École Nationale Supérieure de
Techniques Avancées

Résumé

In the research reported in this paper, we aim at building an emulator for 6 different integrated assessement models : AIM/CGE, GCAM, IMAGE, MESSAGE-GLOBIOM, REMIND-MAGPIE and WITCH for SSP2 scenarios. The study focus on two main outputs which are the emissions abatement rate and the carbon price. We first find a common structure shared by every integrated assessment model for the two outputs. For each output, the six emulators differ on parameters called the first class parameters and giving important behavior characteristics of the model. We then separate these parameters into two components, one scenario specific and another one model specific. The scenario specific component is a function of the mitigation target and the model specific component is a function of parameters called second class parameters. The second class parameters are found with a least square regression to fit best the data. The fact that our emulators share a same model structure allows us to compare them easily and to extract important model behavior characteristics.