Climate model evaluation

Climate models can be extremely complex, and involve interactions and feedbacks among different components of the climate system. The resulting climate predictions are often made using the outputs of 20+ different climate models, which leads to a wide spread of future climate projections. However, since some components are shared among some climate models, the multi-model mean response is not truly independent. ML can help identify and leverage relationships between variables within climate models, which, together with the observed climate changes (i.e., observational constraint) could narrow down the spread in the future climate projections.