Sunday, 20 July 2008

The APS and global warming: What were they thinking?

From the "what were they thinking" department…

The "Physics and Society" Forum of the American Physical Society decided to open up their newsletter to a nice respectful debate on the main conclusion of the IPCC: that "anthropogenic CO2 emissions are very probably likely to be primarily responsible for the global warming that has occurred since the Industrial Revolution". From there, things went downhill quickly!

Two articles appear in the forum's July 2008 newsletter. The "pro" case is A Tutorial on the Basic Physics of Climate Change, by David Hafemeister & Peter Schwartz. The "con" case is Climate Sensitivity Reconsidered, by Christopher Monckton.

Monckton is rather … notorious … for those who follow these debates; and an extraordinary choice for a physics journal. His article has lots of formulae but little insight or competence. It did not take long for things to turn ugly.

In short order, half the blogsphere fell over themselves in triumph that the APS had reversed its long standing recognition of the facts of anthropogenic global warming; and gleefully concluded that the APS with its 50,000 strong membership could now be added to the ranks of the denialists. Fulsome praise was heaped upon Monckton's article as a brilliant mathematical refutation of the IPCC conclusions. It did not take long for the APS to add to its front page a plain statement that there had been no reversal of position; and add in red ink to the top of Monckton's article a notice that it had not been subject to scientific peer review, and drew conclusions that were in "disagreement with the overwhelming opinion of the world scientific community" and with the Council of the American Physical Society. Monckton hit back immediately with a letter demanding an apology and retraction.

How this all plays out will be most interesting to follow. The initial decision by the APS editor was extraordinarily naïve. I don't know what they expected to achieve with this; but whatever happens now it's a big win for Monckton and his fans. He's got a pulpit, and any response will be dismissed as scientific close-mindedness. Treating it as a serious debate is all that the denialists really want to achieve. Firing the editor (as some have suggested) is surely an over-reaction that would only make everything even worse.

Hey ho. I'm going to watch the social developments with interest; and attempt a minor contribution of my own just to indicate some of the errors, in my opinion, in Monckton's article.

There are, by the way, bound to be errors in my analysis as well. I'm posting it because I'll welcome feedback or corrections -- from anyone -- and because I think it is much better to focus on the substance of article, now that it has been published. I'm not an expert, but I co-incidently was reading many of the relevant papers used also by Monckton just recently, and so am willing to risk this attempt at analysis.

Basically, Monckton looks at the matter of "climate sensitivity" and feedbacks. For a useful review paper on the background to this topic, I recommend How Well Do We Understand and Evaluate Climate Change Feedback Processes? by Sandrine Bony and thirteen other authors; in Journal of Climate, vol. 19, issue 15, pp 3445-3482. (37 pages) You can also get a preprint by ftp from University of Washington Earth Observing System. It comes with a very handy little appendix to explain how climate feedbacks are quantified. Monckton also refers to this extensively.

I'll skip down to Monckton's attempt to use "The IPCC’s method of evaluating climate sensitivity".

Part 1. An attempt to use the IPCC's method

It starts out ok. There is a formula used for the temperature change that should be expected for a doubling of the concentration of CO2. It is:
ΔTλ = ΔF2x κ f
The variables here, using Monckton's naming conventions, are
  • ΔTλ This is the change in temperature than should be expected from a doubling of CO2 levels.

  • ΔF2x This is the "forcing", or the change in the energy balance at the top of troposphere, which results from a doubling of CO2 levels. It has units of Watts/m2.

  • κ This is the "base sensitivity", or the expected response of the Earth's temperature, per unit forcing. It has units of K W-1 m2.

  • f This is a dimensionless multiplication factor, capturing the effect of various climate feedbacks to amplify or damp the temperature response.

Monckton then correctly notes (eqn 3) that ΔF2x is about 3.7 W/m2. This follows from some basic physics, albeit obtained with difficult integrations across the spectrum and along the atmospheric column.

