708
Herterich et al.
account additional infonnation from marine sedimentary cores. This is of significant value for vegetation reconstruction within CLIMAP 2000.
Discussion
The advantage of modelling in general is that it
provides more physically and geographically detailed
information on the paleoclimate and a means of
assessing the existing proxy data. It can augment
qualitative classifications of paleoclimatic data such
as "hot / cold" or "dry / humid" with quantitative
assertions. This is accomplished by using the physical and biogeochemical principles embodied in the
mathematical equations. The aim of modelling the
LGM in particular is to depict of what can be
thought of as marking the extreme cold end of climate variability. Within CLIMAP 2000, modelling
will make an essential contribution to a global synthesis ofLGM data and modelling, by probing the
consistency ofthe proxy data, interpolating it into
areas void of observations, and generating complete
global maps. Ideally these maps will include some
sort of error estimates which are required for validating future proxy data or model results.
Modelling the atmosphere as a separate system
is possible as long as sufficiently complete lower
boundary conditions can be specified. In the last
decade, most ofthese lower boundary conditions
have been questioned. Thus within the envisaged
CLIMAP 2000 project, the CLIMAP-SST reconstructions need to be checked: What was the real
cooling of tropical SSTs? Were the Greenland -
Iceland-Norwegian Seas indeed ice-covered during summer? Broccoli and Marciniak (1996) compared LGM-SSTs which were simulated by using
an atmospheric general circulation model coupled
to a simple model of the ocean mixed layer with
LGM-SSTs as reconstructed by CLIMAP. They
found that at low latitudes, the correspondence
between simulated and CLIMAP zonal anomalies
were "excellent in the Atlantic, mediocre in the
Indian, and poor in the Pacific". On the one hand,
it is the land climate conditions, such as surface air
temperature or surface wind, that are closest to
human perception. On the other hand, the potential effect of different SSTs on these land climate
conditions is rather large.
Revised SST data can be assessed by modelling the ocean as a separate system. Careful comparisons of the simulated ocean climate with evidence from marine paleontology can guide in deciding what changes to the CLIMAP-SST reconstructions are meaningful. Hence, for the CLIMAP
2000 project, maps or sections of passive tracers
such as salinity, 8180 or 14C are required at least
for some key regions of the World Ocean.
Equilibrium ocean-alone solutions require, in
addition to SST data, surface wind stress and either fresh water flux or SSS data. There is a subtlety about driving ocean models using SSS estimates for the LGM: The relative error of SSS reconstructions is considerably larger than that of SST
reconstructions, and at high latitudes the ocean circulation is much more sensitive to uncertainties in
the SSS forcing than to those in the SST forcing.
Therefore skill and experience is needed for designing plausible SSS reconstructions (Seidov et al.
1996). By the time atmospheric models prove to
be capable of simulating the fresh water flux of
climates very different from the modem one, they
offer the chance to run ocean models - either in a
coupled or an off-line mode - and suggest an SSS
field which complements the then existing proxy
data. Eventually ocean models, or coupled atmosphere-ocean models, may answer the question
whether the glacial thermohaline circulation with
its associated meridional overturning and northward
heat transport was reduced or as intense as today
(Lynch-Stieglitz et al. 1996; Yu et at. 1996).
Another lower boundary condition for atmospheric models, with an effect on the land climate
simulations comparable to SSTs, is the extent and
height of the ice-sheets. There exist two different
ice-sheet reconstructions for the LGM: the one of
CLIMAP and the one of Peltier (1994), whose
Laurentide and Eurasian ice-sheets are much thinner. The CLIMAP 2000 project should also settle
the question as to which ice-sheet reconstruction
is most probably the right one. In principle, a dynamic ice-sheet model can be coupled asynchronously to an atmospheric model and simulate the
glacial ice-sheets. But first any atmospheric model
must pass a critical test: It must be able to give a
global annual mass balance of precipitation and
ablation positive before, close to zero at, and
Herterich et al.
account additional infonnation from marine sedimentary cores. This is of significant value for vegetation reconstruction within CLIMAP 2000.
