Tuesday
Jul052011
by 
Bishop Hill
Bishop Hill IPCC on climate sensitivity
Jul 5, 2011 Climate: Models Nic Lewis, best known as one of the co-authors of the O'Donnell et al paper on Antarctic temperatures has a must-read post up at Judith Curry's place. The title tells you all you need to know:
The IPCC’s alteration of Forster & Gregory’s model-independent climate sensitivity results.
This is pretty shocking stuff.
Again.
Reader Comments (141)
That was to an earlier post by BBD. Meanwhile, Nic, terrific stuff, thank you.
In response to Jeff Wood, 2:08 PM, here’s my shot. I hope it is not a distraction (especially since I was a bit snooty about the rap stuff earlier), and that others will jump in to improve or correct any infelicities - or perhaps to map it more directly to the matter in hand.
Re the question about confidence intervals. A good one to ask, since they can readily be misunderstood. There are two major types: one using subjective probabilities together with observational data, and one using only the data. In each case, the data is assumed to be a random sample of some larger group about which conclusions are to be drawn. Random samples are not easy to come by, and so this can be an early concern when deciding how seriously to take someone else’s confidence interval. This is key since the whole point of them is to give guidance about the leap from the sample data to at least tentative conclusions about the larger group.
All else being equal, the smaller the sample, the wider the confidence interval. Suppose you wanted to estimate the average (mean) weight of rivets in a box. Take a sample of 5, and you would expect your computed confidence interval for that weight to be wider than if you were to have taken a sample of 50. The interval provides, if you are relying only on the data obtained, lower and/or upper limits spanning a range within which the ‘true’ value would lie in X% of cases were such a sampling procedure to be repeated a great many times. That would be an X% confidence interval for the group value, in this case the mean weight. If you would like X to be larger, say 99% instead of 95%, the calculations will give you a wider interval. Intuitively this also seems plausible. For example, in guessing an interval for the weight of the rivets, the wider the bounds, the more confidence I would have that they include the actual value. This brings in the Bayesian aspect. If I have knowledge of these rivets before taking any sample, I can give an informed estimate of what the mean weight will be. With a bit of work, that knowledge could be expressed as a probability distribution. This is my prior distribution. When the data comes in from my sample of rivets, I will follow a calculation procedure to update my distribution, thereby producing the ‘posterior distribution’ – a distribution reflecting my prior knowledge and the information from the sample. From this distribution, I can select a range of values which include X% of all possible values. This is a Bayesian X% confidence interval, and is to be regarded as a probability that the interval contains the true value – a probability that reflects your new degree of belief about the mean weight. Now if you are very knowledgeable, your prior distribution might well be peaked about your best estimate, and taper off quickly on either side. If you are less sure, you might have a flat distribution (top hat shaped) giving equal probabilities across the range. If your prior range is very wide, then so will your posterior range be unless the sample data is large enough and tight enough to overwhelm the prior. So, choosing a ‘wide enough prior’, will guarantee a wide posterior distribution for any plausible data set. If that posterior distribution includes alarming values within the chosen X% range, then alarm may duly be raised – at least for the mass media, if not necessarily for statisticians.
Now the other confidence interval, the particular one based only on the sample data, is not a probability – if pushed, your view is that the relevant probability is either 1 or 0 since the interval either contains the group mean value or it does not. You are only claiming that if a large number of these samples were taken, then X% of the resulting set of intervals would contain the true value – you are not making that claim for your one particular interval from this set.
Shub said: "Why wont BBD, dana and the consensus-makers comment on this?"
There's a comment from 'barry' @ July 6, 2011 at 2:27 am over at...
http://judithcurry.com/2011/07/05/the-ipccs-alteration-of-forster-gregorys-model-independent-climate-sensitivity-results/
...that Nick has rebutted quite well, so the pro-CAGW guys (barry & dana1981) who were debating with BBD are well aware of this.
