analyzed by separating the impact of the ice caps from the
impact of the reduction of atmospheric greenhouse gas
concentration. This figure refers to the annual average temperature of air 2 m above the surface. The values for this
temperature in the pre-industrial climate simulation are
shown in Fig. 25.5a and the differences between the simulated LGM and pre-industrial climates are shown in
Fig. 25.5d. This map shows an overall global cooling,
moderate (by a few degrees approximately) over the oceans
but very strong (more than 30 °C) over the ice sheets of the
northern hemisphere. Figure 25.5b shows that the contribution of greenhouse gases to this response is less extreme over
the continents and a little weaker over the oceans than the
response to all of the LGM conditions. Figure 25.5c shows
the impact of the ice caps alone. It shows that these are
responsible for a significant cooling over the continents of
the northern hemisphere, but also for a cooling of between 0
and 2 °C over most of the oceans, with the notable exception
of the Southern Ocean. It is worth noting that the sum of the
anomalies shown in Fig. 25.5b and d is not equal to the
difference between the LGM and pre-industrial climates
shown in Fig. 25.5d. This difference is shown in Fig. 25.5e.
This shows that in many regions the impact of the two
factors taken together is greater than the sum of the impacts
of each factor considered separately. This is referred to as the
synergy between the various forcings. In other areas, such as
north of the Nordic Seas, the impact of the two forcings
together is lower than the sum of the impacts of the individual forcings. This shows that to quantify the impact of a
specific forcing within a group of two forcings (as in this
case, the impact of the ice caps and the reduction in greenhouse gases), four simulations must be carried out: a control
simulation (in this case, the pre-industrial climate), one
including all the forcings (in this case, the LGM) and simulations with each factor taken individually. This method is
called the ‘factor separation’ method developed for the
atmospheric sciences by Stein and Alpert (1993).
So far, we have studied sensitivity experiments on the
forcings and boundary conditions imposed on models.
Sensitivity experiments may also be applied to internal
processes of the climate system. To examine the importance
of this processes in the response by the climate system to a
disturbance, its formulation in a model can be modified. For
example, if deep convection at the equator is suspected to
have an important influence on some aspect of the climate in
the mid-latitudes, the formulation for deep convection in the
model can be altered and a sensitivity experiment for this
process can be performed with a modified model, under the
same boundary conditions. This type of sensitivity experiment also makes it possible to evaluate the importance of a
feedback by excluding or activating it.
Outlook
In this section, we mainly describe atmospheric general
circulation models, including some coupled with oceanic
general circulation models. These models, now complemented by vegetation models, carbon cycle models and
atmospheric chemistry models, are becoming increasingly
complex, with more components of the climate system, more
processes and more associated feedbacks being included.
The complex models of the Earth system have sometimes
been described as the ‘biggest’ models that can run on the
‘biggest’ computers, a sort of ‘maximum’ model. These limit
the number of numerical experiments that can be run for
each given problem and the number of sensitivity experiments that can be conducted to better understand the influence of a particular process or mechanism. This situation is
changing, as advances in computing now allow modelers to
carry out more experiments for a given period. These
experiments are essential to improve our understanding of
the importance of forcings or processes within a change in
climate. New developments in complex climate system
models, the improvement in their resolution, the inclusion of
new processes or new components must consider the necessary compromise between the computational time required
for a simulation and the number of simulations that can be
performed with the computer available. With increasing
computing capabilities, it is possible to improve the resolution of the models and increase the number of processes
included. The models used for the IPCC assessment exercises provide a good indication of the progress of climate
modeling over the past two decades (Fig. 25.2). In the
future, we will see new components of the climate system
being integrated as for instance ice caps. The aim of these
developments is mainly to provide a better prediction of the
future climate, but they also contribute greatly to the study of
paleoclimates, which in turn makes it possible to evaluate
these models under climate conditions different to current
ones.
