes. We therefore use this term to interpret the surface temperature changes presented in Fig. 1. It has the advantage of being directly comparable to the other components of the surface energy balance [Eq. (4)]. The patterns of surface temperature changes (Fig. 1) and the longwave emission changes (Fig. 3a) are identical, except for differences related to the cubic temperature, which only implies small relative differences in magnitude, varying mainly with latitude.

The changes in solar radiation reaching the surface under clear-sky conditions (Fig. 3b) are small compared to the other terms, with maximum absolute values of 5 W m−2. The changes are significantly negative everywhere, contributing to a significant global-mean cooling (about −20% of the longwave emission change; see Fig. 4a). The cooling effect is consistent with a greater amount of water vapor under warmer conditions, especially over the oceans, implying a greater absorption of the incoming solar radiation when crossing the atmosphere.

The changes in solar radiation reaching the surface due to changes in cloud properties are shown in Fig. 3c. The difference is significantly positive in most parts of the low and midlatitudes (except in the western Pacific), reaching 20 W m−2, and significantly negative in the high latitudes (and in the western Pacific), reaching −20 W m−2. The changes are more pronounced over the oceans than over the continents. These changes are consistent with the difference in total cloud cover changes shown in Fig. 5a. A reduction (increase) in the cloud coverage implies a greater (smaller) proportion of solar radiation reaching the surface. The cloud cover changes are qualitatively similar to the results found by Ramstein et al. (1998) in the case of a doubling of CO2.

Solar flux absorption changes resulting from albedo changes (Fig. 3d) are strong in the mid- and high latitudes and correspond to regions of sea ice and snow cover retreat (not shown). The oceanic regions concerned with the strongest changes of this flux (up to 50 W m−2) are the ones experiencing the greatest warming globally (Fig. 1b), that is, the Greenland–Iceland–Norway (GIN) and Labrador Seas, Hudson Bay, the Atlantic and Indian sectors around Antarctica, and the whole Arctic Ocean to a lesser extent. Over the continents, the changes are usually weaker than over the oceans. The greatest changes are found in the middle to high northern latitudes over North America, Europe, and Asia, with differences reaching 20 W m−2. On the global mean, the effect of albedo changes is not very large (about 8%, Fig. 4a), but it accounts for about 50% of the longwave emission change in the high latitudes over the oceans (Fig. 4b).

Figure 3e shows the changes in −(a + b × shum)(ΔLWup_clr). It quantifies the direct feedback that relates warmer surface temperatures, hence, longwave surface emissions to greater backward radiation. The pattern is very close to the pattern of surface warming and the values are strong compared to other terms for all regions. On the global mean, its contribution to the warming is the largest of all fluxes (with values about 73% of the longwave surface emission change; see Fig. 4a).

The effect of water vapor changes on the backward longwave radiation (Figs. 3f and 4a) is also relatively strong (maximum values around 20 W m−2 locally, global-mean contribution about 65%) and positive everywhere. Over the oceans, the contribution is even greater (about 73%) and the pattern is similar to the one for surface temperature changes, which is consistent with the Clausius–Clapeyron relationship, which states that the humidity the air can hold before saturation increases nearly exponentially with temperature. It is a robust response of global warming (e.g., Held and Soden 2006). Over the continents, the match between these two patterns is less obvious because humidity supply can be limited by the dryness of the soils. The effect on land is the greatest around the equator, where humidity availability is the strongest. Its contribution to the global warming on land is also important (second contributor after the direct longwave feedback; see Fig. 4a), although it is weaker than over oceans.

The changes of the coefficients in the relationship between LWdnclr, LWupclr, and shum imply only relatively weak radiative effects, ranging between −5 and 7 W m−2 (Fig. 3g). The values are usually positive for most regions and are strongest in the tropics, consistent with an effect of fixed GHG concentration increases on longwave radiation proportional to the surface temperature. Nevertheless, negative values are also found in some regions, which is inconsistent with the expected physics. It usually corresponds to humid regions (equatorial forests, the western Gulf Stream, and Kuroshio regions) where shum changes are the strongest (Fig. 3f). The error in the decomposition resulting from nonlinear effects in the relationship is likely to be the strongest in these regions.

