Perspectives on CMIP5 model performance in the Nile River headwaters regions

Ranking the performance of global climate models (GCMs) is a notoriously difficult exercise. Multi-model comparison studies nearly always show that each model has strengths and weaknesses relative to others, and for many purposes the multi-model ensemble mean delivers better estimates than any individual model. Nevertheless, in regions like East Africa, where there is little consensus between models on the magnitude or sign of 21st century precipitation change, the multi-model ensemble mean approach to climate projection provides little value for adaptation planning. Here, we consider several possible frameworks for model evaluation and ranking, and assess the differences in performance of a subset of models participating in the 5th Coupled Model Intercomparison Project (CMIP5) according to each framework. Our test case is precipitation in the Nile River headwaters regions. We find that there is little consistency in the relative performance of models across frameworks based on amount and seasonality of precipitation, interannual precipitation variability, precipitation teleconnections, and continental scale climate patterns. These analyses offer some guidance on which GCMs are most likely to provide meaningful results for specific applications, but they caution that any effort to select 'best performing' GCMs for the Nile River basin must carefully consider the purposes for which GCMs are being selected.