Everything from this point goes rather pear shaped. He makes a completely different use of the variable, to represent some kind of total anthropogenic forcing associated with a CO2 doubling, using a rather confused set of extrapolations from other effects. Basically, he takes the 3.7, scales it up so that it stands for 75% of a total forcing from other greenhouse gases, subtracts a fixed amount for aerosol cooling, and finally applies a "probability-density function" correction which has me baffled. The probability density functions for combined 1750-2005 forcings are in figure 2.20 of IPCC 4ar; It looks a bit like Monckton has taken the mode of 1.72 for the distributions by adding up best estimates for each individual forcing, and then scaled to get the mean 1.6 of the combined distribution (which is a bit skew) as given in section 2.9.2 of IPCC 4ar. Anyhow, after all of that weirdness, he ends up with about 3.4 as a forcing value; which is no longer a doubling of CO2 forcing but a strange kind of combined forcing not properly associated with any meaningful bench mark.

However, it is a forcing; so let's see what he does with it next.

His value of κ as 0.313 K W-1 m2 is uncontroversial. See the reference to Bony et al (2006) I link above.

The feedback multiplier contains more weirdness. Monckton includes a 0.25 "CO2 feedback", which is actually about the changing rate at which carbon is taken up from the atmosphere into the other reservoirs of the carbon cycle. This is discussed in section 7.3.5 of IPCC 4AR. What it means is that the fraction of emissions removed from the atmosphere reduces as carbon is being taken up and as temperatures increase; so that the same level of emissions results in a greater CO2 concentration.

This is not a feedback in terms of more temperature per unit forcing, and should not treated as such. Adding the 0.25 term is an error here, and it becomes very obvious as an error later in Monckton's article.

In any case, Monckton gets 2.16. He'd have been better to stick with 1.9; which is the actual feedback parameter. The accuracy of this number is low; certainly not enough to justify two decimal places.

The gain is then obtained as (1-2.16*0.313)-1, which is 3.077; far too many figures of accuracy. The errors in the 2.16, combined with the subtraction, mean that this number is only accurate to about +/- 1

Also, it should be (1 - 1.9*0.313)-1, which is more like 2.5.

Finally, he multiplies everything together to obtain 3.405 x 0.313 x 3.077 ≈ 3.28. Using the correct numbers, this would be 3.7 x 0.313 x 2.5 ≈ 2.9.

Monckton congratulates himself for "demonstrating that the IPCC’s method has been faithfully replicated" because his value of 3.28 is close to the central point of the range offered by the IPCC, being from 2 to 4.5.

However, what the IPCC actually says (technical summary) is:
Analysis of models together with constraints from observations suggest that the equilibrium climate sensitivity is likely to be in the range 2°C to 4.5°C, with a best estimate value of about 3°C.

So in fact, if Monckton had simply used the 3.7 forcing and the correct feedback parameter of 1.9, he'd have got much closer to the IPCC conclusion, and would have been following their methods.

So far, the errors don't have a lot of impact, but they demonstrate a level of basic misunderstanding that does not bode well. From here, things go downhill fast.

Part 2.1 Adjusting the numbers. The forcing.

The first and major step is a look at radiosonde data for warming in the troposphere. This is a notoriously difficult area, as the radiosonde record has well known systematic errors, which have been discussed now for decades. A couple of recent papers have come out just this year which address many of the issues by using wind shear information. Specifically:

Basically, the mid-troposphere warming is indeed present, as expected.

Monckton does cite this new research, but dismisses it on the basis of satellite records... another case where measurement and calibration errors are a source of hot dispute. In any case, let it go... because what Monckton does with this is astounding.

He divides the forcing by 3. (equation 17)

That's just surreal. There's no basis to reduce the forcing here. It's the temperature response that is involved. He gives a vague appeal to Lindzen (2007), Taking greenhouse warming seriously, in Energy & Environment 18 (7-8). But that paper does not propose any reducing in forcing; only to sensitivity... on roughly the same dubious basis of limited troposphere warming.

Part 2.2 Adjusting the numbers. The no-feedback sensitivity.

Here I confess to sharing a concern with Monckton. I've been looking at these papers now for a couple of months now for another discussion, and I also have tripped up on how this parameter is defined. I've been reading the same references Monckton gave in his paper (Soden, Bony, Colman etc) and I don't really get how the value of -3.2 is obtained. I can understand the -3.7. If anyone reading this would put a comment or a pointer to help clarify, I'd appreciate it!

