IPCC Fourth Assessment Report: Climate Change 2007
Climate Change 2007: Working Group I: The Physical Science Basis How Realistic Are Results from Climate Model Simulations of the Last Glacial Maximum?

Model intercomparisons from the first phase of the Paleoclimate Modelling Intercomparison Project (PMIP-1), using atmospheric models (either with prescribed SST or with simple slab ocean models), were featured in the TAR. There are now six simulations of the LGM from the second phase (PMIP-2) using Atmosphere-Ocean General Circulation Models (AOGCMs) and EMICs, although only a few regional comparisons were completed in time for this assessment. The radiative perturbation for the PMIP-2 LGM simulations available for this assessment, which do not yet include the effects of vegetation or aerosol changes, is –4 to –7 W m–2. These simulations allow an assessment of the response of a subset of the models presented in Chapters 8 and 10 to very different conditions at the LGM.

The PMIP-2 multi-model LGM SST change shows a modest cooling in the tropics, and greatest cooling at mid- to high latitudes in association with increases in sea ice and changes in ocean circulation (Figure 6.5). The PMIP-2 modelled strengthening of the SST meridional gradient in the LGM North Atlantic, as well as cooling and expanded sea ice, agrees with proxy indicators (Kageyama et al., 2006). Polar amplification of global cooling, as recorded in ice cores, is reproduced for Antarctica (Figure 6.5), but the strong LGM cooling over Greenland is underestimated, although with caveats about the heights of these ice caps in the PMIP-2 simulations (Masson-Delmotte et al., 2006).

Figure 6.5

Figure 6.5. The Last Glacial Maximum climate (approximately 21 ka) relative to the pre-industrial (1750) climate. (Top left) Global annual mean radiative influences (W m–2) of LGM climate change agents, generally feedbacks in glacial-interglacial cycles, but also specified in most Atmosphere-Ocean General Circulation Model (AOGCM) simulations for the LGM. The heights of the rectangular bars denote best estimate values guided by published values of the climate change agents and conversion to radiative perturbations using simplified expressions for the greenhouse gas concentrations and model calculations for the ice sheets, vegetation and mineral dust. References are included in the text. A judgment of each estimate’s reliability is given as a level of scientific understanding based on uncertainties in the climate change agents and physical understanding of their radiative effects. Paleoclimate Modelling Intercomparison Project 2 (PMIP-2) simulations shown in bottom left and right panels do not include the radiative influences of LGM changes in mineral dust or vegetation. (Bottom left) Multi-model average SST change for LGM PMIP-2 simulations by five AOGCMs (Community Climate System Model (CCSM), Flexible Global Ocean-Atmosphere-Land System (FGOALS), Hadley Centre Coupled Model (HadCM), Institut Pierre Simon Laplace Climate System Model (IPSL-CM), Model for Interdisciplinary Research on Climate (MIROC)). Ice extent over continents is shown in white. (Right) LGM regional cooling compared to LGM global cooling as simulated in PMIP-2, with AOGCM results shown as red circles and EMIC (ECBilt-CLIO) results shown as blue circles. Regional averages are defined as: Antarctica, annual for inland ice cores; tropical Indian Ocean, annual for 15°S to 15°N, 50°E to 100°E; and North Atlantic Ocean, July to September for 42°N to 57°N, 35°W to 20°E. Grey shading indicates the range of observed proxy estimates of regional cooling: Antarctica (Stenni et al., 2001; Masson-Delmotte et al., 2006), tropical Indian Ocean (Rosell-Mele et al., 2004; Barrows and Juggins, 2005), and North Atlantic Ocean (Rosell-Mele et al., 2004; Kucera et al., 2005; de Vernal et al., 2006; Kageyama et al., 2006).

The PMIP-2 AOGCMs give a range of tropical ocean

cooling between 15°S to 15°N of 1.7°C to 2.4°C. Sensitivity simulations with models indicate that this tropical cooling can be explained by the reduced glacial greenhouse gas concentrations, which had direct effects on the tropical radiative forcing (Shin et al., 2003; Otto-Bliesner et al., 2006b) and indirect effects through LGM cooling by positive sea ice-albedo feedback in the Southern Ocean contributing to enhanced ocean ventilation of the tropical thermocline and the intermediate waters (Liu et al., 2002). Regional variations in simulated tropical cooling are much smaller than indicated by MARGO data, partly related to models at current resolutions being unable to simulate the intensity of coastal upwelling and eastern boundary currents. Simulated cooling in the Indian Ocean (Figure 6.5), a region with important present-day teleconnections to Africa and the North Atlantic, compares favourably to proxy estimates from alkenones (Rosell-Mele et al., 2004) and foraminifera assemblages (Barrows and Juggins, 2005).

Considering changes in vegetation appears to improve the realism of simulations of the LGM, and points to important climate-vegetation feedbacks (Wyputta and McAvaney, 2001; Crucifix and Hewitt, 2005). For example, extension of the tundra in Asia during the LGM contributes to the local surface cooling, while the tropics warm where savannah replaces tropical forest (Wyputta and McAvaney, 2001). Feedbacks between climate and vegetation occur locally, with a decrease in the tree fraction in central Africa reducing precipitation, and remotely with cooling in Siberia (tundra replacing trees) altering (diminishing) the Asian summer monsoon. The physiological effect of CO2 concentration on vegetation needs to be included to properly represent changes in global forest (Harrison and Prentice, 2003), as well as to widen the climatic range where grasses and shrubs dominate. The biome distribution simulated with dynamic global vegetation models reproduces the broad features observed in palaeodata (e.g., Harrison and Prentice, 2003).

In summary, the PMIP-2 LGM simulations confirm that current AOGCMs are able to simulate the broad-scale spatial patterns of regional climate change recorded by palaeodata in response to the radiative forcing and continental ice sheets of the LGM, and thus indicate that they adequately represent the primary feedbacks that determine the climate sensitivity of this past climate state to these changes. The PMIP-2 AOGCM simulations using glacial-interglacial changes in greenhouse gas forcing and ice sheet conditions give a radiative perturbation in reference to pre-industrial conditions of –4.6 to –7.2 W m–2 and mean global temperature change of –3.3°C to –5.1°C, similar to the range reported in the TAR for PMIP-1 (IPCC, 2001). The climate sensitivity inferred from the PMIP-2 LGM simulations is 2.3°C to 3.7°C for a doubling of atmospheric CO2 (see Section When the radiative perturbations of dust content and vegetation changes are estimated, climate models yield an additional cooling of 1°C to 2°C (Crucifix and Hewitt, 2005; Schneider et al., 2006), although scientific understanding of these effects is very low.