Richard D (Jul 6, 2011 at 11:54 AM) and Phillip B. (Jul 6, 2011 at 12:19 PM)
I don't wish to be disputatious (because I respect you both, greatly) but - particularly in light of Nic's update [Jul 6, 2011 at 7:31 PM] - I hope you will forgive me for concluding that politics appears to have trumped scientific integrity.
IMHO, Shub captured the BIG question that this episode raises, when he wrote:
One cannot help but wonder how many other such ... uh ... liberties the IPCC "author teams" have taken with the peer-reviewed literature they cite and reference (with "happy" or "unhappy" acquiescence of such papers' authors - not to mention unbeknownst to such papers' authors). We know from past experience that they certainly took such liberties in describing the work of McIntyre & McKitrick - and with that of Roger Pielke, Jr.
Hilary
Hilary: Piers Forster has obviously been in touch with Nic, as well as Gabi Hegerl. The moment I read Nic's careful report there seemed to me to be some significant wiggle room: 'part of the author team and not unhappy' (hmm, that could have been stronger) playing into 'he remembers being persuaded by the Oxford group ...'. I don't want to rush to judgement with any of this. Various people need to have time to take in what Nic has shown - and shown with such style. It strikes right at the heart of the matter, at least as the IPCC frames the whole debate.
So, I respect you too: AccessIPCC is a great project. We're all about to see the next stage.
Agree with hro001 that Shub captures the problem perfectly with his “The IPCC is supposed to assess science, not do science... If the IPCC cannot be taken for its word, [on] what it does, how can one believe what it says?”
It’s infinitely more serious than Glaciergate, but how to persuade the mainstream media of this? A dodgy graph, unlike a melting glacier, can’t be accused of risking the lives of millions.
Thanks to the willingness of Nic Lewis and Richard Betts to engage with sceptics, we know who knows what in the scientific community. If we knew that science / environment correspondents in the media knew about this, and if they knew we knew, would they dare continue to support the IPCC so blindly?
However justified the statistical adjustments, the IPCC has no right to make them. This is Johann Hari morality applied to international policy-making.
BBD,
While I would agree with you, in terms of the content of the error in the F&G IPCC case, the matter is entirely the opposite in terms of the nature of the error.
The Mann graph was made by Mann, Bradley and Hughes. It was picked up and highlighted by the IPCC. They did not create it or sketch it. Whatever errors or dishonesty lie behind the graph can be laid at the feet of Mann et al and they can be asked to respond. There is a credible chain of custody.
What the IPCC does with the F&G case, is a different thing. New information is created under the guise of interpretation, and passed off as an assessment. It is this (flawed and dishonest) assessment that then becomes the very basis for your "This doesn't 'refute AGW'". Who can be held responsible for this mis-step? No one.
Thus the non-refutability of AGW is itself 'anthropogenic' - it consists of rearranging and sculpting available evidence in a systematic manner to present a purported whole, flaws in the very parts which go to make up the whole apparently cannot argue against. How is this possible?
Granted, in a proper assessment or evaluation, if one reaches a conclusion and then an error was discovered that affected one small portion in the chain of reasoning, the final conclusion may still remain valid. But if an error is discovered in the representation of evidence in the chain of reasoning, the conclusions reached are on shaky ground.
Think about it: people like James Annan voice their thoughts in the comments section. Then they turn around and carry the water for the IPCC.
How many other places has the IPCC done this type of thing?
You know, there are a lot of skeptics who like to bad-mouth Monckton. But he asked the question about the stupid temperature graph in the IPCC synthesis report (another place where the IPCC makes up its own stuff) where it makes the claim that the temperature change is accelerating. Has the IPCC given an answer?
