Sunday
Apr082012
by 
Bishop Hill
Bishop Hill Hide the incline
Apr 8, 2012 Climate: other Willis Eschenbach has been looking at the Shakun et al paper that was the subject of my Greenhouse reversal post a couple of days ago. It looks as if he has found something important:
I leave the readers to consider the fact that for most of the Holocene, eight
millennia or so, half a dozen different ice core records say that CO2 levels were rising pretty fast by geological standards … and despite that, the temperatures have been dropping over the last eight millennia …
Reader Comments (66)
It's unfortunate that the blogger mentioned in the Intro didn't do what Shakun et al (2012) did and include a [CO2] concentration axis in his figure. Then one would see immediately that the argument is just silly (see below for why). That's not to mention the fact that Shakun et al (2012) is about temporal relationships between [CO2] and temperature during the last deglaciation; the title of their paper is “Global warming preceded by increasing carbon dioxide concentrations during the last deglaciation”. Why should they include data for the mid-late Holocene?
The rather low resolution Taylor Dome data indicates an extremely gradual increase in atmospheric [CO2] from around 265 ppm about 6000 years ago to around 275 ppm 2000 years ago, where the Antarctic CO2 data from Taylor Dome meets the high resolution Law Dome record [see e.g. C. MacFarling Meure et al (2006) Law Dome CO2, CH4 and N2O ice core records extended to 2000 years BP Geophys. Res. Lett, 33, L14810] which indicates an atmospheric CO2 level of 276 ppm 2000 years ago.
So this fuss is about a rise of atmospheric [CO2] of the order of 10 ppm over a period of 4-5000 years which is supposedly "pretty fast" according to some bloke on a blog. That's 4-5 years worth of raised [CO2] by current rates of emission or around 1000 times slower than present day rates of increase of atmospheric [CO2].
Do we know why [CO2] drifted upwards extremely slowly during that period? I'm not sure we do. Bill Ruddiman considers that at least some of this extremely slow and small rise was the result of the onset of agriculture in Eurasia (forest clearance from 8000 years ago; rice growing from 5000 years ago which matches the start of anomalous rice in methane). That seems pretty likely to be at least part of the story. But otherwise we're quibbling over extraordinarily small and slow changes in the context of current emissions and their warming consequences.
Note that there is no necessary incompatibility between the small, slow rise of [CO2] and an accompanying small, slow drop in temperatures during the early/middle to late Holocene. It’s easy to calculate that with a climate sensitivity of 3 oC (surface temperature rise per doubling of [CO2]) that a 265 ppm - 276 ppm rise in [CO2] should contribute a small forcing equivalent to a temperature rise of near 0.18 oC. Spread over 4-5000 years it’s not very surprising that natural contributions (Milankovitch cycles) overcome that tiny greenhouse-forced contribution.
In relation to Carl Wunsch’s work on Milankovitch cycles cited by Spence_UK, it’s easy to see from Wunsch’s later work [*] that his analyses are entirely in line with the dominant role of Milankovitch cycles as the pacemaker for glacial-interglacial transitions. That’s entirely consistent with a huge amount of very detailed and thorough analyses that indicate a pretty rock solid role for Milankovitch cycles.
[*] e.g. Tziperman Eli; Raymo Maureen E.; Huybers Peter; Wunsch, C.
Consequences of pacing the Pleistocene 100 kyr ice ages by nonlinear phase locking to Milankovitch forcing Paleoceanography 21 art. # PA4206
Huybers P; Wunsch C (2005) Obliquity pacing of the late Pleistocene glacial terminations
Nature 434, 491-494
Chris,
If you had read the link from my post (Apr 9, 2012 at 1:12 PM) you would see I've already addressed the Huybers paper. A proper treatment of natural variability shows that this is a far more important component in the glaciations than orbital forcing.
really Spence_UK? You said not one jot about "Huybers paper" (I cited two of Huybers and Carl Wunsch's papers). I pointed these out partly to show that that Dr. Wunsch seems completely in line with the rather vast evidence-base for the dominant role of Milankovitch cycles in pacing glacial cycles (at least that's what his papers say!).
