[Fred Pearce] The result was bad news for the modellers. The two models examined -- one German, the other American -- generated a natural variability of around 0.1 degree C per century. This was less than half that revealed in the proxy data. "Of course we don't have to believe the proxy data. They certainly have problems attached to them. But my belief is that they both models, and proxy data too, underestimate real variability," says Barnett
The models' error was not, perhaps, too surprising. As Barnett points out, they do not include vital "forcing" mechanisms that alter temperature, such as solar cycles and volcanic eruptions. Nor can they yet mimic the strength of the largest year-on-year variability in the natural system, the El Nino oscillation in the Pacific Ocean, which has a global impact on climate.
Nonetheless, the findings should serve as a warning, Barnett says, that "the current models cannot be used in rigorous tests for anthropogenic signals in the real world". If they are they "might lead us to believe that an anthropogenic signal had been found when, in fact, that may not be the case."