Hilary. I don't wish to be disputatious either and agree that politics appears to have trumped scientific integrity. It now appears that F&G were "happy" with the words used by the IPCC. Your words and those of others (shub, Richard, geoff) ring so true. Hats off to Nic for this exemplary exposition of the murky world of the IPCC and its "creative review process".
oldtimer, shub, hilary, I have a list of examples of where IPCC AR4 exaggerates the case for AGW, presents data in a misleading/cherrypicked way, misrepresents the literature, ignores papers that don't support its story, ignores valid reviewer comments, etc.
That's a useful compilation, PaulM.
In any venture on the scale of the IPCC, mistakes and oversights will surely occur. But there is such bias and polarity towards alarmism in so many of them, that I regard the whole venture as not worthy of our trust. The more one learns about the motivations and attitudes of some of the key players involved, the more this view is reinforced.
What a huge shame that the IPCC was not instigated, directed, and led by people of higher calibre and integrity! Instead of providing us with a valuable overview, it has served mainly to degrade, mislead, and coarsen both science and public affairs.
This latest revelation of sleight of hand, of artists intent on creating an emotional or political impact despite the data rather than scientists intent on sharing results and ideas in a straightforward way, is yet another item bringing dismay and disappointment.
@ Richard Drake at 4:33 PM
It is precisely this sort of thing which turned me - and I suspect any reasonable and intelligent people - from agnostic to "sceptic". What sane person would prefer belief according to this green drivel to something reasonably erudite such as Bishop Hill?
@PaulM
Thank you for that link. I will add it to my voluminous correspondence with my MP.
The fact remains that Government policy is driven by acceptance of the work of the IPCC. For example its response to the HoC Science and Technology Committee in Cmd7934 states:
"6. The Government agrees with, and welcomes, the overall assessment of the Science and Technology Committee that the information contained in the illegally-disclosed emails does not provide any evidence to discredit the scientific evidence of anthropogenic climate change. We note that similar findings were returned by both Lord Oxburgh’s and Sir Muir Russell’s reviews. In particular, we note the findings of the Muir Russell Review: that the rigour and honesty of the scientists are not in doubt; that there is no evidence of bias in data selection; that there is no evidence of subversion of peer review and that allegations of misusing the Intergovernmental Panel on Climate Change (IPCC) process cannot be upheld."
Later it states:
"18. We further note that there is a wealth of evidence demonstrating that this warming is already beginning to impact the Earth’s climate, including snow and ice cover, sea levels and the distribution and range of plants and animals . Such findings are summarised in the 2007 Fourth Assessment Report of the Intergovernmental Panel on Climate Change. More recently, evidence has been summarised in the US Academy of Sciences Reports, America’s Climate Choices, 2010."
There is much more along the same lines. The Carbon Plan, signed by Cameron, Clegg and Huhne sets out how they propose to implement their belief in the work of the IPCC.
PaulM, your list of IPCC misdeeds (I am lost for an adequate description as my Thesaurus is currently in storage, so 'misdeeds' will have to suffice) is truly jaw-dropping. Why any politician of any stripe has any kind of faith in the IPCC's pronouncements mystifies me, let alone faith sufficient to base the future of our civilisation upon.
The other factor that is truly disturbing is the utter lack of any kind of real examination of the IPCC by the MSM.
BBD:
Shub:
I would suggest that there are three dimensions of possible badness we should be considering - in fact, we are considering all three, but it may be helpful to articulate these:
1. Badness of the original papers, as science, in isolation
2. Badness of the IPCC process in selecting and summarising them
3. Badness of the overall argument.
Shub is for me right that in dimension 2 the treatment of Forster & Gregory is off-the-scale bad.
But that's not the whole of it, not by any means. The argument that has been put forward to policy-makers is that climate sensitivity to a doubling of CO2 is dangerously. That is why, they've been told, we must act NOW.
Any monkeying around with the evidence for high sensitivity - and what Nic has uncovered sure is monkeying around - is as bad as it gets.
That's after a night to sleep on it. I feel I'm still processing this one.
