Exploiting strength, discounting weakness: combining information from multiple climate simulators

This paper presents and analyses a statistical framework for combining projections of future climate from different climate simulators. The framework recognizes explicitly that all currently available simulators are imperfect; that they do not span the full range of possible decisions on the part of the climate modelling community; and that individual simulators have strengths and weaknesses. Information from individual simulators is automatically weighted, alongside that from historical observations and from prior knowledge. The weights for a simulator depend on its internal variability, its expected consensus with other simulators, the internal variability of the real climate and the propensity of simulators collectively to deviate from reality. The framework demonstrates, moreover, that some subjective judgements are inevitable when interpreting multiple climate change projections: by clarifying precisely what those judgements are, it provides increased transparency in the ensuing analyses. Although the framework is straightforward to apply in practice by a user with some understanding of Bayesian methods, the emphasis here is on conceptual aspects illustrated with a simplified artificial example. A 'poor man's version' is also presented, which can be implemented straightforwardly in simple situations.