SEMINAIRE LABO - Emmanuel Gobet [Ecole Polytechnique]

All the economic sectors and the society experiment climate change and the insurance/banking actors have to better assess the climate risks associated to this change. This is a delicate task due to the complexity of the global system to analyse. In this work, we focus on the climate block that maps trajectories of GHG emissions into gas concentrations in the atmosphere and oceans, radiative forcing, and increase/decrease of temperatures. We discuss how to replace the usual large-scale models (such as those in the Coupled Model Intercomparison Project CMIP6) by a fast and simple metamodel based on Neural Networks, with equal accuracy.