The change in longwave radiation reaching the surface resulting from cloud changes is shown in Fig. 3h. It globally represents a cooling effect (Fig. 4a), which is particularly important over the oceans. The patterns and values are almost identical to the ones found for cloud effects on solar radiation (Figs. 3c and 4a), but with an opposite sign. The response is also consistent with the differences in total cloud cover change shown in Fig. 5a. There is less backward longwave radiation for regions where cloud cover change is reduced, and conversely for regions of increased cloud cover. The total effect of cloud changes on both solar and longwave radiation is shown in Fig. 5b and indicates that the dominant effect depends on the region considered. In the low to midlatitudes, the effect of cloud cover change on shortwave radiation is usually dominant, whereas the effect on longwave radiation is usually greater than for solar radiation in the mid- to high latitudes. Note that although the link between the total cloud cover changes and the effect on solar and longwave fluxes separately is quite direct, the net effect is less directly related to the cloud cover response, at least in this model.

Changes in latent heat fluxes (Fig. 3i) correspond to a cooling in most part of the low and midlatitudes over the oceans, with values reaching about −20 W m−2. It is consistent with a greater evaporation over the oceans resulting from increased radiative fluxes reaching the surface and greater surface temperatures. In the Southern Ocean, significantly positive values are found locally. In the regions of sea ice retreat, evaporation is strongly enhanced, corresponding to negative flux changes. Over the continents, the latent heat flux differences are usually smaller than over the oceans and strongly depend on the regions. The changes in surface latent fluxes on land are consistent with the precipitation differences (Fig. 5c), with more (less) humid regions corresponding to a greater (weaker) evaporation. Note that the match between evaporation and precipitation changes can be connected in both ways through local water recycling. The latent changes represent the main cooling contributor globally (about −30% of the longwave surface emission change; see Fig. 4a), especially over the oceans (−45%).

Changes in sensible heat fluxes (Fig. 3j) are usually positive over the low- and middle-latitude oceans and negative over most continents. It corresponds to a situation where the extra surface heat is transferred from the continents to the oceans. It is consistent with land surfaces warming more than oceanic ones. Over the oceans, the regions where fluxes are the strongest (up to around 15–20 W m−2) are the Southern Ocean and the western sides of the Northern Hemisphere gyres. Over sea ice retreat regions, the difference of surface type results in a greater exchange of heat from the ocean to the atmosphere. Over land, regions of the strongest fluxes (Amazonia, Sahel, and Australia, with values up to −15 W m−2) are usually the ones that experience drier conditions. Drier surfaces lose less energy through evaporation and hence warm more and lose more energy through sensible heat fluxes.

The changes resulting from heat advection differences are shown in Fig. 3k. The pattern is composed of patchy regions of positive and negative values. This term can reach high absolute values locally (over 20 W m−2), especially in the GIN seas, where greater oceanic convection occurs (not shown), and in the Southern Ocean. On the global mean nonetheless, its contribution is opposite to the surface temperature change (Fig. 4a).

Figure 3l represents changes in surface heat content variations (more release or less storage of energy when positive). This term takes into account the trends and multidecadal variations of the different simulations and shows that for most regions, the simulations do not exhibit significant nonequilibration differences. Nevertheless, strong positive values (over 20 W m−2) are found locally in the Kuroshio extension and in the northern part of the GIN seas. The former case is related to changes in the variations of the depth of the mixed layer, whereas the latter is related to a variation in sea ice cover in the fully preindustrial run (not shown). It has little impact on the global mean (Fig. 4a).

The various decompositions of the shortwave and longwave terms in our analysis could result in errors. The residue of the different terms after the decompositions are performed (Fig. 3m) indicates that the errors are in fact very small and do not exhibit coherent patterns.

c. Land–sea warming ratio

In the previous sections, we have shown that the thermal response to the GHG increase implies many surface energy fluxes for both land and oceanic surfaces. The magnitude of the feedbacks is generally weaker over the continents than over the oceans, especially the cooling ones like the latent and cloud-related backward longwave radiation (Fig. 3). Several energy flux terms have a different response in sign or magnitude over the continents and over the oceans. Therefore, the land–sea warming contrast results from a mixed contribution of the different surface energy fluxes.