I'd be surprised if you consider that an unpublished conference presentation by statisticians that do time series analyses (largely of hydrological cycles) "trumps" the entire field of study of glacial cycles! That would be taking special pleading in pursuit of a particular "interpretation" to heroic heights! On the other hand you did say that you had "addressed the Huybers paper" in your post", so maybe that's what you meant...perhaps you could clarify.
While the Greek water cycle statisticians still haven't published their analysis as far as I can see, Huybers has published more recent analyses which further supports a role for obliquity and precession in pacing deglaciations [P. Huybers (2011) Combined obliquity and precession pacing of late Pleistocene deglaciations Nature 480, 229–232]. That seems rather consistent with recent work from the Japanese glacial group on Milankovitch forcing of terminations extended back to pre-Holocene terminations [Kawamura, K. et al. Northern Hemisphere forcing of climatic cycles in Antarctica over the past 360,000 years. Nature 448, 912–916 (2007) and quite a bit of other stuff.
We do need to be careful in overinterpreting statistical analysis of timeseries to attempt to pull out statistically-defined components and their amplitudes, and then assigning elements that don't fit a particular model to "natural variability". A better statistical analysis of glacial cycles ("better" in the sense that it was informed by a mechanistic understanding, I would say!) was published by Lorraine Lisiecki recently [L. E. Lisiecki (2010) Links between eccentricity forcing and the 100,000-year glacial cycle; Nature Geoscience 3, 349 - 352.], which also supports some role for variability that is internal to the climate system (ice sheet dynamics and greenhouse gas variation through the carbon cycle), but which is phase-locked to eccentricity (and suppressed by precession). So again the underlying drivers and pacing of ice age cycles remains the Milankovitch cycles even if natural feedbacks are obviously involved in defining whether thresholds are reached for the switching between climate states and the amplitudes of the transitions.
That's my understanding of the field. There really isn't much doubt that Milankovitch cycles pace the glacial-interglacial transitions - the evidence is simply too strong to consider otherwise. The details by which the variations in patterns of insolation and the feedbacks associated with natural variability affect the precise timing and progression of ice age cycle transitions aren't yet pinned down...
Chris, the link I provided includes a commentary on the Huybers' view of orbital pacing. I can only assume you did not understand the link I provided.
The link notes that there is an influence from orbital forcing, but this influence is small in comparison to Hurst-Kolmogorov dynamics, i.e. the glaciations would have happened anyway even if orbital forcing was not present.
The Hurst-Kolmogorov dynamics and their effect is absolutely crystal clear in the power spectral density of the summer insolation time series.
However, I notice that rather than address the issue raised, you make snarky remarks about the hydrological community. Clearly, the hydrological community has a good grasp of statistics but they are not primarily statisticians. I also find it amusing that you sneer at someone for having some skill at using statistics - statistics is a cornerstone of every scientific discipline, and one who takes such a tone as you do cannot possibly understand science.
But this is shown by your response. You have the opportunity to understand and discuss the poster session that I linked but you find non-scientific reasons for disliking it, perhaps the country of the author, or perhaps the venue at which it was presented. This tells me you have not understood the paper and have no cogent criticism of it.
Kudos to anivegmin for acknowledging that they could not critique the paper. Sad to see that your inability to critique the paper instead results in a childish sneers at hydrologists and statisticians.
If anyone wants to ask further questions about my paper, please do so on the tallbloke's talkshop thread where you will find my thread among the most commented in the left margin.
Not sure how one is supposed to "critique" a poster presentation Spence_UK! Perhaps you aren't aware that a poster is a preliminary summary of research one presents at a scientific meeting as a focus for discussion with colleagues. Once the work is considered sufficiently robust it's submitted for publication. At that point we can assess exactly what they did and how they did it.
Koutsoyiannis and his group presented their poster nearly two years ago and haven't published it yet. So we don't really have any basis for "critiquing" it. When they consider the work sufficiently robust to publish then no doubt it will be interesting to read.
