Coupling of Several Components
In general, depending on the scientific question posed and
the computing power available, the first step is to identify the
part of the system which will be explicitly represented by the
model. If, for example, the intention is to represent the
evolution of the ice caps under the influence of the orbital
parameters of the Earth, it is essential to concentrate on a
model which explicitly represents the dynamic of the ice
caps. However, when the focus is on fast-reacting components (ocean, atmosphere), it is acceptable to fix the
slow-moving components by establishing these as boundary
limits. The converse is not true. Thus, the evolution of the
ice caps only makes sense when its interaction with a
changing climate system is considered. Herein lies one of the
main difficulties with modeling the climate over long periods
of time (both past and future). How can the interactions
(exchanges of energy and matter) between physical objects
with very different time constants be calculated in a meaningful way? Several strategies are possible.
A—If only a state of equilibrium in the system is of
interest, then the equilibrium of one of the systems (for
example, the ice caps) can be found by fixing another (e.g.
giving the atmosphere fixed boundary conditions), then by
performing the inverse operation (the atmosphere is calculated while the ice caps are fixed), then reiterating the process
until the results converge. This is called asynchronous coupling. In essence, the two physical components are coupled,
but in a way that does not reflect (or very inexactly) the real
flow of time. This makes it possible to achieve equilibrium of
the coupled system over a relatively short period of time (for
example, a few decades for the atmosphere), since the
fast-reacting component (the atmosphere is the factor which
demands the most computing time due to the fact that it has
inherent small-scale variations) is always calculated by
assuming an equilibrium with the slow component.
B—Although asynchronous coupling can be a good
method of calculating climate equilibria, it is not a very
rigorous way to calculate a meaningful evolution of the
system. This may provide a useful approximation if the time
constants of the physical systems are very different (e.g.,
atmosphere and ice caps), in which case it amounts to
considering that the slow component is the sole driver of the
evolution of the system with the fast component merely
adapting to the slow component. If physical components
with ‘intermediate’ times are included (such as the ocean,
which reacts more slowly than the atmosphere but faster than
the ice caps), this strategy may fail. Therefore, it is sometimes desirable to simplify the physics of fast-reacting
components (especially the atmosphere) in an attempt to
calculate only the long-term variations that can be used
directly by the slower components. This has led to the
development of models known as ‘intermediate complexity’
models (see Section “Earth system models of intermediate
complexity (EMICS)”).
Comparison with Paleoclimate Data
If our objective is to explain climate variations as reconstructed from paleoclimate records via modeling based as
much as possible on the physics, it is important to set up the
model and experimental design so that this comparison is
easiest. This gives rise to a second major difficulty in paleoclimatology simulations: paleoclimate indicators (proxies)
are never clear-cut in terms of the physical variables simulated by models. These indicators are dependent on particular
climate parameters, but are these the ones being simulated? It
is therefore risky to rely only on the same physical models as
those specially developed for comparison with current
observations, whose important parameters can be measured
by oceanographers, glaciologists and meteorologists. The
ideal situation is to be able to quantitatively compare paleoclimate simulations with paleoclimate indicators. The most
promising strategy is to explicitly simulate these indicators in
the models so that the multiple factors likely to influence them
are taken into account. For example, it is useful to explicitly
simulate the water isotopes (d
18 O and dD) in atmospheric and
ocean models, and the carbon isotopes (d
13 C) in biogeochemical models, in order to have a more direct comparison
between the output and the measurements. An example of the
application of one of these models is given in Chap. 29.
Another important difficulty concerns the chronology of
events. Climate models need boundary conditions and
forcings (e.g. variations of insolation, concentrations of
atmospheric CO 2 , sea level) in order to produce results such
as temperatures or precipitation. It is difficult to enter all of
the forcing parameters at the same chronological scale into
the model. It is also often problematic to compare the results
of these simulations with paleoclimate data whose time scale
is not clearly defined. This is especially true when using
short-term simulations, such as the results of general circulation models (atmospheric or coupled ocean-atmosphere
models). For example, to better understand how a glaciation
starts, simulations of the atmosphere, or of the
ocean-atmosphere system, are performed. But when is the
start of a glaciation? If marine isotopes can be relied upon,
the ice caps began to grow at around 120 or 122 ka BP. This
moment could then be simulated by imposing to the model
the insolation, greenhouse gas forcings etc. which best fit
this time period. In the case of an ‘equilibrium’ simulation
that numerically integrates the atmosphere over a few decades, or the ocean-atmosphere system over 1000–
2000 years, with constant forcings, this strategy is likely to
324
M. Kageyama and D. Paillard
In general, depending on the scientific question posed and
the computing power available, the first step is to identify the
part of the system which will be explicitly represented by the
model. If, for example, the intention is to represent the
evolution of the ice caps under the influence of the orbital
parameters of the Earth, it is essential to concentrate on a
model which explicitly represents the dynamic of the ice
caps. However, when the focus is on fast-reacting components (ocean, atmosphere), it is acceptable to fix the
slow-moving components by establishing these as boundary
limits. The converse is not true. Thus, the evolution of the
ice caps only makes sense when its interaction with a
changing climate system is considered. Herein lies one of the
main difficulties with modeling the climate over long periods
of time (both past and future). How can the interactions
(exchanges of energy and matter) between physical objects
with very different time constants be calculated in a meaningful way? Several strategies are possible.