The land–sea warming ratio shown in Fig. 2c is close to the land–sea ratio of surface longwave emission changes (not shown). By decomposing Δ(−LWup) into the different flux changes (noted ΔFi) according to Eq. (4) for land and oceanic surfaces, we can get
i1520-0442-22-17-4621-e5
where Δland and Δocean represent differences over 100% land and oceanic surface fractions, respectively, and ri the land–sea ratio of flux i. Here, (ri − 1)ΔoceanFiocean(−LWup) represents the contribution of flux i to the departure of the total land–sea ratio from 1. Figure 6 shows the global-mean contribution of each flux in the decomposition of Δ(−LWup). It has been normalized by the total land–sea ratio at each latitude before being averaged and multiplied by 100. It shows that many fluxes indeed contribute to the exact value of the land–sea warming ratio.

The main contributor is the latent heat flux as expected from Sutton et al. (2007). We quantify its role at about 100% of the total land–sea surface longwave emission ratio. Note that the latent effect is not effective to determine the ratio in the high latitudes (not shown). The direct longwave feedback effect obviously contributes positively to the ratio, with a relative magnitude of about 65%. Albedo and advection differences, and cloud changes on longwave radiation have a moderate relative impact ranging between 20% and 40%, whereas GHG changes on shortwave radiation have a smaller role (about 5%).

The main mechanism contributing to decrease the ratio is due to sensible heat flux changes (with a magnitude ranging from −55% to −120%, depending on the boundary conditions). Its influence occurs at all latitudes. Fixed GHG increase effects on longwave radiation tend to decrease the ratio (from about −20% to −30%), which is consistent with globally warmer oceans than continents on an annual average, implying a greater proportion of longwave emission to absorb and reemit toward the surface. Our simulations, not fully equilibrated, tend to decrease the land–sea warming ratio globally (from about −10% to −30%), whereas negative values related to the error terms (Fig. 6, right-most component) indicates that our decomposition of all fluxes only slightly overestimate the total ratio by about 5%. Cloud changes on shortwave radiation and the water vapor effect on longwave radiation do not have a clear coherent impact for both interglacial and glacial boundary conditions.

Comparing the relative role of the different fluxes for both boundary conditions might help us distinguish the robust contributors to the land–sea warming ratio observed when increasing the GHG concentration in the atmosphere. The robust contributing effects are latent heat and the direct longwave feedback effects first, and then albedo, cloud changes on longwave radiation, and advection effects. Sensible heat transfer with the atmosphere is the principal moderating factor. Land–sea warming ratios superior to one and close to each other under both boundary conditions can be explained by similar changes in the fluxes determining the temperature ratio (especially latent fluxes, direct longwave feedback, advection), nevertheless, significant differences between the contribution of several fluxes (e.g., sensible heat fluxes, cloud changes on shortwave radiation) suggest that the almost exact same figure found for the two global-mean ratios (Fig. 2c) should not be regarded as a robust feature.

d. Glacial–interglacial thermal differences

Figure 7a shows the surface flux contributions to the surface longwave emission differences for all surface types (gray bars), land surfaces (black bars), and oceanic surfaces (white bars). On land, the globally warmer conditions under modern boundary conditions are mostly related to the differences in the direct longwave feedback (75%), in albedo differences (38%), and in a greater concentration of water vapor in the atmosphere (30%). The cooling factors on land are mostly evaporation (−28%) and sensible heat fluxes (−27%). Over the oceans, the main contributor to the different warmings under interglacial and glacial conditions is related to advection processes (about 390%), especially in the midlatitudes (not shown). These warming or cooling differences resulting from advection changes are partly balanced by latent (about −280%) and sensible heat fluxes (about −100%). The warmer (the colder) the surface ocean, the greater (weaker) the evaporation and sensible heat fluxes are, lowering the advection effect. The other temperature change differences over the oceans are related mostly to differences in atmospheric water vapor (100%) and to the direct longwave feedback effects (75%).

