8.5.5.2 The last glacial maximum
Figure 8.14: Annual mean tropical cooling at the last glacial maximum:
comparison between model results and palaeo-data. (Centre panel) simulated
surface air temperature changes over land are displayed as a function of
surface temperature changes over the oceans, both averaged in the 30°S
to 30°N latitudinal band, for all the PMIP simulations: models with
prescribed CLIMAP SSTs (circles) and coupled atmosphere-mixed layer ocean
models (squares) (from Pinot et al., 1999). Numbers refer to different models:
circles, 1: LMD4, 2-5: MRI2, ECHAM3, UGAMP, LMD5 (higher resolution), 6-7:
CCSR/NIES1, LMD5, 8: GEN2. Squares : 1: LMD4,2: UGAMP, 3: GEN2, 4: GFDL,
5: HADAM2, 6: MRI2, 7: CCM1, 8: CCC2 (names refer to Table
8.1 and Table 8.4). Results from two EMIC
models including a dynamical ocean model have also been displayed (diamonds):
1-UVIC (Weaver et al., 1998), 2-CLIMBER-2 (Petoukhov et al., 2000).
The comparison with palaeo-data: (upper panel) over land is with estimates
from various pollen data for altitudes below 1,500m (the label “nb
data” refers to the number of data points in three different regions
corresponding to the temperature change estimate plotted in the abscissa)
from (Farrera, et al., 1999); (right panel) the distribution of SST changes
estimated from alkenones in the tropics from the Sea Surface Temperature
Evolution Mapping Project based on Alkenone Stratigraphy (TEMPUS) (Rosell-Melé,
et al., 1998) (nb data: same as upper panel, number of data points for each
temperature change). Caution: in this figure, model results are averaged
over the whole tropical domain and not over proxy-data locations, which
may bias the comparison (e.g., Broccoli and Marciniak, 1996). For example,
for the pollen data, extreme values are obtained for specific regions: weakest
values over the Indonesia-Pacific region and coldest values over South America.
|
Results from the PMIP experiments
The Last Glacial Maximum (LGM) climate involves large changes in ice sheet extent
and height, SSTs, albedo, sea level and CO2 (200 ppm), but only minor changes
in solar radiation. Over the oceans, two sets of experiments have been performed
within PMIP using several atmospheric models, either prescribing SSTs estimated
from macrofossil transfer functions (CLIMAP, 1981) or computing SSTs from a
mixed-layer ocean model. An annual mean global cooling of about -4°C is
obtained by all models forced by the Climate: Long-range Investigation, Mapping
and Prediction (CLIMAP) SSTs, whereas the range of cooling is larger when using
computed SSTs, from -6 to -2°C. This range of 4°C arises both from differences
in the simulated radiative forcing associated primarily with different ice albedo
values, and from differences in model climate sensitivity (see Chapter
9, Section 9.2.1).
Evaluating the consistency between the simulated climate and that reconstructed
from palaeo-data can potentially provide an independent check that model sensitivity
is neither too large nor too small. A detailed analysis of a subset of the PMIP
models (Taylor et al., 2000), shows that their forcing estimates for the LGM
vary from about -4 to -6 Wm-2 and that their global climate sensitivity, given
relatively to a doubling of CO2, ranges from 3.2 to 3.9°C (assuming that
climate sensitivity is independent of the type of forcing, although one model
study shows a slightly stronger sensitivity at the LGM than for a CO2 doubling
(Hewitt and Mitchell, 1997)). A direct evaluation of climate sensitivity is,
however, very difficult since global temperature changes are poorly known. Hoffert
and Covey (1992) estimated a global cooling of -3 ± 0.6°C from CLIMAP
(1981) SST data which would, using the simulated range of forcing, yield to
a global climate sensitivity for a doubling of CO2 ranging from 1.4 to 3.2°C,
which probably gives a lower estimate of climate sensitivity since CLIMAP SSTs
tend to be relatively too warm in the tropics (see below).
An alternative approach to evaluating climate sensitivity is provided by the
detailed comparison of model results with proxy-data over different regions.
The amplitude of the tropical cooling at LGM has long been disputed (Rind and
Peteet, 1985; Guilderson et al., 1994). Compared to a new synthesis of terrestrial
data (Farrera et al., 1999), PMIP simulations with prescribed CLIMAP sea-surface
conditions produce land temperatures that are too warm (Pinot et al., 1999)
(Figure 8.14), which may be due to too-warm prescribed
SSTs, as indicated by new marine data based on alkenone palaeo-thermometry (Rosell-Melé
et al., 1998) (Figure 8.14). Some mixed-layer ocean models
have produced more realistic sea and land temperature cooling (Pinot et al.,
1999), enhancing our confidence in using such models to estimate climate sensitivity
(see Chapter 9, Section 9.3.4) (Figure
8.14). The same conclusion is derived by Broccoli (2000) when accounting
for uncertainties in both the forcing and reconstructed climate from various
proxy data. Over Eurasia, all the models simulate a cooling in fairly good agreement
with proxy data estimates, except over western Europe (Kageyama et al., 2001),
where they all underestimate the winter cooling shown from pollen data (Peyron
et al., 1998). However, such simulations have an important caveat since they
prescribe present day ocean heat transport whereas changes in the North Atlantic
deep water circulation shown by two EMIC models (Ganopolski et al., 1998b; Weaver
et al., 1998) and also inferred by palaeo-oceanographic data (e.g., Duplessy
et al., 1988) may further decrease temperatures over Europe.
Land-surface feedbacks
Vegetation feedbacks at the LGM could have been due to climate-induced shifts
in biomes, CO2-induced changes in vegetation structure (Jolly and Haxeltine,
1997; Street-Perrot et al., 1997; Cowling, 1999), and CO2-induced changes in
leaf conductance (see Chapter 7, Section
7.4.2). Sensitivity experiments (Crowley and Baum, 1997; Levis et al., 1999)
suggest that the first two types dominated. Over much of Eurasia, forests were
replaced by tundra or steppe (Prentice et al., 1998) which may have contributed
to the observed cooling over Europe (Crowley and Baum, 1997; Kubatzki and Claussen,
1998; Levis et al., 1999). Permafrost may also have to be accounted for (Renssen
et al., 2000). In the tropics though, there is yet no systematic improvement
of the simulated cooling, since the models find large areas of warming due to
the simulated deforestation (Crowley and Baum, 1997; Levis et al., 1999). However,
land-surface feedbacks may also have affected climate through mineral aerosol
(dust) concentrations (Mahowald et al., 1999).
8.5.5.3 Summary
Mid-Holocene
Through PMIP experiments, it is now well-established that all atmospheric models
are able to simulate several robust large-scale features of the Holocene climate
but also that they all underestimate these changes. Several complementary simulations
have shown that ocean and vegetation processes introduce important feedbacks
which are necessary to explain the observed monsoon changes. These results urge
for a systematic evaluation of coupled atmosphere-ocean-vegetation models for
the mid-Holocene and for an investigation of the impact of vegetation changes,
such as climate-induced density and land-use cover changes (see Chapter
7, Section 7.4.2), on future climate change projections.
Last Glacial Maximum
The more systematic evaluation of atmosphere alone models conducted within PMIP
confirms that the LGM SST as estimated by CLIMAP (1981) need to be revised.
Some simulations with atmospheric models coupled to mixed-layer models produce
realistic results, especially in the tropics, and enhance our confidence in
the estimates of climate sensitivity used in future climate change studies.
However, such models neglect changes in ocean heat transport as well as land-surface
feedbacks. Moreover, an evaluation of coupled AOGCMs is still needed at the
LGM.
|
