Climate model genealogy: Generation CMIP5 and how we got there

A new ensemble of climate models is becoming available and provides the basis for climate change projections. Here we show a first analysis indicating that the models in the new ensemble agree better with observations than those in older ones, and that the poorest models have been eliminated. Most models are strongly tied to their predecessors, andsome also exchange ideas and code with other models, thus supportingan earlier hypothesis that the models in the new ensemble are not independent of each other, nor independent of the earlier generation. Based on one atmosphere model, we show how statistical methods can identify similarities between model versions and complement process understanding in characterizing how and why a model has changed. We argue that the inter-dependence of models complicates the interpretation of multi model ensembles, but largely goes unnoticed.