Considering the global-mean contributions of the different fluxes for the GHG increase (Fig. 5a) and for the differences between modern and glacial conditions (Fig. 7a), we can consider that the retroactions and feedbacks are almost simply amplified for land surfaces (black bars) in the former conditions. Indeed, the same fluxes are involved in a relatively similar proportion. Nonetheless, the role of cloud-related changes is not directly amplified and albedo changes are stronger under modern boundary conditions, especially in the mid-northern latitudes (not shown), consistent with larger snow retreats under modern conditions in regions covered by ice sheets in the LGM context. Indeed, the ice sheets are imposed to the model and cannot melt, hence preventing a strong albedo change in these regions. In the high latitudes on land, the differences resulting from the different boundary conditions are different from an amplification of the fluxes concerned in the GHG concentration change (cf. Fig. 7b and 4b). The differences result mainly from changes in the cloud properties change (apart from the direct longwave feedback), rather than changes in water vapor concentration. It shows that the cloud changes and their effects on shortwave or longwave radiation are particularly sensitive to the base state of the climate.

Over the oceans, the differences resulting from the different boundary conditions correspond partly to an amplification of the retroactions acting for the GHG increase (cloud changes on shortwave and longwave radiation, the direct longwave feedback effect, and the impact of water vapor; see Figs. 7a and 4a), but along with a major influence of the oceanic advection differences, although they are partly balanced by latent and sensible heat fluxes. Therefore, advection changes seem also to depend strongly on the base state. In the high latitudes, the roles of advection, heat storage, sensible heat flux, cloud influence on longwave radiation, and albedo changes are increased under modern boundary conditions over the oceans (Figs. 7b and 4b). Part of the greater albedo changes is due to the greater oceanic area that can experience sea ice retreat resulting from a higher sea level under modern conditions (not shown).

We can also wonder if the greater temperature of the base state under the modern boundary is responsible for the greater temperature changes. Indeed, the primary effect of increased GHG concentration is to absorb (and reemit) a greater proportion of the upward surface longwave radiation. If this latter condition is greater in magnitude, the value of the extra absorption will be accordingly greater. To test this idea, we normalize the flux changes by the longwave surface emission of the base state at each grid point, and then average them. The normalized fluxes are still greater in magnitude in the case of modern boundary conditions (not shown). It indicates that the changes are partly proportional to the surface longwave emission of the base state, but with a coefficient greater than one due to amplification effects. In fact, the relationship between the changes in the backward longwave radiation and longwave surface emission is more accurately described by Eq. (3) than by a direct proportionality relationship, which shows that other factors relate these fluxes, including water vapor concentration changes.

5. Conclusions

We have performed simulations of the LGM and of the preindustrial climates with GHG concentrations and other boundary conditions (ice sheets, sea level, and orbital parameters) modified separately in IPSL-CM4. The GHG concentration forcing contributes to about 59% to the total LGM forcing (with respect to the sum of the absolute values of the contributions) in terms of global temperature change, with the other boundary conditions (mostly ice sheets) contributing about 29%, whereas the interaction between the two forcing factors is quantified to 12% in this model.

We have studied more specifically the influence of an increase in GHG concentration from LGM to preindustrial values under boundary conditions representing the LGM (glacial) and modern (interglacial) contexts. In both cases, an increase in GHG concentration implies a global warming, more pronounced in the high northern latitudes and over the continents compared to the oceans. The glacial and interglacial climates react qualitatively in the same way. The land–sea warming ratio is similar under both climatic contexts. We decomposed the energy fluxes at the surface in order to isolate and quantify the role of different processes.