But in any case. For this next calculation it become obvious that including the carbon dioxide feedback term of 0.25 as part of the feedback parameter b was an error. Monckton uses his 2.16 feedback parameter for a fixed CO2 forcing taken from observations. But that 2.16 included the curious 0.25 addition intended to account for changes in how carbon is taken up into the carbon cycle. It definitely can't apply here, where direct measurement of CO2 levels are being used.

There is also the bizarre use of a "mean" between two totally conflicting sets of measurements; based on NCDC, and a rather strange halving credited to McKitrick. It should be two alternative values; not a mean. And by using 1.9 rather than 2.16, you should get about 0.31 from NCDC values and 0.22 from the halved temperature.

This is not a sensible way to estimate κ, but in fact using the NCDC it gets close to the original value being used. But now Monckton is "double" dipping, in diverting the number down based on McKitrick... because this is ANOTHER reference to reduced warming trends... already used above to reduce the forcing estimate.

(Hat tip also to Tim Lambert, who notes this same error at the Deltoid blog: Monckton's Triple Counting.)

Part 2.3 Adjusting the numbers. The feedback gain.

Monckton looks immediately to maximum upper bounds here; which conceals another subtle error. The various feedback parameters are not independent of each other.

In particular, the magnitude of a water vapour feedback (positive) tends to track with the magnitude of the lapse rate feedback (negative), since both become stronger with more water vapour in the atmosphere. Water both has a greenhouse effect for a positive feedback, and a weaker lapse rate for a negative feedback. You can't maximize both together; their sum shows less variation than either one by itself.

The actual range of all feedbacks together is available in Bony et al: it is about 1.5 to 2.6

There's no problem with the maximum exceeding the 1/κ value of 3.2

There's also a curious point that Monckton has already proposed a lower value for κ, which raises 1/κ to a bit over 4; but that is a detail. The fundamental error here is in simply adding up the upper bounds of feedbacks. They are not independent values; but are obtained as tuples from a range of models. Details in Bony et al (2006).

Part 3. Conclusion

Monckton's paper looks superficially impressive, but examination of the equations betrays some fundamental confusion on the physics and climate science involved.

  • Dividing a forcing by 3 makes no sense at all, and is directly in conflict with Monckton's own reference to a much more able skeptic, Lindzen. The proper argument from temperature is not that the forcing is wrong, but that the sensitivity is wrong.

  • The climate sensitivity is the temperature response to a certain forcing. Personally, I'd prefer to see people using sensitivity as temperature change per unit forcing; but there is a strong tradition for temperature change per forcing due to doubled CO2. In either case, all Monckton's effort to obtain some kind of mean anthropogenic forcing is precisely the wrong way to consider sensitivity. The forcing used throughout his analysis should be 3.7: finis.

  • The carbon dioxide feedback term is not appropriate in a consideration of sensitivity. The feedback parameter should be 1.9 whereever Monckton uses 2.16.

  • The whole recalculation of κ is flawed. The method is dubious, but if properly applied it actually gives close to the right value, near to 0.31. Holding the feedback parameter fixed as a way of calculating κ is backwards; it is the feedback parameter which is most uncertain.

  • That being said... any commenter who can give me a good explanation of how to get 1/3.2 for κ please do so! I can get about that using a grey-body emissivity relation, based on surface temperature and atmospheric emissions, but that sounds wrong. Holding lapse rate fixed and mean emission height fixed just gives the good old 1/3.7.

Monckton's best case here is simply the alleged lack of mid-troposphere warming. All the maths stuff is so badly flawed that it detracts from the shreds of what argument might be salvaged. The issue of troposphere warming will continue to be a focus of interest and debate; but skeptics invariably fail to take proper account of the large error bars on the old troposphere temperatures they invoke; and with the recent work on wind shear this argument, which was never strong, is looking more and more dubious.

Update: (July 26) Gavin Schmidt at realclimate has a response as well: Once more unto the bray. Gavin, by the way, is the real thing; a scientist active in climatology, and in public communication efforts, and with a daunting record of directly relevant formal scientific publication. He also linked to my little blog! Me and Gavin, yeah, that's the ticket.

Read the full post...