Friday, October 03, 2014

Much ado about sensitivity

Well, perhaps not really very much ado. There's a new paper in Climate Dynamics, by Lewis and Curry, with a central sensitivity estimate of 1.6C with a 90% range of 1-4C, based on energy budget analyses over the instrumental period, updated to the present day, also taking account of the newer AR5 forcing estimates. I don't find it particularly exciting, the authors cite several recent papers with similar results including Aldrin and Otto et al. I wrote about those papers some time ago, and I think these posts (1, 2, 3) still stand. I've commented before on my objections to Lewis' method, and especially the sleight-of-words with which it is described, but (as I've also emphasised) I don't think this substantially affects the results in this application.

Clearly, the longer the relatively slow warming continues, the lower the estimates will go. And despite what some people might like to think, the slow warming has certainly been a surprise, as anyone who was paying attention at the time of the AR4 writing can attest. I remain deeply unimpressed by the way in which this embarrassment has been handled by the climate science insiders, and IPCC authors in particular. Their seemingly desperate attempts to denigrate anything that undermines their storyline (even though a few years ago the same people were using markedly inferior analyses of this very type to bolster it!) do them no credit.
One weakness of these energy-budget type of analyses, that I believe Lewis and others could easily address, is to demonstrate how well it works in application to GCM output. That is, can the method accurately diagnose the sensitivity of a model given equivalent information to that which we have for the real world? Aldrin et al addressed this rather briefly and in a very limited way, using a far stronger forcing scenario (1%pa CO2 enrichment) than what has occurred in reality. It would be easy to investigate the precision of the method, and whether it gives rise to any systematic biases, by using output from the more realistic 20th century simulations. It is also noteworthy that the Aldrin method struggles to cope with hemispheric differences, which may point to some limitations of the energy balance concept. While the climate system certainly does obey the fundamental conservation laws, supposedly “fixed” parameters (in simple models) are not actually constants in reality. And no matter now precisely we can determine the historical transient response to the current radiative imbalance, there will always be a bit of additional uncertainty in extrapolating that to an equilibrium 2xCO2 state.
Finally, it is also amusing to see Judith “we don't know anything” Curry to put her name to this new paper: it is unclear what she might have added, as Nic has been presenting analyses of this nature for some time now. But that's a minor matter.


  1. Hawkins and fellows applied the Otto et al. method to CMIP5 models and it looks like the 2-4.5ÂșC range is well-captured, not sure about individual models. However, the forcing numbers come from Forster et al. 2013 which, if I understand correctly, effectively involves rearranging the energy balance sensitivity equation so getting a sensitivity match from re-rearranging is a bit circular.

    I think the tricky part would be getting forcing numbers, particularly ERF numbers, which are independent of the model sensitivity.

    One interesting aspect of the EBM results is the small ECS/TCR ratio. After adjusting for the obvious bias in using HadCRUT4 I get a central TCR result of about 1.5K, which is only 15-20% smaller than the CMIP5 average of 1.8K, but ECS is still only about 1.9K, about 70% smaller than the CMIP5 mean.

    This would tend to suggest fairly rapid equilibration in the system, which would still mean substantial warming up to 2100, if not much thereafter assuming stabilisation.

    The INMCM3 model provides a good example in the CMIP3 ensemble. It had diagnosed TCR and ECS of 1.6 and 2.1, in good agreement with figures above. Despite the low-end ECS, warming between ~2014 and 2100 for A1B is about the same as the CMIP3 mean.

  2. Lewis & Curry should have used C&W in addition to, or instead of, HadCrut4. C&W warmed about 8-9% more over the instrumental period.

    Most of that is due to coverage bias in the recent period (esp. HadCrut4 omission of Arctic), but some also comes from differences in the 19th century.

    Time_int CW_ver2 HadCrut4
    1859-1882 -0.316 -0.283
    1995-2011 +0.458 +0.427
    Difference +0.781 +0.720

    TCR adjusted = 1.44 (a little lower than Paul S, but close)

    Mind you this is just back of the envelope - I haven't plugged CW into the LC code.

  3. Yes, a direct comparison of HadCRUT to model global data is certainly a mistake. Normal procedure is to mask model output.

    As for CMIP forcing, surely the point would be to use our estimate of real world forcing (which is also uncertain) since the models are supposed to be simulating this.

  4. But masking will still underestimate - just not as badly.

    Shindell proposed using CW, but the "long" series was not available at that time, so his estimate of the necessary correction was too low.

  5. Yes, probably - but at least in that case, the bias will be due to model error (ie not correctly reproducing the warming pattern) rather than a methodological error. And it should surely be smaller.

  6. deepclimate,

    C&W only accounts for coverage bias. Model sensitivities are given as global SAT whereas HadCRUT4 and C&W (mostly) are landSAT+oceanSST. The difference between SST and ocean SAT in models suggests a bias of similar magnitude to what you've identified for C&W-HadCRUT4 difference.

  7. On reading the first draft I suggested C&W. The response I got was satisfactory. A) they want to stick as close to IPCC data as possible. B) they would certainly do this if reviwers questioned it. C) the code was provided for anyone to tinker with

    Satisfied me

  8. "Clearly, the longer the relatively slow warming continues, the lower the estimates will go."

    Not really. The rate of warming is a function of initial state, forcing and natural variability. You would expect a relatively slower rate if there is less forcing or negative phase natural of variability. So (of course) the estimates should only go lower if models are too low after accounting for all involved factors. Two recent works suggest that there is no evidence of that to current date:

    "We see no indication, however, that transient climate response is systematically overestimated in the CMIP5 climate models as has been speculated[8], or that decadal variability across the ensemble of models is systematically underestimated, although at least some individual models probably fall short in this respect."