The two main contributors to the global warming are the direct longwave feedback effect (73% of the changes in surface longwave emission) and the effect of water vapor as a GHG (63%). This latter effect is as large as the former over oceans, and about half of it on land. The effect of cloud and albedo changes on shortwave radiation, of fixed GHG changes on longwave radiation, and sensible fluxes contribute globally to a warming of the surface (between 5% and 20%), although the latter has an opposite effect on land and over the oceans. The factors globally opposing the temperature change are latent heat fluxes (−30% of the changes in surface longwave emission), whose effect is much larger over the oceans than on land, cloud changes on longwave radiation (−24%), GHG changes on shortwave radiation (−22%), and advection (−14%). The relative contributions to the surface warming are different in the high latitudes, especially over the oceans. The surface type changes from snow or sea ice cover to browner ground or open water imply albedo changes that impact the solar absorption by the surface, especially over the oceans. Other fluxes are also largely modified accordingly to the new surface type and also contribute to the final energy budget.

Several fluxes are involved in the different temperature responses over the continents and the oceans, and therefore contribute to the value of the land–sea warming ratio. The dominant flux responsible for the temperature ratio is the latent flux, as suggested by Sutton et al. (2007). The value of its contribution is about 100% of the total ratio in our GHG increase experiments. The direct longwave feedback effect contributes to the ratio at about 65%. Albedo, advection, and cloud changes on longwave radiation also seem robust positive contributors to the land–sea warming ratio. Sensible heat flux changes and fixed GHG increase effects on longwave radiation tend to decrease the ratio. The fact that the values of the land–sea warming ratios are close under modern and glacial boundary conditions can be explained by the common responses to a GHG increase of the main fluxes involved in determining the ratio. Nevertheless, the almost exact same figures obtained under both boundary conditions do not seem a robust feature since our decomposition shows that the contributions of several fluxes differ significantly.

The difference between the responses under glacial and interglacial boundary conditions is only partly due to greater longwave emissions in the modern base state. Albedo changes are greater under interglacial conditions in the mid-northern latitudes because of the different ice sheets imposed to the model over land and partly because there is a greater oceanic area that can experience sea ice retreat resulting from a higher sea level concerning the oceans. Apart from these direct effects of the boundary conditions on albedo, the cloud-related change differences contribute importantly to a greater warming under modern boundary conditions in the high latitudes both on land and over the oceans. The heat oceanic advection change differences are also importantly involved in the temperature change contrast between the two climatic conditions, although they are partly balanced by latent and sensible heat flux differences. Cloud and advection changes are therefore sensitive to the base state. Water vapor and direct longwave emission feedbacks are also involved in the globally stronger temperature change simulated under modern boundary conditions.

The different boundary conditions used suggest that the ice sheets tend to decrease the sensitivity of the climate to a GHG forcing. The ice sheets could therefore play a role in slowing down or accelerating a response to a radiative forcing by growing or melting on millennial time scales.

Our decomposition of the fluxes involved in the surface energy budget change under increased GHG forcing allows a clear quantification of the different processes involved. It would hereafter be interesting to consider the responses for different seasons and also to use this decomposition to compare the sensitivity of different models to a given radiative forcing and eventually compare them to reanalyses.

Acknowledgments

This research has been supported by the MOTIF European EVK2-CT-2002-00153 and the IDEGLACE ANR project JC05-4311.

REFERENCES

  • Alkama, M. R., , M. Kageyama, , G. Ramstein, , O. Marti, , P. Ribstein, , and D. Swingedouw, 2008: Impact of a realistic river routing in coupled ocean-atmosphere simulations of the Last Glacial Maximum climate. Climate Dyn., 30 , 855869.

    • Search Google Scholar
    • Export Citation
  • Braconnot, P., and Coauthors, 2007a: Results of PMIP2 coupled simulations of the Mid-Holocene and Last Glacial Maximum—Part 1: Experiments and large-scale features. Climate Past, 3 , 261277.

    • Search Google Scholar
    • Export Citation
  • Braconnot, P., and Coauthors, 2007b: Results of PMIP2 coupled simulations of the Mid-Holocene and Last Glacial Maximum—Part 2: Feedbacks with emphasis on the location of the ITCZ and mid- and high latitudes heat budget. Climate Past, 3 , 261277.

    • Search Google Scholar
    • Export Citation
  • Cai, M., 2005: Dynamical amplification of polar warming. Geophys. Res. Lett., 32 , L22710. doi:10.1029/2005GL024481.