Your over-the-top false precis of my post points to a certain defensiveness on your part! It's useful to understand the background of any piece of work and I don't see a problem with stating that these guys are time series statisticians who work largely on hydrological cycles. They've never published on glacial cycles before. And personally I think statisticians are great; not sure what you gain by imagining things in my post that I simply didn't say.
Although one can't really critique Koutsoyiannis's work until they publish it we can point out a problem that is inherent in statistical analyses of this type. The authors assess a time series (paleodata on parameters associated with glacial/interglacial transitions) in relation to a model (Milankovitch cycles), to assess the proportion of the variance in the data that can be explained by the model; they lump the part that doesn't fit as "natural variability". However, that's only valid in a strict statistical sense. Because, of course, their "Natural variability" isn't necessarily natural variability at all. I expect a large part of it is in fact a result of inaccuracies in the dating of the paleo-parameters that make up the observational time series. That's rather indicated in the fact that the "natural variability" is reduced rather dramatically if the statistical analysis is performed on orbitally-tuned paleodata.
So if I was going to consider whether Koutsoyiannis's work (if or when published) had anything particularly interesting to say, I would want to see how their analysis performed on the more recent paleo-data in which timing issues have been further resolved [e.g.H. Cheng (2009) et al. Ice Age Terminations; Science 326, 248-232.; Huybers (2011) and Kawamura et al (2007) cited in my post above, S. Barker et al (2009) 800,000 Years of Abrupt Climate Variability, Science 334, 348-351 and the more recent improved ice core dating by Kawamura et al (2012) and so on.
The very large body of recent work on ice age cycles and terminations makes it increasingly difficult, in my opinion, to maintain opposition to the evidence for a dominant role for Milankovitch cycles in pacing ice age cycles; the evidence is simply far too strong. You cited Dr. Wunsch earlier on this thread, but his more recent work strongly supports Milankovitch pacing; it would be more useful for this thread if you were to "critique" his work, and that of Lisiecki, and Kawamara, and Barker and Cheng and Huybers, rather than demanding that we critique an unpublished poster. The idea that one can reject a vast body of evidence on the basis of an unpublished conference poster presentation seems to smack of desperation to me...
Wow. There is some analysis in their Chris. Tell us what is wrong about the analysis.
Sheesh, this isn't rocket science.
If I say 2+2=5, you can critique that without it being peer reviewed, am I right? You just point to the arithmetic error. It isn't that difficult.
The work you link to fails to correctly apply a stochastic model to assess the consequences of the dominant stochastic component of the glaciations. This is simple enough to understand, and something underscored in Wunsch's original article.
But then, if you don't understand why 2+2 isn't 5, then you can't very well critique it, can you?
And sadly, that's where your knowledge of Hurst-Kolmogorov dynamics runs into problems.
The one thing we see persist throughout proxies of all ages and all types is the presence of long-term persistence in the time series autocorrelation structure. It is visible in power spectral density estimates from the order of days out to the order of hundreds of millions of years.
There are two other reasons your argument here makes no sense. Firstly, the natural variability is the dominant term - it isn't a second order term, but the primary source of variability in the record. So there is nothing for it to be confused with unless the entire record is junk. I do not think this is the case (the Huybers reconstruction is actually very good). Secondly, if there were dating issues, we would expect the orbital peaks to be smeared - but they are anything but, they are sharp and clean.
This made me laugh a lot. If you tune a core to the conclusion you want, you get results which are more consistent with said tuning.
Oh dear. Putting your conclusions in your assumptions is not a good plan, old chap. You end up drinking your own bath water. What is more, even WITH orbital tuning - which by definition must be the maximum possible effect the orbital cycle can have - Hurst-Kolmogorov dynamics alone can completely change the timing of the glaciations. And with high confidence the orbital cycle has less effect than this.
Not really Spence. You don't seem to be grasping the points. The first is that Koutsoyiannis and his group haven't published their work. We don't know what they did or how they did it, so we can't really critique it. That's obvious isn't it? Perhaps they haven't published it since they don't consider it sufficiently robust to publish. Who can say...
