The paper by Steven Sherwood has been agitating those of a green disposition in recent days, with all sorts of wailing on Twitter about how we're going to hit four degrees of warming by the end of the century. This is certainly the story that Nature gave out in its press release:
Global average temperatures will rise at least 4°C by 2100 and potentially more than 8°C by 2200 if carbon dioxide emissions are not reduced according to new research published in Nature. Scientists found global climate is more sensitive to carbon dioxide than most previous estimates.
The research also appears to solve one of the great unknowns of climate sensitivity, the role of cloud formation and whether this will have a positive or negative effect on global warming.
As readers here know, climate models all run far too hot. As far as the latest CMIP5 generation go, this is at least partly because they use estimates of aerosol forcing that are much higher than observations suggest is the case. And if the latest hypothesising about the hiatus in surface temperature rises is correct, the models are all missing a key climate subsystem too, namely transport of heat to the deep oceans.
I haven't got hold of a copy of the Sherwood paper as yet, but from what I have been able to glean from the abstract, the press release, and from conversations around the web, he and his colleagues looked at climate models to see how well these reproduced observations of clouds, finding that the best match came from the models that ran hottest.
In other words, the models that had the most realistic simulations of clouds had the least realistic representations of temperature changes.
If I've understood what was done correctly, this is an interesting conundrum for climate scientists to explore. What it is not is any reason to think that the output of such models is policy-relevant or reason to think that we will warm by 4°C by the end of the century.
Nature really has sunk to the level of the Sunday Sport.