Thanks to Drudge, all the right-wing blogs have been touting a story alleging the American Physical Society has reversed its stance on global warming. Joe Romm has the sordid details. The basis for the story is an article published in an APS newsletter (not jornal) by our old friend Christoper Monckton. Monckton's article now carries a disclaimer saying:

The following article has not undergone any scientific peer review. Its conclusions are in disagreement with the overwhelming opinion of the world scientific community. The Council of the American Physical Society disagrees with this article's conclusions.

That's probably good enough for most people, but here at Deltoid we go that extra mile, so I've read Monckton's article and can explain what's wrong with it.

First, I should disclose that I am not a physicist and only did first-year physics and an honours level course in mathematical physics at uni. But that's way more than Monckton ever did, and more than enough to see where he went wrong.

Monckton is trying to make a case that climate sensitivity, the amount that the global average temperature increases if CO2 doubles is much less than the IPCC estimate of 3°C. Monckton reckons sensitivity is just 0.58K. (Actually he says °K, which is wrong -- it's Kelvins, not degrees Kelvin.)

How does he come up with such a number?

He starts with an equation for forcing ΔT_λ

ΔT_λ = ΔF_2xκf

where ΔF_2x is the radiative forcing (in Watts per square metre) from doubling CO₂, κ is the sensitivity (ignoring feedbacks) in units KW^-1m², and f is the feedback multiplier that takes account of feedbacks in the climate system. So far so good.

Then Monckton claims that the supposedly missing hotspot means that ΔF_2x has to be reduced by a factor of three:

Since the great majority of the incoming solar radiation incident upon the Earth strikes the tropics, any reduction in tropical radiative forcing has a disproportionate effect on mean global forcings. On the basis of Lindzen (2007), the anthropogenic-ear radiative forcing as established in Eqn. (3) are divided by 3 to take account of the observed failure of the tropical mid-troposphere to warm as projected by the models

ΔF_2x ≈ 3.405 / 3 ≈ 1.135 Wm^-2.

But Lindzen (2007) (which was published in Energy and Environment rather than in a proper journal) does not say that CO2 radiative forcing is too high by a factor of three. In fact, he specifically says that ΔF_2x "is about 3.5 watts per square meter". As far as I can tell, Monckton has misunderstood this statement from Lindzen:

we can reasonably bound the anthropogenic contributions to surface warming since 1979 to a third of the observed warming, leading to a climate sensitivity too small to offer any significant measure of alarm

This is a statement about sensitivity not CO2 forcing.

Next Monckton turns his attention to κ and argues it's too high as well:

We assume that Chylek (2008) is right to find transient and equilibrium climate sensitivity near-identical; that all of the warming from 1980-2005 was anthropogenic; that the IPCC's values for forcings and feedbacks are correct; and, in line 2, that McKitrick is right that the insufficiently-corrected heat-island effect of rapid urbanization since 1980 has artificially doubled the true rate of temperature increase in the major global datasets.

With these assumptions, κ is shown to be less, and perhaps considerably less, than the value implicit in IPCC (2007).

Did you spot what he just did? If you assume that there is no delay in warming (which is wrong) and McKitrick is right (which is also wrong), then you get a low value of sensitivity. If you also assume that the IPCC values for ΔF_2x and f are correct, then their value of κ must be too high -- Monckton comes up with a number 20% less. But in the previous section Monckton argued that the IPCC value of ΔF_2x was too high by a factor of three. If instead you use Monckton's number, the IPCC value of κ is too low.

What Monckton is doing is double counting his (dubious) evidence that sensitivity is lower than the IPCC number. If he had two pieces of evidence that sensitivity is half the IPCC number he would multiply them together to claim that sensitivity is one quarter the IPCC number. This is not correct.

Too put it another way, in this case, by making some unrealistic assumptions he came up with a sensitivity estimate 20% less than the IPCC number i.e. 2.4K. Logically he should have stopped there -- he has an estimate of sensitivity. Instead he uses this estimate of sensitivity in a chain of reasoning that leads him to conclude that sensitivity is 0.58K.

Anyway, Monckton goes on to pull the same stunt with f -- using arguments that sensitivity is lower than the IPCC number to argue that f must also be lower than the IPCC number. So that's triple counting. Then he multiplies all his improved factors together to come with his final sensitivity of 0.58K.

The editor of the APS newsletter, Jeffrey Marque, actually invited Monckton to contribute this piece.

See original blog post and the most interesting comments here:

http://scienceblogs.com/deltoid/2008/07/moncktons_triple_counting.php