The latest global climate models produce a 'fingerprint' that aligns well with actual temperature observations, and underscores the human influence on climate through the release of greenhouse gases and ozone-depleting chemicals.

Ben Santer of the Lawrence Livermore National Laboratory in Livermore, California, and his group analysed simulations from 20 climate models at the core of the Intergovernmental Panel on Climate Change's fifth assessment, and compared the results with satellite temperature records dating back to 1979. The team found general agreement with observed global-warming patterns, although the models typically overestimate warming in the lower atmosphere while underestimating cooling trends higher up, in the stratosphere. The analysis suggests that climate modellers could improve model performance by incorporating more realistic treatments of ozone chemistry and aerosols.

Proc. Natl Acad. Sci. USA (2012)