Earth System Models of Intermediate
Complexity (EMICS)
Basic Principles and History
We have seen that climate models developed from general
circulation models were intended to be as comprehensive as
possible in their representation of the climate system. This
requires considerable computing power and calculation times,
which in practice forces the modeler to limit the number of
simulations performed. These simulations are also quite short
334
M. Kageyama and D. Paillard
impact of the reduction of atmospheric greenhouse gas
concentration. This figure refers to the annual average temperature of air 2 m above the surface. The values for this
temperature in the pre-industrial climate simulation are
shown in Fig. 25.5a and the differences between the simulated LGM and pre-industrial climates are shown in
Fig. 25.5d. This map shows an overall global cooling,
moderate (by a few degrees approximately) over the oceans
but very strong (more than 30 °C) over the ice sheets of the
northern hemisphere. Figure 25.5b shows that the contribution of greenhouse gases to this response is less extreme over
the continents and a little weaker over the oceans than the
response to all of the LGM conditions. Figure 25.5c shows
the impact of the ice caps alone. It shows that these are
responsible for a significant cooling over the continents of
the northern hemisphere, but also for a cooling of between 0
and 2 °C over most of the oceans, with the notable exception
of the Southern Ocean. It is worth noting that the sum of the
anomalies shown in Fig. 25.5b and d is not equal to the
difference between the LGM and pre-industrial climates
shown in Fig. 25.5d. This difference is shown in Fig. 25.5e.
This shows that in many regions the impact of the two
factors taken together is greater than the sum of the impacts
of each factor considered separately. This is referred to as the
synergy between the various forcings. In other areas, such as
north of the Nordic Seas, the impact of the two forcings
together is lower than the sum of the impacts of the individual forcings. This shows that to quantify the impact of a
specific forcing within a group of two forcings (as in this
case, the impact of the ice caps and the reduction in greenhouse gases), four simulations must be carried out: a control
simulation (in this case, the pre-industrial climate), one
including all the forcings (in this case, the LGM) and simulations with each factor taken individually. This method is
called the ‘factor separation’ method developed for the
atmospheric sciences by Stein and Alpert (1993).
So far, we have studied sensitivity experiments on the
forcings and boundary conditions imposed on models.
Sensitivity experiments may also be applied to internal
processes of the climate system. To examine the importance
of this processes in the response by the climate system to a
disturbance, its formulation in a model can be modified. For
example, if deep convection at the equator is suspected to
have an important influence on some aspect of the climate in
the mid-latitudes, the formulation for deep convection in the
model can be altered and a sensitivity experiment for this
process can be performed with a modified model, under the
same boundary conditions. This type of sensitivity experiment also makes it possible to evaluate the importance of a
feedback by excluding or activating it.
Outlook
In this section, we mainly describe atmospheric general
circulation models, including some coupled with oceanic
general circulation models. These models, now complemented by vegetation models, carbon cycle models and
atmospheric chemistry models, are becoming increasingly
complex, with more components of the climate system, more
processes and more associated feedbacks being included.
The complex models of the Earth system have sometimes
been described as the ‘biggest’ models that can run on the
‘biggest’ computers, a sort of ‘maximum’ model. These limit
the number of numerical experiments that can be run for
each given problem and the number of sensitivity experiments that can be conducted to better understand the influence of a particular process or mechanism. This situation is
changing, as advances in computing now allow modelers to
carry out more experiments for a given period. These
experiments are essential to improve our understanding of
the importance of forcings or processes within a change in
climate. New developments in complex climate system
models, the improvement in their resolution, the inclusion of
new processes or new components must consider the necessary compromise between the computational time required
for a simulation and the number of simulations that can be
performed with the computer available. With increasing
computing capabilities, it is possible to improve the resolution of the models and increase the number of processes
included. The models used for the IPCC assessment exercises provide a good indication of the progress of climate
modeling over the past two decades (Fig. 25.2). In the
future, we will see new components of the climate system
being integrated as for instance ice caps. The aim of these
developments is mainly to provide a better prediction of the
future climate, but they also contribute greatly to the study of
paleoclimates, which in turn makes it possible to evaluate
these models under climate conditions different to current
ones.
Earth System Models of Intermediate
Complexity (EMICS)
Basic Principles and History
We have seen that climate models developed from general
circulation models were intended to be as comprehensive as
possible in their representation of the climate system. This
requires considerable computing power and calculation times,
which in practice forces the modeler to limit the number of
simulations performed. These simulations are also quite short
334
M. Kageyama and D. Paillard