ARTICLE IN PRESS
Quaternary Science Reviews 25 (2006) 2082–2102
Last Glacial Maximum temperatures over the North Atlantic, Europe and western Siberia: a comparison between PMIP models, MARGO sea–surface temperatures and pollen-based reconstructions M. Kageyamaa,, A. Laıˆ ne´a, A. Abe-Ouchib, P. Braconnota, E. Cortijoa, M. Crucifixc, A. de Vernald, J. Guiote, C.D. Hewittc, A. Kitohf, M. Kucerag, O. Martia, R. Ohgaitoh, B. Otto-Bliesneri, W.R. Peltierj, A. Rosell-Mele´k, G. Vettorettij, S.L. Weberl, Y. Yum, MARGO Project Members a
LSCE/IPSL, CE Saclay, L’Orme des Merisiers, Baˆtiment 701, 91191 Gif-sur-Yvette Cedex, France b Center for Climate System Research, The University of Tokyo Kashiwa, 277-8568, Japan c Hadley Centre for Climate Prediction and Research, Met Office, FitzRoy Road, Exeter EX1 3PB, Devon, UK d GEOTOP, Universite´ du Que´bec Montre´al, C.P. 8888, succursale Centre Ville, Montre´al, Que´., Canada H3C 3P8 e CEREGE, UMR CNRS/Universite´ Paul Ce´zanne 6635, BP 80, F13545 Aix-en-Provence cedex 4, France f Climate Research Department, Meteorological Research Institute, 1-1 Nagamine, Tsukuba, Ibaraki 305-0052, Japan g Institut fu¨r Geowissenschaften, Eberhard-Karls Universita¨t Tu¨bingen, Sigwartstrasse 10, DE-72076 Tu¨bingen, Germany h Frontier Research Center for Global Change (FRCGC), JAMSTEC, Yokohama City 236-0001, Japan i Climate Change Research, National Center for Atmospheric Research, 1850 Table Mesa Drive/P.O. Box 3000, Boulder, CO 80307, USA j Department of Physics, University of Toronto, 60 St. George Street, Toronto, Ont., Canada M5S 1A7 k ICREA and Institute of Environmental Science and Technology (ICTA), Autonomous University of Barcelona (UAB), 08193 Bellaterra, Catalonia, Spain l Climate Variability Research, Royal Netherlands Meteorological Institute (KNMI), P.O. Box 201, 3730 AE De Bilt, The Netherlands m LASG, Institute of Atmospheric Physics, Chinese Academy of Sciences, P.O. Box 9804, Beijing 10029, People’s Republic of China Received 1 June 2005; accepted 9 February 2006
Abstract Evaluating the ability of models to simulate climates different from the modern one is important for climate prediction. Here we present a first comparison between results from simulations of the Last Glacial Maximum climate and continental and surface ocean reconstructions for the North Atlantic, Europe and western Siberia. The simulations include prescribed sea surface temperature (SST) and slab-ocean atmospheric general circulation model runs performed within the PMIP1 project, and atmosphere–ocean fully coupled runs performed after PMIP1 and within the PMIP2 project. The surface ocean reconstructions are from the MARGO project. Continental reconstructions are based on pollen data. Over the North Atlantic, most models simulate the strengthening of the SST meridional gradient suggested by the reconstructions, but most do not reproduce the LGM–control SST anomaly at the right location, nor with the right amplitude. Over western Siberia, the model results are much improved when a new ice-sheet reconstruction (ICE-5G) is used to force the models. The main discrepancy remains for western Europe winter temperatures, for which LGM–control anomalies are significantly underestimated by all models. All models indicate that this region during the LGM experienced significantly higher interannual variability in coldest-month temperatures compared to the control runs. This increased variability could have conspired to bias the apparently extremely cold pollen-based temperature reconstructions. Crown Copyright r 2006 Published by Elsevier Ltd. All rights reserved.
Corresponding author. Tel.: +33 1 69 08 87 06.
E-mail address:
[email protected] (M. Kageyama). 0277-3791/$ - see front matter Crown Copyright r 2006 Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.quascirev.2006.02.010
ARTICLE IN PRESS M. Kageyama et al. / Quaternary Science Reviews 25 (2006) 2082–2102
1. Introduction: lessons from atmospheric general circulation model simulations of the Last Glacial Maximum climate (PMIP1) During the first phase of the Palaeoclimate Modelling Intercomparison Project (Joussaume and Taylor, 1995, 2000; Braconnot, 2004), atmospheric general circulations models (AGCMs) were used to simulate the climate of the Last Glacial Maximum (LGM, 21,000 years ago). The results were then evaluated over the continents against paleoclimatic reconstructions. The performance of the models in terms of simulating the climate over Europe and western Siberia has been mostly compared to pollenbased climatic reconstructions (Kageyama et al., 2001). This comparison pointed to severe discrepancies between model results and climatic reconstructions over two areas. In western Europe and around the Mediterranean northern coast, the models underestimated the winter cooling and the year-round drying. On the other hand, over northwestern Siberia, the models overestimated the cooling, especially in summer. The boundary conditions and experimental design chosen for the PMIP1 LGM simulations could partly explain these discrepancies:
The sea–surface temperatures (SST) were either prescribed to the CLIMAP (1981) reconstructions or computed via a slab-ocean model, under the hypothesis that the oceanic meridional fluxes were identical to their modern values. Both methods represent strong assumptions. The CLIMAP (1981) very extensive winter sea-ice over the North Atlantic is a significant constraint on the atmospheric circulation and therefore, potentially, on European temperatures. Recent reconstructions based on assemblages of planktonic foraminifera (as in CLIMAP, 1981) (e.g. Pflaumann et al., 2003), on dinoflagellate cyst assemblages (e.g. De Vernal et al., 1997) and on geochemical proxies (Barker et al., 2005; Rosell-Mele´ et al., 2004) have since suggested a less extensive sea-ice over the North Atlantic. The impact of warmer SSTs and a less extensive sea-ice has been investigated in two studies with results varying from a significantly warmer European coast (Pinot et al., 1999) to a very moderate change in the same region (Toracinta et al., 2004). The PMIP1 slab-ocean experiments generally displayed SSTs warmer than CLIMAP (1981) and a less extensive sea-ice. The simulated continental temperature over western Europe is on average warmer in those runs, compared to the prescribed SST ones, especially north of 45 N (Kageyama et al., 2001). This is not the case over central and eastern Europe, nor over western Siberia, where there is no systematic relationship between the LGM-pre-industrial anom-
2083
aly in North Atlantic SST and that of the continental temperature, neither in winter, nor in summer. This dispersion could be linked to the differences in the SSTs computed by the slab-ocean coupled to each AGCM, but also to the model-dependent reaction of the stationary waves to the SST difference (clearly illustrated in Pinot et al., 1999). This shows that the relationship between the LGM North Atlantic cooling/sea-ice expansion and the corresponding continental temperature anomalies over Europe and western Siberia is far from being completely understood and is probably quite modeldependent. The ice-sheet reconstruction used in these experiments was the Peltier (1994) ICE-4G reconstruction, in which the Fennoscandian ice-sheet extended far east over northwestern Siberia. This clearly was not consistent with pollen data being available in this region (Tarasov et al., 1999). The resolution of the PMIP1 models was, to present standards, low (48 points in longitude 36 in latitude for the coarsest resolution model) to medium (T42, equivalent to 2:8 2:8 , for the highest resolution model). Given the local character inherent to the continental paleo-records, some of the model–data discrepancies could be due to the rather low resolution of the models, especially in regions of complex topography or coastlines such as western Europe and the Mediterranean region. Pollard and Barron (2003) and Jost et al. (2005) have investigated this possibility, but found that in all but one of the models, the discrepancy between models and data remained at higher resolution. The Hadley Centre Regional Model simulated temperatures of the coldest month as cold as suggested by the reconstructions (Jost et al., 2005) for western Europe but this appeared to be related to very cold temperatures over the adjacent sea-ice rather than the better resolution of local features on the continent. The PMIP1 simulations did not include any change to the land surface properties of the ice-free areas. The effect from changing the vegetation from forests to tundra or steppe over Europe and Siberia has a cooling effect of a few degrees (Kubatzki and Claussen, 1998; Levis et al., 1999; Wyputta and McAvaney, 2001). However, in some of the PMIP1 modern climate simulations, the land cover is not the potential vegetation but the actual one, i.e., mostly, agriculture. In this case, the effect of imposing an LGM vegetation could be less. Furthermore, the impact of permafrost, which covers large fractions of the northern hemisphere mid and high latitudes at LGM, was not taken into account in the PMIP1 experiments. Its amplitude could be less than suggested by Renssen et al. (2000) for the Younger Dryas case (Poutou, 2003). The PMIP1 simulations did not include the impact of dust, which could also result in a few C cooling (Mahowald et al., 1999; Werner et al., 2002).
