摘要：Abstract. The rationale for using multi-model ensembles in climate changeprojections and impacts research is often based on the expectation thatdifferent models constitute independent estimates; therefore, a range of modelsallows a better characterisation of the uncertainties in the representationof the climate system than a single model. However, it is known that researchgroups share literature, ideas for representations of processes,parameterisations, evaluation data sets and even sections of model code.Thus, nominally different models might have similar biases because ofsimilarities in the way they represent a subset of processes, or even benear-duplicates of others, weakening the assumption that they constituteindependent estimates. If there are near-replicates of some models, thentreating all models equally is likely to bias the inferences made using theseensembles. The challenge is to establish the degree to which this might betrue for any given application. While this issue is recognised by many in thecommunity, quantifying and accounting for model dependence in anything otherthan an ad-hoc way is challenging. Here we present a synthesis of the rangeof disparate attempts to define, quantify and address model dependence inmulti-model climate ensembles in a common conceptual framework, and provideguidance on how users can test the efficacy of approaches that move beyondthe equally weighted ensemble. In the upcoming Coupled Model IntercomparisonProject phase 6 (CMIP6), several new models that are closely related toexisting models are anticipated, as well as large ensembles from some models.We argue that quantitatively accounting for dependence in addition to modelperformance, and thoroughly testing the effectiveness of the approach usedwill be key to a sound interpretation of the CMIP ensembles in futurescientific studies.
