Estimates of the nature of future climate change typically come from a range of different climate models based at different institutes around the world. Having information from more than one model is useful, since there are many uncertainties in climate projections, and using multiple models (as well as a range of simulations from each of them) gives us a collection of independent projection estimates.

But how independent are these estimates? How can we tell?

Answers to these questions determine whether or not agreement between models is a sign of robustness.

In this review paper, we contextualise the broad and seemingly disparate range of attempts to define and address model dependence within climate model ensembles, and offer concrete advice on how best to avoid overconfidence.