Kind reader, please insert one word as you consider my last post:
As I say, I'm still processing :)
Richard Drake,
I tend to Shub's point of view too, but it seems a bit futile and also sadly child-like for us to be discussing levels of 'badness'. Should not us 'grown-ups' have some authority to appeal to over this outrageous crime against civilisation as we know it. Somehow, retractions and excuses don't seem to cut it at all, given the magnitude of this offence.
I'm beginning to understand how the citizens of Nottingham felt when they were being robbed by their sherriff.
Like you, I am still processing this, which is probably why I am reaching back into my childhood to find satisfying metaphors for wickedness.
Alexander K: There's a paradox here. Nic's own presentation on Climate Etc (which I thought was a model of clarity) is portrayed as follows by Richard Knights on WUWT earlier today:
Then as you, I and various others try and think through the implications it gets labelled it 'child-like'.
In fact I have no problem with being childlike. I think of Einstein as childlike, in many ways. His spectacular insights and intuitions about the real world were closely connected, for me, with this childlike quality. (Helped no doubt by the fact he wasn't a tenured academic, just a lowly patent clerk, when he transformed physics with three breakthrough papers in 1905.)
I also find this muddled at best:
Excuses won't cut it, sure. But retractions from the right people could go an awfully long way.
I think the issue and the nature of the badness Nic Lewis has uncovered here is key. I knew I wasn't satisfied by BBD's formulation. I make no apology for trying to clarify.
Richard Drake,
Sorry to go on at length about what I see as the moral and ethical implications of the chicanery Nic has shone his spotlight on, but I do not understand how any criminal can 'retract' a crime. I don't accept that my thinking is muddled at all, but I view this act as a crime and cannot conceive as to how retractions or apologies can suffice.
As I see it, you and I are on the same side of the moral and ethical fence but are differing as to the magnitude of the 'badness'. When I think of the ramifications of fuel poverty in the UK alone, plus all the other associated evils the IPCC's manufactured doomsaying has wrought, the 'badness' is beyond description.
AK: No need to apologise. What I found muddled last time around was including 'excuses' with 'retractions'. This time it's an issue of division of labour. Whatever Gabi Hegerl and Francis Zwiers did, say, as Coordinating Lead Authors of WG1 Chapter 9, to Forster & Gregory (did to them as paper and as people), it doesn't make H&Z responsible for all fuel poverty in the UK. Or even that part of fuel poverty attributable to 'CAGW mitigation'. There are clearly many other people involved in the decision chain. But what Hegerl and Zwiers appear to have done is bad. It's important to remain patient and sober, not lurch between the two extremes: "Move on, there's nothing to see here" to the equally unhelpful "UK pensioner deaths have increased because of YOU!"
Mind you, I agree that the social impact of CAGW policies, especially in the developing world, should be much more present in everyone's mind than it seems to be. But that's too much for this thread to bear. Nic has done the heavy lifting - and he's aware of more to do, for this to become suitable for peer review and publishing. We smaller fish (to mangle metaphors) should seek to do our bit with equivalent care.
Shub, Richard
I had chapter 11 of THSI in mind: 'The Hockey Stick and the IPCC'. Plenty of shaping the consensus going on there. I see this as essentially the same thing as the F&G business.
I'm not actually sure it's worth debating the minutiae. The general picture is clear. See July 6 7:12pm.
Richard Drake Jul 7, at 2:59 PM |
"It's important to remain patient and sober, not lurch between the two extremes"
Couldn't aggree more. Whilst the chicanery of the warmists does, I admit, make me intensely angry, one must refute this nonsense and virulent rhetoric in a calm manner, and not descend into the green-type of hysterics. Leave that to Paul Nurse.
Richard Drake,
I take your point. I shall calm down now. But this thing outrages me so! Having said that, I realise that when my outrage is up and running it tends to bolt.
I have a picture in my minds eye of academics I worked with in a university department many years ago. I can see some wondering what is exciting me and others muttering that something should be done about it, but nobody actually willing to grasp the nettle.
