Comparison of near-surface air temperature variability in 11 coupled global climate models

The common variance between 100-yr-long control runs from 11 coupled global climate models (CGCMs) has been studied by use of common empirical orthogo nal functions (EOFs). The results suggest that there is a considerable disp arity between the CGCMs estimates of internal variability. About one-half o f this difference can be attributed to model drift dr other low-frequency v ariations in several of the models. However, even after accounting for this effect it was found that the models can easily differ by a factor of 2 or more for the energy levels in different EOF mode (wave) numbers, Comparison with observations showed that no one model consistently reproduced the obs erved partial eigenvalue spectrum. Again, differences between observed and model energy levels were commonly a factor of 2 or more. It is speculated t hat at least some of the disagreement is due to the relative coarse resolut ion of the models used in this study. Separate analysis of a 1000-yr control run of the Geophysical Fluid Dynamic s Laboratory model suggested that intramodel variability is much smaller th an intermodel variability. It was also found that an estimate of the anthro pogenic signal due to greenhouse gases and aerosols from the Max Planck Ins titute had strong spatial similarities to the leading modes of the models' common EOFs. This fact complicates the detection/attribution problem.