25
Modeling and Paleoclimatology
Masa Kageyama and Didier Paillard
Why Develop Paleoclimate Models?
Today, climate models are widely discussed because of the
climate predictions they produce for the next century (see
Chap. 31), particularly when the IPCC (Intergovernmental
Panel on Climate Change) assessment reports are published.
The models used for climate prediction are developed first
and foremost based on the data from observations of recent
decades. Before being applied to forecasting, these models
are evaluated to assess their ability to reproduce the present
climate and its recent variations over the last decades. They
represent our understanding of the current climate and of the
mechanisms that play important roles in recent variations,
but also our ability to translate this understanding into digital
codes which is necessarily limited by computing capabilities. Thus, even though the most powerful computers are
currently used to produce simulations with the most complex
climate models, they are limited by computing power in
terms of resolution and system complexity. We will return
later to this crucial question of finding the best compromise
between the desired timespan for a climate simulation and
the complexity of the climate model used, a question that is
even more critical in paleoclimatology, where the time scales
are much longer than for the IPCC predictions (which last a
few centuries at most).
It is easy to understand why models are used for the
forecasting of future climates: this approach, based on our
understanding of the physics of the climate system, provides
the only means of obtaining these forecasts. But modeling
can also contribute a lot to our understanding of current and
past climates which are characterized through observations
and reconstructions. A first justification is the evaluation of
the models used for future forecasts. During the
development phase of a model, the first step in evaluation is
to compare it to the available observations for the current
climate and its short-term variability over time scales ranging from a few years to a few decades. In recent decades,
anthropogenic disruption has created previously unknown
levels of atmospheric greenhouse gas concentrations, so
models need to be tested in climate configurations different
to the current climate. There is no perfect paleoclimate
equivalent for the forcing of anthropogenic disturbance, but
paleoclimates, even if they are not recorded with the same
precision as the present climate, offer examples of climates
very different from the current one and of transitions of
varying rapidity between states. If we have correctly
understood the climate system and if we want to use this
understanding to predict climate in the future, we must be
able to reproduce the variations in past climates.
An example of an evaluation based on paleoclimates is
one dealing with entry into the last glacial period, also called
last glacial inception. This occurred about 115,000 years
ago, and the major cause of this disturbance, external to the
climate system, is the difference in insolation received by the
Earth. In principle, a climate model, even if it does not
include an ice sheet model, should be able to simulate
perennial snow cover at the formation sites of the first ice
caps located north of present-day Canada. Following various
trials, it was found that very few atmospheric general circulation models were capable of simulating this perennial
cover. Furthermore, those that did manage to do so were
often ones that simulated a current climate that was much
colder than observations. These trials were therefore far from
satisfactory. However, as components (ocean, vegetation)
were added one by one to the climate models, it was found
that these components could play an important role in
amplifying the initial insolation signal. Thus, over a longer
period, the disappearance of forests or the appearance of sea
ice, initiated by changes in insolation, support the development of perennial snow cover by modifying the albedo. It
can therefore be concluded that it would be difficult to
M. Kageyama (&) Á D. Paillard
Laboratoire des Sciences du Climat et de l’Environnement,
LSCE/IPSL, CEA-CNRS-UVSQ, Université Paris-Saclay,
91190 Gif-Sur-Yvette, France
e-mail: masa.kageyama@lsce.ipsl.fr
© Springer Nature Switzerland AG 2021
G. Ramstein et al. (eds.), Paleoclimatology, Frontiers in Earth Sciences,
https://doi.org/10.1007/978-3-030-24982-3_25
319
Modeling and Paleoclimatology
Masa Kageyama and Didier Paillard
Why Develop Paleoclimate Models?
Today, climate models are widely discussed because of the
climate predictions they produce for the next century (see
Chap. 31), particularly when the IPCC (Intergovernmental
Panel on Climate Change) assessment reports are published.
The models used for climate prediction are developed first
and foremost based on the data from observations of recent
decades. Before being applied to forecasting, these models
are evaluated to assess their ability to reproduce the present
climate and its recent variations over the last decades. They
represent our understanding of the current climate and of the
mechanisms that play important roles in recent variations,
but also our ability to translate this understanding into digital
codes which is necessarily limited by computing capabilities. Thus, even though the most powerful computers are
currently used to produce simulations with the most complex
climate models, they are limited by computing power in
terms of resolution and system complexity. We will return
later to this crucial question of finding the best compromise
between the desired timespan for a climate simulation and
the complexity of the climate model used, a question that is
even more critical in paleoclimatology, where the time scales
are much longer than for the IPCC predictions (which last a
few centuries at most).
It is easy to understand why models are used for the
forecasting of future climates: this approach, based on our
understanding of the physics of the climate system, provides
the only means of obtaining these forecasts. But modeling
can also contribute a lot to our understanding of current and
past climates which are characterized through observations
and reconstructions. A first justification is the evaluation of
the models used for future forecasts. During the
development phase of a model, the first step in evaluation is
to compare it to the available observations for the current
climate and its short-term variability over time scales ranging from a few years to a few decades. In recent decades,
anthropogenic disruption has created previously unknown
levels of atmospheric greenhouse gas concentrations, so
models need to be tested in climate configurations different
to the current climate. There is no perfect paleoclimate
equivalent for the forcing of anthropogenic disturbance, but
paleoclimates, even if they are not recorded with the same
precision as the present climate, offer examples of climates
very different from the current one and of transitions of
varying rapidity between states. If we have correctly
understood the climate system and if we want to use this
understanding to predict climate in the future, we must be
able to reproduce the variations in past climates.
An example of an evaluation based on paleoclimates is
one dealing with entry into the last glacial period, also called
last glacial inception. This occurred about 115,000 years
ago, and the major cause of this disturbance, external to the
climate system, is the difference in insolation received by the
Earth. In principle, a climate model, even if it does not
include an ice sheet model, should be able to simulate
perennial snow cover at the formation sites of the first ice
caps located north of present-day Canada. Following various
trials, it was found that very few atmospheric general circulation models were capable of simulating this perennial
cover. Furthermore, those that did manage to do so were
often ones that simulated a current climate that was much
colder than observations. These trials were therefore far from
satisfactory. However, as components (ocean, vegetation)
were added one by one to the climate models, it was found
that these components could play an important role in
amplifying the initial insolation signal. Thus, over a longer
period, the disappearance of forests or the appearance of sea
ice, initiated by changes in insolation, support the development of perennial snow cover by modifying the albedo. It
can therefore be concluded that it would be difficult to
M. Kageyama (&) Á D. Paillard
Laboratoire des Sciences du Climat et de l’Environnement,
LSCE/IPSL, CEA-CNRS-UVSQ, Université Paris-Saclay,
91190 Gif-Sur-Yvette, France
e-mail: masa.kageyama@lsce.ipsl.fr
© Springer Nature Switzerland AG 2021
G. Ramstein et al. (eds.), Paleoclimatology, Frontiers in Earth Sciences,
https://doi.org/10.1007/978-3-030-24982-3_25
319