Caractère unique des sciences du climat

Séance organisée avec le soutien de l’école doctorale UNITER (Sciences de l’Univers, de l’Espace, de la Terre et du Climat).

Daniel Williamson

Daniel Williamson est statisticien (University of Exeter). Il a effectué des recherches sur la prise de décision et la quantification des incertitudes en rapport avec les problèmes ayant recours aux simulations numériques. Dans son travail, il s’est particulièrement intéressé aux questions liées au changement climatique, à la modélisation climatique, au calibrage et au paramétrage, et à l’utilisation des statistiques bayésiennes dans le contexte de la modélisation climatique.

Résumé

Through the Coupled Model Inter-comparison projects (CMIP) and a host of other satellite “MIPs”, climate science has developed its own unique approach to scientific discovery. By comparing the world’s different climate models and their component parts across a series of centrally designed experiments on tuned runs, the field tests hypotheses and offers uncertainty quantification that is used to support policy makers in a way that is different from any other scientific field that uses computational models to understand the world. In this talk I will explore the question of whether or not this is sensible. Starting with the question of what climate models are and proceeding from there, I will explore the structure of climate science. I will focus particularly on uncertainty quantification : what it is, why it is important and how it is approached in climate science, including model tuning and estimation of model error. Finally, I will explore how MIPs might be optimally designed and used to quantify uncertainty.