What was done
Using newly-developed advanced methods from statistical physics, i.e., wavelet techniques and detrended fluctuation analysis - which the authors say "are able to distinguish between trends and persistence" - they compare the output of several atmosphere-ocean (AO) general circulation models (GCMs) against real-world characteristics of these latter two phenomena, i.e., climatic trends and persistence.

What was learned
Based on their earlier work in this area, the authors determined that "a universal long range power law correlation may exist which governs atmospheric variability at all spatiotemporal scales."  In their current paper, this conclusion was vindicated using real-world temperature data from a number of places around the world, enabling them to conclude that "the power law behavior can serve as an ideal test for climate models."  When several prominent AOGCMs were given this test, however, they displayed "wide performance differences and actually fail[ed] to reproduce the universal power law behavior of the persistence."

What it means
The fact that the authors found "the models tend to underestimate persistence while overestimating trends" implies, in their words, "that the models exaggerate the expected global warming of the atmosphere," and that it therefore "cannot be excluded that the global warming in the next 100 yr will be less pronounced than predicted by the models," which is a kinder way of saying that the models are probably way off-base.