The role of data and priors in estimating climate sensitivity

Abstract

In Bayesian theory, the data together with the prior produce a posterior. We show that it is also possible to follow the opposite route, that is, to use data and posterior information (both of which are observable) to reveal the prior (which is not observable). We then apply the theory to equilibrium climate sensitivity as reported by the Intergovernmental Panel on Climate Change in an attempt to get some insight into the prior beliefs of the IPCC scientists. It appears that the data contain much less information than one might think, due to the presence of correlation. We conclude that the prior in the fifth IPCC report was too low, and in the sixth report too high.