Monday, May 24, 2010

Equilibrium Climate Sensitivity

I recently became interested in observational constraints to Equilibrium Climate Sensitivity (ECS), and as a result learned something new about the conclusions of the 2007 IPCC report. The reason for my interest is that I am finishing up the edits on the paper about the 1983 NAS report Changing Climate that was chaired by father. In almost every case that the report covered the scientific conclusions were virtually identical to the 2007 IPCC report. The most interesting difference was that they concluded that climate sensitivity to a doubling of CO2 was more likely in the bottom half of the 1.5-4.5C range. They based this on the observational record, and assumptions about CO2.

Quoting from the chapter authored by Weller.

"If the preindustrial CO2 concentration was near 300 ppm, the sensitivity of climate to CO2 (expressed as projected temperature increase for a doubling of CO2 concentration) might be as large as suggested by the upper half of the range of thestudy of the CO2/Climate review panel (1982), i.e., up to perhaps 4.5°C. However, if thepreindustrial concentration was well below 300 ppm, and other forcing factors did not intervene, the sensitivity must be below about 3°C to avoid inconsistency with the available record."

UPDATE: By well below the Weller panel was talking about something between 250ppm and 300ppm. Separately Machta reported that the pre-industrial concentration was in the range of 260-280ppm. Combining these findings it made sense that based on observation climate sensitivity might be in the lower half of the range.

ANOTHER UPDATE: Page 19 of the report synthesis shows that the 260-280 figure is "preferred." Page 33 of the report synthesis provides the .5C increase over preindustrial levels. Note that the monitoring panel didn't yet feel that this could clearly be attributed to CO2. These are both summarized from the relevant chapters.

For a long time while I was working on the paper, I just assumed that this conclusion had simply turned out not to be in the scientific mainstream. After all I had never heard of anything like that, and it had been 30 years. Lot's of time for new observations and scientific work. But then I noticed there was a section in the IPCC WG1 report on the subject of observational constraints on climate sensitivity. You can read the section for yourself (9.6.4 page 726), but here is the bottom line.

"Results from studies of observed climate change … indicate that ECS [Equilibrium Climate Sensitivity] is very likely larger than 1.5°C with a most likely value between 2°C and 3°C.” (The go on to point out that this supports the overall estimate of sensitivity as being 2-4C.)

That's exactly the same as saying that observations would put ECS in the lower part of the range. Which means that the 1983 report had it right. The only difference is that in 1983 the NAS was willing to include this in the executive summary, but for some reason in 2007 the IPCC wasn't.

Personally I can't imagine a scientific report summary which would place so little value on observation versus modeling that they wouldn't even mention that the observations would put sensitivity in the lower half of the range. It is fine with me if they would make that comment and then explain why they think the most likely figure of 3C is still correct, but to say nothing is really weird.

To me it seems pretty well buried, which feels more like politics than science, but to be fair I haven't asked anyone involved with the IPCC about it. I'm not even sure who to ask as the chapter conclusion speaks for itself.


  1. (quickly): no, the 1983 report was wrong, becuase it said "However, if the preindustrial concentration was well below 300 ppm...". Pre-industrial CO2 was ~280 ( So whilst their argument might have been sound, their predicates were false, thus the conclusions they drew from it are void.

  2. Gee I think you need to hold off on the quick comments. By well below they were talking about the 270-280 range, which turned out to be correct. Others at the time had estimated 290 or so.

    To see this go to page 242 where Leonard Machta reports that the 1983 pre-industrial estimate is 260-280ppm

    To see the effect look at page 308 figure 5.5. The entire range of possible values discussed is 250-300.

    Also you miss the main point of my post, which is that the 2007 IPCC report came to the same conclusion about the observational evidence.

  3. Err well hold on a minute.

    However, if the preindustrial concentration was well below 300 ppm...

