Reference
Sheffield, J., Barrett, A.P., Colle, B., Fernando, D.N., Fu, R., Giel, K.L., Hu, Q., Kinter, J., Kumar, S., Langenbrunner, B., Lombardo, K., Long, L.N., Maloney, E., Mariotti, A., Meyerson, J.E., Mo, K.C., Neelin, J.D., Nigam, S., Pan, Z., Ren, T., Ruiz-Barradas, A., Serra, Y.L., Seth, A., Thibeault, J.M., Stroeve, J.C., Yang, Z. and Yin, L. 2013. North American climate in CMIP5 experiments. Part I: Evaluation of historical simulations of continental and regional climatology. Journal of Climate 26: 9209-9245.
Background
As time progresses, so too does work on climate models also progress, from one stage to another, with the most recent step forward being that which was taken in advancing from the group of CMIP3 models to the group of CMIP5 models.
What was done
Focusing on a core set of 17 CMIP5 models that "represent a large set of climate centers and model types," Sheffield et al. evaluated the abilities of these models to reproduce in retrospect a number of observational data sets.
What was learned
The 27 researchers report that "the performance of the CMIP5 models in representing observed climate features has not improved dramatically compared to CMIP3." They note, for example, that "there are some models that have improved for certain features (e.g., the timing of the North American monsoon)," but they say that others "have become worse" in terms of the more basic "continental seasonal surface climate." And, "furthermore," as they conclude, "the uncertainty in the future projections across models can also be of the same magnitude [as] the model spread for the historic period."
What it means
Even a little progress is better than none, but not much better, especially when one considers all of the time, funds and person-power that went into the enterprise.