ARTICLE IN PRESS M. Kageyama et al. / Quaternary Science Reviews 25 (2006) 2082–2102
2084
Since the PMIP1 project completion, several fully coupled atmosphere–ocean models have been used to simulate the climate of the LGM. The first simulations have used the same ice-sheet reconstruction as in PMIP1 (Hewitt et al., 2001, 2003; Kitoh et al., 2001; Kitoh and Murakami, 2002; Kim et al., 2002, 2003; Peltier and Solheim, 2002, 2004; Shin et al., 2003a,b) but slightly different experimental setups between the different models. In 2002, the second phase of PMIP was launched (Harrison et al., 2002), with the objective of evaluating fully coupled ocean–atmosphere general circulation models which employ a common experimental methodology. The present study aims at reassessing the model performance in simulating the temperatures over Europe and western Siberia, and as such is a follow-up of Kageyama et al. (2001). Given the possible sensitivity of the simulated climate over Europe to SST/sea-ice conditions over the adjacent ocean, we also include a comparison of the model results over the North Atlantic and the Nordic Seas to the available reconstructions from the MARGO project (Kucera et al., 2005a, http://www.pangaea.de/Projects/ MARGO/). 2. Models and experimental design The present study includes the PMIP2 fully coupled atmosphere–ocean general circulation model (AOGCM)
experiments available in the PMIP2 database as of January 15, 2006 (http://www-lsce.cea.fr/pmip2/). These are the very first results from the PMIP2 project, put together in order to comply with the Third IPCC evaluation exercise. We compare these results to the PMIP1 (http:// www-lsce.cea.fr/pmip/) AGCM results and to the results from several of the post-PMIP1 AOGCM simulations forced by the Peltier (1994) ice-sheet reconstruction: CCSM1.4-UToronto (Peltier and Solheim, 2002, 2004), MRI-CGCM1 (Kitoh et al., 2001; Kitoh and Murakami, 2002) and HadCM3 (Hewitt et al., 2001, 2003). Table 1 gives an exhaustive list of the models included in our comparison. The boundary conditions for each model/project are summarised in Table 2. The main differences in the experimental design are: (1) the change from prescribed SSTs to slab-ocean (both in PMIP1) and finally fully coupled ocean, (2) the change in the radiative forcing related to the imposed concentrations in greenhouse gases, (3) the use of the Peltier (2004) ICE-5G ice-sheet reconstruction in the PMIP2 experiments for the LGM experiment. This reconstruction has been constrained by the results from the Quaternary Environment of the Eurasian North (QUEEN) (Svendsen et al., 2004) project for ice-sheet margins in the Eurasian sector. As a result, the Fennoscandian ice-sheet does not extend as far eastwards (Fig. 1). The Laurentide ice-sheet is also significantly higher than in the ICE-4G reconstruction, which could
Table 1 General circulation models compared in this study Model (Abbreviation)
Intercomparison Project
Ocean
Length of run analysed (Years)
Reference climate
CCC2.0
PMIP1
CCM1 CCSR1 ECHAM3 GENESIS1 (gen1) GENESIS2 (gen2)
PMIP1 PMIP1 PMIP1 PMIP1 PMIP1
GFDL LMCELMD4ter (lmcelmd4) LMCELMD5.3 (lmcelmd5) MRI2 UGAMP
PMIP1 PMIP1
CLIMAP (1981) Slab-ocean Slab-ocean CLIMAP (1981) CLIMAP (1981) Slab-ocean CLIMAP (1981) Slab-ocean Slab-ocean CLIMAP (1981) Slab-ocean
10 10 10 10 10 10 10 10 10 10 10
Modern Modern Modern Modern Modern Modern Modern Modern Modern Modern Modern
UKMO MRI-CGCM1
PMIP1 –
CLIMAP (1981) Slab-ocean CLIMAP (1981) Slab-ocean Slab-ocean Fully coupled
10 10 10 10 10 Control: 40 LGM: 50
Modern Modern Modern Modern Modern Modern
CCSM1.4-UToronto (CCSM1.4-UT) HadCM3 CCSM3 MIROC3.2 HadCM3M2 ECBILTCLIO IPSL-CM4-V1-MR FGOALS-1.0g
– – PMIP2 PMIP2 PMIP2 PMIP2 PMIP2 PMIP2
Fully Fully Fully Fully Fully Fully Fully Fully
150 50 100 100 30 100 100 100
Modern Pre-industrial Pre-industrial Pre-industrial Pre-industrial Pre-industrial Pre-industrial Pre-industrial
PMIP1 PMIP1 PMIP1
coupled coupled coupled coupled coupled coupled coupled coupled
ARTICLE IN PRESS M. Kageyama et al. / Quaternary Science Reviews 25 (2006) 2082–2102
2085
Table 2 Boundary conditions and experimental design for the LGM simulations analysed in the present paper Experiment
Ice-sheets
Ocean
Greenhouse gases
PMIP1fix
Peltier (1994)
CLIMAP (1981)
PMIP1cal
Peltier (1994)
Slab-ocean
Peltier and Solheim (2004)
Peltier (1994)
Fully coupled ocean
Kitoh et al. (2001) Hewitt et al. (2003) PMIP2
Peltier (1994) Peltier (1994) Peltier (2004)
Fully coupled ocean Fully coupled ocean Fully coupled ocean
ðCO2 Þ ¼ 200 ppm ð345 ppmÞ ðCH4 Þ ¼ 350 ppb ðN2 OÞ ¼ 190 ppb (Raynaud et al., 1993; Leuenberger and Siegenthaler, 1992) ðCO2 Þ ¼ 200 ppm ð355 ppmÞ ðCH4 Þ ¼ 400 ppb ð1714 ppbÞ ðN2 OÞ ¼ 275 ppb ð311 ppbÞ (Petit et al., 1999) ðCO2 Þ ¼ 200 ppm ð345 ppmÞ ðCO2 Þ ¼ 200 ppm ð280 ppmÞ ðCO2 Þ ¼ 185 ppm ð280 ppmÞ ðCH4 Þ ¼ 350 ppb ð760 ppbÞ ðN2 OÞ ¼ 200 ppb ð270 ppbÞ (Monnin et al., 2001; Dallenbach et al., 2000; Fluckiger et al., 1999)
In addition to the boundary conditions and experimental design features described in the table, the following characteristics are valid in all cases: the orbital parameters have been set to their 21 ka BP values and land surface characteristics for ice-free land are equal to their modern values. The greenhouse gases concentrations for the control run have been indicated in brackets. Note that the control corresponds either to the present climate or to the preindustrial climate.
45
45 -120
120
60
-120
60
75
75
60
-60
60
-60
0
0
0
120
500
1000
1500
2000
2500
3000
3500
Fig. 1. Last Glacial Maximum topography (in m above sea-level) and land-ice (thick black lines) limits from the Peltier (1994) ICE4G (left hand side) and Peltier (2004) ICE5G (right hand side) reconstructions.
have a remote influence on the European climate via modified stationary waves and storm-tracks (Kageyama and Valdes, 2000). There are also differences in the control simulations. Most post-PMIP1 models use modern greenhouse gases (GHG) concentrations for their control run, while all PMIP2 experiments and the post-PMIP1 HadCM3 run use pre-industrial values. Contrary to the PMIP1 models, which often differed from the IPCC models of that time (their version was older and/or resolution lower), the PMIP2 models are generally those that have been used for the most recent IPCC simulations.
3. Sea surface temperatures 3.1. Sea surface temperature reconstructions for the North Atlantic and the Nordic Seas MARGO (multiproxy approach for the reconstruction of the Glacial Ocean surface, http://www.pangaea.de/ Projects/MARGO/) is an international program launched in September 2002 with the objectives of improving the pioneering results of the CLIMAP group (climate longrange investigation, mapping, and prediction, CLIMAP, 1976, 1981) in reconstructing ocean surface conditions
ARTICLE IN PRESS M. Kageyama et al. / Quaternary Science Reviews 25 (2006) 2082–2102
2086
during the LGM. Whereas CLIMAP mainly used planktonic foraminifer data as paleoindicators of the LGM sea surface conditions, the MARGO project follows a multiproxy approach as well as the use of different transfer function techniques and better quality-assessed datasets. In addition, the CLIMAP LGM time slice was not precisely defined and in some cases probably included Heinrich events, whereas the MARGO project uses a better constrained chronostratigraphy and defines the LGM as the interval between 19 and 23 ka, which excludes Heinrich events (Mix et al., 2001). In this paper, we use the reconstructions of the North Atlantic and the Nordic Seas SSTs from various transfer function approaches based on census counts of assemblages of planktonic foraminifera (Kucera et al., 2005b), Mg/Ca ratios of planktonic foraminifera (Barker et al., 2005; Meland et al., 2005), dinoflagellate cyst assemblages (De Vernal et al., 2005) and alkenone unsaturation ratios (Rosell-Mele´ et al., 2004). The reconstruction based on census counts of assemblages of planktonic foraminifera uses different transfer function techniques (artificial neural networks, revised analog method and SIMMAX with and without distance weighting). The SSTs used in this study and the error bars associated with them in Figs. 2 and 3 consist of the average
and the standard deviation calculated from the results of the four techniques. This foraminifera-based reconstruction and the one based on census counts of dinoflagellate cysts provide summer (July–August–September; jas), winter (January–February–March; jfm) and annual glacial SSTs. The error bars for the dinocyst-based reconstructions represents the calibration method uncertainty. Although the whole annual cycle must be important in determining the relative distribution of the different species, reconstructions of SSTs might be more robust for the summer season, during which planktonic populations experience maximum development. In the present work, we have assigned the freezing value ð1:8 CÞ to all samples in regions where both dinocysts and foraminifera indicate perennial sea-ice. This essentially concerns the region off the east coast of Greenland (see Kucera et al., 2005b, figure 25; De Vernal et al., accepted for publication). The Mg/Ca temperatures are calcification temperatures of planktonic foraminifera shells whose production nowadays spikes in spring and summer at mid to high latitudes (Barker et al., 2005). The estimates for the northern North Atlantic and the Nordic Seas (Meland et al., 2005) have all been obtained from N. pachyderma sin. and calibrated using jas temperatures, since this species mainly calcifies in summer. They will therefore be represented on the model–data
Control 25 20 15 10 5 0 25 30 35 40 45 50 55 60 65 70 75 80 85 LGM
Control ccc2.0 ccm1 gen1 gen2 gfdl lmcelmd4 mri2 ugamp ukmo HadCM3 MRI-CGCM1 CCSM1.4-UT CLIMAP Climatology
0
5 0 25 30 35 40 45 50 55 60 65 70 75 80 85
foraminifera MgCa alkenone dinocyst
LGM 20
15
5
15 10
25
20
10
CCSM3 HadCM3M2 MIROC3.2 ECBILTCLIO IPSL-CM4-V1-MR FGOALS-1.0g
25 20
foraminifera MgCa alkenone dinocyst
15 10 5 0
25 30 35 40 45 50 55 60 65 70 75 80 85
25 30 35 40 45 50 55 60 65 70 75 80 85
LGM - Control 10 5 0 -5 -10 -15 -20 25 30 35 40 45 50 55 60 65 70 75 80 85 latitude (a)
LGM - Control 10 5 0 -5 -10 -15 -20 25 30 35 40 45 50 55 60 65 70 75 80 85 (b) latitude
Fig. 2. (a) Sectorial averages (35 W–20 E) of the summer (July–August–September) sea surface temperatures over the North Atlantic and the Nordic Seas, compared with sea–surface temperature reconstructions, for the PMIP1 and post-PMIP1 simulations. The Mg/Ca estimates are from Meland et al. (2005). (b) Same as (a) for the PMIP2 models, in coloured solid lines. The PMIP1 and post-PMIP1 results are indicated in grey.
ARTICLE IN PRESS M. Kageyama et al. / Quaternary Science Reviews 25 (2006) 2082–2102 Control
Control 20 15 10 5 0 25 30 35 40 45 50 55 60 65 70 75 80 85 LGM
ccc2.0 ccm1 gen1 gen2 gfdl lmcelmd4 mri2 ugamp ukmo HadCM3 MRI-CGCM1 CCSM1.4-UT CLIMAP Climatology
20
CCSM3 HadCM3M2 MIROC3.2 ECBILTCLIO IPSL-CM4-V1-MR FGOALS-1.0g
20 15 10 5 0 25 30 35 40 45 50 55 60 65 70 75 80 85
foraminifera MgCa alkenone dinocyst
LGM 20
15
15 foraminifera MgCa alkenone dinocyst
10 5
10 5 0
0 25 30 35 40 45 50 55 60 65 70 75 80 85
25 30 35 40 45 50 55 60 65 70 75 80 85
LGM - Control
LGM - Control 5
5 0
0
−5
−5
−10
−10
−15
−15 25 30 35 40 45 50 55 60 65 70 75 80 85
25 30 35 40 45 50 55 60 65 70 75 80 85
(a)
2087
latitude
(b)
latitude
Fig. 3. (a) Same as Fig. 2a but for the annual averages. The Mg/Ca estimates are from Barker et al. (2005). (b) Same as Fig. 2b but for the annual averages.
comparisons for summer in the present paper. The error bars associated with these reconstructions are based on duplicate measurements which gave a 1 sigma error of 0:08 mmol=mol (Meland, personal communication). These values have been used to determine the error bars for each temperature estimate via the formula from Meland et al. (2005). On the other hand, the five estimates for the southern North Atlantic (Barker et al., 2005) were obtained from different species of foraminifera (G. ruber and G. bulloides). The reconstructed temperatures will be considered to represent annual mean SSTs in our comparison, and not summer temperatures, because at these latitudes and for those species, the bloom does not strictly occur in summer. Barker et al. (2005) also report that these reconstructions have been mostly interpreted as annual mean SSTs. Finally, the comparison of the reconstructions obtained from core tops with modern data (top panels, Fig. 3) shows a good agreement for annual mean SSTs. The error bars on these reconstructions are based on the same considerations as those from Meland et al. (2005). The alkenone proxy is based on the measurement of the relative abundance of long-chain alkenones (Brassell et al., 1986). These molecules are produced by a few, but ubiquitous, species of marine algae. To convert data into temperature values, two alternative empirical calibration
approaches have been used. The first one was based on 0 measuring the UK 37 ratio in laboratory cultures (Prahl et al., 1988). This calibration was later confirmed from a second calibration based on marine surface sediments (e.g. Mu¨ller et al., 1998), which is valid for the globe except in some coastal sites and in Arctic waters. This study showed that 0 global sedimentary values of UK 37 have a higher correlation factor with annual ocean temperatures (from World Ocean Atlas, 1998) at 0 m depth than for any other depths and 0 seasons. Therefore, SST estimates from UK 37 derived from this calibration can be interpreted as reflecting annual mean values at the ocean surface. At high latitudes, however, this is unlikely to be the case given the strong seasonal contrasts in phytoplankton production, and alkenone-SSTs may better reflect summer temperatures (Haug et al., 2005). Given this uncertainty on the season recorded by the alkenone proxy, we have chosen to represent the core-top and LGM estimates (for which the same calibration was used) on the annual and summer model–data comparison figures (Figs. 2 and 3). These are discussed in the following sections. The LGM–control anomalies, also represented on these figures, are computed from the LGM and core-top estimates. The error bars on the LGM reconstructed temperatures represent the combined uncertainties related to the analytical procedure and the calibration method.