    UPDATE: By well below the Weller panel was talking about something between 250ppm and 300ppm

    No: by "well below" 300 they did not mean "something between 250 and 300". Quite what they meant is somewhat obscure. So we need to look into their pix; and fig 5.5 does seem to be a good choice. From that, I think they are saying, that *if* CO2_1850 was 250ppm, *then* Cl Sens of 4.5 oC is unlikely, cos delta-T from 1850 would have to be too high. But the same pic shows that if CO2_1850 is 280, then to get (by their figures) Cl Sens of 4.5 you only delta-T of about 0.8 oC since 1850, which is well within the bounds of possibility.

    So I think that with interpretation, 250 could could as "well below" but 280 can't. And 280 is the answer.

    As to your bottom line: no, I think you are misrepresenting IPCC ( Below your bottom line is Although upper limits can be obtained by combining multiple lines of evidence, remaining uncertainties that are not accounted for in individual estimates (such as structural model uncertainties) and possible dependencies between individual lines of evidence make the upper 95% limit of ECS uncertain at present. Nevertheless, constraints from observed climate change support the overall assessment that the ECS is likely to lie between 2°C and 4.5°C with a most likely value of approximately 3°C (Box 10.2). Box 10.2 fig 1 is also instructive

  4. Well at least we agree that 5.5 is a good choice.

    You are ignoring the fact that they had concluded that 280 was correct. See chapter 3. So given that they were reasonably smart they probably understood how that related to what they were concluding.

    Now how about some elementary chart reading. Draw a horizontal line across from .5 which was their gross estimate of the temperature increase. Then draw a vertical line up from 280 which would be the preindustrial CO2 level. Those two lines will intersect somehwere between the 1.5 and 2.25 sensitivity lines. Which as it turns out puts it in the lower half of the range.

    On the second point. There is no confusion about what their overall assessment is. I put that right in my original post. My point had to do with the conclusions based on observation. I am quoting directly from the end of the chapter so it simply says what it says.

    The graph adds nothing as the language is clear.

  5. I'm not sure you're being entirely open and honest here. And given that this is essentially a private conversation between the two of us, that is disappointing.

    1) "You are ignoring the fact that they had concluded that 280 was correct." - not true, for the purposes of that figure, and the discussion we're having. See p307 (from the same chapter): mid-19th C CO2 is thought to have been between 250 and 290.

    2) "Now how about some elementary chart reading" - please don't talk down to me like this or I'll get offended and go away. ".5 which was their gross estimate of the temperature increase" - hold on there. Who says so? P 307 says that the expected CO2-induced T change since 1850 may be between a few tenths to 1.5 oC - that doesn't sound like they are sure of 0.5. Fig 5.5 says *if* the T change is taken to be 0.5 oC... Browsing misc graphs in that chapter, most of the T ones don't even go back to 1850. Where are you getting your 0.5 oC from?

  6. First I am being completely open and honest. Maybe it isn't communicated well. It is a big complicated document written by multiple people along with a summary. I don't believe that every conclusion was integrated into every chapter. Some parts of chapters reference ranges where other parts of chapters have proposed answers.

    You are right about the tone. But you aren't always that nice to me BTW.

    I should point out that your original premise was that 280 wasn't enough below to justify the conclusion. I have shown that it is enough below if you assume a .5 degree increase by 1983 or so. I believe that this would match the current estimates of the increase through 1983.

    For the best picture, I think, look at the beginning of Machta's chapter 3.4. It shows the most up to date ranges for preindustrial CO2. As you can see in June 1983, right before changing climate was published, the WMO put the range between 260-280. I would therefore think that it was reasonable for them to use that in forming their conclusion about sensitivity.

    Well I did get it from their "if" statement in the referenced figure. I read that differently than you. I see it as a proposed gross estimate.

    This is backed up by table 5.5 on page 314 which shows a number of studies along with the amount of warming each attributes to CO2. In particular see Hansen et al, and Vinnikov et al. In fact those appear to be at the high end of the range.

