water vapor are redistributed within the atmosphere, cannot
be represented in a useful way in the radiative-convective
models. Thus, even rough estimates of the changes in the
global mean temperature require the atmospheric dynamics
to be taken into consideration and, for more precise calculations, three-dimensional models representing the general
circulation of the global atmosphere of the Earth are
needed.
General Circulation Models: Progress
and Limitations
The Evolution of Climate Models
A general circulation climate model is a simplified representation of the climate system, but including as best as
possible most processes that influence the climate. It is based
on a preliminary physical analysis, in order to reduce the
number of processes to be incorporated, as well as on
appropriate mathematical and numerical formulations.
Numerical modeling of the climate began in the 1970s and
has since greatly progressed, thanks to the steady increase in
the processing power of computers. The general philosophy
behind this development, established by Charney and his
collaborators in the 1950s, was to understand the problem on
a global scale, even if this meant making very general
approximations initially, with the aim of gradually improving the model by identifying its drawbacks and its limitations. The first models only described the atmosphere and
the continental surfaces. In order to reduce model complexity, the oceanic surface temperature was imposed: even
if the energy balance of the ocean surface is very different
from the observed measurement, the surface temperature is
maintained at its prescribed value. However, to investigate
the variations in climate accounting for oceanic temperatures
is required.
The first studies of the impact of a CO 2 doubling using
this type of model were carried out in the 1970s at the GFDL
(Geophysical Fluid Dynamics Laboratory, Princeton, USA)
with a representation of the ocean with no circulation and
zero heat capacity, so that equilibrium with the atmosphere
could be achieved quickly. In this model, there was no
diurnal or annual variation in insolation, and ad hoc corrections were applied to heat fluxes at the air-sea interface to
keep the ocean surface temperature close to observations.
The use of this type of model became widespread during the
1980s, with a gradual increase in sophistication and realism.
For example, both annual and diurnal insolation variations
were incorporated, modeling of cloud formation processes
began, etc. At the same time, new satellites were being used
to estimate global cloud cover and radiative fluxes at the top
of the atmosphere, which contributed to the improvement
and evaluation of atmospheric models.
In parallel, ocean general circulation models were
developed to simulate heat transport and to study the role of
the ocean in the Earth’s energy balance. Progressively, they
included sea-ice models and, from the 1990s onwards, they
were coupled with atmospheric models to simulate the
atmosphere-ocean-sea ice interactions. These early models
did not simulate the heat and water fluxes at the air-sea
interface, resulting in strong biases in the simulated oceanic
surface temperatures. To fix these shortcomings, the fluxes at
the air-sea interface were corrected in an ad hoc way, before
these corrections were gradually removed after the end of the
1990s, thanks to continuous model improvement.
These coupled atmosphere-ocean models gradually
became the basic tool for studying both past and future
climate variations. For example, in preparation for the
Fourth IPCC Assessment Report (IPCC 2007), some twenty
of these coupled models performed a whole series of climate
change simulations, and only six of them were based on flux
correction at the air-sea interface. These models can simulate
a natural climate variability that can then be compared to
observations at different time scales: a few days, a few years
(interannual variability, the best known of which is El Niño),
or a few tens, or even several hundred years.
In a schematic way, climate models simulate the energy
and water cycles. Progressively, representations of chemical
reactions in the atmosphere, biogeochemical cycles and the
transport of species were introduced into modeling so as to
study new aspects of climate variations: the effect of aerosols, coupling between climate change and the chemical
composition of the atmosphere, and between climate change
and the carbon and methane cycles. This required advances
in our understanding of each of the components of the
system: atmosphere, ocean, vegetation, continental surface,
sea ice. Numerical climate models progressively incorporate,
in a coherent way, a wide range of physical processes
governing the climate variations and the interactions
between the different climate components. On the other
hand, the inclusion of a growing number of physical processes has made the models more complex and more difficult
to develop and evaluate.
As ice sheets interact with atmosphere, ocean and vegetation, the next key challenge for the climate modeling
community is to incorporate ice-sheet models in general
circulation climate models. This is a necessary step to obtain
a comprehensive representation of the climate system for
past, present and future time periods.
What Are the Uncertainties Inherent in Climate
Models?