ARTICLE IN PRESS 2088
M. Kageyama et al. / Quaternary Science Reviews 25 (2006) 2082–2102
The dinoflagellate and alkenone proxies are related to primary productivity and reflect conditions in the photic zone, i.e., in the upper part of the mixed layer (0–50 m), whereas planktonic foraminifers may inhabit different depths in the water column and some species, such as the polar taxon N. pachyderma sin. often occupy mesopelagic layers, below the pycnocline (e.g. Be´ and Tolderlund, 1971; Kohfeld et al., 1996; Simstich et al., 2003; Hillaire-Marcel et al., 2004). Since growth and calcification of N. pachyderma sin. in the glacial North Atlantic may have occurred at depths greater than 50–100 m, the transfer function and Mg/Ca temperatures based on foraminifera could represent subsurface conditions. Despite uncertainty in the precise depth of the planktonic populations, all calibrations in the MARGO project, for the different proxies, used the seasonal means of surface temperature at 10 m from World Ocean Atlas (1998), and the reconstructed temperatures thus represent the upper layer of the ocean. Further studies will investigate the simulated LGMmodern ocean temperature differences in the upper layers of the ocean, but for the present study, we have chosen to only compare the MARGO SST reconstructions with the SST simulated by the models. Those SST reconstructions from different proxies and methods are used for comparison with the results of the different models (Table 2) for the summer season (jas) and the annual mean. This is a first-step comparison which gives a broad view of the range of the model results and the paleoreconstructions (this latter dispersion being more specifically discussed in De Vernal et al., accepted), and the differences between them. We did not perform the comparison for the winter season, given the potentially less accurate reconstructions and the important impact of the air temperature over sea ice rather than of SSTs for Europe and western Siberia winter temperature (see Section 4).
3.2. Comparison to the model results 3.2.1. Comparison for the modern situation The SST pattern in the North Atlantic is at first order zonal. Figs. 2 and 3 show the sectorial averages of the simulated SSTs over the eastern part of the North Atlantic (between 35 W and 20 E) as a function of latitude, along with the SST reconstructions, for the pre-industrial runs (control, upper panels), the LGM runs (middle panels) and the anomalies between the two (LGM–control, lower panels) for the summer and annual mean, respectively. We chose to perform the zonal average over the eastern part of the North Atlantic ocean because it is the sector which is the most densely covered by proxy data, especially foraminifera (Fig. 4). The upper panels of Figs. 2 and 3 indicate that most models stand within 2–3 1C around the climatological HadISST curve (Rayner et al., 2003), especially between 25 N and 55 N. The dispersion of models increases north of 55 N. The models’ bias w.r.t. the
climatology and the inter-model differences are very similar for the summer and annual averages. The results of the slab models (dashed lines in Figs. 2a and 3a) have low spreading between each other compared with fully coupled models and generally show good consistency with the climatology (although slightly colder on average), except for GENESIS1 which simulates significantly colder temperatures south of 40 N and between 55 N and 67 N and CCM1 SSTs which are very cold north of 55 N. This general tendency to low spreading and close match to climatology can be explained by the modern oceanic meridional fluxes used in the slab models. The range of the results of the fully coupled models is larger, especially for the summer season (Figs. 2a (postPMIP1 runs) and b (PMIP2 runs), upper panels), although the warm temperatures simulated by MRI-CGCM1 and CCSM1.4-UT can be attributed to the use of modern concentrations of GHG rather than pre-industrial ones. The upper panels of Figs. 2 and 3 also show the modern SSTs at the sites included in the foraminifera and dinoflagellate datasets. These are very close to the zonally averaged climatological curve, which indicates that the location of those cores does not exhibit strong regional signals but rather are good indicators of the mean zonal temperature at the given latitudes. The alkenone-based SST and the annual reconstructions derived from Mg/Ca concentrations for the control are not the climatological value obtained at the core location (as it is the case for the foraminifera and dinoflagellate), but are a reconstruction of the ocean temperature from coretop sediments with the same calibration used for the LGM reconstruction. The comparison with the climatological curve therefore does not indicate the degree of zonal dispersion at the core locations but a degree of confidence in the SST reconstruction. For instance, this SST reconstruction yields particularly cold temperatures between 35 N and 45 N in summer (Fig. 2). The same alkenone-based reconstruction plotted in comparison with annual means obtained from other proxies and models (Fig. 3, upper panel) shows that it may indeed better represent annual mean SSTs rather than summer SSTs in this range of latitudes. Nevertheless, the alkenone-based reconstruction further north is closer to the summer climatology than to the annual-mean climatology. The season represented by this proxy seems to depend on the conditions found at different latitudes and might be difficult to predict in a very different context like the LGM case before further investigation is carried out to better understand this behaviour. 3.2.2. LGM SSTs and LGM–control SST anomalies The dispersion between the LGM runs is larger than in the CTRL ones (Figs. 2 and 3, middle panels). For the slabocean models the ranking of the results in terms of temperature is usually conserved. For example, GENESIS1, which was among the coldest slab models for the control, is even colder than the other slab models for the LGM, with sea ice forming as far south as 52 N (SSTs at
ARTICLE IN PRESS M. Kageyama et al. / Quaternary Science Reviews 25 (2006) 2082–2102
the freezing point of 1:8 C). MRI2, which was also among the coldest models between 25 N and 55 N in the control run, simulates much colder SSTs than most slab models at those latitudes under LGM conditions. In the same way, the relative ranking of coupled models in temperature is generally similar under preindustrial and LGM conditions. For instance, IPSL-CM4-V1-MR, which is one of the coldest models in the low to mid-latitudes in the control run keeps this characteristics in the glacial run. The cold temperatures simulated by CCSM3 between 45 N and 50 N is also present in both runs. At high latitudes, the relatively warm and cold behaviours of HadCM3M2 and
2089
FGOALS, respectively, are also preserved under LGM conditions. Despite this large dispersion, the models suggest a stronger meridional SST gradient north of 40 N in the LGM runs, associated with a moderate cooling in the tropics compared to present conditions, but with cold temperatures extending far south in the extra-tropics (the sea-ice edge reaches 50 N for the coldest models). This feature of a stronger SST gradient is also suggested by the foraminifera-based reconstructions between 40 and 50 N, with roughly uniform temperatures farther north, and by the Mg/Ca reconstructions, also giving a similar magnitude 80 70 60 50 40 30
CCSM1.4-UT
MRI-CGCM1
HadCM3
80 70 60 50 40 30
ccc2.0
70 60 50 40 30
ccm1
lmcelmd4
80 70 60 50 40 30
gen1
ukmo
80 70 60 50 40 30
mri2
gen2
80 70 60 50 40 30
ugamp
-80 (a)
gfdl
-15
-13
-60
-40 -11
-20 -9
0
CLIMAP
-80 -7
-5
-60
-40 -3
-20 -1
0 1
3
Fig. 4. LGM-control anomalies in the computed summer sea–surface temperatures in the PMIP1 slab-ocean and fully coupled post-PMIP1 ocean– atmosphere models. The CLIMAP (1981) reconstructions are also given for comparison. Coloured squares represent the reconstructions from the assemblages of planktonic foraminifera (Kucera et al., 2005b), on the same colour scale as the model results. The reference sea surface temperatures for the reconstructions are the 10 m temperatures from World Ocean Atlas (1998). (b) Same as Fig. 4a but for the PMIP2 simulations.