    It certainly matches the graph on page 326, which while it doesn't go back to 1850, does go back to 1880 and supports something around .5C.

  7. Also reading the sentence in context on page 307 they are simply saying that this is the full possible range. The sentence ends by saying "estimates in the lower part of this range appear more consistent with the climatic record."

  8. Let me preface this by thanking you for setting the record straight on your father's 1983 report. I wasn't even aware of its existence and the bits I read were historically informative.

    You do not seem to have read the IPCC report carefully enough. The following caveat is used: "and other forcing factors did not intervene". As is well known, other forcing factors and most notably aerosols did intervene. The uncertainty regarding the forcing caused by these aerosols, especially during the 20th century, is probably the biggest reason why contemporary climate change is not as useful as you might think to constrain sensitivity (the main contender being the uncertainty regarding the absorption of heat by the deep ocean in my opinion). As a result of these uncertainties, observed climate change is mainly useful to determine the lower bound on sensitivity, not the upper bound.
    You shouldn't assume the IPCC summaries are politically motivated without a better understanding of the physical issues. That said, some publications in leading journals are still recklessly dismissing other forcing factors even though their authors and reviewers should know better by now. So I wouldn't want to sound like I'm blaming your father's committee for failing to take them into account in their summary.

  9. Anonymous,

    Thank you for your comments on the NAS paper.

    In the interest of positive discussion I don't particularly like the statement "you don't seem to have read the IPCC report carefully enough." Leaving that statement out, and perhaps providing a page reference to the quote would have been more productive.

    I don't find your quote in either the summary at the start of 9.6, or in the conclusions at the end of 9.6 (or anywhere else for that matter).

    Of course uncertainties in forcings is a factor in the overall uncertainty, but this is true of climate models as well.

    As to my statement about politics, I don't mean politics like Republicans and Democrats, but rather an organizational desire to stay fixed on a particular result. In this case estimates of overall climate sensitivity has barely budged in thirty years. Many people are concerned that the actual figure will be higher than the current ranges. There is almost certainly incredible resistance to presenting anything that would tend to lower that figure.

  10. I believe it's productive to speak frankly. Apologies if you were offended, as that was not my intent. I confess not to read these reports carefully myself (my interest in the field isn't professional and I don't publish).

    My quote comes from your short quote of the Nierenberg report, not from the IPCC. Sorry for being unclear. I thought you'd recognize it.
    In case you need a page reference for the uncertainties in the forcings according to the IPCC, check AR4 WG1 SPM page 4 for a convenient graphical summary. Note that the combined error bars on aerosols are larger than the whole CO2 forcing.
    As to how the IPCC arrived at its sensitivity numbers, check box 10.2 of AR4 WG1. The section 9.6.4. which you quoted refers to it.

    The uncertainties about the forcings are not such a problem for models. You could in principle run a model which doesn't include aerosols and get a good sensitivity value. But you can't observe the temperature changes which would have taken place if there hadn't been changes in aerosols. We don't know much about the aerosols which were in the atmosphere during the different parts of the 20th century so that makes it difficult to explain the observed temperature changes.
    The main uncertainty when trying to determine the "actual" sensitivity from models is completely different: it seems to come from the simulation of clouds. Clouds are not a problem for observations because the effect of clouds is of course included in temperature measurements.
    Note that sensitivity is a theoretical tool which is most useful when comparing models. It's not some kind of physical constant. It can be easily obtained from a model but deriving a sensitivity from observations is not straightforward. Most (if not all) of the studies referenced by the IPCC derived sensitivites from observations using climate models.

    I would hope that everyone would be glad to be able to justify lowering the sensitivity figure but maybe I'm naive.
    In any case, my point was that it's a mistake to think that one ought to be deriving sensitivity from observations as opposed to models. Weller seems to have derived a reasonable number from the limited data that was available but it might have been more out of luck than anything else (I don't know what he or his committee did exactly).

