One of the most hilarious mantras that climate modellers intone in their media appearances is that the models must be right "because they've got physics in". The fact that huge swathes of the physics are parameterised - in other words they are not modelled at all, but are summarised down to a single number that the scientists hope will do the job - is quietly overlooked. Or at least, it's quietly overlooked by the climate modellers. Sceptics, of course, have been yelling about the deficiencies of this approach for years, to complete media indifference.
One of the most egregious parameterisations in the climate models is that of clouds, which are one of the most important feedbacks to the climate system. There has now been an attempt to rectify this failing - Steve McIntyre has picked up a remarkable report from a few years back that raises some interesting questions.
A team of climate modellers decided to take their model that one step further and intead of parameterising their clouds, they built a simple cloud model into each gridcell column (imagine vertical atmospheric columns rising up from each point on the earth's surface). The result was that the Earth turned out to be much less sensitive to carbon dioxide than previously advertised. The clouds essentially stopped some of the heat reaching the Earth in the first place.
Crisis over then? Perhaps. We'll have to see.
In the meantime we can wonder why these results, which were available to the IPCC during its Fourth Assessment of the world's climate, barely warranted a mention in the report. As someone once said, "Hey, it's climate science".