ARTICLE IN PRESS M. Kageyama et al. / Quaternary Science Reviews 25 (2006) 2082–2102
2090
80 70 60 50 40 30
CCSM3
80 70 60 50 40 30
HadCM3M2
−15
IPSL-CM4-V1-MR
80 70 60 50 40 30
MIROC3.2
(b)
ECBILTCLIO
−13
−11
−9
−7
FGOALS-1.0g
−5
−3
−1
1
3
Fig. 4. (Continued)
in the cooling for summer (Kucera et al., 2005b; Meland et al., 2005). North of 50 N, LGM reconstructions based on dinoflagellate cyst assemblages exhibit about 10 1C warmer SSTs than the foraminifera ones. At these latitudes, the latter reconstructions are mainly based on N. pachyderma sin., which dominates the assemblages and which is also used for Mg/Ca measurements. As seen in Section 3.1, the different proxies could indicate temperatures at different depths. This could explain some of the differences between the reconstructions. However, dinoflagellate assemblages also show quite uniform temperatures north of 50 N, which translates in a positive northward gradient in the temperature anomalies. At these latitudes, despite a spread of the order of 5 1C in the model anomalies and specific modeldependent behaviours, the overall behaviour of the models also consists in an SST anomaly weakening with latitude (Figs. 2 and 3, lower panels). Figs. 4a and b show the spatial distribution of the temperature anomalies for the different models for the summer season, along with the planktonic foraminifera SST reconstructions. For the slab-ocean models (Figs. 4a, lower panels), the structure of the SST anomaly is very zonal, with the largest anomaly between 50 and 70 N. Since these models do not include any change in heat transport associated with differences in ocean currents and/ or deep ocean circulation, this mainly represents a thermal response to the atmospheric cooling. When the dynamics of the ocean is taken into account, as it is the case for the fully coupled models (Fig. 4a, upper panels and Fig. 4b), the temperature anomalies are usually more regional, the strongest cooling being generally located off New Foundland and extending into the mid-Atlantic. Changes in the Gulf Stream and the Labrador currents, along with the surface wind forcing, should be further analysed in each model to better understand this feature. Also, changes in the strength of the thermohaline circulation and in the
location of the convection sites, resulting in a change in the northward ocean heat transport, have to be considered in order to understand the spatial distribution of the SST anomalies. For example, PMIP2 ocean models (Fig. 4b) continue to transport heat northward in the eastern part of the North Atlantic, explaining the small temperature differences in this region. 4. Continental temperatures over Europe and western Siberia 4.1. Temperature reconstructions based on pollen data As in Kageyama et al. (2001), we compare the model results to pollen-based temperature reconstructions. The LGM vegetation in Europe and western Siberia is very different from the present vegetation (dominated by agriculture) or the potential vegetation for the present climate (Ramankutty and Foley, 1999). It mainly consists of steppe with significant elements of tundra (steppetundra), except in Siberia where taiga is also present. This vegetation difference is consistent with a significant decrease in winter temperature and annual precipitation, especially in western Europe and on the northern Mediterranean coast. The quantification of these climatic anomalies has been performed for Europe and western Siberia by Peyron et al. (1998) and Tarasov et al. (1999), who used a neural network technique coupled with a plant functional type approach. The reconstructions for Europe and the northern Mediterranean coast by Peyron et al. (1998) have since been updated through the use of a more extensive calibration dataset (Peyron et al., 2005). The new reconstructions of the temperature of the coldest month are included in the confidence intervals of the previous reconstructions, but show a slightly less intense cooling over western Europe (compare figure 7 with figure 2 from
ARTICLE IN PRESS M. Kageyama et al. / Quaternary Science Reviews 25 (2006) 2082–2102
Kageyama et al. (2001) for the temperature of the coldest month). The temperature of the warmest month, which had only been reconstructed for central and eastern Europe and Siberia (Tarasov et al., 1999), has also been assessed for western Europe and the Mediterranean coast. Both the temperatures of the coldest month and of the warmest month have been used in the high resolution model–data comparison by Jost et al. (2005). The reconstructions (coloured squares and diamonds in Figs. 5 and 6 for the exact values and associated confidence intervals) show that the LGM temperature of the warmest month is around 10 1C cooler than the modern value for the European sector. The cooling (between 5 and 0 1C) is smaller for western Siberia. For the temperature of the coldest month (Fig. 7), there is a strong gradient in the LGM–control climate anomalies, from the western sector, where they reach 20 C, to western Siberia, where they remain at around 10 C. These low values are consistent, at least for western Europe, with the climatic reconstruction based on beetles (Guiot et al., 1993) and those based on permafrost limits (Renssen and Vandenberghe, 2003). Indeed these authors showed that permafrost reached the latitude 50 N in its continuous form and 45 N in its discontinuous form. Using the temperature limits they give for these types of permafrosts, the absolute annual temperature should have been between 4 and 8 C and the coldest month temperature between 20 and 25 C. In southern Italy, Allen et al. (1999) evaluated the winter cooling to 20 C (during the LGM) up to 12 C during the warmer episodes of the glacial times. Furthermore, Guiot et al. (2000) have used a process-based vegetation model (BIOME3) in an inverse mode to reconstruct from pollen data the most probable climate under precipitation seasonality change and under lowered CO2 atmospheric concentration. Applied to LGM pollen data from Greece and Italy, they showed that winter was ca 15–20 1C colder than the present, in agreement with the above climate reconstruction, and that CO2 effects do not bias winter climate reconstructions in southern Europe. For summer, Guiot et al. (1999) showed that, in Europe as well as in Siberia, CO2 does not appear to be responsible of any bias in the above reconstructions. 4.2. Model–data comparison The new temperature reconstructions used in Jost et al. (2005) are reconstructions of the temperatures of the coldest and the warmest months. We have therefore computed these variables from the model results by first computing the mean annual cycle and then taking the warmest/coldest month from this climatology. This is consistent with the calibration dataset used for the reconstructions from pollen assemblages. The warmest month for the modern climate simulations is July on the European continent, except near the Atlantic ocean and the Mediterranean Sea where it can be August (not shown). The size of this zone of oceanic influence is model
2091
dependent, it is very small in HadCM3M2, CCSM3, MIROC3.2, IPSL-CM4-MR and FGOALS-1.0g but spreads out over most of western Europe and the Mediterranean area in CCSM1.4-UT. In the LGM simulations, most models simulate their temperature maxima over Europe in July as well. However, the warmest LGM month simulated by MIROC3.2 over present-day Northern France, Germany and Poland is August, and the zone in which CCSM1.4-UT simulates its warmest temperatures in August is reduced to France and Spain. Therefore, although most models simulate a temperature maximum over Europe in July, there can be deviations from this situation according to the location (in particular near the coasts) and the time period (modern climate vs LGM). As far as the thermal minimum is concerned, most models simulate it in January over the continent, but for large zones, which are dependent on the model and the time period, the minimum is simulated either in February (especially in the modern climate simulations) or in December (especially in the LGM simulations). Hence, to get the largest possible consistency with the calibration of the pollen reconstructions, it is necessary to compute the actual temperature of the warmest/coldest month rather than assuming they are the months of July/January, respectively. 4.2.1. Temperature of the warmest month The analysis of the PMIP1 simulations (Kageyama et al., 2001) showed that the main factor driving the LGM– control differences in the temperature of the warmest month over Europe and western Siberia was the presence of the Fennoscandian ice-sheet. The largest anomalies, in the prescribed SST as well as in the slab ocean experiments, were indeed located over the ice-sheet. The second factor would be the change in SSTs over the North Atlantic and the Mediterranean Sea, but the relationship between these SSTs and continental temperature is not straightforward. For instance, the slab-ocean simulation that yields the largest LGM-modern SST anomaly in the mid-latitudes is GENESIS1 (Figs. 4 and 2, lower panel). This model is far from simulating the coldest anomaly in the temperature of the warmest month in the adjacent western Europe sector (Fig. 6, upper panel), nor over the other two sectors, except over the ice-sheet east of 15 E. Furthermore, most slabocean AGCMs simulated a North Atlantic SST anomaly smaller than the CLIMAP (1981) one, but the entanglement of the simulated temperatures of the warmest month anomalies for the fixed SST and slab-ocean simulations shows how difficult it is to relate SSTs with continental temperatures, even for regions close to the North Atlantic ocean such as western Europe. Fig. 5 (left column) shows the anomaly in the temperature of the warmest month simulated by the fully coupled models over the North Atlantic, Nordic Seas, Europe and western Siberia, along with the corresponding continental reconstructions. The coldest anomalies (a cooling larger than 18 1C) are reached over the Fennoscandian
ARTICLE IN PRESS M. Kageyama et al. / Quaternary Science Reviews 25 (2006) 2082–2102
2092
80 70 60 50 40 30
CCSM3
80 70 60 50 40 30
HADCM3M2
80 70 60 50 40 30
MIROC3.2
80 70 60 50 40 30
ECBILTCLIO
80 70 60 50 40 30
IPSL_CM4_MR
80 8 70 60 6 50 40 30
FGOALS-1.0g
80 70 60 50 40 30
CCSM1.4-UT
80 8 70 60 50 40 30
HADCM3
MRI-CGCM1
−60 −30
−30
0 −24
30
60 −18
90
80 70 60 50 40 30 −60 −12
−30
0 −6
30 −3
60 0
90 3
Fig. 5. Surface air temperature over the North Atlantic, the Nordic Seas, Europe and western Siberia for the fully coupled ocean–atmosphere models forced by the Peltier (2004) ICE-5G reconstruction for the top four lines, and the Peltier (1994) ICE-4G reconstruction for the bottom three lines. Lefthand side: LGM–CTRL anomaly in the temperature of the warmest month. Right-hand side: same but for the coldest month. The minimum (left column) and maximum (right-hand side) sea-ice extents are indicated (orange line for the pre-industrial run, red line for the LGM run). Pollen-based reconstructions are shown with the same colour scale. Diamonds correspond to the Peyron et al. (1998) reconstructions, updated in Jost et al. (2005), squares correspond to the Tarasov et al. (1999) reconstructions. The CTRL for the pollen-based reconstructions is the 1960–1990 average.
ice-sheet and the North Atlantic/Nordic Seas sea-ice if present (see for instance the CCSM1.4-UToronto results). The cooling decreases south of the ice-sheet. Over Europe, for longitudes between 10 W and 50 E, the simulated
cooling amounts to 6–12 1C for most models, in broad agreement with the reconstructions. HadCM3, MRICGCM1 and ECBILTCLIO simulate a more modest cooling (3–6 1C), which is in agreement with the reconstructions
ARTICLE IN PRESS M. Kageyama et al. / Quaternary Science Reviews 25 (2006) 2082–2102
2093
temperature anomaly
temperature anomaly
temperature anomaly
Twarm, LGM-CTRL 5 0 -5 -10 -15 -20 -25 -30 30 5 0 -5 -10 -15 -20 -25 -30 30 5 0 -5 -10 -15 -20 -25 -30 30
longitudes [-10:15], latitudes [30:70]
35
40
45
50
55
60
65
70
65
70
65
70
longitudes [15:50], latitudes [30:70]
35
40
45
50
55
60
longitudes [50:90], latitudes [30:70]
35
40
45
(a)
50
55
60
Jost et al 2005 Tarasov et al 1999 ccsr1 gen2 echam3 mri2 ccc2.0 ugamp lmcelmd4 lmcelmd5 ccm1 gen2 gen1 mri2 ccc2.0 ugamp ukmo lmcelmd4 gfdl HadCM3 CCSM1.4-UT MRI-CGCM1
latitude
temperature anomaly
temperature anomaly
temperature anomaly
Twarm, LGM-CTRL
(b)
5 0 -5 -10 -15 -20 -25 -30 30 5 0 -5 -10 -15 -20 -25 -30 30 5 0 -5 -10 -15 -20 -25 -30 30
longitudes [-10:15], latitudes [30:70]
35
40
45
50
55
60
65
70
65
70
65
70
Jost et al 2005 Tarasov et al 1999 HadCM3M2 MIROC3.2 FGOALS-1.0g IPSL-CM4-V1-MR CCSM3 ECBILTCLIO
longitudes [15:50], latitudes [30:70]
35
40
45
50
55
60
longitudes [50:90], latitudes [30:70]
35
40
45
50
55
60
latitude
Fig. 6. (a) Sectorial averages (land points only) of the simulated temperature of the warmest month for three sectors: (10 W–15 E) (top), (15 E–50 E) (middle), (50 E–90 E) (bottom), along for the temperature reconstructions for each sectors (dots with confidence interval). Dotted lines: PMIP1 prescribed SST simulations. Dashed lines: PMIP1 slab-ocean simulations. Dot-dashed lines: fully coupled models forced by the Peltier (1994) ICE-4G reconstruction. (b) Same as a for the PMIP2 models (solid coloured lines). PMIP1 and post-PMIP1 model results are indicated in grey.
for two data points in France. As noticed from the previous analysis of the PMIP1 fixed SST and slab-ocean simulations, it is difficult to establish a relationship between the temperature of the warmest month over the ocean and over the adjacent continent. For instance, MIROC3.2 simulates a 6–12 1C cooling over southwestern Europe but an oceanic cooling over the adjacent North Atlantic between 3 and 6 1C. On the other hand, MRI-CGCM1 simulates a
moderate cooling over southwestern Europe (3–6 1C) but a larger one (between 6 and 12 1C) over the whole North Atlantic. This lack of correlation is further discussed in Section 5.1. The anomalies simulated by the models over western Siberia are very sensitive to the ice-sheet reconstruction used to force the model (Fig. 1): CCSM1.4-UT, HadCM3 and MRI-CGCM1 (lower three maps) use the Peltier
ARTICLE IN PRESS M. Kageyama et al. / Quaternary Science Reviews 25 (2006) 2082–2102
2094
temperature anomaly
temperature anomaly
temperature anomaly
Tcold, LGM-CTRL longitudes [-10:15], latitudes [30:70]
0 -10 -20 -30 -40 30
35
40
45
50
55
60
65
70
65
70
65
70
longitudes [15:50], latitudes [30:70]
0 -10 -20 -30 -40 30
35
40
45
50
55
60
longitudes [50:90], latitudes [30:70]
0 -10
Jost et al 2005 Tarasov et al 1999 ccsr1 gen2 echam3 mri2 ccc2.0 ugamp lmcelmd4 lmcelmd5 ccm1 gen2 gen1 mri2 ccc2.0 ugamp ukmo lmcelmd4 gfdl HadCM3 CCSM1.4-UT MRI-CGCM1
-20 -30 -40 30
35
40
45
(a)
50
55
60
latitude
temperature anomaly
temperature anomaly
temperature anomaly
Tcold, LGM-CTRL
(b)
longitudes [-10:15], latitudes [30:70]
0 -10 -20 -30 -40 30
35
40
45
50
55
60
65
70
65
70
65
70
Jost et al 2005 Tarasov et al 1999 HadCM3M2 MIROC3.2 FGOALS-1.0g IPSL-CM4-V1-MR CCSM3 ECBILTCLIO
longitudes [15:50], latitudes [30:70]
0 -10 -20 -30 -40 30
35
40
45
50
55
60
longitudes [50:90], latitudes [30:70]
0 -10 -20 -30 -40 30
35
40
45
50
55
60
latitude
Fig. 7. (a) Same as Fig. 6a but for the temperature of the coldest month. (b) Same as Fig. 6b but for the temperature of the coldest month.