Going back to BBDs list:
As for what I think about the apparent misrepresentation of F&G, it is this:
- This is not evidence against AGW:
Not in itself, but the original F&G paper, along with others based on empirical science rather than models, tend to support the lukewarmers case rather than the catastrophists (and my suspicion is that more empirical data will tend to move the calculated sensitivity value lower, perhaps to the point where AGW is recognised as both a real and trivial process)
- This is strongly suggestive of IPCC WG1 bias in emphasising the likelyhood of the most severe outcomes from AGW:
Given that the reinterpretation has been undertaken by the IPCC rather than the paper's original authors, that the results are biased upwards and that there is a great deal of 'chartmanship' involved in making the observational results and model outputs appear to agree, that seems a fair comment.
- It's not as bad as the Mannean Hockey Stick in AR3
In the sense that it is not based on fundamentally bad data and the same level of statistical manipulation, then true. Also it is a lower profile entry to the IPCC report, although is substance is actually far more significant to the AGW case than the hockey stick.
- It's the same general sort of thing though
I think it's more similar to the Mann 08 issue with the Tiljander sediment series - the data as originally published stands well on its own, but overly aggressive statistical processing (the 'prior' in this case and the PCA automatic inversion of the series in the Tiljander case) have led to inaccurate interpretations. Whether this is wilful, or out of a lack of understanding of the effect of the statistical processes applied could only be answered by those who undertook the process
- Breaking the handle of the Hockey Stick doesn't 'refute AGW':
True, but it does increase the likelihood (by showing that natural variations have historically been greater) that the AGW CO2 sensitivity could be much lower than currently determined.
- This doesn't 'refute AGW'
As above, although the other issue in this case is that it has hidden the fact that empirical evidence is suggesting a much lower climate sensitivity than derived from GCMs (leaving aside the science philosophy question of whether you actually can derive anything from computer models).
- Both the above suggest zeal bordering on desperation within the IPCC to get the point across. The old 'Schneider balance' between honesty and effectiveness again.
Agree and add to that the question of whether the IPCC are actually exceeding their remit in undertaking, at least on a small scale, original work in undertaking this reinterpretation of previously published data.
Ian Blanchard
Thank you for your comments. I'm a lukewarmer ;-)
Ian Blanchard:
That's the key point. We mustn't lose sight of that.
I just uploaded this graph to my blog - *Two Distributions*. It is from my post-graduate thesis. The details are immaterial for our purposes, but compare the two distributions shown.
The blue curve is obtained when an expert observer's manual, visual estimate of the quantity of the entity under study, is plotted. The red curve is the same entity measured by an objective method - computer-aided morphometry, to be precise.
In other words, experts take a quantity distributed as the red curve, and interpret it to fall as the blue curve.
What does the graph tell us? What is immediately obvious is that a more constrained distribution is overestimated - both in the estimate of the median, and the standard deviation. In my study's case, it is simply proof that the human eye, expert though it may be, is very poor at quantitative assessment, and indeed makes systematic errors.
The more insidious aspect is, that experts constantly overestimate the quantity under study and this has informed their view of the pathophysiology of the process itself (analogous to sensitivity). Indeed this has occurred, and is in the process of being corrected (if at all).
Compare the above graph to situation to this figure from Nic Lewis. Again, what is immediately obvious is that the IPCC transmogrifies one thing into something else - there is error, there is misrepresentation (of sorts). More insidiously, it is such (mis)representation that has given rise to the claim that 'higher sensitivities are not ruled out in all estimates of S'.
In other words, the shape of a distribution has deep implications on the inferences drawn from the data. There is no getting away from this fact.
Small point Shub, I think Fig 4 of the initial blog post makes your point. Fig 4 of the followup post is comparing 2 different studies, not the IPCC modification.
True rc. I should have been more explicit.
What I was drawing attention toward was the blue and the red curves (both derived from Gregory 2002) in Nic's latest graph - a situation that seems analogous to what I encountered in a different context in my own work.
