Geoscience Reference
In-Depth Information
new environmental conditions, which would lead to the disappearance
of some biotopes, thus reinforcing the local greenhouse effect in some
regions (for example this is what occurred in France during the heat
wave of 2003).
The question of what climatic warming is attributable to is crucial
and has been explored with the help of coordinated experiments with
numerical simulation. To do this, numerical climate models are used,
which describe the evolution of the two fluids, the ocean and the
atmosphere, taking account of their exchanges with ice and
vegetation, under the imposed boundary conditions. An initial group
of experiments is carried out by assessing the evolution conditions of
solar forcing and volcanic eruptions since 1850. Then a second group
of experiments follows the same experimental protocol with the
addition of the radiative contribution of GHG and aerosols. We note
here the important work of coordinated experiments that has been in
place for the last 15 years within the framework of the
Climate
Modeling Intercomparison Program
(CMIP) group of the World
Climate Research Programme. The climate models used over the last
20 years have steadily improved over the course of time.
Although they show the main characteristics of regional climates,
biases with observation persist in certain regions and show the
difficulty of correctly representing certain phenomena. These biases
are difficult to reduce, notably because they often result from
interactions between different processes. For example, the choice of
parameters to represent clouds can vary from one modeling
group to another, leading to significant regional differences.
Moreover, the atmosphere and the ocean are turbulent fluids and
divergent trajectories can result rapidly from the behavior of small
structures. To work around these difficulties, we do not only consider
one simulation but rather an ensemble of several simulations carried
out with slight alterations in their initial state or in the choice of
parameters. Each model produces a set of several simulations of
which it is hoped the average follows the observed trajectory. By
combining the average trajectories of several models, it is possible to
estimate the range of uncertainty resulting from the biases of the