    "Accounting for these adjusted trends we show that a climate model of reduced complexity with a transient climate response of about 1.8 °C is consistent with the temperature record of the past 15 years, as is the ensemble mean of the models in the Coupled Model Intercomparison Project Phase 5 (CMIP5)"

    Where are the evidence and arguments for a GCM low bias? If there is I'm genuinely interested in knowing about it.

    Lewis&Curry is another energy balance paper which, as already has been mentioned, are biased low according to current knowledge. And when including uncertainties in forcing, ocean heat content and model simplifications, they don't help very much in constraining sensitivity, just narrowing the normal interval a bit.

  9. William Connolley,

    It is Sir Mosher.

    A isn't really true since the IPCC give sensitivity estimates which they chose to ignore.

    B suggests they knew their results were biased low and underestimated uncertainty but made a deliberate attempt to push the paper through without disclosing this. I recall someone recently arguing such a thing could reasonably be described as fraud.

    C is irrelevant to the point.

  10. Paul S,

    You say:

    "The difference between SST and ocean SAT in models suggests a bias of similar magnitude to what you've identified for C&W-HadCRUT4 difference."

    Does that bias include the Arctic Ocean? C&W estimates Arctic SAT, not SST. So the SST/SAT bias should be evaluated for open ocean only (no sea ice), right? Would that appreciably affect the magnitude of the SST/SAT bias?

    In other words, what would be your adjusted estimate of TCR, accounting for both coverage bias and SST/SAT bias?

    It sounds like we're up to ~1.6K.

  11. Nic Lewis at CA:

    "It has been claimed that incomplete coverage of high-latitude zones in global temperature datasets biases down their estimate of the rate of increase in GMST. However, over the long periods involved in this study there is no evidence of any such bias. The increase in GMST per the published HadCRUT4v2 global dataset, used in the study, exceeds rather than underestimates the area-weighted average of the calculated increases for ten separate latitude zones, which method gives a full weighting to each zone."

    Robert Way:

    "There are three available datasets that go back to 1850. CW2014, Hadcrut4v2 and BEST. Both CW2014 and BEST show about 10% more warming using your preferred base and final periods (1859-1882; 1995-2011). Is the rationale for not including coverage bias or either of these datasets discussed in more detail anywhere?"

    Also see:

    Robert Way again:

    "If you haven’t read the papers or tested the methods then its hard to accept your reasoning for not including. Your intuition is not the same as actually looking at the data. I just do not understand how from a communications standpoint it makes sense to even open yourself up to the cherry-picking accusation from those you’re trying to convince of your papers credibility."

  12. deepclimate,

    There isn't just a double-counting issue. I think because of losing sea ice insulation the SAT/SST ratio in the Arctic is very large, and this wouldn't apply to observations since they haven't ever been below sea ice. My SAT/SST bias estimate is based on SAT vs SST with Arctic and Antarctic regions cut out. It's about 5-9%.

    My overall bias estimate is based on comparing model 60S-60N landSAT+oceanSST with model globalSAT, with an adjustment to reflect that HadCRUT4 global shows about 6% more warming than HadCRUT4 60S-60N. The result suggests an overall 15% bias. I've also tried 55S-55N and 50S-50N with negligible change in result.

  13. Oh, so based on HadCRUT4 deltaT being 0.72, my bias-adjusted estimate is 0.83 which works out to TCR being 1.55K.

    However, this 0.72 figure seems to come from a just-released new version of HadCRUT4. My bias estimate is based on the previous version which gave 0.71. Based on that the TCR comes to 1.525K, so about 1.5K.

  14. In my comment 4/10/14 4:42 pm "low" should be "high" at two places (sorry for the stupid errors).

    Trenberth has commented on L&C, referred here:

    Aside the temperature coverage issues, he claims similar errors in their selection of ocean heat content values and concludes:

    "The result is that the Lewis and Curry estimates are perhaps 50% too low, and their uncertainties are much too low."

  15. Slight correction (sigh): The deltaT for HadCrut is 0.71, as Paul S stated.

    So the TCR adjustment for the main "best" estimate is 1.33 -> 1.45.

    RC has a post on LC by the way:

  16. Hi, I found your comment on another blog that you know about post docs in Japan. I am a Ph.D. molecular medicine student considering doing a post doc in Japan. I work with a person who is from Japan, and he said that he worked 80 to 100 hours a week in Japan and that he doesn't want to go back and strongly hinted that I would be expected to work 80 hours a week if I went there. Help! I would be okay with 50 to 60 hours a week, but 80 to 100!? Please, I've tried looking up information online but I can't find anything and no one's responding to me. I very desperately would like to hear a first person account from someone who has worked in Japan. How many hours did you work per week? Did your colleagues work per week? Was 9 am to 9 pm normal? My work personal e-mail address is Please give me advice.


  17. Erica - in Japan, non-Japanese are treated completely differently from Japanese. The Japanese post doc expects to be told exactly what to do, and the boss expects them to do as they are told. The foreigner, however, does not usually get treated this way. In fact quite the opposite; they are often left almost entirely to their own devices. This means that the most important thing for a foreigner going to do a post doc in Japan is that they are self-motivated and independent. Hours are usually very flexible in universities and research institutes.

    Of course, Japan is a big place, and so behaviour may vary in different places... Some bosses are much more non-Japanese friendly than others.

  18. Erica, Eli has some friends who post-doced in Japan and had a great time. The problem is that it will be about impossible to get a job from there (it is far away from the US and Europe) and you will have to take another post doc or something temporary when you return in order to job search

  19. Curry had an op-ed in the Wall Street Journal touting her paper

    "The Global Warming Statistical Meltdown
    Mounting evidence suggests that basic assumptions about climate change are mistaken: The numbers don’t add up."