Using Model Hierarchies to Better Understand Past Climate Change

International audienceAs soon as atmospheric general circulation models (AGCMs) have been operational for present or future climate studies, they have also been used to simulate past climates. The goal of this type of exercise has always been twofold: 1) to validate the models' ability to simulate climates different from the present one; 2) to understand the mechanisms that lead to past climate changes depicted by the numerous marine, continental and glaciological records. In the Paleoclimate Modelling Intercomparison Project (PMIP), the simulations from up to 18 different AGCMs have been compared to climate reconstructions for two periods: the Mid-Holocene, 6,000 years ago, and the Last Glacial Maximum, 21,000 years ago. Examples of the PMIP results for each period are presented here. Then, taking these PMIP results as a basis, we discuss the increasing complexity that can be taken into account in paleoclimate modelling: • Increase in the number of components of the climate system taken into account: complex models can now include not only the atmosphere, but also the oceans and the vegetation, allowing analyses of the feedback from these components onto the atmosphere, of their role in the establishment of climates different from ours. Taking a larger number of climate system components into account allows using more data for the validation of the models rather than for their boundary conditions. • Increase in the duration of the numerical experiments to study climate transitions rather than only equilibria: models including simpler representations of some components of the climate system are less expensive to be run than the comprehensive general circulation models mentioned above and give the opportunity of studying not only climate equilibria, but also climate transitions. The simplicity of this type of models and their low running cost allow the modellers to carry numerous sensitivity experiments to explore the importance of one forcing factor or one parameterisation. • Increase in the spatial scale of the simulations: the resolution of the models used in paleoclimate studies has improved, following the increase of super-computer peformances. However, the spatial scale of global circulation *The whole palaeoclimate modelling team from Laboratoire des Sciences du Climat et de l'Environnement contributed to the results presented here