| Abstract |
In order for our society to adapt to or mitigate the future climate change, it is very important to have a reliable future prediction. For this purpose, we should evaluate climate models and also have probabilistic predictions by considering that models are not perfect.For the probabilistic prediction, we have to create “ensembles” of model simulations by sampling uncertainties included in climate models. In this study, we evaluate the reliability of state-of-the-art climate model ensembles using a new method. Here we use the “Coupled Modeling Intercomparison Project Phase 3 (CMIP3)” ensemble, which was created by the world climate research centers for the IPCC forth assessment report.We compare the results of historical simulations and observations in order to assess the reliability of the climate model ensembles. We find that the CMIP3 ensemble is reliable enough. Our analysis also reveals that it is very important for climate model ensembles to have a structural variety for the reliability. A great deal of recent research on climate change is based on the results of CMIP3 ensembles, and thus our finding supports the reliability of these studies. |
