past and future polar ampliﬁcation of climate change ...changes have been used to estimate...

ORIGINAL ARTICLE

V. Masson-Delmotte Æ M. Kageyama Æ P. BraconnotS. Charbit Æ G. Krinner Æ C. Ritz Æ E. Guilyardi

J. Jouzel Æ A. Abe-Ouchi Æ M. Crucifix

R. M. Gladstone Æ C. D. Hewitt Æ A. KitohA. N. LeGrande Æ O. Marti Æ U. Merkel Æ T. Motoi

R. Ohgaito Æ B. Otto-Bliesner Æ W. R. Peltier

I. Ross Æ P. J. Valdes Æ G. Vettoretti Æ S. L. Weber

F. Wolk Æ Y. YU

Past and future polar amplification of climate change: climate modelintercomparisons and ice-core constraints

Received: 31 May 2005 / Accepted: 22 September 2005 / Published online: 20 December 2005� Springer-Verlag 2005

Abstract Climate model simulations available from thePMIP1, PMIP2 and CMIP (IPCC-AR4) intercompari-son projects for past and future climate change simula-tions are examined in terms of polar temperaturechanges in comparison to global temperature changesand with respect to pre-industrial reference simulations.For the mid-Holocene (MH, 6,000 years ago), themodels are forced by changes in the Earth’s orbital

parameters. The MH PMIP1 atmosphere-only simula-tions conducted with sea surface temperatures fixed tomodern conditions show no MH consistent response forthe poles, whereas the new PMIP2 coupled atmosphere–ocean climate models systematically simulate a signifi-cant MH warming both for Greenland (but smaller thanice-core based estimates) and Antarctica (consistent withthe range of ice-core based range). In both PMIP1 and

V. Masson-Delmotte (&) Æ M. Kageyama Æ P. BraconnotS. Charbit Æ E. Guilyardi Æ J. Jouzel Æ O. MartiLaboratoire des Sciences du Climat et de l’Environnement,(LSCE/IPSL, UMR CEA-CNRS 1572) L’Orme des Merisiers,Batiment 701, CEA Saclay, 91 191 Gif-sur-Yvette Cedex, FranceE-mail: [email protected]

G. Krinner Æ C. RitzLaboratoire de Glaciologie et de Geophysique de l’Environnement,(UMR 5183 CNRS-UJF), Domaine Universitaire, St Martind’Heres, France

A. Abe-OuchiCenter for Climate System Research, The University of Tokyo,Kashiwa, 277-8568 Japan

M. Crucifix Æ C. D. HewittHadley Centre for Climate Prediction and Research, Met Office,FitzRoy Road, Exeter, EX1 3 PB Devon, UK

R. M. Gladstone Æ I. Ross Æ P. J. ValdesSchool of Geographical Sciences,University of Bristol,University Road, Bristol, BS8 1SS UK

A. Kitoh Æ T. MotoiClimate Research Department,Meteorological Research Institute,1-1 Nagamine, Tsukuba,Ibaraki, 305-0052 Japan

A. N. LeGrandeNASA Goddard Institute for Space Studies andCenter for Climate Systems Research, Columbia University,New York, NY, USA

U. MerkelIFM-GEOMAR,Duesternbrooker Weg 20,24105 Kiel, Germany

R. Ohgaito Æ A. Abe-OuchiFrontier Research Center for Global Change (FRCGC),JAMSTEC, Yokohama City, 236-0001 Japan

B. Otto-BilesnerClimate Change Research,National Center for Atmospheric Research,1850 Table Mesa Drive, P.O. Box 3000,Boulder, CO, 80307 USA

W. R. Peltier Æ G. VettorettiDepartment of Physics, University of Toronto,60 St. George Street, Toronto,ON, M5S 1A7 Canada

S. L. WeberClimate Variability Research,Royal Netherlands Meteorological Institute (KNMI),P.O. Box 201, 3730 AE De Bilt, The Netherlands

F. WolkInstitut d’Astronomie et de Geophysique G. Lemaıtre,Universite catholique de Louvain, Chemin du cyclotron, 2,1348 Louvain-la-Neuve, Belgium

Y. YuLASG, Institute of Atmospheric Physics,Chinese Academy of Sciences,P.O. Box 9804, Beijing, 10029People’s Republic of China

Climate Dynamics (2006) 26: 513–529DOI 10.1007/s00382-005-0081-9

PMIP2, the MH annual mean changes in global tem-perature are negligible, consistent with the MH orbitalforcing. The simulated last glacial maximum (LGM,21,000 years ago) to pre-industrial change in globalmean temperature ranges between 3 and 7�C in PMIP1and PMIP2 model runs, similar to the range of tem-perature change expected from a quadrupling of atmo-spheric CO2 concentrations in the CMIP simulations.Both LGM and future climate simulations are associ-ated with a polar amplification of climate change. Therange of glacial polar amplification in Greenland isstrongly dependent on the ice sheet elevation changesprescribed to the climate models. All PMIP2 simulationssystematically underestimate the reconstructed glacial–interglacial Greenland temperature change, while someof the simulations do capture the reconstructed glacial–interglacial Antarctic temperature change. Uncertaintiesin the prescribed central ice cap elevation cannot ac-count for the temperature change underestimation byclimate models. The variety of climate model sensitivitiesenables the exploration of the relative changes in polartemperature with respect to changes in global tempera-tures. Simulated changes of polar temperatures arestrongly related to changes in simulated global temper-atures for both future and LGM climates, confirmingthat ice-core-based reconstructions provide quantitativeinsights on global climate changes.

1 Introduction

The Arctic (Corell et al. 2004) and the AntarcticPeninsula (Jacka and Budd 1998; Vaughan et al. 2003)are experiencing some of the most rapid temperaturerises on Earth. Since the 1960s, temperatures have risenby more than 2�C in some areas, several times morerapidly than the average global temperature (IPCC2001; Moritz et al. 2002). Due to possible changes insnow and ice cover and associated albedo feedbacks,polar regions are likely to act as amplifiers of climatechange, however modulated by local processes generat-ing multi-decadal variability (Bengtsson et al. 2004;Polyakov et al. 2002). Future climate change simulationsin the Arctic indeed show a polar amplification with 1.5–4.5 times the global warming (Holland and Bitz 2003).These contrasted results apparently result from threemain processes in the climate models: different changesin cloud cover (and cloud radiative feedbacks); differentchanges in heat advection (Alexeev et al. 2005); and,most important, different response in sea–ice. The levelof future polar amplification was indeed shown to bedependent on the modern sea–ice thickness and snowcover simulations (Dixon et al. 2003; Holland and Bitz2003). Early studies have revealed that models with coldmodern biases, when forced with increased greenhousegas concentrations, had a larger sensitivity due to alarger ice feedback (Dixon et al. 2003; Rind et al. 1997).Reducing the uncertainties of future climate change

predictions is extremely important for high latitudeareas because (1) local eco-systems will be extremelyvulnerable to large warmings because of competitionwith living organisms from lower latitudes more adaptedto warm conditions, and impossibility of migration tocolder places (Corell and coauthors 2004); (2) ice sheetsare susceptible to respond to a fast warming (Rignot andThomas 2002), either by increased accumulation (Hu-ybrechts et al. 2004) or coastal thinning (Krabill et al.2004) and modify the freshwater supply to the highlatitude oceans; (3) the warming of the Arctic may in-crease the release of carbon from soils and act as a po-sitive feedback on climate change (Knorr et al. 2005).

In order to assess the capacity of state-of-the-art cli-mate models to simulate realistically the high latitudeclimate changes, past climate simulations are comparedhere with ice-core estimates for last glacial maximum(21,000 years ago, hereafter LGM) and mid-Holocene(6,000 years ago, hereafter MH) temperature changes.These simulations enable the test of the model responsesto a variety of forcings. MH boundary conditions includepre-industrial greenhouse gas levels together with modi-fied orbital parameters (Berger 1978) which induce smallchanges in low to high latitude annual mean insolation(effect of obliquity change), but with a seasonal cycleincreased (respectively reduced by locally �5%) in thenorthern (respectively southern) hemisphere (in responseto precession changes). In this respect, polar MH simu-lations provide a test of climate model response mainly toa modified seasonal cycle forcing (Renssen et al. 2005a,b). LGM boundary conditions include reduced atmo-spheric greenhouse gas levels (CO2, CH4 and N2O),modified surface boundary conditions including thepresence of massive ice sheets over northern NorthAmerica and Fennoscandia (Peltier 1994, 2004) andsmall changes in orbital parameters (Table 1).