The climate is characterized by a very wide range of both
spatial (from the micrometer to several thousand kilometers)
and temporal (from the second to several thousand years or
more) scales. The processes at these different scales interact
460
S. Charbit et al.
be represented in a useful way in the radiative-convective
models. Thus, even rough estimates of the changes in the
global mean temperature require the atmospheric dynamics
to be taken into consideration and, for more precise calculations, three-dimensional models representing the general
circulation of the global atmosphere of the Earth are
needed.
General Circulation Models: Progress
and Limitations
The Evolution of Climate Models
A general circulation climate model is a simplified representation of the climate system, but including as best as
possible most processes that influence the climate. It is based
on a preliminary physical analysis, in order to reduce the
number of processes to be incorporated, as well as on
appropriate mathematical and numerical formulations.
Numerical modeling of the climate began in the 1970s and
has since greatly progressed, thanks to the steady increase in
the processing power of computers. The general philosophy
behind this development, established by Charney and his
collaborators in the 1950s, was to understand the problem on
a global scale, even if this meant making very general
approximations initially, with the aim of gradually improving the model by identifying its drawbacks and its limitations. The first models only described the atmosphere and
the continental surfaces. In order to reduce model complexity, the oceanic surface temperature was imposed: even
if the energy balance of the ocean surface is very different
from the observed measurement, the surface temperature is
maintained at its prescribed value. However, to investigate
the variations in climate accounting for oceanic temperatures
is required.
The first studies of the impact of a CO 2 doubling using
this type of model were carried out in the 1970s at the GFDL
(Geophysical Fluid Dynamics Laboratory, Princeton, USA)
with a representation of the ocean with no circulation and
zero heat capacity, so that equilibrium with the atmosphere
could be achieved quickly. In this model, there was no
diurnal or annual variation in insolation, and ad hoc corrections were applied to heat fluxes at the air-sea interface to
keep the ocean surface temperature close to observations.
The use of this type of model became widespread during the
1980s, with a gradual increase in sophistication and realism.
For example, both annual and diurnal insolation variations
were incorporated, modeling of cloud formation processes
began, etc. At the same time, new satellites were being used
to estimate global cloud cover and radiative fluxes at the top
of the atmosphere, which contributed to the improvement
and evaluation of atmospheric models.
In parallel, ocean general circulation models were
developed to simulate heat transport and to study the role of
the ocean in the Earth’s energy balance. Progressively, they
included sea-ice models and, from the 1990s onwards, they
were coupled with atmospheric models to simulate the
atmosphere-ocean-sea ice interactions. These early models
did not simulate the heat and water fluxes at the air-sea
interface, resulting in strong biases in the simulated oceanic
surface temperatures. To fix these shortcomings, the fluxes at
the air-sea interface were corrected in an ad hoc way, before
these corrections were gradually removed after the end of the
1990s, thanks to continuous model improvement.
These coupled atmosphere-ocean models gradually
became the basic tool for studying both past and future
climate variations. For example, in preparation for the
Fourth IPCC Assessment Report (IPCC 2007), some twenty
of these coupled models performed a whole series of climate
change simulations, and only six of them were based on flux
correction at the air-sea interface. These models can simulate
a natural climate variability that can then be compared to
observations at different time scales: a few days, a few years
(interannual variability, the best known of which is El Niño),
or a few tens, or even several hundred years.
In a schematic way, climate models simulate the energy
and water cycles. Progressively, representations of chemical
reactions in the atmosphere, biogeochemical cycles and the
transport of species were introduced into modeling so as to
study new aspects of climate variations: the effect of aerosols, coupling between climate change and the chemical
composition of the atmosphere, and between climate change
and the carbon and methane cycles. This required advances
in our understanding of each of the components of the
system: atmosphere, ocean, vegetation, continental surface,
sea ice. Numerical climate models progressively incorporate,
in a coherent way, a wide range of physical processes
governing the climate variations and the interactions
between the different climate components. On the other
hand, the inclusion of a growing number of physical processes has made the models more complex and more difficult
to develop and evaluate.
As ice sheets interact with atmosphere, ocean and vegetation, the next key challenge for the climate modeling
community is to incorporate ice-sheet models in general
circulation climate models. This is a necessary step to obtain
a comprehensive representation of the climate system for
past, present and future time periods.
What Are the Uncertainties Inherent in Climate
Models?
The climate is characterized by a very wide range of both
spatial (from the micrometer to several thousand kilometers)
and temporal (from the second to several thousand years or
more) scales. The processes at these different scales interact
460
S. Charbit et al.