694
Herterich et al.
et al. (1994) SSS data (corrected for the global
salinity increase due to the glacial sea level reduction) and paleo-SSS reconstructions by Duplessy
et al. (1991), Sarnthein et al. (1992, 1994) or
Schafer-Neth (1998). Methods and problems in
reconstructing paleo salinities are also discussed by
Wolff et al. (this volume). It must be noted that
paleo-SSS reconstructions are only available for the
North Atlantic and a few locations in the Southern
Ocean.
The equilibrium solution for the glacial circulation is highly sensitive to possible errors in the salinity forcing (Fichefet et al. 1994): When the restoring salinity is increased or decreased by 0.4 psu
at high latitudes, the meridional overturning changes
dramatically. Nevertheless, two features of the
glacial circulation as modelled by various ocean
models seem to be robust: (1) The deep water formation in the Greenland-Iceland-Norwegian Seas
either ceased or was strongly reduced. (2) There
was an active thermohaline conveyor belt in the
Atlantic.
The sensitivity to possible errors in the temperature forcing is still to be tested. But if the CLIMAP
tropical SST reconstruction were too high and the
CLIMAP Greenland - Iceland - Norwegian Seas
SST reconstruction were too low, then the true
meridional SST gradient in the Atlantic would be
smaller and a lower meridional overturning could
be anticipated.
The need for global and reliable SST reconstructions will only be superseded by the advent of efficient and accurate coupled atmosphere - ocean
models. In the case of SSS reconstructions, the
situation might be different: Reliable estimates for
some key areas of the World Ocean could possibly suffice for the purpose of model validation or
data assimilation. This is because recent atmospheric models such as the ECHAM in its version
4 (Roeckner et al. 1996) seem to represent the
hydrological cycle better, thus the fresh water flux
calculated from these models becomes more suited
for driving ocean models. However, in high latitudes
sea ice has a strong impact on the ocean due to
the salinity difference between sea ice and the
ocean (Stossel et al. 1996). Substantial brine release occurs wherever sea ice is formed and fresh
water is added to the surface layer after melting.
In the Southern Ocean, in particular, these processes play an important role in the formation of Antarctic Bottom Water (AABW), with a direct effect on the thermohaline circulation. Furthermore,
many parameterizations used in atmospheric models are tuned to the modern climate and lead to considerable uncertainty in results on climates very different from the modern one. Hence an accurate
representation of sea ice and atmospheric model
parameterizations less dependent on the modern
climate are prerequisites for simulating rather than
prescribing SSS fields. Besides salinity, there are
other tracers which would also allow to assess the
glacial circulation, e.g. & 18 0 (Lehman et al. 1993)
and &13C (Winguth et al. 1996).
Figure 4 stems from a simulation ofthe glacial
circulation in a coarse-resolution global ocean
model. It shows the depth down to which the convection in the North Atlantic penetrates at modern
times and at the LGM. The formation of North
Atlantic Deep Water (NADW) at the LGM is
stopped in the Greenland - Iceland - Norwegian
Seas and occurs further to the south at about SOON.
The ocean model used is based on the Hamburg
LSG (Maier-Reimer et al. 1993). The most substantial difference from the original model is the new
advection scheme derived from the work of
Farrow and Stevens (1995). In this particular case,
the LSG ocean model is forced by surface wind
stress and surface air temperature from the atmospheric model run of Lorenz et al. (1996), and its
upper layer is restored to composite SSS data. In
the northern North Atlantic the SSS is reconstructed
from foraminiferal & 18 0 in a manner similar to
Schafer-Neth (1998), but strictly using the original
CLIMAP-SST data. Figs. 5 and 6 show the corresponding Atlantic meridional overturning and the
heat transport for the global ocean, the Atlantic
basin and the Pacific and Indian basins combined.
The formation ofNADW is reduced from 19 to 14
Sv (by about 26%) and the export ofNADW to
the Southern Ocean is reduced from 14 to 10 Sv
(by about 30%). There is no significant increase in
AABW inflow, which is in conflict with most reconstructions from benthic &13C records (e.g.
