Sunday, March 14, 2010

Green, Armstrong and Soon (2010): Yes, the IPCC (1992) prediction was skillful

A new paper from Amstrong and colleagues (and yes, it's *that* Soon) has made its way onto the blogosphere. For his previous self-embarrassment, see here. Of course they don't actually express their result quite so clearly as I've written it, because their goal for some time now has clearly been to denigrate the IPCC wherever possible, but they cannot avoid the simple truth:
We found that averaged across all 17 forecast horizons, the 1992 IPCC projected warming rate errors for the period 1992 to 2008 were 16% smaller than forecast errors from our benchmark
Full paper here. There's a lot of dubious stuff in there too, so don't take any of it too seriously. Eg, even though the IPCC did mention a warming rate of 0.3C/decade, they also said (on-line here):
Because GCMs do not yet include possible opposing anthropogenic influences, including the forcing from sulphate aerosols and stratospheric ozone depletion, the net rate of increase in surface temperature is expected to be less, at least during the period for which sulphur emissions continue to increase, than would be expected from greenhouse gas forcing alone.
So the 0.3C/decade was believed to be an overestimate back in 1992. Green et al used it anyway and found out it was better than their baseline "no change" forecast.

5 comments:

Anonymous said...

That would be Green the economist.

Tom Fid said...

This looks like a slight rehash of their IJF paper, which I took apart here
http://blog.metasd.com/2009/07/unprincipled-forecast-evaluation/ (second half)

Joshua Stults said...

Does it matter? Is mean surface temp useful for informing policy?

(I read your Bayes D&A commment, so I've got a pretty low entropy prior on your answer, looking to see if I'll be surprised)

James Annan said...

Tom, I think it's exactly the same paper.

jstults, I don't think that global mean SAT is particularly useful for local adaptation on the decadal time scale, but I do think it is useful for global policy looking at the long term (at least though links to other effects). Regardless, the only point I'm trying to make here is that the model forecast was skillful, which has a clear technical definition.

Paul Baer said...

Wow... I actually just now found the paper, thanks to a trail that started with one of my friends who posted a link on facebook to denialdepot. I admit not to having read all of the classics of the "skeptics" literature, but this has got to rank high on the list of "worst papers ever."