ľ¹Ï¸£ÀûÓ°ÊÓ

The role of model-to-model differences and internal variability in causing uncertainty in climate projections

The value of single model initial-condition large ensembles (SMILEs) comes from the ability to precisely quantify both the forced response and internal variability in a single model. The use of multiple SMILEs allows the quantification of model-to-model differences in the forced response as well as model-to-model differences in the internal variability itself.
We first investigate short-term trends (15-years) to determine the range of possible future surface temperature trends that we could observe. We find that cooling could be observed at any individual point on the globe due to internal variability, although we are more likely to observe warming. We then quantify the uncertainty due to model-to-model differences, the forcing scenario used, and internal variability. We find that the short-term trends are dominated by internal variability with little influence of which model or scenario is used. This result is robust to using the largest and smallest estimate of internal variability from the models. Next we investigate long-term projections under strong forcing (RCP8.5). We quantify the magnitude of uncertainty due to model-to-model differences and internal variability for temperature, precipitation, and their temporal variability. We find that model-to-model differences are larger than internal variability for both temperature and precipitation changes, but that internal variability is larger than model-to-model differences for the temporal variability of both quantities in the extra-tropics. We then provide a new method that produces similar results using CMIP5. Finally we investigate model-to-model differences in the forced response in six SMILEs in Europe, Australia, south-east Asia and the tropical Pacific and show that even when the sign of the change agrees across models, the magnitude of the change can be widely different.