Abstract
Choices of adaption strategies to climate change are often based on the knowledge summarized in the Intergovernmental Panel on Climate Change ( IPCC ) Assessment Reports. For climate impact research, precipitation is of high relevance and often used as a criterion for climate model evaluation. Therefore, we assess simulated precipitation of 24 models of Phase 3 and 26 models of Phase 5 of the Coupled Model Intercomparison Project ( CMIP ) of the latest Assessment Reports of the IPCC , altogether 149 individual simulations. These model simulations and the two multi‐model means are evaluated against five observational datasets from different sources over the late 20th century. Monthly, seasonal and annual means are analysed based on five selected study areas, leading to a comparative regional assessment of model quality. For the more recent CMIP ensemble mean, the overall bias has been reduced. Aside from some underestimation in summer and overestimation in winter in wet regions, there is still a considerable model bias (40–150% compared to observational data) especially in the arid areas of our study. None of the single models is superior in all five regions, although for all regions models can be identified that show excellent results. Additionally, the relationship between horizontal and vertical model resolution and the model bias is evaluated. Despite some tendencies, the results show no significant dependency of model performance on the range of resolution in CMIP3 and CMIP5 . Thus, more sophisticated metrics have to be applied to construct subensembles which have higher skill and lower spread of simulated precipitation.