Séminaires : Documents

Climate prediction: where are we?
François Massonnet (UCL)

The evolution of the atmosphere beyond a few days is governed, on the one hand, by variations in the so-called "slow" components of the climate system (oceans, sea ice, ice sheets, land) and, on the other hand, by changes in the atmospheric composition -- from natural or anthropogenic origins. Climate prediction accounts for these two sources of predictability and attempts to produce skillful and useful climate information at time scales ranging from a few weeks to a few decades.

Numerical models of the general circulation of climate (GCMs) are primary tools in this emerging area of research. They allow (1) exploring the very notion of predictability (what are the physical mechanisms that underpin potential predictions) and (2) delivering the predictions themselves. The use of GCMs is, however, not a cakewalk. These models are expensive in terms of computational and storage requirements, they feature systematic biases and their (ensemble) predictions are not always straightfoward to verify.

During this seminar, I will review recent advances in the area of climate prediction and will introduce several concrete examples linked to my research. In particular, I will aim at answering four questions: (1) What is a climate prediction system? (2) What is a good climate prediction? (3) Are today's climate predictions good? (4) How can these predictions be improved?

All welcome!