Climate models offer unique frames to understandthe spatial representativity of polar ice core sites. Inparticular, glacial–interglacial Antarctic temperaturechanges have been used to estimate paleoclimatic con-straints on climate sensitivity, assuming that they arerepresentative of global temperature changes (Genthonet al. 1987; Lorius et al. 1990; Manabe and Broccoli1985). In this paper, we review the global and polartemperature response of new past and future climatesimulations conducted with state-of-the-art coupled cli-mate models with standardized forcings. We have usedthe database of Coupled Models Intercomparison Pro-ject (CMIP, http://www.pcmid.llnl.gov/ipcc, database asof 15.04.2005) and second phase of the PaleoclimateModelling Intercomparison Project (PMIP2, http://www.lsce.cea.fr/pmip2 database of 15.09.2005) (Harri-son et al. 2002). These recent model intercomparisonresults are currently used as part of the fourth IPCCassessment. They are also compared to earlier simula-tions conducted with atmosphere alone models (Paleo-climate Modelling Intercomparison Project, PMIP,http://www.lsce.cea.fr/pmip, Joussaume and Taylor1995).

514 Masson-Delmotte et al.:Past and future polar amplification of climate change: climate model intercomparisons and ice-core constraints

Table

1Boundary

conditionsusedforthepast

andfuture

clim

ate

simulations

Sim

ulations

Atm

ospheric

composition

(greenhouse

gases)

Landsurface

Oceansurface

Orbitalparameters

Years

analysed

2·C

O2

1ppm/yearincrease

untilCO

2doubling

(70years)then

stabilization

Modern

Calculated

Modern

Years

120–170

Controlrun

Pre-industrial

�280ppm

CO

2

First

50years

of

thesimulations

4·C

O2

1ppm/yearincrease

untilCO

2quadrupling

(140years)then

stabilization

Modern

Calculated

Modern

Years

190–240

Controlrun

Pre-industrial

�280ppm

CO

2

First

50years

of

thesimulations

PMIP1MH

fixed

SST

‘‘fix’’

Pre-industrial

Modern

Fixed

tomodern

6,000BP

Allyears

indatabase

(equilibrium

simulations)

Controlrun

Modern

Modern

PMIP2MHCoupled

models‘‘cpl’’

Pre-industrial

Modernorinteractively

calculated(O

AV

models)

Calculated

6,000BP

Allyears

indatabase

(equilibrium

simulations)

Controlrun

Modern

PMIP1LGM

fixed

SST‘‘fix’’

Glacial200ppmv

Vegetation:modern,

Icesheets,

coastline:(Peltier

1994)

Fixed

to(C

LIM

AP1981)

21,000BP

Allyears

indatabase

(equilibrium

simulations)

Controlrun

Modern

Modern

Modern

Modern

PMIP1LGM

Computed

SST‘‘slab’’

Glacial200ppmv

Vegetation:modern,

Icesheets,

coastline:(Peltier

1994)

Slabocean

21,000BP

Allyears

indatabase

(equilibrium

simulations)

Controlrun

Modern

Modern

Modern

Post-PMIP1LGM

Coupledmodels‘‘cpl’’

Glacial185ppmv

Vegetation:control,

Icesheets,(Peltier

1994,

2004)Coastline:either

unchanged

or�105m

Calculated

21,000BP

Allyears

indatabase

(equilibrium

simulations)

Controlrun

Modernexcept

HadCM3

(Pre-industrial)

Modern

Modern

PMIP2LGM

Coupled

models‘‘cpl’’

Glacial185ppmv

Vegetation:control,

Coastline:�105m,

Icesheets,(Peltier

2004)

Calculated

21,000BP

Allyears

indatabase

(equilibrium

simulations)

Controlrun

Pre-industrial

Modern

Modern

Forthe

description

ofthe

differentmodelsplease

referto

the

CMIP

and

PMIP

database

description

(http://w

ww.lsce.cea.fr/pmip/,

http://w

ww.lsce.cea.fr/pmip2/,

http://

www.pcm

di.llnl.gov/ipcc/about_ipcc.php)

Masson-Delmotte et al.:Past and future polar amplification of climate change: climate model intercomparisons and ice-core constraints 515

The key questions we address are: what is the per-formance of the models in simulating past climatechanges at the poles? Is there a polar amplification underdifferent forcings? How do relationships between globaland polar temperature changes vary for radically dif-ferent climates such as the LGM and increased CO2

experiments?

2 LGM and MH climate change: polar ice core data

Deep polar ice cores offer unique archives of past cli-mate and environmental changes extending over severalclimatic cycles in East Antarctica, at Vostok (Petit et al.1999), Dome F (Watanabe et al. 2003) and Dome C(EPICA-community-members 2004) and almost a fullclimatic cycle in Greenland at Summit (Cuffey and Clow1997; Johnsen et al. 1997) and even further at North-GRIP (NorthGRIP-community-members 2004).

The progressive distillation of air masses during themoisture advection from mid to high latitudes is classi-cally used to estimate past temperature changes frompolar ice core water isotopic composition in either d18Oor dD (Dansgaard 1964; Lorius et al. 1969). This ‘‘iso-topic thermometer’’ can however be biased by pastchanges in snowfall seasonal deposition (Krinner et al.1997; Werner et al. 2000); moisture origin (e.g. Boyle1997; Cuffey and Vimeux 2001a, b) or other factors(inversion strength, short term covariance between pre-cipitation and temperature...). The full isotopic compo-sition of snow and ice itself brings constraints on changesin moisture origin, owing to the deuterium excessparameter (d=dD – 8 d18O) (Masson-Delmotte et al.2005a, b; Stenni et al. 2001). Climate models are usefultools to help estimate past changes in deposition condi-tions (seasonal cycle of precipitation, condensationtemperature versus surface temperature) which may alterthe water stable isotope to temperature relationships.

In Antarctica, water stable isotopes only have beenused to quantify glacial-interglacial temperature changes.The consistency of ice core dating with an ice flow modelfed by the temperature and accumulation derived fromwater stable isotope fluctuations warrants the realism ofthese reconstructions, within 20–30% (Jouzel et al. 2003;Parrenin et al. 2001; Stenni et al. 2001; Watanabe et al.2003). LGM to present surface air temperature change isestimated to be 9±2�C; this error bar is estimated bymodelling of precipitation isotopic composition changeswith an air mass distillation model, taking into accountchanges in moisture origin derived frommeasurements ofdeuterium excess in the ice (Jouzel et al. 2003; Stenni et al.2001; Watanabe et al. 2003). This temperature changemay partly arise from an ice sheet elevation differencebetween modern and glacial conditions, with variableranges of estimates (Ritz et al. 2001; Peltier 2004). Suchelevation effects may also partly influence local AntarcticMH temperature change; most inland ice core isotopicrecords suggest thatMH temperatures were either similaror warmer by at most 0.8�C (Masson et al. 2000). A few

coastal records such as Taylor Dome or Law Domesuggest that MH temperatures could have been warmerby more than 1�C (Steig et al. 1998).

Although associated with 20–30% uncertainty inAntarctica, temperatures derived from water stable iso-tope measurements in Greenland have been showed tounderestimate glacial temperature changes in Greenland.Alternative paleothermometry methods have beendeveloped and are particularly suited for Greenlandwhere the accumulation rate is typically two to ten timeslarger than for central Antarctica. Precise measurementsof the vertical temperature profiles in the boreholes whenthe drilling operations are completed provide directreconstructions of past temperatures owing to the heatdiffusion in ice sheets. However, the inversion of the heatdiffusion equations requires a priori knowledge of theshape of the temperature variations so there are stilluncertainties on MH and LGM changes, typically within2�C (Cuffey et al. 1995; Johnsen et al. 1995; Cuffey andClow 1997; Dahl-Jensen et al. 1998). These boreholetemperature reconstructions show that MH tempera-tures were typically 1.5–2.5�C warmer than now andLGM temperatures 19–22�C colder than now. By con-trast, temperature reconstructions based on water stableisotopes suggest a smaller MH warming (0.5–0.9�C)(Masson-Delmotte et al. 2005a, b) and only a 10�Camplitude for the glacial interglacial change. This canonly be reconciled with the LGM borehole temperatureprofile when taking into account dramatic changes in theprecipitation seasonal cycle: the lack of winter snowfallduring glacial times, in response to changes in sea–iceand land–ice cover, could be responsible for a seasonalbias as suggested by simulations conducted with atmo-spheric general circulation models (Krinner et al. 1997;Krinner and Werner 2003; Werner et al. 2000).