(1994) ICE-4G reconstruction, which extends far east over western Siberia. Of these three models, only HadCM3 simulates a weaker cooling over western Siberia than over western Europe, which is the broad gradient indicated by the reconstructions. To the extreme north of the sector, the ice-sheet imposes very cold temperatures while the reconstructions indicate a very mild change. In the six simulations forced by the new Peltier (2004) ICE-5G reconstruction, all models except FGOALS-1.0g simulate weaker cooling over western Siberia than over western
Europe. Two of the models even simulate a slight warming over northwestern Siberia. To account for the model dispersion and the confidence interval of the reconstructions, we have represented sectorial averages for the same sectors as in Kageyama et al. (2001), i.e. western Europe (10 W–15 E), central and eastern Europe (15–50 E) and western Siberia (50–90 E). All PMIP1, 2 and additional fully coupled model results are represented in Fig. 6. The dispersion of the fully coupled models is slightly larger than that of the PMIP1
ARTICLE IN PRESS M. Kageyama et al. / Quaternary Science Reviews 25 (2006) 2082–2102
simulations, over the western sector, as is the case for the SST anomalies (Fig. 2, lower panel). ECBILTCLIO stands on the ‘‘warm’’ side, its simulated anomaly standing between 2 and 5 C outside the ice-covered area. CCSM1.4-UT stands on the ‘‘cold’’ side, especially between 40 and 55 N, with temperature anomalies reaching 15 C. This dispersion of the model results is in fact comparable to the confidence interval of the reconstructions. Most models are in broad agreement with the reconstructions over the western European sector. For the central and eastern sector (Fig. 6, middle and lower panels), the dispersion of the fully coupled model results is slightly smaller (although comparable) to that of the PMIP1 models, indicating that the source of intermodel differences is more related to differences in the atmosphere or the land-surface treatment in the models than to the differences in SSTs. For the central and eastern Europe sector (Fig. 6, middle panel), apart from a few points between 55 N and 60 N, model results are in good agreement with reconstructions. The discrepancy might not be as large as it appears on this figure because these three points are located near the eastern edge of the sector, where temperature anomalies would be smaller than the average over the sector, as shown in Fig. 5 for the fully coupled models. This is especially the case for the models forced by the ICE-5G reconstruction. The impact of the ICE-5G reconstruction used in the PMIP2 simulation is very clear on the lower panel of Fig. 6, which shows the model–data comparison for western Siberia. South of 60 N, apart from the CCM1 which simulates too cold anomalies, model results compare well with the reconstructions. North of 60 N, the only agreement is reached by five of the six PMIP2 models, which used the ICE-5G reconstructions. All other models simulated temperature anomalies 5–20 1C too cold compared to the reconstructions. The hypothesis for model– data discrepancy that was proposed in our PMIP1 study (Kageyama et al., 2001, and Section 1), i.e. that the simulated temperatures of the warmest month over western Siberia were too cold because of the imposed ice sheet in this region, is therefore confirmed through the simulations forced by the ICE-5G reconstructions. 4.2.2. Temperature of the coldest month The reconstructions of the anomalies of the temperatures of the coldest month show a large SW–NE gradient, from western Europe where they stand around 20 C, to western Siberia where they equal ca 10 C. The results from the PMIP1 model–data comparison showed that the models were not able to simulate such a gradient in the temperature anomalies, mainly because they were not able to simulate the large cooling depicted by the reconstructions for western Europe and the Mediterranean area. The updated reconstructions over this area result in ca 5 1C smaller anomalies, but they still suggest a significant SW– NE gradient, as shown in Figs. 5 and 7. The results from the PMIP1 runs analysed in Kageyama et al. (2001)
2095
showed that the largest cooling of the temperature of the coldest month was associated with extensive sea-ice and by the ice-sheets. The influence of the sea-ice is clearly shown by the fixed SST experiments, which were forced by the extensive CLIMAP (1981) sea-ice for the LGM runs. In the slab-ocean experiments, the cooling was smaller, and the relative influence of the ice-sheet larger (Kageyama et al., 2001, Figure 6). Fig. 5 (right column) shows the simulated anomalies of the temperature of the coldest month for the fully coupled models. These anomalies can easily be compared to the simulated anomalies of the temperatures of the warmest month (left column). Values of the pollen-based reconstructions are represented with the same colour scale. For all but one of the models, the simulated cooling in the temperature of the coldest month is much larger than the cooling of the temperature of the warmest month, as suggested by reconstructions. Also, most models simulate a cooling over western Siberia weaker than over western Europe, as suggested by the reconstructions. This gradient in the simulated cooling is stronger in the PMIP2 simulations forced by the Peltier (2004) ICE-5G reconstruction. This could result from the direct influence of the different location of the ice-sheet, but also from the influence of the Fennoscandian topography on the atmospheric general circulation. The remote impact of a much higher Laurentide ice-sheet in ICE-5G compared to ICE4G cannot be excluded either, the European climate being sensitive to the state of the Laurentide ice-sheet via the influence of stationary waves and storm-tracks (Kageyama and Valdes, 2000). Over the western Europe sector (Fig. 7, upper panel), the anomalies in the temperature of the coldest month simulated by the fixed SST models, forced by the CLIMAP (1981) SST and sea-ice reconstruction, are colder than those simulated by the slab-ocean models, which simulated SST anomalies significantly warmer than CLIMAP (1981), except for the GENESIS1 model. This is not the case in the central Europe and western Siberian sectors, for which there is no systematic grouping of the models according to the type of PMIP1 simulation. This impact of the North Atlantic sea-ice/SST anomalies is also clear in the comparison of the coupled simulation results. CCSM1.4UT simulates sea-ice as far south as 35 S, which is further south than the CLIMAP (1981) reconstruction. The associated anomalies of the temperatures of the coldest month are extremely cold, over western Europe as well as further East, all the way to 90 E. CCSM3 also simulates quite an extensive sea-ice maximum for the LGM, although not as large as CCSM1.4-UT. This appears to be associated with a smaller anomaly of the temperature of the coldest month over the northern Mediterranean coast. Furthermore, the anomalies on the western Siberia sector are also much smaller, which results in the pattern of this anomaly being much less zonal than in CCSM1.4-UT. The example of ECBILTCLIO or HadCM3 would further prove this relationship between the presence of extensive
ARTICLE IN PRESS 2096
M. Kageyama et al. / Quaternary Science Reviews 25 (2006) 2082–2102
sea-ice in the North Atlantic and the amplitude of the difference in the temperature of the coldest month, at least over western Europe. On the other hand, the case of the HadCM3M2 model, in comparison with the CCM3 results, shows that a model can simulate similar anomalies in the temperature of the coldest month over land, while yielding a much less extensive maximum sea-ice extent and a smaller winter SST anomaly over the North Atlantic (not shown). Similarly, MIROC3.2 and ECBILTCLIO stand on the warm side of the model results as far as SST anomalies are concerned (this can be seen in Fig. 3), but MIROC3.2 simulates colder temperatures than ECBILTCLIO (for instance) over land, especially over the ice-sheet. The difference between the models could also be linked to the land-surface schemes and their resolution, particularly over the land-ice and topographic features. For instance, ECBILTCLIO, being run at a much lower resolution than MIROC3.2, does not ‘‘see’’ the Fennoscandian ice-sheet as well as the latter model, which could explain part of the temperature anomaly difference between those models. Figs. 5 and 7 show that among all the models (PMIP1 and fully coupled results are shown in Fig. 7), only one (CCSM1.4-UT) can simulate temperature anomalies as cold as suggested by the reconstructions for western Europe and the Mediterranean area. However, this model overestimates the cooling over western Siberia by more than 20 1C and overestimates the SST cooling over the North Atlantic, at least on an annual average basis (Fig. 3, lower panel). The simulated sea-ice over the North Atlantic reaches 35 N in winter, 10 to the south of the CLIMAP (1981) maximum extension limit. None of the MARGO reconstructions suggests such cold conditions, since they yield conditions warmer than the initial CLIMAP (1981) reconstructions. In fact, in this model, the very extensive sea-ice and its very zonal limit over the North Atlantic ocean are probably related to a strong zonal atmospheric circulation, as in the fixed SST PMIP1 experiments forced by the CLIMAP (1981) SST and sea-ice (Kageyama et al., 1999). On the other hand, the other models show a much better agreement on western Siberia, but underestimate the cooling over western Europe by ca 10 1C and over some parts of the Mediterranean area by a few C. Most of these models simulate North Atlantic SST anomalies that are compatible with reconstructions for the annual mean. Thus, from the models’ point of view, it seems difficult to reconcile the reconstructed SSTs over the North Atlantic, very cold anomalies in the temperatures of the coldest month over the western Europe and Mediterranean sectors and the smaller anomalies in the temperature of the coldest month further to the North and the East, over Central Europe and western Siberia. These results show the diversity of the model responses, in terms of the temperature of the coldest month, to LGM boundary conditions. Shin et al. (2003a) placed their fully coupled model (CCSM1.4 forced by Peltier (1994)’s ICE4G reconstructions) results w.r.t. the PMIP1 models and found that their model was in the range of these PMIP1
models. This is not true for all fully coupled models: CCSM1.4-UT simulates much colder temperatures anomalies than any of the other models, while MRI-CGCM1 simulates smaller anomalies than most PMIP1 models. Thus, as expected, the fully coupled model results show a wider dispersion than the PMIP1 prescribed SST or slabocean results, which was also the case for the annual mean SST anomalies (Fig. 3). 5. Summary and discussion 5.1. Relationship between regional and global changes The LGM represents a large anomaly with respect to the control climate not only for the North Atlantic, European and western Siberian regions, but also at a global scale. Table 3 gives the global annual averages of the 2-m air temperature for the control climate, the LGM climate and the anomaly between LGM and CTRL for each model
Table 3 Global annual mean temperature for all models analysed in this study, for the control climate (the asterisk indicates models for which the preindustrial one is the control climate, all others use the modern climate), for the LGM climate and the anomaly between the LGM and the control climates Database
Model
T control ð CÞ
T LGM ð CÞ
DT ð CÞ
PMIP1fix PMIP1fix PMIP1fix PMIP1fix PMIP1fix PMIP1fix PMIP1fix PMIP1fix
CCSR1 GEN2 ECHAM3 MRI2 CCC2.0 UGAMP lmcelmd4 lmcelmd5
14.39 13.53 14.34 15.02 13.35 13.51 15.15 15.37
9.64 9.33 10.05 11.44 9.21 9.82 11.90 11.98
4.75 4.20 4.29 3.59 4.13 3.69 3.26 3.39
PMIP1cal PMIP1cal PMIP1cal PMIP1cal PMIP1cal PMIP1cal PMIP1cal PMIP1cal PMIP1cal
CCM1 gen2 gen1 MRI2 CCC2.0 UGAMP UKMO lmcelmd4 GFDL
12.15 14.15 13.05 12.49 14.71 13.71 13.12 13.09 12.79
5.79 10.06 7.60 7.83 8.45 9.79 8.21 11.24 8.79
6.36 4.09 5.45 4.66 6.27 3.92 4.91 1.85 4.00
post-PMIP1* post-PMIP1 post-PMIP1
HADCM3 CCSM1.4-UT MRI-CGCM1
12.50 15.53 15.50
8.66 6.37 11.56
3.85 9.17 3.94
PMIP2* PMIP2* PMIP2* PMIP2* PMIP2* PMIP2*
HadCM3M2 MIROC3.2 FGOALS-1.0g IPSL-CM4-V1-MR CCSM3 ECBILTCLIO
13.08 12.40 11.72 11.76 11.89 13.11
7.84 7.88 6.26 8.11 7.46 9.71
5.24 4.53 5.46 3.65 4.43 3.40
PMIP1fix indicates models run in the PMIP1 project and using CLIMAP (1981) boundary conditions. PMIP1cal indicates models run in the PMIP1 models coupled to a slob ocean. Post-PMIP1 indicates coupled models run with PMIP1 boundary conditions. PMIP2 indicates models run in the PMIP2 project.
