Figure 8.13: Annual mean precipitation changes (mm/yr) over Africa (20°W to 30°E) for the mid-Holocene climate: (upper panel) Biome distributions (desert, steppe, xerophytic and dry tropical forest/savannah (DTF/S)) as a function of latitude for present (red circles) and 6,000 yr BP (green triangles), showing that steppe vegetation replaces desert at 6,000 yr BP as far north as 23°N (vertical blue dashed line); (middle panel) 6000 yr BP minus present changes as simulated by the PMIP models. The black hatched lines are estimated upper and lower bounds for the excess precipitation required to support grassland, based on present climatic limits of desert and grassland taxa in palaeo-ecological records, the intersection with the blue vertical line indicates that an increase of 200 to 300 mm/yr is required to sustain steppe vegetation at 23°N at 6,000 yr BP (redrawn from Joussaume et al., 1999); (lower panel) same changes for the IPSL atmosphere-alone (A), i.e., PMIP simulation, the coupled atmosphere-ocean (OA), the atmosphere-alone with vegetation changes from OA (AV) and the coupled atmosphere-ocean-vegetation (OAV) simulations performed with the IPSL coupled climate model. The comparison between AV and OAV emphasises the synergism between ocean and land feedbacks (redrawn from Braconnot et al., 1999).
Accurate simulation of current climate does not guarantee the ability of a model to simulate climate change correctly. Climate models now have some skill in simulating changes in climate since 1850 (see Section 8.6.1), but these changes are fairly small compared with many projections of climate change into the 21st century. An important motivation for attempting to simulate the climatic conditions of the past is that such experiments provide opportunities for evaluating how models respond to large changes in forcing. Following the pioneering work of the Co-operative Holocene Mapping Project (COHMAP-Members, 1988), the Paleoclimate Modeling Intercomparison Project (PMIP) (Joussaume and Taylor, 1995; PMIP, 2000) has fostered the more systematic evaluation of climate models under conditions during the relatively well-documented past 20,000 years. The mid-Holocene (6,000 years BP) was chosen to test the response of climate models to orbital forcing with CO2 at pre-industrial concentration and present ice sheets. The orbital configuration intensifies (weakens) the seasonal distribution of the incoming solar radiation in the Northern (Southern) Hemisphere, by about 5% (e.g., 20 Wm-2 in the boreal summer). The last glacial maximum (21,000 years BP) was chosen to test the response to extreme cold conditions
Atmosphere alone simulations
Within PMIP, eighteen different atmospheric general circulation models using different resolutions and parametrizations have been run under the same mid-Holocene conditions, assuming present-day conditions over the oceans (Joussaume et al., 1999). In summer, all of the models simulate an increase and northward expansion of the African monsoon; conditions warmer than present in high northern latitudes, and drier than present in the interior of the northern continents. Palaeo-data do not support drying in interior Eurasia (Harrison et al., 1996; Yu and Harrison, 1996; Tarasov et al., 1998), but they clearly show an expanded monsoon in northern Africa (Street-Perrott and Perrott 1993; Hoelzmann et al., 1998; Jolly et al., 1998a, 1998b), warming in the Arctic (Texier et al., 1997), and drying in interior North America (Webb et al., 1993).
Vegetation changes reconstructed from pollen data in the BIOME 6000 project (Jolly et al., 1998b; Prentice and Webb III, 1998) provide a quantitative model-data comparison in northern Africa. The PMIP simulations produce a northward displacement of the desert-steppe transition, qualitatively consistent with biomes, but strongly underestimated in extent (Harrison et al., 1998). At least an additional 100 mm/yr of precipitation would be required for most models to sustain grassland at 23°N, i.e., more than twice as much as simulated in this area (Joussaume et al., 1999) (Figure 8.13). The increased area of lakes in the Sahara has also been quantified (Hoelzmann et al., 1998) and, although the PMIP simulations do produce an increase, this latter is not large enough (Coe and Harrison, 2000). A similar underestimation is obtained at high latitudes over northern Eurasia, where PMIP simulations produce a northward shift of the Arctic tree-line in agreement with observed shifts (Tarasov et al., 1998) but strongly underestimated in extent (Kutzbach et al., 1996b; Texier et al., 1997; Harrison et al., 1998). Model-data discrepancies may, however, be due to missing feedbacks in the simplified PMIP experimental design.
Recent experiments with asynchronous (Kutzbach and Liu 1997; Liu et al., 1999) and synchronous (Hewitt and Mitchell, 1998; Braconnot et al., 2000) coupling of atmospheric models to full dynamical ocean models have been performed for the mid-Holocene. They all produce a larger enhancement of the African monsoon than shown in their PMIP atmosphere only experiments, resulting from the ocean thermal inertia and changes in the meridional ocean heat transport (Braconnot et al., 2000). However, the changes are not sufficient to reproduce the observed changes in biome shifts over northern Africa.
Coupled models are also beginning to address the issue of changes in interannual to inter-decadal variability under conditions of large differences in the basic climate. Some palaeo-environmental evidence has suggested that short-term climate variability associated with El Niño-Southern Oscillation (ENSO) was reduced during the early to mid-Holocene (Sandweiss et al., 1996; Rodbell, 1999). Up to now, ENSO variability has only been analysed in the CSM simulation, exhibiting no significant change at the mid-Holocene (Otto-Bliesner, 1999).
Land-surface changes also provide an additional important feedback. During the mid-Holocene, vegetation changes over northern Africa have indeed favoured a larger increase in monsoon precipitation as shown through sensitivity experiments (Kutzbach et al., 1996a; Brostrom et al., 1998; Texier et al., 2000) as well as through coupled atmosphere-vegetation experiments (Claussen and Gayler, 1997; Texier et al., 1997; Pollard et al., 1998; Doherty et al., 2000; de Noblet et al., 2000). Including the observed occurrence of large lakes and wetlands (Coe and Bonan, 1997; Brostrom et al., 1998) also intensifies monsoon rains. Vegetation feedbacks also amplify the effects of orbital forcing at high latitudes where they led to greater and more realistic shifts of vegetation cover over northern Eurasia (Foley et al., 1994; Kutzbach et al., 1996b; Texier et al., 1997). The importance of land-surface feedbacks has further been emphasised in the IPSL AOGCM coupled to a vegetation model (Braconnot et al., 1999). The IPSL simulation shows that combined feedbacks between land and ocean lead to a closer agreement with palaeo-data (Figure 8.13). The ocean feedback increases the supply of water vapour, while the vegetation feedback increases local moisture recycling and the length of the monsoon season. The importance of land-surface feedbacks has also been shown by an EMIC (Ganopolski et al., 1998a), further emphasising that vegetation feedbacks may explain abrupt changes in Saharan vegetation in the mid-Holocene (Claussen et al., 1999).
Other reports in this collection