Although obtained at individual locations, the mostreliable ice core based temperature estimates (boreholedata for Greenland, stable isotopes for East Antarctica)are representative not only of their local drilling site butof central ice caps, as evidenced by the convergent re-cords from a variety of ice cores (Fig. 1). In the fol-lowing sections, we compare model results averagedover inland ice caps with these ice core based estimates.

3 Model experimental design

For future climate change we consider the two scenarioswhere atmospheric CO2 levels were respectively stabi-lised at a doubling of pre-industrial levels (hereafter2·CO2) and a quadrupling of pre-industrial levels(hereafter 4·CO2) after a 1% per year increase from pre-industrial levels. We only consider recent model simu-lations performed for the IPCC fourth assessmentavailable from the CMIP database. In these two cases,we have calculated the difference between a 50-yearperiod of the stabilisation period (centred on the timeperiod of 75 years after the end of atmospheric CO2

increase) and a 50-year period of the control run


(Table 1). For a consistent comparison with the paleo-climate results, we need to compute the response of theclimate models closest to the equilibrium response, whilekeeping a maximum of models for our analysis. A set often model results was available for the 2·CO2 experi-ment (nine for the 4·CO2 experiment), including severalsimulations conducted either with different initial con-ditions or different model versions from the same groups(Table 2).

For past climates, we have used simulations per-formed with the same coupled model (‘‘cpl’’) versionsused in CMIP, as part of PMIP2, with a total of sevenmodels for MH and five for LGM. Earlier simulationsconducted with three coupled models but with forcingsdifferent from the standard protocol of PMIP2 have alsobeen included (post-PMIP1) (Table 1). In addition, wehave also considered atmospheric alone simulations(‘‘fix’’) of the first phase of PMIP (16 models for MH and

Fig. 1 Maps showing thelocation of ice core records usedto infer estimates of LGMtemperature changes in aGreenland and b Antarctica.From south to north, inGreenland, open circles locatethe sites of Dye 3, Summit(GRIP and GISP2) andNorthGRIP ice cores. InAntarctica, from 90�W in thetrigonometric directions, opencircles correspond to Byrd,Siple Dome, Taylor Dome,Dome C, Vostok,Komsomolskaia, Dome B andDome F ice cores


Table

2Listofsimulationsandmodel

versionsusedforthisintercomparison

Model

name

inthedatabase

PMIP1MH

Fixed

SST

PMIP2MH

Coupledmodels

PMIP1LGM

Fixed

SST

PMIP1LGM

CalculatedSST

Post-PMIP1LGM

Coupledmodels

PMIP2LGM

Coupledmodels

CMIP

2·C

O2

Coupledmodels

CMIP

4·CO2

Coupledmodels

bmrc

Xccc2.0

XX

Xccm1

Xccm3

Xccsr1

XX

csiro

Xclim

ber-2

Xcnrm

-2X

echam3

XX

gfdl

XX

gen1

Xgiss-iip

Xlm

celm

d4

XX

Xlm

celm

d5

XX

mri2

XX

Xugamp

XX

Xuiuc11

Xukmo

XX

yonu

XMIR

OC3.2

(medres)

XX

XX

NCAR_CCSM3

XX

XX

FOAM

XIPSL-C

M4-V

1-M

RX

XX

XUBRIS-H

ADCM3M2

XOA

andOAV

GISSmodelE-R

XX

UKMO-H

ADCM3M2

XECBIL

TCLIO

XMRI-CGCM1

XHADCM3

XUToronto-C

CSM1.4

XCNRM-C

M3

XGFDL-C

M2.0

XX

GFDL-C

M2.1

XX

GISS-EH

XGISS-ER

XX

FGOALS-g1.0

XX

X(3

simulations)

INM-C

M3.0

XX

ECHAM5/M

PI-OM

XX

MRI-CGCM2.3.2

XX

(2simulations)

Number

ofsimulations

16

87

93

416

11

Forthedescriptionofthedifferentmodelsplease

referto

theCMIP

andPMIP

databasesdescriptions

(http://w

ww.lsce.cea.fr/pmip/,http://w

ww.lsce.cea.fr/pmip2/,http://w

ww.pcm

di.llnl.gov/ipcc/about_ipcc.php)


7 for LGM) as well as nine atmospheric models coupledwith slab ocean models (‘‘cal’’) for LGM (Tables 1, 2).The ‘‘fix’’ and ‘‘cal’’ PMIP runs are equilibrium simula-tions and all years of simulated climate available in thedatabase were averaged (Table 1); for PMIP2, modellinggroups provided 30 to 100 year long outputs, obtainedafter major adjustments to the forcings (but withoutconstraints on deep ocean equilibrium). These 100 yearlong outputs where averaged for our analyses.

The most recent simulations (CMIP and PMIP2) usecoupled ocean–atmosphere models and their referencesimulations are long pre-industrial control runs. In thecase of PMIP1 atmosphere and atmosphere with slabocean models, the reference simulation is the modernclimate. In between these two standardized procedures,post-PMIP1 simulations used either modern or pre-industrial reference simulations; these differences are notcritical for our analyses.

We have extracted from the CMIP, PMIP1 andPMIP2 databases the available monthly model outputs(temperature, orography and for the paleoclimate sim-ulations also precipitation). In all the figures and the restof this paper, all the temperature anomalies will be dis-cussed as control minus LGM, control minus MH, and2·CO2 or 4·CO2 climate minus control climate, forconsistency of time direction. Global and regional re-sults were calculated by weighting each model grid pointby the actual area of these grid points. For each avail-able simulation, we have calculated the global meantemperature change. For each model, Central Greenlandwas defined as the model grid points located between 65and 80�N, 60 and 25�W, with an elevation above1,300 m. The same approach was used for CentralAntarctica with grid points between 90 and 70�S, 0 and180�E, and an elevation above 2,500 m.

4 Results and discussion

4.1 Model-data comparisons: preliminary remarks

4.1.1 Last glacial maximum ice cap elevations

The representation of the orography in atmosphericmodels strongly depends on their resolution and theirdiscretisation method (spectral or grid-point). Theaverage control run central Greenland orography variesbetween 1,315 and 2,376 m. This dispersion could ex-plain the large range of model results for the controlsimulation. We therefore do not compare absolute val-ues for each period but anomalies with respect to thecontrol runs.

The ICE-4G LGM ice cap prescribed in PMIP1 andpost-PMIP1 (Peltier 1994) is characterized by an average800 m elevation increase compared to the present alti-tude (between 640 and 924 m). In contrast, the ICE-5GLGM ice cap reconstruction prescribed for the PMIP2simulations (Peltier 2004) have a much reduced centralGreenland elevation anomaly (106–275 m). However,

both ice-core air content information (Raynaud et al.1997) and ice sheet modelling (Tarasov and Peltier 2003)suggest that changes in central Greenland elevation werelimited (and even in the opposite direction). We there-fore face the problem that different LGM simulationswere conducted with different glacial Greenland eleva-tion changes, significantly larger than those derived fromice core information and ice sheet modelling. Whenconsidering the typical polar vertical temperature lapserate (circa 10�C per 1,000 m) (Krinner and Genthon,1999), such discrepancies in elevation changes (up to800 m) may account for up to 8�C temperature differ-ences (without taking into account atmosphericdynamics related to glacial ice cap topographies).

In central Antarctica, the modern elevation repre-sented in the models varies between 2,956 and 3,448 m.For PMIP1, the prescribed ice sheet change (Peltier1994) results in an average elevation increase of 368 m(from 120 to 430 m depending on the models). For post-PMIP1, all the models analysed here show changes ofelevation between 392 and 399 m. For PMIP2 with theICE-5G ice caps, similar orders of magnitude are alsoobtained (Peltier 2004) (elevation changes ranging be-tween 343 and 400 m above modern levels). Again, thiscentral Antarctic ice cap elevation increase at the LGMis too high when compared with ice core air contentchanges or with dynamical ice cap modelling forced byice-core based climate histories (Ritz et al. 2001).

Because of the hundreds of meters of maximum dif-ferences both in modern polar elevations and in LGM tomodern changes, we have compared model outputs di-rectly, as well as with a correction for changes in ele-vation. Observations of modern polar vertical lapse rateand spatial elevation-temperature gradients show lapserates varying between 6�C per 1,000 m and 15�C per1,000 m depending on the season and the location(Ohmura and Reeh 1991; Krinner and Genthon 1998,1999). In order to account for large glacial to moderncentral ice sheet elevation changes, we have used aconstant first order lapse rate of 10�C per 1,000 m. Theuse of this lapse rate provides a rough estimate of theeffect of the prescribed orography changes; we estimatethat this elevation correction is associated with a maxi-mum uncertainty of 50%, given the range of observedmodern polar lapse rates.