  • Crucifix, M., 2006: Does the Last Glacial Maximum constrain climate sensitivity? Geophys. Res. Lett., 33 , L18701. doi:10.1029/2006GL027137.

    • Search Google Scholar
    • Export Citation
  • Hall, A., 2004: The role of surface albedo feedback in climate. J. Climate, 17 , 15501568.

  • Held, I. M., , and B. J. Soden, 2006: Robust responses of the hydrological cycle to global warming. J. Climate, 19 , 56865699.

  • Holland, M. M., , and C. M. Bitz, 2003: Polar amplification of climate change in coupled models. Climate Dyn., 21 , 221232.

  • Hourdin, F., and Coauthors, 2006: The LMDZ4 general circulation model: Climate performance and sensitivity to parameterized physics with emphasis on tropical convection. Climate Dyn., 27 , 787813.

    • Search Google Scholar
    • Export Citation
  • Kim, S., 2004: The effect of atmospheric CO2 and ice sheet topography on LGM climate. Climate Dyn., 22 , 639651.

  • Marti, O., and Coauthors, 2006: The new IPSL climate system model: IPSL-CM4. IPSL Note du Pôle de Modélisation 26, 332 pp. [Available online at http://dods.ipsl.jussieu.fr/omamce/IPSLCM4/DocIPSLCM4/HTML/].

    • Search Google Scholar
    • Export Citation
  • Masson-Delmotte, V., and Coauthors, 2005: Past and future polar amplification of climate change: Climate model intercomparisons and ice-core constraints. Climate Dyn., 26 , 513529. doi:10.1007/s00382-005-0081-9.

    • Search Google Scholar
    • Export Citation
  • Otto-Bliesner, B. L., , E. C. Brady, , G. Clauzet, , R. Tomas, , S. Levis, , and Z. Kothavala, 2006: Last Glacial Maximum and Holocene climate in CCSM3. J. Climate, 19 , 25262544.

    • Search Google Scholar
    • Export Citation
  • Peixoto, J., , and A. H. Oort, 1992: Physics of Climate. American Institute of Physics, 520 pp.

  • Peltier, W. R., 2004: Global glacial isostasy and the surface of the ice-age Earth: The ICE-5G (VM2) model and GRACE. Annu. Rev. Earth Planet. Sci., 32 , 111149. doi:10.1146/annurev.earth.32.082503.144359.

    • Search Google Scholar
    • Export Citation
  • Ramstein, G., , Y. Serafini-Le Treut, , H. Le Treut, , M. Forichon, , and S. Joussaume, 1998: Cloud processes associated with past and future climate changes. Climate Dyn., 14 , 233247.

    • Search Google Scholar
    • Export Citation
  • Shin, S., , Z. Liu, , B. Otto-Bliesner, , E. Brady, , J. Kutzbach, , and S. Harrison, 2003: A simulation of the Last Glacial Maximum climate using the NCAR-CCSM. Climate Dyn., 20 , 127151.

    • Search Google Scholar
    • Export Citation
  • Solomon, S., , D. Qin, , M. Manning, , Z. Chen, , M. Marquis, , K. B. Averyt, , M. Tignor, , and H. L. Miller, Eds. 2007: Climate Change 2007: The Physical Science Basis. Cambridge University Press, 996 pp.

    • Search Google Scholar
    • Export Citation
  • Sutton, R. T., , B. Dong, , and J. M. Gregory, 2007: Land/sea warming ratio in response to climate change: IPCC AR4 model results and comparison with observations. Geophys. Res. Lett., 34 , L02701. doi:10.1029/2006GL028164.

    • Search Google Scholar
    • Export Citation
  • Swingedouw, D., , P. Braconnot, , P. Delecluse, , E. Guilyardi, , and O. Marti, 2007: Quantifying the AMOC feedbacks during a 2 × CO2 stabilization experiment with land-ice melting. Climate Dyn., 29 , 521534.