  11. Well I was wondering if you were quoting "Changing Climate", but it didn't make sense in context of your comment about not reading the IPCC report carefully enough. It's not a matter of speaking frankly, just clearly and avoiding things we don't know like whether I read the report carefully. (I think I did, I guess it's a matter of opinion.)

    Look I'm just reading what the document says. The people who worked on AR4 section 9.6 are well aware of forcing uncertainties.

    Of course you can easily derive sensitivity from a model. In that case the question is whether the model matches reality. So knowing the sensitivity of a model is interesting but not relevant in and of itself.

    Section 9.6 is perfectly clear. The central estimate of sensitivity from observations is between 2 and 3 degrees. The lower limit is well constrained, and the upper limit isn't.

    As you mention they synthesize this information on page 798, but even here they use the range 2 to 3.5 for observational ECS even though 9.6 puts the most likely range at 2 to 3.

    I'll also note that all future projections were based on the models, not the constraints from observations, so it seems to me in the end that the fact that the most likely range for observations was in the range of 2 to 3 made no difference to their overall assessment.

    As to whether they would institutionally like to lower the range I seriously doubt it. In fact their was a minor ruckus when some people interpreted their sea level rise estimates as being a little lower than the last assessment. They felt the need to carefully explain that this wasn't true. Even so there has been a great effort since then to show that the consensus estimate was too low.

    I can't even imagine the discussion if someone suggested that they lower the most likely figure on sensitivity.

  12. Seems to me that some of this paper is "citations needed" and some of it is argument over what words mean.

    Somewhere there the paper says Oreskes "paraphrased" something as, followed by words in direct quotes, then the statement that none of those words appear in the source. Well, they wouldn't, if it's a paraphrase rather than a direct quotation; the complaint should be to the lack of a citation to a source. Now lack of a cite is a fair complaint, and a list of claims inadequately cited is a good idea, that's how progress is made in contending papers. Just want to be clear what "paraphrase" means -- it means using different words.

    Speaking of that, another example:

    > sensitivity from observations

    This would be, I think, a short term subset of the total sensitivity. Someone better up on climatology can clarify, I hope. As I read it,
    "observations" doesn't include the paleo work -- observations are changes observed in the instrumental record thus far -- and given the known lag time for the clilmate system to go to even the shorter term, let alone the longer term, stage of equilibrium, the sensitivity 'from observations' is a part of the expected change, not the whole thing. Add that we're pushing the system far faster than in the paleo record, so it's way too early to see the full effect. The human push is faster, but the ocean circulation for example isn't any faster in response; adding CO2 faster doesn't, over the short term thus far, rush the climate response to the same rate of change.

    The other place I see a mismatch is that the discussion seems to jump back and forth between discussion of a "consensus" -- the old one, at the time, that climate change was maybe possible but not assured -- and an "emerging consensus" (which wasn't yet the consensus when the reports were written, and is still being argued against by blog scientists. The report stayed with the then consensus, it didn't go with the emerging consensus.

  13. Hi Hank,

    Welcome to my blog!

    On your first paragraph could you provide a reference? What paper are you referring to and what page number? As a general observation if something is in quotes it should be a quote even if it uses elipses etc. But without a reference I don't know what we are talking about.

    On your second point. There are lots of issues with using observational data. They are discussed in 9.6. There are also lots of issues with determining climate sensitivity from models. (In fact one of them is that the models are validated against observational data, as well as paleo data.) All of this is covered in the chapter and the conclusions stated include those issues. I am not substituting my judgment.

    On your last point again a reference would help but in my opinion you misstate the meaning of what the "emerging consensus" would be. Oreskes claims that there was an emerging consensus for policy action by scientists in 1983 that was contradicted by "Changing Climate." In my opinion this is not backed up by a review of the documents from the period. Oreskes view also contradicts Weart as I cited.