For reference, we have applied this methodology tocompare the control simulations (Table 1) with themodern temperatures at deep ice core sites. The range ofcontrol simulation elevations are displayed together withtemperatures ‘‘uplifted’’ at the elevation of ice core sites(typically 3,000 m for central Greenland and 3,400 m forEast Antarctic Plateau). It can be observed that, inaverage, climate models capture the correct expectedrange of elevation-corrected temperatures in centralGreenland. By contrast, there is a tendency to observean underestimation of elevation-corrected temperatureson the East Antarctic Plateau by a few degrees (Table 3);this may be due to different representations of theAntarctic inversion strength.


4.1.2 Comparing model outputs with ice core isotopepaleothermometry

Water stable isotopes record past temperature changesthrough the progressive depletion of atmospheric va-pour of heavy molecules along their transport path frommoisture sources to the poles. The relationship betweenisotopic composition and temperature therefore resultsfrom the control of cloud temperature on condensation.Changes in moisture source origin, in the gradient be-tween surface and condensation temperatures may in-duce changes in the temporal relationships betweenpolar precipitation isotopic composition and tempera-ture. Deuterium excess data from both poles suggest thatchanges in moisture source is not a major bias on glacialinterglacial ice core based temperature reconstructions(Cuffey and Vimeux 2001a, b; Stenni et al. 2003, 2001).By contrast, changes in the seasonal cycle of precipita-tion remain one plausible explanation for the mismatchbetween Greenland glacial temperature change derivedfrom borehole measurements versus temperature derivedfrom the classical spatial isotope/temperature slope(Masson-Delmotte et al. 2005, b). In the present work,we have not assessed changes in the vertical temperaturegradient, some of the databases which we use missingthis type of data. However, we have attempted to esti-mate the impact of precipitation seasonality changes onpast temperatures recorded in ice cores. For this pur-pose, we have taken into account the impact of changesin the seasonal cycle of the precipitation by comparingthe change in annual mean temperature (arithmeticaverage of 12 monthly mean temperatures) with thechange in precipitation-weighted annual mean temper-ature, for each LGM simulation (calculated frommonthly outputs of temperature and precipitation;Fig. 4a, b), following the methodology of (Krinner et al.

1997; Krinner and Werner 2003; Werner et al. 2001).When calculated from all the control simulations avail-able, this seasonality effect accounts for a 2.3±1.7 and3.3±2.9�C difference between the modern precipitationweighted temperatures and the annual mean tempera-tures, respectively for Greenland and East Antarctica.This is the consequence from the fact that climatemodels simulate slightly more precipitation today duringthe warmest months in both polar regions than duringthe coldest months; this effect is reinforced at the LGM.Indeed, PMIP2 results show systematic shifts towardsmore precipitation provided during the warm season atLGM. If this change of seasonal cycle is real, this effectmay account for an underestimation of LGM annualmean temperature changes as recorded in stable isotopesfrom ice cores, which remains however limited. Thedifferences between weighted LGM and annual meanLGM temperatures increase to 4.0±0.8 and 4.6±1.3�Cfor Greenland and Antarctica, accounting for a relativeuncertainty of 8–23% of the full glacial-control range,depending on the hemisphere and on the model.

4.2 Global mean temperature changes

In order to assess the range of global temperaturechanges simulated by climate models under 2·CO2,4·CO2, MH and LGM boundary conditions, we havecompared on Fig. 2 the distribution statistics (full range,25th, 50th, 75th quartiles and the mean of all models) ofmodel results for these various experiments. This pre-sentation is well suited when limited numbers of simu-lations are available (in our case from 4 to 16) and helpsvisualizing the range (maximum to minimum values),the average, median and dispersion of model results. Inparticular, extrema levels significantly different from the

Table 3 « Central Greenland » and « East Antarctic Plateau » control simulations of elevation and temperature respectively uplifted to3,000 and 3,400 m elevation (taking into account a vertical lapse rate of 10�C per 1,000 m)

Control runs for experiments Central Greenland correctedtemperature (‘‘uplifted’’at 3,000 melevation, �C)

Central Greenlandelevation (m)

East Antarctic Plateau correctedtemperature (‘‘uplifted’’ at3,400 m elevation, �C)

East Antarctic Plateausimulated / observedelevation (m)

PMIP1 6fix �29.6±5.7 1,921±239 �47.3±8.0 3,095±120PMIP1 21 fix �28.4±8.2 1,942±321 �44.8±10.1 3071±70PMIP1 21 cal �30.5±4.3 1,899±189 �45.3±7.6 3,115±85Pre-PMIP2 �32.9±5.4 1,835±212 �49.5±7.2 3,059±30PMIP2 6 k cpl �34.1±2.9 1,826±152 �50.3±3.0 3,053±72PMIP2 21 k cpl �33.5±3.4 1,883±233 �48.9±5.5 3,057±71IPCC 2·CO2 �32.1±2.9 2,007±144 �48.7±2.7 3,050±41IPCC 4·CO2 �32.0±3.2 2,005±150 �50.4±2.5 3,065±24GRIP �29.7 3,200NorthGRIP �32.4 2,917Vostok �54.6 3,490Dome C �55.5 3,200Dome B �55.0 3,650Komsomolskaia �51.6 3,500

For comparison, mean characteristics of various drilling sites (see Fig. 1) are also displayed. Uncertainties in the vertical lapse rate usedfor elevation corrections should induce uncertainties on corrected temperatures within 1–2�C. For each type of control simulation, themean value of all model outputs available is displayed together with the inter-model standard deviation


25th to 75th ranges clearly identify the occurrence ofoutliers among the different model results.

For the MH, 6,000 years ago, climate models docapture reconstructed seasonal and regional changessuch as wetter conditions in northern Africa (Joussa-ume et al. 1999; Braconnot et al. 1999) or warmerconditions in some areas of the mid-latitudes in sum-mer (Cheddadi et al. 1997; Masson et al. 1999; Wohl-fahrt et al. 2004). This results from a seasonal andlatitudinal redistribution of solar radiative forcing dueto changes in precession and obliquity of the Earth’sorbit (Berger 1978; Joussaume and Braconnot 1997)modulated by the differential heat capacity of the oceanand land masses as well as by the sea–ice response athigh latitudes (Crucifix et al. 2002). In some regions,the seasonal changes do have a signature in the annualmean; this is for instance the case for annual meanprecipitation in monsoon regions. Globally, the meantemperature change between control and MH simula-tions is between �0.05 to +0.4�C with a median valueof respectively +0.05 and +0.08�C for PMIP1 atmo-spheric models and PMIP2 coupled climate models. Inaverage, climate models simulate slightly warmer con-trol global temperatures compared to the MH simula-tions (Fig. 2). Note that in this MH case, when globalmean temperature changes are small, the notion ofglobal climate sensitivity computed from global andannual averages is not suitable because it does notaccount for the seasonal aspects both in terms oforbital forcing and climate response.

The median range of control minus LGM globaltemperature change is fairly constant for PMIP1 simu-lations conducted with CLIMAP ocean surface bound-ary conditions (‘‘fix’’) and atmospheric models alone,PMIP1 simulations with atmospheric models coupledwith mixed layer ocean models (‘‘cal’’) and PMIP2simulations with fully coupled ocean-atmosphere cli-

mate models (‘‘cpl’’). The median of model results israther stable from PMIP1 to PMIP2 (3.7�C for fixedSST, 4.9�C for calculated SST, 3.9�C for post-PMIP1and 4.1�C for PMIP2 results). With the exception of twooutliers (LMCELMD4 for PMIP1 cal and UTORON-TO-CCSM1.4 for post-PMIP1), all model results rangebetween 3.1 and 6.4�C in terms of glacial to controlglobal temperature change

The range of global temperature change for 4·CO2

experiments, with a median of 4.3�C, is comparable tothe LGM to modern change. In the case of 2·CO2

experiments, the global temperature range has a medianof 2.0�C. The dispersion range between various modelresults (indicated by the vertical dashed line in Fig. 2) issignificant when compared to the full amplitude of thechanges. In the case of future climate change scenarios,climate models have a range of dispersion of the sameorder of magnitude as the median change: 2.3�C for2·CO2 and 2.7�C for 4·CO2. A similar 2�C dispersion isalso observed for PMIP2 LGM to modern changes. Thedifference among individual models forced with similarboundary conditions is therefore about 100% of themedian global temperature change in the case of 2·CO2,and 50% of the median global temperature change in thecase of 4·CO2 and LGM simulations.