    • Search Google Scholar
    • Export Citation
  • Weaver, A. J., , M. Eby, , A. F. Fanning, , and E. C. Wiebe, 1998: Simulated influence of carbon dioxide, orbital forcing and ice sheets on the climate of the Last Glacial Maximum. Nature, 394 , 847853.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Annual-mean surface temperature change for an increase in GHG concentration from LGM to preindustrial values, for (a) LGM (“glacial”) and (b) modern (“interglacial”) boundary conditions, and (c) the difference between the temperature change under interglacial and glacial boundary conditions. Blue contours represent the ice sheets imposed in the model. Units: °C. Blank areas are regions for which the null hypothesis cannot be rejected at 95%.

Citation: Journal of Climate 22, 17; 10.1175/2009JCLI2771.1

Fig. 2.
Fig. 2.

Zonal- and annual-mean (a) surface temperature change (°C), (b) surface temperature change divided by the global-mean temperature change, and (c) the ratio between the temperature change over land and ocean (for grid cells representing 100% of the surface type) for modern (black) and LGM (gray) boundary conditions. Horizontal lines represent global averages.

Citation: Journal of Climate 22, 17; 10.1175/2009JCLI2771.1

Fig. 3.
Fig. 3.

Annual-mean surface energy flux [see Eq. (4)] differences (GHGpreind − GHGlgm) (W m−2) for modern boundary conditions: (a) longwave surface emission changes, (b) solar absorption changes due to GHG concentration changes, (c) solar absorption changes due to cloud changes, (d) solar absorption changes due to albedo changes, (e) direct longwave feedback differences, (f) effect of water vapor concentration changes on longwave incoming radiation, (g) effect of fixed GHG concentration changes on longwave incoming radiation, (h) effect of cloud changes on longwave incident radiation, (i) changes in latent heat fluxes, (j) changes in sensible heat flux, (k) heat advection changes, (l) heat storage differences, and (m) errors resulting from the decomposition. Positive fluxes point downward, that is, correspond to a warming of the surface. Blank areas are regions for which the null hypothesis of the results cannot be rejected at the 95% level.

Citation: Journal of Climate 22, 17; 10.1175/2009JCLI2771.1

Fig. 4.
Fig. 4.

Mean contributions of the different surface fluxes from Eq. (4) to the changes in longwave surface emission for modern boundary conditions, for all types of surfaces (gray bars), continental surfaces only (grid cells with 100% land, black bars), and for oceanic surfaces (100% oceanic grid cells, white bars). Fluxes are normalized by the changes in longwave emission at each grid cell (a) before being averaged globally and (b) for high latitudes (superior to 60° in both hemispheres).

Citation: Journal of Climate 22, 17; 10.1175/2009JCLI2771.1

Fig. 5.
Fig. 5.

Annual-mean difference for modern boundary conditions in (a) cloud cover change (in percentage of the sky covered by clouds), (b) total effect of cloud changes on solar and longwave radiation for modern boundary conditions (W m−2), and (c) precipitation changes (mm day−1). Blank areas are regions for which the null hypothesis cannot be rejected at 95%.

Citation: Journal of Climate 22, 17; 10.1175/2009JCLI2771.1

Fig. 6.
Fig. 6.

Global-mean contributions of the different surface fluxes from Eq. (4) to the land–sea longwave emission change ratio [see Eq. (5)] for interglacial boundary conditions (white bars) and glacial boundary conditions (black bars). Ratios are calculated over 100% land and oceanic grid cells. Contributions are calculated for each latitude before being averaged.

Citation: Journal of Climate 22, 17; 10.1175/2009JCLI2771.1

Fig. 7.
Fig. 7.

Same as in Fig. 4, but for the differences between the changes under interglacial and glacial boundary conditions.

Citation: Journal of Climate 22, 17; 10.1175/2009JCLI2771.1

Table 1.

Global-mean surface temperature (°C) for preindustrial and LGM GHG concentrations (first and second column, respectively) and the difference between the two (last column); and for other boundary conditions (ice sheets, orbital parameters, and sea level) from the preindustrial and LGM climate (first and second rows, respectively) and the difference between the two (last row).

Table 1.
Save