ARTICLE IN PRESS M. Kageyama et al. / Quaternary Science Reviews 25 (2006) 2082–2102
Tcold vs. Atl Tcold
0.8
0.8
0.6
0.6 correlation
correlation
Tcold vs. Global annual mean
0.4 0.2 0
0.4 0.2 0
Atl1Cold
Atl2Cold
WEurCold
CEEurCold
WSibCold
WEurCold
Twarm vs. Global annual mean
CEEurCold
WSibCold
Twarm vs. Atl Twarm 0.6
0.8
0.4 correlation
0.6 correlation
2097
0.4
0.2
0.2
0
0 Atl1Warm
Atl2Warm
WEurWarm CEEurWarm WSibWarm
WEurWarm
CEEurWarm
WSibWarm
Fig. 8. (Top left) Correlations between regional averages (regions indicated on the x-axis) of the 2-m air temperature of the coldest month and the global mean annual 2-m air temperature. (Bottom left) Same as above for the temperatures of the warmest month. (Top right) Correlations between regional averages of the temperature of the coldest month over the European continent and the average over the Atlantic region (35–70 N; 35 W–20 E). (Bottom right) Same as top right except for temperatures of the warmest month. Abbreviations: Atl1: (35–50 N; 35 W–20 E); Atl2: (35–70 N; 35 W–20 E); WEur: (35–50 N; 15 W–10 E); CEEEur: (35–70 N; 15–50 E); WSib: (35–70 N; 60–90 E).
analysed in the present study. Global annual mean temperatures range between 12.15 and 15.53 1C for those models whose control climate is the modern one, and between 11.72 and 13.11 1C for models using the preindustrial climate as their control runs. The anomalies between the LGM and control global annual mean temperatures range between 9:17 and 1:85 C and there is no relationship between the value of these anomalies and the type of control climate (modern or pre-industrial). Masson-Delmotte et al. (2005) show that there is a significant relationship between temperature changes at the poles and the global mean annual temperature changes. Here, to put the LGM climate anomalies over the North Atlantic and Europe into a global context, we examine the relationship between the temperature changes over the regions on which we have focussed so far, namely the eastern North Atlantic, western Europe, central and eastern Europe and western Siberia, for winter and summer, on the one hand, and on the other hand the global annual mean temperature change. For each model, we have computed the global annual mean average LGM– CTRL anomaly (Table 3, third column) and the averages over several regions: Atl1 (air temperatures over the
Atlantic mid-latitude ocean): (35–50 N; 35 W–20 E); Atl2 (air temperatures over the Atlantic mid- and high latitude ocean): (35–70 N; 35 W–20 E); WEur (air temperature over continental western Europe): (35–50 N;15 W–10 E); CEEur (air temperature over continental central and eastern Europe): (35–70 N; 15– 50 E); WSib (air temperature over western Siberia): (35– 70 N; 60–90 E). The first two averages (Atl1 and Atl2) are computed from ocean points only, while the last three (WEur, CEEur and WSib) are computed from land points only. We therefore obtain global and regional temperature averages for the 26 models considered in this study, and can compute the correlations between regional and global results, and between averages over the Atlantic ocean and averages over the above continental regions. The main results from this study are summarised in Fig. 8. The confidence intervals on the computed correlations were obtained using a bootstrap methodology with 1000 experiments. The 5th and 95th percentiles are shown on the figure, together with the median value obtained through this calculation (open circle). The top left graph in Fig. 8 shows the correlations between the averages of LGM–CTRL anomalies of the
ARTICLE IN PRESS 2098
M. Kageyama et al. / Quaternary Science Reviews 25 (2006) 2082–2102
temperature of the coldest month and the global annual mean temperature anomaly for the five regions defined above. There is no significant correlation between the temperature changes over the North Atlantic and the global temperature changes. The correlation is stronger over the continent, but the relationship between regional and global averages in this case is very loose, as shown by the large confidence interval and a fifth percentile lower than 0.5. Fig. 8, bottom left, shows that there is no significant relationship between the anomaly of temperature of the warmest month over the North Atlantic and the global annual mean anomaly. Again, the relationship is stronger over the continent, particularly for western Europe. Therefore, winter and summer temperature changes over the North Atlantic, Europe and western Siberia do not relate very well to global temperature changes. We have also examined the relationship between LGM– CTRL anomalies over western Europe, Central and Eastern Europe and Western Siberia on the one hand, and the changes in North Atlantic temperatures for the same season on the other. The results are the same whether the Atl1 or Atl2 regions are considered. Fig. 8 (top right for the temperatures of the coldest month and bottom right for the temperatures of the warmest month) therefore only show the correlations between the continental temperature anomalies and the Atl2 anomaly. The only significant relationship between the continental anomalies and the North Atlantic anomalies is for the temperatures of the coldest month and the western European region. In all other cases there is no significant correlation between continental and oceanic changes, the fifth percentile of the distribution obtained using the bootstrap method being very close or sometimes lower than zero. This shows that on the whole, the North Atlantic, Europe and western Siberia can behave quite differently from the global mean average, when the responses of different models are taken into account. For the North Atlantic temperature, this could be due to the different behaviours of the oceanic circulation under global climate changes. The relationship between European and western Siberian temperature changes and North Atlantic temperature anomalies is not straightforward either, apart from the correlation between changes in the western Europe temperatures of the coldest month and those in the North Atlantic temperatures, as could be expected from the dominant, westerly, circulation in winter. This reinforces our conclusion about the incompatibility between SST reconstructions over the North Atlantic on the one hand, and temperatures of the coldest month on the other, from the point of view of global general circulation models. 5.2. The possible influence of interannual variability One of the reasons for the model–data discrepancies over southwestern Europe and the Mediterranean area might
reside in a change in the characteristic variability in an LGM climate, compared to the present climate. Fig. 9 illustrates the change in the variability of the temperature of the coldest month in the models for which we had long series of monthly data. We have computed the temperature of the coldest month for each year and then summarised the variability of this temperature via its standard deviation. Fig. 9 (left column) shows the ratio of the LGM standard deviation of the temperature of the coldest month to its value for the CTRL simulations. The same calculation has been performed for the temperature of the warmest month (right column). In most models, the variability of the temperature of the coldest month increases over western Europe and the Mediterranean area. This appears to be partly related to sea-ice cover variability and variability of the temperature over the seaice (the sea-ice limit is indicated in Fig. 5). This amplification is largest over western Europe and the Mediterranean/Black Sea/Caspian Sea area for the variability of the temperature of the coldest month. Such an amplification is not simulated for other regions or for the variability of the temperature of the warmest month. The significant amplification over the ocean appears to be consistent with the reconstruction of sea–surface conditions that suggests highly variable sea-ice, salinity and SSTs, especially along the continental margins of northwestern and northeastern North Atlantic (De Vernal et al., 2000). The present reconstructions are calibrated w.r.t. climatological averages. The vegetation can be more sensitive to extreme events than to the climatological average whereas the reconstruction method implicitly assumes that the typical variability at one site does not change between the present and the LGM. The models clearly show that the variability in the temperature of the coldest month is significantly amplified at LGM over the regions where the model–data discrepancy is largest. It would be worth investigating the impact of this larger variability on vegetation, which could be done, for instance, with a dynamical vegetation model. The interannual/interdecadal variability over the ocean, and particularly over the sea-ice cover, could also partly explain the difference in the SST reconstructions. If a given proxy reacts to particularly warm years and other proxies to different years, the resulting reconstructions could potentially be biased to those events. 5.3. Towards a finer comparison of the ocean temperatures In the present work, we have chosen to perform a simple comparison of the sea–surface temperatures, i.e. we have chosen to compare reconstructions from the different proxies to simulated SSTs only, and not to temperatures at different depths under the surface. This focus on SSTs is justified by the fact that one of our objectives was also to examine the relationship between the temperature changes over sea w.r.t. to changes over the continents. As stated in
ARTICLE IN PRESS M. Kageyama et al. / Quaternary Science Reviews 25 (2006) 2082–2102
2099
70 60 50 40 30
CCSM3
70 60 50 40 30
HADCM3M2
70 60 50 40 30
MIROC3.2
70 60 50 40 30
ECBILTCLIO
70 60 50 40 30
IPSL_CM4_MR
70 60 50 40 30
FGOALS-1.0g
CCSM1.4-UT
−60 0.0
0.25
−30 0.5
0 0.75
30 1.0
60 1.25
70 60 50 40 30 90 −60 1.5
−30
1.75
0 2.0
30 2.25
60 2.5
90 2.75
3.0
Fig. 9. Ratio (LGM/CTRL) of the interannual variabilities (measured through the standard deviations) of the temperatures of the coldest and the warmest months.
Section 3.1, some planktonic organisms used to reconstruct SSTs can live at greater depths (below 50 m, and deeper at some locations where the mixed layer is thicker), while others stay at the very surface of the ocean. Furthermore, the depth at which they live could have been different in a context where the stratification of the ocean is modified. The presence of seasonal sea-ice and the existence of stratified upper water masses at LGM may account for a higher sensitivity and variability in sea–surface conditions in the upper water layer (Hillaire-Marcel et al., 2001; De Vernal et al., 2002). Further work on the model–data comparison will therefore examine the changes in the upper ocean compared to surface changes and the changes in the vertical stratification of the upper ocean. We have also kept the comparison simple in terms of the choice of the season (July–August–September and the annual average), which is constrained by the MARGO reconstructions available at present. However, the season during which each biological indicator lives might be
slightly different in the modern and glacial oceans. From the model results, we could examine the timing of the seasonal cycle and its possible changes to help with the data interpretation. De Vernal et al. (submitted for publication) explain the possible reasons for the differences in the reconstructions from the different proxies in more detail. One last hypothesis for reconstructed SST differences is the possible transport of the lightest proxy carriers by ocean currents. These currents and the advection of the North Atlantic water masses are quite different in LGM simulations compared to present day (e.g. Hewitt et al., 2003). Such changes could be re-examined in the different fully coupled PMIP LGM experiments. 5.4. Experimental design The PMIP2 experiments, by using fully coupled atmosphere–ocean models, imply model–data comparisons different from PMIP1. SST reconstructions are not used
ARTICLE IN PRESS 2100
M. Kageyama et al. / Quaternary Science Reviews 25 (2006) 2082–2102
to force atmospheric models anymore, but to evaluate ocean–atmosphere models. By progressing from atmospheric models to coupled atmosphere–ocean models, we avoid the difficult problem of prescribing glacial SSTs in AGCM simulations. This was one of the reasons invoked for model–data discrepancies in Section 1. Another improvement is the use of the Peltier (2004) ICE-5G icesheet reconstruction for the LGM, which, with the constraint of the QUEEN project results on the eastern limit of the Fennoscandian ice-sheet, yields a much more favourable model–data comparison over western Siberia. This was another reason for discrepancies given in Section 1. However, some of the reasons for model–data discrepancies listed in Section 1 have not been examined through these post-PMIP1 numerical experiments. In particular, the influence of vegetation, permafrost and dust are not examined here. It is planned, in the MOTIF European project, to perform coupled atmosphere–ocean–vegetation simulations of the LGM climate. This will lead to new comparisons, for the temperatures as well as for the vegetation.
western Europe, over which all but one of the models systematically underestimate the cooling depicted by the data. The only model which does simulate temperatures as cold as the reconstructions does not yield a good agreement with the continental reconstructions over Central Europe and western Siberia, nor over the North Atlantic, where it overestimates the cooling compared to the reconstructions. This suggests an inconsistency between the reconstructions over western Europe in winter on the one hand and North Atlantic SSTs and Central Europe/western Siberia temperatures of the coldest month on the other. One of the possible reasons for this discrepancy, which will have to be investigated in future work, is the significant increase in the atmospheric and oceanic variability over this region. Over the ocean as well as over land, proxies could be more sensitive to climatic extremes than to changes in the average climate, and a change in the variability (e.g. interannual variability, which is shown in the present work) indicates that the relationship between the proxy and the mean climate might have been different during the Last Glacial Maximum.