This comparison shows that the global temperatureresponse is very similar under 4·CO2 and LGM forc-ings, although occurring in response to different spa-tially-distributed forcings (greenhouse gases in one case,greenhouse gases plus ice sheets in the other case).

4.3 The particular case of MH polar temperaturechange: model-data comparison

For the MH, it is impossible to discuss a polar amplifi-cation of climate change, defined as the ratio between

0

1

2

3

4

5

6

7

8

9

10

Glo

bal t

empe

ratu

re c

hang

e (˚

C)

75th

50th

25th

mean

PMIP1 PMIP2 IPCC IPCC PMIP1 PMIP1 pre-PMIP2 PMIP20-6 cpl 0-6 cpl 2xCO2-0 4xCO2-0 0-21 fix 0-21 cal 0-21 cal 0-21 cpl

Fig. 2 Annual mean globaltemperature changes simulatedby a variety of climate modelsrun under similar boundaryconditions. The full range(dashed line), mean (crosssymbol), 25th (lower bold dashsymbol), 50th (grey squaresymbol), 75th (upper bold dashsymbol) percentiles of thevarious model results arecalculated from the distributionof the various model results (seeTable 2). The ‘‘fix’’, ‘‘slab’’ and‘‘cpl’’ abbreviations refer todifferent configurations ofmodels used and are describedin Sect. 3 and Table 1


the temperature change in central polar regions to globaltemperature changes, due to the negligible global meantemperature changes.

For central Greenland, simulations conducted withfixed ocean surface conditions within PMIP1 show alarge dispersion of temperature change between MH andcontrol conditions (range between �0.8 and 1.5�C withhalf of the models simulating warmer and half of themodels simulating colder conditions than present). Thisdispersion is probably related to different responses ofthe atmospheric stationary waves and the associatedheat transport to Greenland; note that the range oftemperature changes is quite small compared to thetypical temperature interannual variability range (Mas-son et al. 1998). By contrast, all the PMIP2 coupledmodel simulations performed with an interactive oceanand sea–ice model simulate a temperature decrease fromMH to modern conditions varying from 0.35 to 0.90�C,consistent with paleoclimatic estimates of warmer MHconditions, but weaker than borehole temperaturereconstructions (Dahl-Jensen et al. 1998; Masson-Del-motte et al. 2005a, b). The use of fully coupled climatemodels therefore improves the consistency of the modelresults over Greenland. Indeed, coupled climate modelsimulate a shaping of the seasonal cycle by the oceanresponse associated with a late summer warming, a re-duced sea–ice cover and an Arctic warming (Vavrus andHarrison 2003). Although PMIP2 models do capture thecorrect sign of the reconstructed temperature change,they underestimate the amplitude of the MH Greenlandtemperature change in average by a factor of 3. Last,these simulations enable us to explore changes in theprecipitation seasonal cycle as a clue to reconcile bore-hole and water isotope temperature reconstructions. Allthe PMIP2 MH simulations reveal that precipitation-weighted temperature changes should be larger than thearithmetic annual mean of monthly temperatures,resulting in a range of precipitation-weighted tempera-ture changes from 1.0 to 1.2�C (due to more snowfallduring the warmest months). This is not consistent withthe different ranges of temperature changes estimatedfrom borehole versus stable isotope thermometry, andfurther suggests that the change in moisture advectionsimulated by PMIP2 coupled models may not be correct.It has to be noted that the PMIP2 simulations did notsystematically include changes in vegetation. Earlierstudies had shown that changes in high northern latitudevegetation enhanced the local warming (Wohlfahrt et al.2004).

We have performed the same analysis of MH modelresults for central Antarctica. PMIP1 simulations showno model consistency and result in changes in Antarctictemperatures varying from �0.6 to +0.7�C. Again,PMIP2 simulations performed with coupled modelssystematically show a temperature decrease from MHto modern levels, by 0.1–0.7�C. As for Greenland, thislarger consistency among models is attributed to sea-sonal ocean and sea–ice response in the surroundingareas. For central Antarctica, coupled models are

qualitatively and quantitatively in agreement withtemperature reconstructions based on water isotopesfrom polar ice cores, which suggest a range of tem-perature change from 0 to +0.8�C (Masson et al.2000). When precipitation seasonality is considered, thesimulated temperature range shifts by 0.3–0.6�C, con-firming the previous range of temperature change andfurthermore the fact that changes of Antarctic precip-itation seasonal cycle should not be a major source ofuncertainty when reconstructing local past temperaturechanges from ice core water stable isotopes, evenwithin the current interglacial period (Jouzel et al.2003).

4.4 Polar amplification: LGM versus future climatesimulations

Here we discuss the ratio of polar versus global meantemperature changes. Figure 3a displays the compari-son of LGM and future climate change range of modelresults for CMIP and PMIP1 and PMIP2 intercom-parison exercises. Increasing CO2 concentrations im-posed in 2·CO2 and 4·CO2 induce a central Greenlandtemperature change of respectively 1.37 and 1.40 timesabove the simulated global temperature increase; incentral Antarctica, this amplification factor is respec-tively 1.16 and 1.25. This intercomparison thereforeconfirms previous studies conducted as a function oflatitude but only for the northern hemisphere (Hollandand Bitz 2003). Because the median of all climatemodels is systematically above 1 (Fig. 3), we can rea-sonably argue that both past and future climate chan-ges are indeed associated with a polar amplification ofclimate change.

For the LGM, part of this polar amplification may bedue to the local elevation increase imposed by ice sheetreconstructions used as boundary conditions (Peltier1994, 2004) (see Sect. 4.1.1). In order to assess the pro-portion of local temperature changes which could beaccounted for by local topography changes, we haveadded ‘‘elevation corrected’’ results where LGM simu-lated temperature changes are corrected from changes inelevation using a slope of 10�C per 1,000 m elevation(these calculations appear as ‘‘corr’’ labels in Fig. 3).

Glacial boundary conditions induce a very largepolar amplification for Greenland (Fig. 3a). PMIP1boundary conditions result in a factor of 5 betweenGreenland and global temperature changes, and afactor of 3 when corrected for large elevation changes.Such a large amplification is clearly the result of theimposed CLIMAP surface conditions (including a largesea–ice coverage in the North Atlantic plus ratherwarm tropical temperatures). By contrast, PMIP2simulations conducted with a different ice sheettopography (Peltier 2004) and with fully coupledocean–atmosphere–sea ice models suggest a range ofamplification of 2.6–2.1 (respectively without and withtopography correction) (Fig. 4a). These results suggest


that Greenland temperature changes are deeply ampli-fied under glacial conditions, due to (1) imposed localtopography changes (up to half of the signal), (2)northern hemisphere albedo, atmospheric and oceancirculation effects. The impact of large sea–ice changesin the North Atlantic on Greenland temperature is

most obvious from the dispersion of PMIP1 slab oceansimulations. Many of these processes should be activeboth in the warming from glacial to modern conditionsand in the warming from modern to future conditions.Note that future climate change simulations analysedhere do not account for the possible polar ice sheet

0

1

2

3

4

5

6

ratio

of

Gre

enla

nd to

glo

bal

tem

pera

ture

cha

nge

slop

e.

75th

50th

25th

mean

ratio of 1

IPCC IPCC PMIP1 PMIP1corr PMIP1 PMIP1corr pre-PMIP2 pre-PMIP2corr PMIP2 PMIP2 corr 2xCO2-0 4xCO2-0 0-21 fix 0-21 fix 0-21 cal 0-21 cal corr 0-21 cpl 0-21 cpl 0-21 cpl 0-21 cpl

0

0.5

1

1.5

2

2.5

3

ratio

of

Ant

arct

ic to

glo

bal

tem

pera

ture

cha

nge

slop

e.

75th

50th

25th

mean

ratio of 1

IPCC IPCC PMIP1 PMIP1corr PMIP1 PMIP1corr pre-PMIP2 pre-PMIP2corr PMIP2 PMIP2 corr 2xCO2-0 4xCO2-0 0-21 fix 0-21 fix 0-21 cal 0-21 cal corr 0-21 cpl 0-21 cpl 0-21 cpl 0-21 cpl

a

b

Fig. 3 a Central Greenland polar amplification (defined as the ratiobetween central Greenland and global annual mean temperaturechanges) simulated by climate models. The full range (dashed line),mean (cross symbol), 25th (lower bold dash symbol), 50th (greysquare symbol), 75th (upper bold dash symbol) percentiles of the

various model results are calculated from the distribution of thevarious model results (see Table 2). ‘‘corr’’ stands for elevation-corrected temperature values (see text). b Same as (a) but forcentral eastern Antarctica. Note that the vertical scale is half assmall as for Greenland


response in terms of elevation change. Preliminary re-sults obtained with an earth climate model of inter-mediate complexity fully coupled to Greenland andAntarctica ice-sheet models (S. Charbit et al., personalcommunication) suggest that such effects should beimportant for central Greenland, similar to the findingsof (Huybrechts et al. 2004).