Duplessy et al. 1988; Sarnthein et al. 1994). In the
South Atlantic, the glacial heat transport is still directed northward and only slightly lower than
Herterich et al.
et al. (1994) SSS data (corrected for the global
salinity increase due to the glacial sea level reduction) and paleo-SSS reconstructions by Duplessy
et al. (1991), Sarnthein et al. (1992, 1994) or
Schafer-Neth (1998). Methods and problems in
reconstructing paleo salinities are also discussed by
Wolff et al. (this volume). It must be noted that
paleo-SSS reconstructions are only available for the
North Atlantic and a few locations in the Southern
Ocean.
The equilibrium solution for the glacial circulation is highly sensitive to possible errors in the salinity forcing (Fichefet et al. 1994): When the restoring salinity is increased or decreased by 0.4 psu
at high latitudes, the meridional overturning changes
dramatically. Nevertheless, two features of the
glacial circulation as modelled by various ocean
models seem to be robust: (1) The deep water formation in the Greenland-Iceland-Norwegian Seas
either ceased or was strongly reduced. (2) There
was an active thermohaline conveyor belt in the
Atlantic.
The sensitivity to possible errors in the temperature forcing is still to be tested. But if the CLIMAP
tropical SST reconstruction were too high and the
CLIMAP Greenland - Iceland - Norwegian Seas
SST reconstruction were too low, then the true
meridional SST gradient in the Atlantic would be
smaller and a lower meridional overturning could
be anticipated.
The need for global and reliable SST reconstructions will only be superseded by the advent of efficient and accurate coupled atmosphere - ocean
models. In the case of SSS reconstructions, the
situation might be different: Reliable estimates for
some key areas of the World Ocean could possibly suffice for the purpose of model validation or
data assimilation. This is because recent atmospheric models such as the ECHAM in its version
4 (Roeckner et al. 1996) seem to represent the
hydrological cycle better, thus the fresh water flux
calculated from these models becomes more suited
for driving ocean models. However, in high latitudes
sea ice has a strong impact on the ocean due to
the salinity difference between sea ice and the
ocean (Stossel et al. 1996). Substantial brine release occurs wherever sea ice is formed and fresh
water is added to the surface layer after melting.
In the Southern Ocean, in particular, these processes play an important role in the formation of Antarctic Bottom Water (AABW), with a direct effect on the thermohaline circulation. Furthermore,
many parameterizations used in atmospheric models are tuned to the modern climate and lead to considerable uncertainty in results on climates very different from the modern one. Hence an accurate
representation of sea ice and atmospheric model
parameterizations less dependent on the modern
climate are prerequisites for simulating rather than
prescribing SSS fields. Besides salinity, there are
other tracers which would also allow to assess the
glacial circulation, e.g. & 18 0 (Lehman et al. 1993)
and &13C (Winguth et al. 1996).
Figure 4 stems from a simulation ofthe glacial
circulation in a coarse-resolution global ocean
model. It shows the depth down to which the convection in the North Atlantic penetrates at modern
times and at the LGM. The formation of North
Atlantic Deep Water (NADW) at the LGM is
stopped in the Greenland - Iceland - Norwegian
Seas and occurs further to the south at about SOON.
The ocean model used is based on the Hamburg
LSG (Maier-Reimer et al. 1993). The most substantial difference from the original model is the new
advection scheme derived from the work of
Farrow and Stevens (1995). In this particular case,
the LSG ocean model is forced by surface wind
stress and surface air temperature from the atmospheric model run of Lorenz et al. (1996), and its
upper layer is restored to composite SSS data. In
the northern North Atlantic the SSS is reconstructed
from foraminiferal & 18 0 in a manner similar to
Schafer-Neth (1998), but strictly using the original
CLIMAP-SST data. Figs. 5 and 6 show the corresponding Atlantic meridional overturning and the
heat transport for the global ocean, the Atlantic
basin and the Pacific and Indian basins combined.
The formation ofNADW is reduced from 19 to 14
Sv (by about 26%) and the export ofNADW to
the Southern Ocean is reduced from 14 to 10 Sv
(by about 30%). There is no significant increase in
AABW inflow, which is in conflict with most reconstructions from benthic &13C records (e.g.
Duplessy et al. 1988; Sarnthein et al. 1994). In the
South Atlantic, the glacial heat transport is still directed northward and only slightly lower than