On the Level of Agreement Between Climate-Change Projections from Same-Centre Models

English Abstract:

Climate models developed by a same research group are prone to share structural similarities. It is also known that such dependencies may induce resembling climatic features in the models' simulations. The "same-centre" hypothesis, which combines these two assertions into a single one, suggests that two models from a same centre should lead to climate-change projections that agree to some extent. We explore this idea throughout an analysis of climate-change projections by considering several groups of same-centre climate models.

With some exceptions, the same-centre hypothesis appears as an efficient rule to filter out non-informative agreements between models and thus clarify the message conveyed by an ensemble of opportunity. Minor modelling differences often lead to non-informative agreements while non-generalized structural differences (e.g. changing the ocean component) may reveal disagreements that are limited to specific regions (e.g. in surface air temperature over the Hudson Bay). However, there are also cases where the same-centre hypothesis fails. One typical example is that of slight iterative changes between two versions of a same model that lead to highly differing climate sensitivities. Maybe more interestingly, major changes in all of the model components may correspond to highly similar projections. The latter result suggests two outcomes. First, the two models may have reach an informative agreement, i.e. a robust result that deserves a high confidence. But conversely, this might also reflect some higher-level dependencies, such as in institutional decisions made in model development or through similar practices for model validation and tuning.