In Central Antarctica (Fig. 3b), the range of polaramplification is smaller than for central Greenland bothfor future and glacial climate changes; this is probablydue to the larger albedo effect in the northern hemi-sphere due to ice caps and snow cover over the conti-nents, and to changes in northern hemisphere stormtracks. PMIP1 boundary conditions result in an ampli-fication factor of 2.1 (without orography correction) to1.2 (with orography correction). PMIP2 simulationsconducted with coupled climate models give rathersimilar ranges of changes (median model amplificationfactors of respectively 1.9 and 1.2). The sea–ice changessimulated by PMIP1 slab ocean runs, post-PMIP1 andPMIP2 models is systematically less extensive than pre-viously prescribed from CLIMAP reconstructions, con-sistent with new austral sea–ice reconstruction efforts(Gersonde et al. 2005). When orography effects oncentral Antarctic temperature changes are roughly ac-counted for (Fig. 4b), the order of magnitude of polaramplification are very comparable to the range of futureAntarctic amplification (around 1.20); some models evenshow no Antarctic amplification. The coherency ofAntarctic temperature change with global temperaturechange may result from a dominant role of greenhousegases on southern high latitude radiative budget, onlyslightly modulated by the internal feedbacks associatedwith sea–ice extent, cloudiness, atmospheric heat andmoisture advection (including water vapour feedback,more important at high latitudes). By contrast, we canspeculate that Greenland polar amplification should bemore influenced by other processes such as dust forcing,changes in vegetation in the northern mid and highlatitudes, large icy surfaces and associated feedbackson atmospheric circulation and heat transport toGreenland.

4.5 Ice-core constraints on global temperature changes

When taking into account all available 2·CO2 and4·CO2 simulations, a linear regression relating Green-land to global temperature change leads to a slope of1.46 and a determination coefficient (R2) of 0.9 (Fig. 4c).Therefore, in the world of model results, in average,future changes in Greenland temperature are linearlydetermined by the global temperature range (‘‘climatesensitivity’’) but amplified by a factor of 46% (‘‘polaramplification’’).

When considering glacial to control climate chan-ges, we restrict here our analysis to the coupled modelsimulations performed under the same boundary con-ditions and with models involved in the IPCC fourth

assessment (PMIP2 simulations). In this case, centralGreenland temperature change is also linearly relatedto global temperature change with a slope of 2.71(without topography correction) and a determinationcoefficient of 0.81 (Fig. 4a). This suggests that paleo-climatic polar temperature reconstruction provide aquantitative constraint on the range of global tem-perature change. In fact, the range of simulated centralGreenland temperature changes lies between 6.2 and14.5�C, significantly below the borehole temperatureestimates of 19–22�C for 21,000 years before present.As for the MH climate, coupled model results docapture the glacial cooling but underestimate by afactor from 1.3 to 3.5 the reconstructed temperaturechange. Part of this inconsistency could be attributedto the uncertainty linked with the effect of increaseddust concentrations on climate during glacial times,probably underestimated when taken into accountwith atmosphere-only models (Werner 2002). Othermissing factors include the lack of LGM vegetation(Crowley and Baum 1997) and soil (Poutou 2003)feedbacks, more important in the northern hemispherebut which should be rather limited for Greenland(Crucifix and Hewitt 2005). When extrapolated, thismodel-data comparison however suggests that thesimulations most consistent with the paleoclimaticconstraints are associated with large glacial to controlglobal temperature changes (above 4�C) (Fig. 4b). Theprecipitation seasonality effect remains negligible(within a few �C) compared to the model-data dis-crepancy, and even acts against any reconciliation ofthe isotope to the borehole thermometry (all modelresults give a smaller amplitude of changes whenprecipitation weighted). The consistency of model re-sults could question the interpretation of proxy re-cords; however, borehole temperature profiles incentral Greenland are based on a physical mechanism(diffusion of heat into the ice).

For central Antarctica, a similar analysis can beperformed. When a linear regression is performed on allthe range of 2·CO2 and 4·CO2 simulations, this leads toa slope of 1.16 and a determination coefficient of 0.80,suggesting again that future Antarctic temperaturechange is, in the models, related to the global tempera-ture change, and in average amplified by about 16%above the global mean temperature change (Fig. 4d).When applied to PMIP2 last glacial maximum to controlresults, the same analysis leads to a slope of 2.09 and asmaller determination coefficient of 0.70 (Fig. 4b). Thisresult suggests that Antarctic temperature changes arerepresentative of large scale temperature changes (up tothe global scale), with an amplification factor of 2. Thisis a model confirmation that the early hypothesis of(Genthon et al. 1987) and (Lorius et al. 1990) remainsvalid. Antarctic glacial–interglacial temperature changescan be considered to constrain global climate sensitivityto greenhouse gases.

The range of simulated glacial to control centralAntarctic temperature change varies between 2.4�C for


ECBILT-CLIO (an intermediate complexity climatemodel) to 10.8�C. The impact of elevation changesaccounts for 2.9–3.5�C of the glacial to pre-industrial

warming. Without orography correction, all of the fourgeneral circulation coupled model results fall within therange of ice-core based temperature range (9±2�C).

y = 2.64x

R2 = 0.81

0

5

10

15

20

25

0 2 4 6 8 10

Global temperature change (˚C)

Gre

enla

nd te

mpe

ratu

re c

hang

e (˚

C)

HadCM3M2 HadCM3M2 HadCM3M2MIROC3.2 MIROC3.2 MIROC3.2CCSM CCSM CCSMIPSLCM4 IPSLCM4 IPSL CM4ECBILTCLIO ECBILTCLIO slope 1slope 2 slope 3 range of reconstructionsrange of reconstructions

y = 2.07x

R2 = 0.70

0

5

10

15

20

25

0 1 2 3 4 5 6 7 8 9 10


Ant

arct

ic te

mpe

ratu

re c

hang

e (˚

C)

HadCM3M2 HadCM3M2 HadCM3M2MIROC3.2 MIROC3.2 MIROC3.2CCSM CCSM CCSMIPSLCM4 IPSLCM4 IPSLCM4ECBILTCLIO slope 1 slope2slope 3 range of reconstructions range of reconstructions

a

b

Fig. 4 a Comparison of LastGlacial Maximum to controlcentral Greenland annual meantemperature change simulatedby climate models (PMIP2coupled ocean-atmospheresimulations only) with the rangeof paleoclimaticreconstructions. Filled blacksquares show direct modelresults. Open black squaresshow model results correctedfrom LGM to control ice sheetelevation changes (‘‘elevationcorrected’’ results). Greysquares show model resultscorrected from elevationchanges and precipitation-weighted (‘‘seasonalitycorrected’’ results). Horizontallong-dashed lines reflect therange of temperature changederived from Greenlandborehole thermometry. Short-dashed lines correspond toslopes of 1, 2 and 3 forreference. A linear regressioncalculated on the results ofthese four models is alsodisplayed (solid black line andregression result). Values belowzero are not displayed (resultsof ECBILT CLIO withcorrections. b Same as (a) butfor central Antarctica.Horizontal long-dashed linesreflect the range of temperaturechange derived from Antarcticice core water stable isotopes. cSame as (a) but for futureclimate change simulations.Open black squares represent4·CO2 simulation anomalies,and filled black rhomboids2·CO2 simulation anomalies.The solid line is a linearregression on all the simulationresults. The black dashed linesrepresent lines with slopes of 1and 2. d Same as (b) but forfuture climate changesimulations. Open black squaresrepresent 4·CO2 simulationanomalies, and filled blackrhomboids 2·CO2 simulationanomalies. The solid line is alinear regression on all thesimulation results. The blackdashed lines represent lines withslopes of 1 and 2


This suggests that the range of global glacial to controltemperature scale given by these models (3.7–5.1�C) isconsistent with the Antarctic proxy data. If the icesheet elevation change is taken into account, then,again, models do underestimate fixed-elevation centralAntarctica LGM temperature change by a factor of1.2–2.3.

