Lack of scaling in global climate models

Detrended fluctuation analysis is used to test the performance of global climate models. We study the temperature data simulated by seven leading models for the greenhouse gas forcing only (GGFO) scenario and test their ability to reproduce the universal scaling (persistence) law found in the real records for four sites on the globe: (i) New York, (ii) Brookings, (iii) Tashkent and (iv) Saint Petersburg. We find that the models perform quite differently for the four sites and the data simulated by the models lack the universal persistence found in the observed data. We also compare the scaling behaviour of this scenario with that of the control run where the CO₂ concentration is kept constant. Surprisingly, from the scaling point of view, the simple control run performs better than the more sophisticated GGFO scenario. This comparison indicates that the variation of the greenhouse gases affects not only trends but also fluctuations.