A—If only a state of equilibrium in the system is of
interest, then the equilibrium of one of the systems (for
example, the ice caps) can be found by fixing another (e.g.
giving the atmosphere fixed boundary conditions), then by
performing the inverse operation (the atmosphere is calculated while the ice caps are fixed), then reiterating the process
until the results converge. This is called asynchronous coupling. In essence, the two physical components are coupled,
but in a way that does not reflect (or very inexactly) the real
flow of time. This makes it possible to achieve equilibrium of
the coupled system over a relatively short period of time (for
example, a few decades for the atmosphere), since the
fast-reacting component (the atmosphere is the factor which
demands the most computing time due to the fact that it has
inherent small-scale variations) is always calculated by
assuming an equilibrium with the slow component.
B—Although asynchronous coupling can be a good
method of calculating climate equilibria, it is not a very
rigorous way to calculate a meaningful evolution of the
system. This may provide a useful approximation if the time
constants of the physical systems are very different (e.g.,
atmosphere and ice caps), in which case it amounts to
considering that the slow component is the sole driver of the
evolution of the system with the fast component merely
adapting to the slow component. If physical components
with ‘intermediate’ times are included (such as the ocean,
which reacts more slowly than the atmosphere but faster than
the ice caps), this strategy may fail. Therefore, it is sometimes desirable to simplify the physics of fast-reacting
components (especially the atmosphere) in an attempt to
calculate only the long-term variations that can be used
directly by the slower components. This has led to the
development of models known as ‘intermediate complexity’
models (see Section “Earth system models of intermediate
complexity (EMICS)”).
Comparison with Paleoclimate Data
If our objective is to explain climate variations as reconstructed from paleoclimate records via modeling based as
much as possible on the physics, it is important to set up the
model and experimental design so that this comparison is
easiest. This gives rise to a second major difficulty in paleoclimatology simulations: paleoclimate indicators (proxies)
are never clear-cut in terms of the physical variables simulated by models. These indicators are dependent on particular
climate parameters, but are these the ones being simulated? It
is therefore risky to rely only on the same physical models as
those specially developed for comparison with current
observations, whose important parameters can be measured
by oceanographers, glaciologists and meteorologists. The
ideal situation is to be able to quantitatively compare paleoclimate simulations with paleoclimate indicators. The most
promising strategy is to explicitly simulate these indicators in
the models so that the multiple factors likely to influence them
are taken into account. For example, it is useful to explicitly
simulate the water isotopes (d
18 O and dD) in atmospheric and
ocean models, and the carbon isotopes (d
13 C) in biogeochemical models, in order to have a more direct comparison
between the output and the measurements. An example of the
application of one of these models is given in Chap. 29.
Another important difficulty concerns the chronology of
events. Climate models need boundary conditions and
forcings (e.g. variations of insolation, concentrations of
atmospheric CO 2 , sea level) in order to produce results such
as temperatures or precipitation. It is difficult to enter all of
the forcing parameters at the same chronological scale into
the model. It is also often problematic to compare the results
of these simulations with paleoclimate data whose time scale
is not clearly defined. This is especially true when using
short-term simulations, such as the results of general circulation models (atmospheric or coupled ocean-atmosphere
models). For example, to better understand how a glaciation
starts, simulations of the atmosphere, or of the
ocean-atmosphere system, are performed. But when is the
start of a glaciation? If marine isotopes can be relied upon,
the ice caps began to grow at around 120 or 122 ka BP. This
moment could then be simulated by imposing to the model
the insolation, greenhouse gas forcings etc. which best fit
this time period. In the case of an ‘equilibrium’ simulation
that numerically integrates the atmosphere over a few decades, or the ocean-atmosphere system over 1000–
2000 years, with constant forcings, this strategy is likely to
324
M. Kageyama and D. Paillard