Gregory 2002 is in the IPCC 9.2 graph as well. Updating G02 with the latest (for the AR4 2007 deadline) data for ocean heat loss and radiative forcing estimates makes it look more like GF06 - a tighter distribution with a neglible tail toward the higher sensitivities.
Imagine if the the IPCC had published an updated Gregory 2002 and an unmodified GF2006.
The situation would have been totally different - the models would have said one thing and the observational data based studies would have said another. The contrast would have been cute. 'S' would have been assigned a different pdf - and the mantra would have been something else - "Scientists agree that some mild warming to occur. Possibility of scorched earth remote".
Spot on, Shub. Once again, with emphasis, this monkeying around is not in a backwater but the heart of the IPCC's argument for policy makers. If they'd been no alteration to FG06 (which was, remember, honest enough to say that use of OLS meant it would be overstating the sensitivity) but completely justifiable bringing up to date of G02 with the latest data, the world's politicians and public would have been given something of immense import to chew on - the difference between what real world data was saying and the models.
It's not that I think we yet have enough real world data. But to fiddle what we have, in the light of the awesome influence the IPCC has had on world opinion, takes the biscuit. Every policy maker on the planet needs to be made aware of this, just like the hockey stick, but more so.
I missed a day, but if John Shade is still looking in, very many thanks John for the response to my question on confidence intervals. It will take me a few attempts to understand it properly, but that is my problem.
This is one of the richer threads that our host has presided over. The quality of a blog is not measured alone by its host, but also by the quality of its commenters.
Shub
Interesting post.
You do make me smile though. A short while ago, you said this:
It's all about climate sensitivity. Always was.
Shub, brilliantly put, I think that is it in a nutshell. Had the IPCC not smeared the data then the narrative would have been quite different. And the world would be all lukewarm eh BBD ;-)
No BBD, I agree with you - in the end, 'it' is always about the sensitivity. My point in that thread had more to do with whether we need to discuss it all the time.
Which is why, this is such an important thing.
You know that if the probability distribution function of S from Gregory 02 and the F&G papers had matched the models, they would not have modified it in any way. These guys, are f***ing around with the one most important pillars of the anthropogenic debate. If you indeed argue for a lower sensitivity, wouldn't your hand have been strengthened by the IPCC itself, if it had printed fig 9.2 in a more honest manner?
Incidentally, I stumbled across this paper which shows how catastrophists just love 'fat-tailed' distributions.
(Fat-Tailed Uncertainty in the Economics of Catastrophic Climate Change, Martin L. Weitzman
February 23, 2011. REEP Symposium on Fat Tails.)
Josh, we just crossposted!
I wasn't portraying Nic's presentation as unclear, as the statement above implies ("model of clarity"), but rather characterizing the topic of his paper (Bayesian statistics) as "technical" and abstruse--which is correct, and explains why WUWTers have been hesitant to weigh in so far--which is what I was trying to account for in my comment.
Shub and BBD exchanged above about whether it was "all about sensitivity". This reminds me that I commented twice on this thread - once in response to Philip B about whether sensitivity is a meaningful thing to worry about, once to try to paraphrase Nic L's presentation in less technical form. So, is sensitivity a useful thing to worry about? The case against, as I understand it, is that there is no fundamental reason why the sensitivity values Y and S should be constants - they might vary in time (depending on the state of the planet at the time that forcings are changed), with starting temperature, or with the nature of the forcing (perhaps changes with the same 'forcing' in insolation/CO2/methane/etc do not all lead to the same change in temperature). Indeed, some of these effects are known and included in models. Also, each forcing will not have an effect on the global temperature that is purely instantaneous - the effect may set in over a month, a year, a decade or even much more (transient vs. equilibrium climate sensitivity). All of these considerations mean that giving one number to Y or S is going to be difficult, so the famous Fig. 2 in FAQ 2.1 of Chap 2 of the IPCC WG1 report (p. 8 of the linked pdf) showing all the forcings and their known or unknown magnitude is in some respects a simplification. My view, though, is that sensitivity is a useful concept: the uncertainties linked with all the aspects I've just mentioned are probably smaller, with the current state of knowledge, than the intrinsic uncertainty in sensitivity. Think of the probability distribution functions for S discussed in Nic's essay: they cover our lack of knowledge of what S is, but also perhaps the ranges of S values that might occur at different points of time. So even those who don't think S is a fixed number should care about whether it behaves like a number that is mostly small rather than mostly large.