6. Conclusions
Acknowledgements
We have performed a first comparison of North Atlantic and Nordic Seas simulated and reconstructed Last Glacial Maximum sea surface temperatures on the one hand, for the summer season and the annual mean, and on the other hand a comparison between the simulated and reconstructed temperatures of the coldest and the warmest months over the Europe and western Siberia. Model results are from the initial PMIP1 experiments, using fixed SST and slab-ocean atmospheric general circulation models, from post-PMIP1 fully coupled atmosphere–ocean experiments, and from the PMIP2 coordinated fully coupled experiments. The SSTs reconstructed from different proxies, i.e. planktonic foraminifera assemblages, planktonic foraminifera Mg/Ca, dinocyst assemblages and alkenones, show significant inter-proxy differences but point to a stronger meridional gradient at LGM in the North Atlantic. This strengthening of the SST meridional gradient is generally captured by the models, although not always at the right location nor with the amplitude suggested by the reconstructions. Some models overestimate the mid-latitude cooling corresponding to this steepening of the SST meridional gradient, others underestimate it, there is no consistent difference between the models and the reconstructions. Pollen-based reconstructions show a large continental cooling over western Europe in winter. This cooling decreases eastward and is reduced in summer. The models show a good agreement with the reconstructions in summer, especially when using the Peltier (2004) ICE-5G reconstruction, very important for the western Siberian sector. There is also reasonable agreement for this region in winter. The largest model–data discrepancy remains over
This work was funded by the MOTIF European Project (EVK2-2001-00263), and stimulated by discussions during the PMIP meeting in Giens, April 2005. This meeting was funded by CNRS, CEA, PAGES and CLIVAR. The participants are thanked for the discussions around this work. Marius Meland Claire Waelbroeck Isabel Cacho, Kazuyo Tachikawa and Stephen Barker are thanked for the discussion about the model–data comparisons. Pascal Yiou is thanked for his advice on computing the correlations reported in section 5.1. Jean-Yves Peterschmitt offered great help with handling the data and producing the graphics. M. K. is funded by CNRS. A. L.’s Ph.D. thesis is funded by CEA and Meteo-France. This is LSCE contribution # 1807. References Allen, J.R.M., Brandt, U., Brauer, A., Hubberten, H.-W., Huntley, B., Keller, J., Kraml, M., Mackensen, A., Mingram, J., Negendank, J.F.W., Nowaczyk, N.R., Oberha¨nsli, H., Watts, W.A., Wulf, S., Zolitschka, B., 1999. Rapid environmental changes in southern Europe during the last glacial period. Nature 400, 740–743. Barker, S., Cacho, I., Benway, H., Tachikawa, K., 2005. Planktonic foraminiferal Mg/Ca as a proxy for past oceanic temperatures: a methodological overview and data compilation for the Last Glacial Maximum. Quaternary Science Reviews 24, 821–834. Be´, A.W.H., Tolderlund, D.S., 1971. Distribution and ecology of living planktonic foraminifera in surface waters of the Atlantic and Indian oceans. In: Funnel, B.M., Riedel, W.R. (Eds.), The Micropaleontology of Oceans. Cambridge University Press, Cambridge, pp. 105–149. Braconnot, P., 2004. Mode´liser le dernier maximum glaciaire et l’Holoce`ne moyen. C. R. Geoscience 336, 711–719. Brassell, S.C., Eglinton, G., Marlowe, I.T., Pflaumann, U., Sarnthein, M., 1986. Molecular stratigraphy: a new tool for climatic assessment. Nature 320, 129–133. CLIMAP, 1976. The surface of the ice-age earth. Science 191, 1138–1144.
ARTICLE IN PRESS M. Kageyama et al. / Quaternary Science Reviews 25 (2006) 2082–2102 CLIMAP, 1981. Seasonal Reconstructions of the Earth’s Surface at the Last Glacial Maximum. Map Chart Series MC-36. Geological Society of America, Boulder, CO. Dallenbach, A., Blunier, T., Fluckiger, J., Stauffer, B., Chappellaz, J., Raynaud, D., 2000. Changes in the atmospheric CH4 gradient between Greenland and Antarctica during the Last Glacial and the transition to the Holocene. Geophysical Research Letters 27, 1005–1008. De Vernal, A., Rochon, A., Turon, J.-L., Matthiessen, J., 1997. Organicwalled dinoflagellate cysts: palynological tracers of sea–surface conditions in middle to high latitude marine environments. Geobios 30, 905–920. De Vernal, A., Hillaire-Marcel, C., Turon, J.-L., Matthiessen, J., 2000. Reconstruction of sea–surface temperature, salinity, and sea-ice cover in the northern North Atlantic during the last glacial maximum based on dinocyst assemblages. Canadian Journal of Earth Sciences 37, 725– 750. De Vernal, A., Hillaire-Marcel, C., Peltier, W.R., Weaver, A., 2002. The structure of the upper water column in the northwest North Atlantic: modern vs. Last Glacial Maximum conditions. Paleoceanography 17, 1050. De Vernal, A., Eynaud, F., Henry, M., Hillaire-Marcel, C., Londeix, L., Mangin, S., Matthiessen, J., Marret, F., Radi, T., Rochon, A., Solignac, S., Turon, J.-L., 2005. Reconstructions of sea–surface conditions at middle to high latitudes of the Northern Hemisphere during the Last Glacial Maximum (LGM) based on dinoflagellate cyst assemblages. Quaternary Science Reviews 24, 897–924. De Vernal et al. Multiproxy reconstruction of lgm sea–surface conditions in the northern North Atlantic. Quaternary Science Reviews, accepted for publication. Fluckiger, J., Dallenbach, A., Blunier, T., Stauffer, B., Stocker, T.F., Raynaud, D., Barnola, J.-M., 1999. Variations in atmospheric N2O concentration during abrupt climatic changes. Science 285, 227–230. Guiot, J., deBeaulieu, J.-L., Cheddadi, R., David, F., Ponel, P., Reille, M., 1993. The climate in western Europe during the last glacial/interglacial cycle derived from pollen and insect remains. Palaeogeography, Palaeoclimatology, Palaeoecology 103, 73–94. Guiot, J., Torre, F., Cheddadi, R., Peyron, O., Tarasov, P., Jolly, D., Kaplan, J.O., 1999. The climate of the Mediterranean Basin and of Eurasia of the Last Glacial Maximum as reconstructed by inverse vegetation modelling and pollen data. Ecologia Mediterranea 25, 193–204. Guiot, J., Torre, F., Jolly, D., Peyron, O., Boreux, J., Cheddadi, R., 2000. Inverse vegetation modelling: a tool to reconstruct palaeoclimates under changed CO2 conditions. Ecological Modelling 127, 119–140. Harrison, S.P., Braconnot, P., Joussaume, S., Hewitt, C.D., Stouffer, R.J., 2002. Comparison of paleoclimate simulations enhances confidence in models. EOS 83, 447. Haug, G., Ganopolski, A., Sigman, D., Rosell-Mele´, A., Swann, G.E.A., Tiedemann, R., Jaccard, S., Bollmann, J., Maslin, M.A., Leng, M.J., Eglinton, G., 2005. North Pacific seasonality and the glaciation of North America 2.7 million years ago. Nature 433, 821–825. Hewitt, C.D., Broccoli, A.J., Mitchell, J.F.B., Stouffer, R.J., 2001. A coupled model study of the last glacial maximum: was part of the North Atlantic relatively warm? Geophysical Research Letters 28, 1571–1574. Hewitt, C.D., Stouffer, R.J., Broccoli, A.J., Mitchell, J.F.B., Valdes, P.J., 2003. The effect of ocean dynamics in a coupled GCM simulation of the Last Glacial Maximum. Climate Dynamics 20, 203–218. Hillaire-Marcel, C., de Vernal, A., Candon, L., Bilodeau, G., Stoner, J., 2001. Changes of potential density gradients in the northwestern North Atlantic during the last climatic cycle based on a multiproxy approach. In: Seidov, D., et al. (Eds.), The Oceans and Rapid Climate Changes: Past, Present and Future. Geophysical Monograph Series, vol. 126, pp. 83–100. Hillaire-Marcel, C., de Vernal, A., Polyak, L., Darby, D., 2004. Sizedependent isotopic composition of planktic foraminifers from Chukchi Sea vs. NW Atlantic sediments—implications for the Holocene
2101
paleoceanography of the western Arctic. Quaternary Science Reviews 23, 245–260. Jost, A., Lunt, D., Kageyama, M., Abe-Ouchi, A., Peyron, O., Valdes, P.J., Ramstein, G., 2005. High resolution simulations of the last glacial maximum climate over Europe: a solution to discrepancies with continental paleoclimatic reconstructions? Climate Dynamics, DOI 10.1007/s00382-005-0009-4. Joussaume, S., Taylor, K., 1995. Status of the Paleoclimate Modeling Intercomparison Project (PMIP). In: Proceedings of the First International AMIP Scientific Conference, vol. WRCP-92, 15–19 May 1995, Monterrey, CA, USA, pp. 425–430. Joussaume, S., Taylor, K.E., 2000. The Paleoclimate Modeling Intercomparison Project. In: Braconnot, P. (Ed.), Paleoclimate Modelling Intercomparison Project (PMIP). Proceedings of the Third PMIP Workshop, vol. WCRP-111, WMO/TD-No. 1007, 4–8 October 1999, Canada, pp. 9–24. Kageyama, M., Valdes, P.J., 2000. Impact of the North American icesheet orography on the Last Glacial Maximum eddies and snowfall. Geophysical Research Letters 27, 1515–1518. Kageyama, M., Valdes, P.J., Ramstein, G., Hewitt, C.D., Wyputta, U., 1999. Northern hemisphere storm-tracks in present day and last glacial maximum climate simulations: a comparison of the European PMIP models. Journal of Climate 12, 742–760. Kageyama, M., Peyron, O., Pinot, S., Tarasov, P., Guiot, J., Joussaume, S., Ramstein, G., PMIP participating groups, 2001. The Last Glacial Maximum climate over Europe and western Siberia: a PMIP comparison between models and data. Climate Dynamics 17, 23–43. Kim, S.-J., Flato, G.M., Boer, G.J., McFarlane, N.A., 2002. A coupled climate simulation of the Last Glacial Maximum, part 1: transient multi-decadal response. Climate Dynamics 19, 520–537. Kim, S.-J., Flato, G.M., Boer, G.J., 2003. A coupled climate model simulation of the Last Glacial Maximum, part II: approach to equilibrium. Climate Dynamics 20, 635–661. Kitoh, A., Murakami, S., 2002. Tropical pacific climates at the midHolocene and the Last Glacial Maximum simulated by a coupled ocean—atmosphere GCM. Paleoceanography 17. Kitoh, A., Murakami, S., Koide, H., 2001. A simulation of the Last Glacial Maximum with a couplaged atmosphere–ocean GCM. Geophysical Research Letters 28, 2221–2224. Kohfeld, K.E., Fairbanks, R.G., Smith, S.L., Walsh, I.D., 1996. Neogloboquadrina pachyderma (sinistral coiling) as paleoceanographic tracers in polar oceans: evidence from northeast water polynya plankton tows, sediment traps, and surface sediment. Paleoceanography 11, 679–699. Kubatzki, C., Claussen, M., 1998. Simulation of the global biogeophysical interactions during the Last Glacial Maximum. Climate Dynamics 14, 461–471. Kucera, M., Rosell-Mele´, A., Schneider, R., Waelbroeck, C., Weinelt, M., 2005a. Multiproxy approach for the reconstruction of the glacial ocean surface (MARGO). Quaternary Science Reviews 24, 813–819. Kucera, M., Weinelt, M., Kiefer, T., Pflaumann, U., Hayes, A., Weinelt, M., Chen, M.-T., Mix, A.C., Barrows, T.T., Cortijo, E., Duprat, J., Juggins, S., Waelbroeck, C., 2005b. Reconstruction of sea–surface temperatures from assemblages of planktonic foraminifera: multitechnique approach based on geographically constrained calibration data sets and its application to glacial Atlantic and Pacific Oceans. Quaternary Science Reviews 24, 951–998. Leuenberger, M., Siegenthaler, U., 1992. Ice-age atmospheric concentration of nitrous oxide from an antarctic ice core. Nature 360, 449–451. Levis, S., Foley, J.A., Pollard, D., 1999. CO2, climate, and vegetation feedbacks at the Last Glacial Maximum. Journal of Geophysical Research 104, 31191–31198. Mahowald, N., Kohfeld, K., Hansson, M., Balkanski, Y., Harrison, S.P., Prentice, I., Schulz, M., Rodhe, H., 1999. Dust sources and deposition during the last glacial maximum and current climate: a comparison of model results with paleodata from ice cores and marine sediments. Journal of Geophysical Research 104, 15895–15916.