5 Conclusions

We have performed model–model comparisons of pastand future climate change at the poles, and model-datacomparisons of past polar temperature changes both forthe MH and the LGM.

gfdl_cm2_0

ncar_ccsm3_0

mpi_echam5

miroc3_2_medres

picntrl_run1

gfdl_cm2_0

gfdl_cm2_1

giss_model_e_h

giss_model_e_r

iap_fgoals1_0_g iap_fgoals1_0_g iap_fgoals1_0_g

inmcm3_0

ipsl_cm4

miroc3_2_medres

ncar_ccsm3_0 mpi_echam5

mri_cgcm2_3_2a gfdl_cm2_1

giss_model_e_r

inmcm3_0

ipsl_cm4

mri_cgcm2_3_2a

y = 1.46x

R2 = 0.91

0

2

4

6

8

10

12

0 1 2 3 4 5 6 7 8 9 10Global temperature change (˚C)

Gre

enla

nd te

mpe

ratu

re c

hang

e (˚

C)

mri_cgcm2_3_2a

ipsl_cm4

ncar_ccsm3_0

mri_cgcm2_3_2amri_cgcm2_3_2a

mpi_echam5

ipsl_cm4

cnrm_cm3

gfdl_cm2_0

gfdl_cm2_1

giss_model_e_hgiss_model_e_r

iap_fgoals1_0_g

iap_fgoals1_0_giap_fgoals1_0_g miroc3_2_medres

mpi_echam5

ncar_ccsm3_0gfdl_cm2_0

gfdl_cm2_1giss_model_e_r

inmcm3_0

miroc3_2_medres

y = 1.16x

R2 = 0.80

0

1

2

3

4

5

6

7

8

9

10

0 1 2 3 4 5 6 7 8 9 10


Ant

arct

ic te

mpe

ratu

re c

hang

e (˚

C)

.

c

d

Fig. 4 (Contd.)


The main outcome of the model–model comparison isthat the range of global temperature change is verysimilar for 4·CO2 experiments and glacial experimentscompared to control simulations. The range of inter-model dispersion has the same order of magnitude forglacial and 2·CO2 experiments. Moreover, the polartemperature changes are linearly related to the globaltemperature change for both future and glacial climatechanges, with different polar amplifications at the twopoles. In the world of simulated climate change, tem-perature changes in central Greenland and centralAntarctica are therefore representative of global climatesensitivity. The polar amplification in Greenland ismuch larger than for central Antarctica, possibly due toalbedo effects and changes in atmospheric storm tracks.

Paleoclimatic simulations make possible quantitativemodel-data comparisons with reconstructions based onice core analyses. For the MH, PMIP1 atmosphere-onlymodels show a large variety of temperature changes. Bycontrast, coupled climate model simulations performedwithin PMIP2 all simulate MH temperatures warmerthan the control runs in central Greenland and centralAntarctica, consistent with ice-core signals, stressing therole of the ocean and sea–ice components of the climatesystem in integrating the seasonal orbital forcing into anannual mean response. However, they strongly under-estimate the amplitude of the central Greenland change,possibly due to the lack of vegetation feedbacks. It hasto be noted that these MH polar changes occur with anegligible change in global temperatures. Further effortswill be required to assess the validity of ice core tem-perature reconstructions in terms of larger scale climatechange for other interglacial periods.

The model-data comparison for the LGM is ham-pered by uncertainties of polar ice sheet topographychanges. First, the different central Greenland and cen-tral Antarctica ice caps prescribed in the PMIP1 (ICE-4G) and PMIP2 (ICE-5G) exercises are a few hundredmeters above modern elevations. Second, these pre-scribed topographies are in contradiction with ice-corebased estimates (either from air content indications, orfrom ice sheet models forced by ice-core climatic sig-nals). In this paper, we have circumvented this difficultyby comparing direct temperature changes and elevation-corrected temperature changes (with an estimated errorof 50% in this correction due to the range of observedmodern polar lapse rates). We are aware that this firstorder correction does not take into account atmosphericdynamical aspects related to the full geometry of theprescribed ice caps. These comparisons show that (1) asignificant part of the Greenland and Antarctic coolingof the PMIP1 and PMIP2 simulations is caused by theprescribed local elevation increase at the last glacial; (2)when corrected for elevation changes, models stronglyunderestimate Greenland temperature changes, al-though they are rather consistent with Antarctic tem-perature changes. A direct representation of water stableisotopes inside coupled climate models would help tounderstand the reasons for such inconsistencies, which

here are shown not to arise from changes in precipitationseasonal cycle only. The model-data quantitative dis-crepancy for central Greenland has to be placed in thelarger northern hemisphere context and should be re-lated to changes in stationary waves and storm tracktrajectories. The response of polar ice caps to futureclimate change may also act as a positive feedback onthe polar amplification, not taken into account in theCMIP experiments discussed here.

The comparison of PMIP1 and PMIP2 modelsclearly shows that changes in ocean surface conditions(in particular sea–ice extent) are strongly involved inpast polar amplification. In order to fully exploit thepaleoclimatic constraints on climate sensitivity, futurework will be needed to disentangle the local radiativeperturbation in response to changes in greenhouse gasesfrom the internal feedbacks involving in particulartropical sea surface temperatures (Lea 2004), sea–iceextent and atmospheric circulation.

6 DATABASE access

http://www.lsce.cea.fr/pmip/http://www.lsce.cea.fr/pmip2/http://www.pcmdi.llnl.gov/ipcc/about_ipcc.php

Acknowledgments We acknowledge the international modellinggroups for providing their data for analysis, the Program for Cli-mate Model Diagnosis and Intercomparison (PCMDI) for col-lecting and archiving the model data, the JSC/CLIVAR WorkingGroup on Coupled Modelling (WGCM) and their Coupled ModelIntercomparison Project (CMIP) and Climate Simulation Panel fororganizing the model data analysis activity, and the IPCC WG1TSU for technical support. The IPCC Data Archive at LawrenceLivermore National Laboratory is supported by the Office of Sci-ence, US Department of Energy. This work was funded by theFrench Project PNEDC IMPAIRS, European Projects EPICA-MIS and MOTIF (EVK2-2001-00263), and stimulated by discus-sions during the PMIP meeting in Giens, April 2005. This meetingwas funded by CNRS, CEA, PAGES and CLIVAR. The partici-pants are thanked for the discussions around this work. Jean-YvesPeterschmitt offered great help with handling the data and pro-ducing the graphics. V.M.D. is funded by CEA and M.K. byCNRS. We thank 3 anonymous reviewers for helpful comments.This is LSCE publication 1770.

References

Alexeev VA et al (2005) Polar amplification of surface warming onan aquaplanet in ‘‘ghost forcing’’ experiments without sea icefeedbacks. Clim Dyn 24:655–666

Bengtsson L et al (2004) The early twentieth-century warming onthe Arctic—a possible mechanism. J Clim 17:4045–4057

Berger AL (1978) Long-term variations of daily insolation andquaternary climatic change. J Atmos Sci 35:2362–2367

Boyle EA (1997) Cool tropical temperatures shift the global d18O-Trelationship: an explanation for the ice core ?d18O boreholethermometry conflict? Geophys Res Lett 24:273–276

Braconnot P et al (1999) Synergistic feedbacks from ocean andvegetation on the African monsoon response to mid-Holoceneinsolation. Geophys Res Lett 26:2481–2484

Cheddadi R et al (1997) The climate of Europe 6000 years ago.Clim Dyn 13:1–10


CLIMAP (1981) Seasonal reconstructions of the Earth’s surface atthe last glacial maximum. Geol Soc Am Map and Chart Series,MC-36, 18 p

Corell R et al (2004) Impacts of a warming Arctic: Arctic climateimpact assessment. Cambridge University Press, Cambridge, p144

Crowley TJ, SK Baum (1997) Effect of vegetation on an ice-ageclimate model simulation. J Geophys Res 102:16463–16480

Crucifix M, Hewitt CD (2005) Impact of vegetation changes on thedynamics of the atmosphere at the Last Glacial Maximum.Clim Dyn 25:447–459

Crucifix M et al (2002) Climate evolution during the Holocene: astudy with an Earth system model of intermediate complexity.Clim Dyn 19:43–60

Cuffey KM, Clow GD (1997) Temperature, accumulation, andelevation in central Greenland through the last deglacial tran-sition. J Geophys Res 102:26383–26396

Cuffey KM et al (1995) Large Arctic temperature change at theWisconsin-Holocene glacial transition. Science 270:455–458

Cuffey KM, Vimeux F (2001a) Covariation of carbon dioxide andtemperature from the Vostok ice core after deuterium-excesscorrection. Nature 412(6846):523–527

Cuffey KM, Vimeux F (2001b) Covariation of carbon dioxyde andtemperature from the Vostok ice core after deuterium-excesscorrection. Nature 421:523–527

Dahl-Jensen D et al (1998) Past temperatures directly from theGreenland ice sheet. Science 282:268–271