Discussion
The advantage of modelling in general is that it
provides more physically and geographically detailed
information on the paleoclimate and a means of
assessing the existing proxy data. It can augment
qualitative classifications of paleoclimatic data such
as "hot / cold" or "dry / humid" with quantitative
assertions. This is accomplished by using the physical and biogeochemical principles embodied in the
mathematical equations. The aim of modelling the
LGM in particular is to depict of what can be
thought of as marking the extreme cold end of climate variability. Within CLIMAP 2000, modelling
will make an essential contribution to a global synthesis ofLGM data and modelling, by probing the
consistency ofthe proxy data, interpolating it into
areas void of observations, and generating complete
global maps. Ideally these maps will include some
sort of error estimates which are required for validating future proxy data or model results.
Modelling the atmosphere as a separate system
is possible as long as sufficiently complete lower
boundary conditions can be specified. In the last
decade, most ofthese lower boundary conditions
have been questioned. Thus within the envisaged
CLIMAP 2000 project, the CLIMAP-SST reconstructions need to be checked: What was the real
cooling of tropical SSTs? Were the Greenland -
Iceland-Norwegian Seas indeed ice-covered during summer? Broccoli and Marciniak (1996) compared LGM-SSTs which were simulated by using
an atmospheric general circulation model coupled
to a simple model of the ocean mixed layer with
LGM-SSTs as reconstructed by CLIMAP. They
found that at low latitudes, the correspondence
between simulated and CLIMAP zonal anomalies
were "excellent in the Atlantic, mediocre in the
Indian, and poor in the Pacific". On the one hand,
it is the land climate conditions, such as surface air
temperature or surface wind, that are closest to
human perception. On the other hand, the potential effect of different SSTs on these land climate
conditions is rather large.
Revised SST data can be assessed by modelling the ocean as a separate system. Careful comparisons of the simulated ocean climate with evidence from marine paleontology can guide in deciding what changes to the CLIMAP-SST reconstructions are meaningful. Hence, for the CLIMAP
2000 project, maps or sections of passive tracers
such as salinity, 8180 or 14C are required at least
for some key regions of the World Ocean.
Equilibrium ocean-alone solutions require, in
addition to SST data, surface wind stress and either fresh water flux or SSS data. There is a subtlety about driving ocean models using SSS estimates for the LGM: The relative error of SSS reconstructions is considerably larger than that of SST
reconstructions, and at high latitudes the ocean circulation is much more sensitive to uncertainties in
the SSS forcing than to those in the SST forcing.
Therefore skill and experience is needed for designing plausible SSS reconstructions (Seidov et al.
1996). By the time atmospheric models prove to
be capable of simulating the fresh water flux of
climates very different from the modem one, they
offer the chance to run ocean models - either in a
coupled or an off-line mode - and suggest an SSS
field which complements the then existing proxy
data. Eventually ocean models, or coupled atmosphere-ocean models, may answer the question
whether the glacial thermohaline circulation with
its associated meridional overturning and northward
heat transport was reduced or as intense as today
(Lynch-Stieglitz et al. 1996; Yu et at. 1996).
Another lower boundary condition for atmospheric models, with an effect on the land climate
simulations comparable to SSTs, is the extent and
height of the ice-sheets. There exist two different
ice-sheet reconstructions for the LGM: the one of
CLIMAP and the one of Peltier (1994), whose
Laurentide and Eurasian ice-sheets are much thinner. The CLIMAP 2000 project should also settle
the question as to which ice-sheet reconstruction
is most probably the right one. In principle, a dynamic ice-sheet model can be coupled asynchronously to an atmospheric model and simulate the
glacial ice-sheets. But first any atmospheric model
must pass a critical test: It must be able to give a
global annual mass balance of precipitation and
ablation positive before, close to zero at, and