Here's a way to sharpen the point of the knife. Ask defenders of the IPCC's putting its thumb on the scale about climate sensitivity, "What would the Inter-Academy Council (IAC) think of this fancy footwork? Would it buy your arguments?"
I must be very naive, in addition to ignorant, but it seems as bizarre that anyone would think there could be a sensitivity number as to suppose a global temperature number. Surely with the many temperature, humidity, atmospheric constituency, ocean state,and others of the thousands or maybe millions of processes which might react in different ways to any of these forcings, the reduction to a single number seems preposterous.
What am I missing?
j ferguson Jul 11 @11:03
Not naive - these global numbers have little local meaning on a given day so their meaning is hard to get ones head around in an intuitive way. But the average temperature is something that can easily be defined mathematically, and that turns out to be a useful number for looking at the magnitude of climate change. Here's an analogy: think about a house with 100 equally sized rooms. If the temperature within each room is the same in each part of the room, the global temperature of the house is the sum of the temps in all the rooms divided by 100. If the temperature in each room goes up by 1 degree, the average goes up by one degree. But if the temperature in 99 rooms goes down by ten degrees, while in one room it gets A LOT hotter (by 1090 degrees), then the average also goes up by 1 degree, though that does not really do a good job of describing what happened in any one of the rooms. In practice, though, a house with such a central heating or airconditioning system and with such insulation between rooms might be hard to build... Concerning the earth, the "rooms" are interconnected, so it will by and large be the case that if the global temperature goes up by one degree, most places will see their average temperature over a year go up by something close to one degree.
There are lots of such global numbers that mean little to individuals - gross national product can storm ahead in a country, meaning that everyone - on average - is better off than last year, but that doesn't stop some people losing their job in that year. The climate temperature sensitivity is another such global number.
Thanks for taking a shot at this Jeremy, but my thoughts were directed to the effects of change in temperatures where the reaction or sensitivity effects might be from different processes and not only might not be linear with temperature rise, but might not trigger at all except at specific temperatures and might not always be signed positive or negative. there might be hundreds of such reactions, physical and biological.
the single sensitivity number would sum these reactions but because of their variance might be a useful approximation in only the range of a degree or two C.
I agree generally agree with your view of the global temperature, but the rooms are not the same and don't have the same stuff in them, so the sensitivity calculation can only be a wild shot at the summed effect of agglomerating a whole lot of unknown and perhaps unsuspected processes.
does that make sense?
j ferguson: yes, your points make excellent sense. The climate is very complicated and as I mentioned above (Jul 9, 2011 at 8:49 PM) it seems to be correct to worry that sensitivity is not constant with time, nor with temperature, nor is it the same in response to each type of forcing. So clearly saying that sensitivity = exactly 1.283 degrees per doubling of CO2 is not possible.
But the question is, how important are all these complications? If they have the effect of spreading sensitivity all over the place, then the concept loses all meaning. But provided there are limits to how complicated the climate is, then it still makes sense to talk about a typical sensitivity value. You might not trust models in a quantitative sense (I know I don't!) but even pretty simple models (I quite like the discussion on Science of Doom, see e.g. this set of posts) can give you a reasonable qualitative understanding of how climate works. Those models could be said, I think, to suggest that sensitivity is not very variable, so it does at least make sense to talk about it as a useful number that tells us something about climate.