After filling in our questionnaire, our dear sceptic readers in this blog should devote a certain amount of time in reading a few papers that have appeared recently - authored by very much respected mainstream climatologist - and that in substance question the ability of climate models to reproduce the evolution of the observed hemispheric and global mean temperatures. The most recent of these papers aims to characterize a sudden change in the difference between the Northern and Southern Hemisphere annual mean temperatures. Common wisdom so far is inclined to think that the temperature difference between both Hemispheres in the 20th century was been mainly due to the larger thermal inertia of the Southern Hemisphere and the effect of the aerosol radiative forcing due to industrial atmospheric pollution, which is supposed to have been stronger in the Northern Hemisphere. The magnitude of this forcing is quite uncertain, as are the mechanism by which industrial aerosols may affect the radiative balance directly or indirectly by modulating the characteristics and life time of clouds. All in all, aerosol forcing may have almost completely offset the potential warming due to anthropogenic greenhouse gases or, alternatively, be almost negligible. The most recent estimations I am aware of tend to indicate that the cooling effect of aerosols may have been smaller that previously thought, and therefore the warming effect of CO2 on the 20th century temperatures would have been also smaller. But what the paper by Thompson et al now tell us is that a large portion of the cooling or lack of warming observed in the 1970's has been due to oceanic natural variability. They reach this tentative conclusion by looking at the patterns of temperature difference between the Northern Hemisphere and the Southern Hemisphere, and so they do not really investigate what are those mechanisms of natural variability may be moreprecisely. Further analysis will be undoubtedly follow.
The same authors had published previously another paper that tried to explain an abrupt change in temperature observed around 1945. This time the ultimate reason for this sudden change seems to lie on the observing systems, and not on the temperatures themselves. During those years a major switch occurred in the proportion of US and UK ships reporting temperature data occurred. As each Navy applied different measurements methods, the varying proportion of the reports originating from one or the other could explain this rather abrupt change.
The lesson for climate modellers is that it becomes increasingly difficult to defend that climate models can reproduce the observed temperature evolution in the 20th century, even after adjustments to the data, as in the first paper, or after concluding that natural variations have played a more important role. If models can reproduce the 20th century temperatures, irrespective of whether they have been caused by external forcing and natural variability, I would conclude that the discriminative and predictive power of the models would not be as solid.
A third paper published in Geophysical Research Letters this year by some of the same authors, Fyfe, Gillet Thompson, and that has been, as far as I know, totally overlooked, deals precisely with the comparison of model and observed recent temperature trends. The ongoing debate about the skill of the models to replicate the observed temperature trend of the last 20 years is , as one can assume, quite lively. Even more surprising is that this paper has not received much attention. The authors apply a statistical method to subtract the influence of known natural variations (ENSO, volcanism,..) in the global temperature record. They argue that the signal of the external forcing should be more apparent in the resulting temperature residuals than in the full temperature data. And, indeed, the uncertainties in the estimation of the 20th century temperature trends seem to be reduced more or less by half. Good news, one would be tempted to think. However, they also conclude that
We have also shown that the observed and simulated uncertainty in 1950–2000 trends drops by about half when the natural signals are removed, making clearer where the anthropogenic response in some models deviates significantly from observed. The simulated and observed global mean temperature trends are statistically indistinguishable in 12 of 24 models for the raw data, but in 8 of 24 models for the residual data
Using the residual temperature, only one third of the IPCC models are found to be compatible with observations. Perhaps the dawn of paper democracy will usher in the end of model democracy