ARTICLE IN PRESS 2102
M. Kageyama et al. / Quaternary Science Reviews 25 (2006) 2082–2102
Masson-Delmotte, V., Kageyama, M., Braconnot, P., Charbit, S., Krinner, G., Ritz, C., Guilyardi, E., Jouzel, J., Abe-Ouchi, A., Crucifix, M., Gladstone, R.M., Hewitt, C.D., Kitoh, A., LeGrande, A.N., Marti, O., Merkel, U., Motoi, T., Ohgaito, R., Otto-Bliesner, B., Peltier, W.R., Ross, I., Valdes, P.J., Vettoretti, G., Weber, S.L., Wolk, F., Yu, Y., 2005. Past and future polar amplification of climate change: climate model intercomparisons and ice-core constraints. Climate Dynamics, DOI 10.1007/s00382-005-0081-9. Meland, M.Y., Jansen, E., Elderfield, H., 2005. Constraints on SST estimates for the northern North Atlantic/Nordic Seas during the LGM. Quaternary Science Reviews 24, 835–852. Mix, A.E., Bard, E., Schneider, R., 2001. Environmental processes of the ice age: land, ocean, glaciers (EPILOG). Quaternary Science Reviews 20, 627–657. Monnin, A., Indermuhle, E., Dallenbach, A., Fluckiger, J., Stauffer, B., Stocker, D., Raynaud, T.F., Barnola, J.-M., 2001. Atmospheric CO2 concentrations over the last glacial termination. Science 291, 112–114. Mu¨ller, P.J., Kirst, G., Ruhland, G., Von Storch, I., Rosell-Mele´, A., 0 1998. Calibration of the alkenone paleotemperature index UK 37 based on core-tops from the eastern South Atlantic and the global ocean (60 N–60 S). Geochimica et Cosmochimica Acta 62, 1757–1772. Peltier, W.R., 1994. Ice age paleotopography. Science 265, 195–201. Peltier, W.R., 2004. Global glacial isostasy and the surface of the ice-age Earth: the ICE-5G (VM2) model and GRACE. Annual Review of Earth and Planetary Science 32, 111–149. Peltier, W.R., Solheim, L.P., 2002. Dynamics of the ice-age Earth: solid mechanics and fluid mechanics. Journal of Physics IV, Pr10-95–Pr10104. Peltier, W.R., Solheim, L.P., 2004. The climate of the Earth at Last Glacial Maximum: statistical equilibrium state and a mode of internal variability. Quaternary Science Reviews 23, 335–357. Petit, J.-R., Jouzel, J., Raynaud, D., Barkov, N.I., Barnola, J.-M., Basile, I., Bender, M., Chappellaz, J., Davis, M., Delaygue, G., Delmotte, M., Kotlyakov, V.M., Legrand, M., Lipenkov, V.Y., Lorius, C., Pe´pin, L., Ritz, C., Saltzman, E., Stievenard, M., 1999. Climate and atmospheric history of the past 420,000 years from the Vostok ice core, Antarctica. Nature 399, 429–436. Peyron, O., Guiot, J., Cheddadi, R., Tarasov, P., Reille, M., de Beaulieu, J.-L., Bottema, S., Andrieu, V., 1998. Climatic reconstruction in Europe for 18,000 yr B.P. from pollen data. Quaternary Research 49, 183–196. Peyron, O., Be´geot, C., Brewer, S., Heiri, O., Magny, M., Millet, L., Ruffaldi, P., Van Campo, E., Yu, G., 2005. Lateglacial climate in the Jura Mountains (France) based on different quantitative reconstruction approaches from pollen, lake-levels and chironomids. Quaternary Research 64, 197–211. Pflaumann, U., Sarnthein, M., Chapman, M., d’Abreu, L., Funnell, B., Huels, M., Kiefer, T., Maslin, M., Schulz, H., Swallow, J., van Kreveld, S., Vautravers, M., Vogelsang, E., Weinelt, M., 2003. Glacial North Atlantic: sea–surface conditions reconstructed by GLAMAP 2000. Paleoceanography 18. Pinot, S., Ramstein, G., Marsiat, I., de Vernal, A., Peyron, O., Duplessy, J.-C., Weinelt, M., 1999. Sensitivity of the European LGM climate to North Atlantic sea–surface temperature. Geophysical Research Letters 26, 1893–1896. Pollard, D., Barron, E.J., 2003. Causes of model–data discrepancies in European climate during oxygen isotope stage 3 with insights from the last glacial maximum. Quaternary Research 59, 108–113. Poutou, E., 2003. Etude nume´rique du roˆle des interactions entre la surface et l’atmosphe`re dans le cadre d’un changement climatique aux hautes latitudes nord. Ph.D. Thesis, Joseph Fourier University of Grenoble 1, France, 336pp.
Prahl, F.G., Muehlhausen, L.A., Zahnle, D.L., 1988. Further evaluation of long-chain alkenones as indicators of paleoceanographic conditions. Geochimica et Cosmochimica Acta 52, 2303–2310. Ramankutty, N., Foley, J., 1999. Estimating historical changes in global land cover: croplands from 1700 to 1992. Global Biogeochemical Cycles 13, 997–1027. Raynaud, D., Jouzel, J., Barnola, J.M., Chappellaz, J., Delmas, R.J., Lorius, C., 1993. The ice record of greenhouse gases. Science 259, 926–934. Rayner, N.A., Parker, D.E., Horton, E.B., Folland, C.K., Alexander, L.V., Rowell, D.P., Kent, E.C., Kaplan, A., 2003. Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. Journal of Geophysical Research 108. Renssen, H., Vandenberghe, J., 2003. Investigation of the relationship between permafrost distribution in NW Europe and extensive winter sea-ice cover in the North Atlantic Ocean during the cold phases of the Last Glaciation. Quaternary Science Reviews 22, 209–223. Renssen, H., Isarin, R.F.B., Vandenberghe, J., Lautenschlager, M., Schlese, U., 2000. Permafrost as a critical factor in paleoclimate modelling: the Younger Dryas case in Europe. Earth and Planetary Science Letters 176, 1–5. Rosell-Mele´, A., Bard, E., Emeis, K.-C., Grieger, B., Hewitt, C., Mu¨ller, P.J., Schneider, R.R., 2004. Sea surface temperature anomalies in the 0 oceans at the LGM estimated from the alkenone-UK 37 index: comparison with GCMs. Geophysical Research Letters 31. Shin, S.-I., Liu, Z., Otto-Bliesner, B., Brady, E.C., Kutzbach, J.E., Harrison, S.P., 2003a. A simulation of the Last Glacial Maximum climate using the NCAR-CCSM. Climate Dynamics 20, 127–151. Shin, S.-I., Liu, Z., Otto-Bliesner, B., Kutzbach, J.E., Vavrus, S.J., 2003b. Southern ocean sea-ice control of the glacial North Atlantic thermohaline circulation. Geophysical Research Letters 30. Simstich, J., Sarnthein, M., Erlenkeuser, H., 2003. Paired d18 O signals of Neogloboquadrina pachyderma (s) and Turborotalita quinqueloba show thermal stratification structure in nordic seas. Marine Micropaleontology 48, 107–125. Svendsen, J.I., Alexanderson, H., Astakhov, V., Demidov, I., Dowdeswell, J.A., Funder, S., Gataullin, V., Henriksen, M., Hjort, C., HoumarkNielsen, M., Hubberten, H.W., Ingo´lfsson, O´., Jakobsson, M., Kjær, K.H., Larsen, H., Lokrantz, E., Lunkka, J.-P., Lysa˚, A., Mangerud, H., Matiouchkov, A., Murray, A., Mo¨ller, P., Niessen, F., Nikolskaya, O., Polyak, L., Saarnisto, M., Siegert, C., Siegert, M.J., Spielhagen R., Stein, R., 2004. Late Quaternary ice sheet history of northern Eurasia. Quaternary Science Reviews 23, 1229–1271. Tarasov, P.E., Peyron, O., Guiot, J., Brewer, S., Volkova, V.S., Bezusko, L.G., Dorofeyuk, N.I., Kvavadze, E.V., Osipova, I.M., Panova, N.K., 1999. Last Glacial Maximum climate of the former Soviet Union and Mongolia reconstructed from pollen and macrofossil data. Climate Dynamics 15, 227–240. Toracinta, E., Oglesby, R., Bromwich, D., 2004. Atmospheric response to modified CLIMAP ocean boundary conditions during the Last Glacial Maximum. Journal of Climate 17, 504–522. Werner, M., Tegen, I., Harrison, S.P., Kohfeld, K.E., Prentice, I.C., Balkanski, Y., Rodhe, H., Roelandt, C., 2002. Seasonal and interannual variability of the mineral dust cycle under present and glacial climate conditions. Journal of Geophysical Research 107, 4744. World Ocean Atlas, 1998. Version 2, hhttp://www.nodc.noaa.gov/oc5/ wod98v2.htmli. Technical Report, National Oceanographic Data Center, Silver Spring, Maryland. Wyputta, U., McAvaney, B.J., 2001. Influence of vegetation changes during the last glacial maximum using the BMRC atmospheric general circulation model. Climate Dynamics 17, 923–932.