When presenting the results of climate models to the public and to policy makers, their ability to reproduce the general features of historic changes in global mean temperature are highlighted as a demonstration of their reliability and their relevance for informing us about the future. I will present results which imply that achieving this consistency with observations in the context of model errors and biases, inherently limits the ability of these models to inform us about the consequences of nonlinear behaviour and thus provides a conservative view of possible futures. It also restricts the potential for these models, and for multi-model ensembles, to inform us about the regional changes consistent with any particular global mean change.
If time allows I will touch on the friction between the demand for policy relevant, local information about future climate and the ability of climate science to provide such information from models. The recent launch of the 2018 UK Climate Projections provides a context for reflexions on this issue.