    In terms of certainty about future climate change, "Changing Climate" was more specific and covered more ground than anything before it. It also remains remarkably in sync with the current consensus thirty years later. It is difficult to pick a significant scientific conclusion from "Changing Climate" that one would disagree with today.

  14. I was remembering the word "paraphrase" from your draft, now that I look--

    In your published version, the word isn't used. Do you have a diff file for changes?
    Well, I'll read the published version and try to forget what I read in the draft, that's only fair.

    On 'consensus' -- I was thinking about these words, where I think you make a leap and didn't reply to what they wrote:

    "... the first major report on climate science issued by the National Academy of Sciences that challenged the emerging consensus view on global warming.”114 We fail to see what aspect of the then-current consensus the report challenged...."

    They refer to the 'emerging consensus view' -- hard to cite, except to personal recollection; emerging ideas are exchanged in hallway gossip at meetings long before they see first publication. Nothing is a 'consensus' until well after it finishes emerging, and gets fought out in the journals, and eventually displaces the then current consensus. By which time, new hallway gossip is starting to bubble. This stuff deserves to be dug out and written down in histories, but you and Oreskes and maybe Weart would likely be the first to the job if you pursue this.
    Climate hasn't had formal "consensus statements" til the first IPCC, as far as I know. William will know more of the history.

    Compare that to medicine, where consensus statements are published annually in many areas of interest, and much debated.
    Or nutrition, where things like the "food pyramid" are fought over by lobbying groups.

    "Consensus means that lots of people say collectively what nobody believes individually." -- Abba Eban

    (Aside -- I get what may be the same Blogger comment bug other blogs were experiencing recently.
    The 'Post a Comment' line appears just as plain text; to comment, I have to View Source on the main page, extract the URL for the comment field, paste that into a new page, get the comment window as a whole separate page; then preview (and everything disappears)
    Then PageBack, and it will be showing the text and security question as it ought to in your page.
    Then Post, and everything disappears.

    I gather Blogger may be fixing pages for blog owners who ask for help, but I don't know more.)

  15. The published version is vastly different than the earlier one. And I believe is a great example of how peer review, and a good editor can help. Take a look and let me know if you think the comments still apply.

    As to consensus Oreskes definitely makes an attempt to pin down evidence for this emerging consensus. But I believe that this was incorrect as I point out in the paper. But again take a look and let me know what you think. If you think that she meant some type of general thing with nothing to reference then I don't think it is appropriate in historical writing.

  16. An emergency prevented me from answering your comment dated "August 2, 2010 2:16 PM" in a timely fashion.

    What I meant by talking about Weller's caveat in the same breath as the IPCC report is that the IPCC report explains why Weller's caveat is actually a fatal flaw. With hindsight, I realize my meaning was practically impossible to discern. Sorry about that.

    I don't think 9.6 is all that clear. 9.6 does not say sensitivity is most likely between 2 and 3 C per doubling. Instead, 9.6 reaffirms that the most likely value is "approximately 3C". I don't think the studies about observed climate change were supposed to determine the upper bound of sensitivity (see table 9.3 for the very loose upper bounds they support). If you approached the final paragraph of 9.6 with the notion that sensitivity can be deduced from observations, the I can see how you'd get the wrong impression.

    Yes, one needs to verify whether a model matches reality. But this isn't principally done by comparing sensitivities because sensitivity is not observable. It is a property of models, not of reality (except theoretically). Reality is a lot richer than that! See 8.6.4 for a summary of more relevant aspects of models to compare with reality.
    Note that sensitivity values are not quite forecasts about what would happen to temperatures if the atmospheric CO2 concentration was doubled so you really want to use models for forecasting anyway.

    Hopefully I've now adequately explained why I don't think that the upper bound of the 2-3 range should have been given much weight in the overall assessment. I'm not seeing any evidence of some kind of institutional bias at work.