On the subject of validation of climate models, I noticed an interesting post on the "Climate Skeptic" blog a few days ago, which updates a simple model originally devised in 2007 for reproducing the global average temperature anomaly since 1870:
My guess at the background to this is that in the mid-00s, the AGWers were keen on a line of argument that as complex climate models could 'postdict' the past, they must be able to predict the future. There was a BBC documentary in 2006 which featured a presenter called Paul Rose (I think the documentary, for anybody interested, was this one: http://docuwiki.net/index.php?title=Meltdown:_A_Global_Warming_Journey ), where Rose was initially portrayed as a sceptic, but after talking to various experts was converted to the cause, and was particularly convinced by climate models being able to produce a fairly good match to historical temperature data. In response to this line of argument some AGW sceptic bloggers pointed out that they could get good matches to historical data using much simpler approaches, and at no expense to the public.
The Climate Skeptic blog model is actually showing pretty good predictive behaviour for the six years following on from 2007, up to mid-2013, and is outperforming the 'professional' climate models. The model predicts that global cooling will occur over the next twenty years.