Dansgaard W (1964) Stable isotopes in precipitation. Tellus16:436–468

Dixon KW et al (2003) A comparison of climate change simula-tions produced by two GFDL coupled climate models. GlobalPlanet Change 37:81–102

EPICA-community-members (2004) Eight glacial cycles from anAntarctic ice core. Nature 429:623–628

Genthon C et al (1987) Vostok ice core: climatic response to CO2

and orbital forcing changes over the last climatic cycle. Nature329:414–418

Gersonde R et al (2005) Sea-surface temperature and sea ice dis-tribution of the Southern Ocean at the EPILOG Last GlacialMaximum—a circum-Antarctic view based on siliceous micro-fossil records. Quaternary Sci Rev 24:869–896

Harrison SP et al (2002) Comparison of palaeoclimate simulationsenhance confidence in models. EOS Trans 83:447–447

Holland MM, Bitz CM (2003) Polar amplification of climatechange in coupled models. Clim Dyn 21:221–232

Huybrechts P et al (2004) Modelling Antarctic and Greenlandvolume changes during the 20th and 21st centuries forced byGCM time slice integrations. Global Planet Change 42:83–105

IPCC (2001) Climate change 2001: impacts, adaptation and vul-nerability–contribution of working group II to the thirdassessment report of IPCC, Cambridge University Press, UK, p1000

Jacka TH, Budd WF (1998) Detection of temperature and sea-ice-extent changes in the Antarctic and Southern Ocean. AnnGlaciol 27:553–559

Johnsen SJ et al (1997) The d18O record along the Greenland icecore project deep ice core and the problem of possible Eemianclimatic instability. J Geophys Res 102:26397–26410

Johnsen SJ et al (1995) Greenland palaeotemperatures derivedfrom GRIP borehole temperature and ice core isotopic profiles.Tellus 47B:624–629

Joussaume S, Braconnot P (1997) Sensitivity of paleoclimate sim-ulation results to the definition of the season. J Geophys Res102:1943–1956

Joussaume S, Taylor KE(1995) Status of the paleoclimate modelingintercomparison project (PMIP). In: Gates WL (ed) Proceed-ings of the first international AMIP scientific conference,Monterey, pp 425–430

Joussaume S et al (1999) Monsoon changes for 6000 years ago:Results of 18 simulations from the paleoclimate modelingintercomparison project (PMIP). Geophys Res Lett 26:859–862

Jouzel J et al (2003) Magnitude of the isotope/temperature scalingfor interpretation of central Antarctic ice cores. J Geophys Res108:1029–1046

Knorr W et al (2005) Long-term sensitivity of soil carbon turnoverto warming. Nature 433:298–301

Krabill W et al (2004) Greenland ice sheet: increased coastalthinning. Geophys Res Lett 31:L24402

Krinner G, Genthon C (1998) GCM simulations of the Last Gla-cial Maximum surface climate of Greenland and Antarctica.Clim Dyn 14:741–758

Krinner G, Genthon C (1999) Altitude dependence of the ice sheetsurface climate. Geophys Res Lett 26:2227–2230

Krinner G et al (1997) GCM analysis of local influences on ice cored signals. Geophys Res Lett 24:2825–2828

Krinner G, Werner M (2003) Impact of precipitation seasonalitychanges on isotopic signals in polar ice cores. Earth Planet SciLett 216:525–538

Lea DW (2004) The 100,000-Yr cycle in tropical SST, greenhouseforcing, and climate sensitivity. J Clim 17:2170–2179

Lorius C et al (1990) Greenhouse warming, climate sensitivity andice core data. Nature 347:139–145

Lorius C et al (1969) Variation in the mean deuterium content ofprecipitations in Antarctica. J Geophys Res 74:7027–7031

Manabe S, Broccoli A (1985) A comparison of climate modelsensitivity with data from the last glacial maximum. J AtmosSci 42:2643–2651

Masson-Delmotte V et al (2005a) Deuterium excess reveals mil-lennial and orbital scale fluctuations of Greenland moistureorigin. Science 309:118–121

Masson-Delmotte V et al (2005b) Holocene climatic changes inGreenland: different deuterium excess signals at GRIP andNorthGRIP. J Geophys Res 110, Art. No. D14102

Masson V et al (1999) Mid-Holocene climate in Europe: what canwe infer from PMIP model-data comparisons? Clim Dyn15:163–182

Masson V et al (1998) Impact of parameterizations on simulatedwinter mid-Holocene and Last Glacial Maximum climaticchanges in the Northern Hemisphere. J Geophys Res 103:8935–8943

Masson V et al (2000), Holocene climate variability in Antarcticabased on 11 ice cores isotopic records. Quaternary Res 54:348–358

Moritz RE, et al. (2002) Dynamics of recent climate change in theArctic. Science 297:1497–1502

NorthGRIP-community-members (2004) High resolution climaterecord of the northern hemisphere reaching into last interglacialperiod. Nature 431:147–151

Ohmura A, Reeh N (1991) New precipitation and accumulationmaps for Greenland. J Glaciol 37:140–148

Parrenin F et al (2001) Dating of the Vostok ice core by an inversemethod. J Geophys Res 106:31837–31851

Peltier WR (1994) Ice age paleotopography. Science 265:195–201Peltier WR (2004) Global glacial isostasy and the surface of the ice-

age Earth: the ICE-5G (VM2) model and GRACE. Ann RevEarth Planet Sci 32:111–149

Petit JR et al. (1999) Climate and atmospheric history of the past420000 years from the Vostok Ice Core, Antarctica. Nature399:429–436

Polyakov IV et al (2002) Observationally based assessment of polaramplification of global warming. Geophys Res Lett 29(18):1878

Poutou E (2003) Etude numerique du role des interactions entre lasurface et l’atmosphere dans le cadre d’un changement clima-tique aux hautes latitudes nord, Grenoble, p 318

Raynaud D et al (1997) Air content along the GRIP core: a recordof surface climatic parameters and elevation in Central Gree-land. J Geophys Res 102:26607–26613

Renssen H et al (2005a) Simulating the Holocene climate evolutionat northern high latitudes using a coupled atmosphere-sea ice-ocean-vegetation model. Clim Dyn 24:23–43

Renssen H et al (2005b) The Holocene climate evolution in the highlatitude southern hemisphere simulated by a coupled atmo-sphere sea ice ocean vegetation model. Holocene (in press)


Rignot E, Thomas RH (2002) Mass balance of polar ice sheets.Science 297:1502–1506

Rind D et al (1997) The role of sea-ice in 2·CO2 climate modelsensitivity. 2. Hemispheric dependencies. Geophys Res Lett24:1491–1494

Ritz C et al (2001) Modeling the evolution of Antarctic ice sheetover the last 420000 years: implications for altitude changes inthe Vostok region. J Geophys Res 106(D23):31943–31964

Steig E et al (1998) Synchronous climate changes in Antarctica andthe North Atlantic. Science 282:92–95

Stenni B et al (2003) A high resolution site and source late glacialtemperature record derived from the EPICA Dome C isotoperecords (East Antarctica). EPSL 217:183–195

Stenni B et al (2001) An oceanic cold reversal during the lastdeglaciation. Science 293:2074–2077

Tarasov L, Peltier WR (2003) Greenland glacial history, boreholecontraints and Eemian extent. J Geophys Res 108. DOI 101029/2001JB001731

Vaughan DG et al (2003) Recent rapid regional climate warmingon the Antarctic Peninsula. Clim Change 60:243–274

Vavrus S, Harrison SP (2003) The impact of sea ice dynamics onthe Arctic climate system. Clim Dyn 20:741–757

Watanabe et al (2003) A first comparison of the Dome Fuji andVostok isotopic climate records. Memoirs of the NationalInstitute for Polar Research, Tokyo, Japan

Werner M et al (2001) Isotopic composition and origin of polarprecipitation in present and glacial climate simulations. Tellus53B:53–71

Werner M, Tegen I, Harrison SP, Kohfeld KE, Colin IC, BalkanskiY, Rodhe H, Roelandt C (2002) Seasonal and interannualvariability of the mineral dust cycle under present and glacialclimate conditions. J Geophys Res 107. DOI10.1029/2002JD002365

Werner M et al (2000) Borehole versus isotope temperatures onGreenland: Seasonality does matter. Geophys Res Lett 27:723–726

Wohlfahrt J et al (2004) Synergetic feedbacks between ocean andvegetation on mid- and highlatitude climates during the mid-Holocene. Clim Dyn 22:223–238


past and future polar ampliﬁcation of climate change ...changes have been used to estimate...

Documents