I'll describe the second point a thrid time since you don't seem to get it. In a statistical study one assigns significance of observations according to a model with a self-constructed null hypothesis, or in this case one can extract amplitudes for components according to a simple model (e.g. fitting paleo-observations to predictions from Milankovitch cycles to extract the contributions to variance from, in this case, the Milankovitch cycles). Koutsoyiannis lump the residuals into "natural variance". However that obviously doesn't mean that everything that isn't described by the cycles is due to "natural variance" since poor fitting of observation to model will clearly also have a greater or lesser contribution from to the quality of the observation. As we've seen, if the analyses is done on orbitally-tuned data the contribution from "natural variance" plummets. Of course that's not a reason for preferring orbitally-tuned data for this analysis (orbitally-tuned data is useful for some analysis and less usful for others). However that analysis does show that with an improved timing of paleo-data, the role of Milankovitch cycles in statistical analyses might well be improved (it might not be too, of course!).
We now have considerable recent improvement in the timing of paleo-data, and so one would like to see Koutsoyiannis analysis done on that, if they ever consider their data worth publishing.
Not sure there's much point in continuing this discussion since you seem to be determined to play semantic games and to argue by assertion - you don't seem very interested in the science, and that makes your assertions less than interesting. One simply can't get away from the fact that the data on Milankovitch cycles is pretty robust now; I've cited a whole slew of recent papers, including a couple of papers by Cal Wunsch who seems (at least according to his more recent published work) to be rather in agreement with the dominant role for Milankovitch pacing of glacial cycles. You really need to convince us why Wunsch, and Huybers, and Lisiecki and Kawamura, and Cheng and Barker and all their colleagues are wrong.
Anyway, I'm pretty relaxed about the evidence and don't feel like having a dull argument you (your attempts at "gotcha"-style arguing are tedious) ....I'm sure there are natural contributions to ice age cycles - it's obvious that slow ice sheet dynamics and its roles in feedback through albedo effects and effects on thermohaline circulation play a role in ice age cycles. So must the ocean sequestration and re-recruitment of CO2 since that's a dominant part of the paleo-record and we know that CO2 is a strong greenhouse gas. However it's pretty incontrovertible that the Earth orbital properties pace the ice age transitions - that's what the evidence shows...perhaps we can agree on that and agree to disagree on the science that is supported by a vast evidence base but which you don't like..
What? The time series used are published time series, and I assume are available. The analysis used is straight forward, with equations and everything, plus references to how those equations are derived. Unlike (say) Lonnie Thompson's published work, for which raw data is *not* available making detailed critique difficult if not impossible, anyone who understands how to manipulate equations can understand, replicate and therefore critique the work.
Your claim that something is not published and therefore not possible to critique is a bizarre non-sequitur.
You clearly DO NOT understand hypothesis testing and are falling into the traps of basic fallacies.
1. The null hypothesis is not "self-constructing". It requires a model for handling characteristics of the time series, such as serial correlation. In hypothesis testing, you DO NOT simply get to hand wave serial correlation away as you are doing - that is guaranteed to lead to many type I errors.
2. Classical statistics typically handle serial correlation by assuming short-term persistence. This is fine for many time series but unsuitable for climate time series. Many researchers have checked for the presence of long-term persistence and confirm its presence.
No. The model fitted by Koutsoyiannis here maximises that attributable to the orbital cycles. You can see from the power spectrum that the resultant is a continuum of variability beneath that, and that it ultimately dominates the time series. Furthermore, this component is present and verified in all proxies and even high-quality instrumental series. So to hand wave it away is to ignore the elephant in the room.
Plummets!?!?! It still dominates! You have got to be kidding me! Plus orbitally tuned data will obviously inflate the amount represented by the orbital terms, a point Wunsch makes clearly in his article!
And it might get worse as well. We can all hand-wave about how things might be different. That isn't science.
They have published many papers on the topic, and many papers are referenced in their poster, a point you are notable ignoring.
(Cont'd)
No, I've linked to an analysis using real data and equations that can be discussed. You're the one making basic fallacies and errors about hypothesis testing. Believing null hypotheses are "self-constructing" is one of the funniest things I've heard for ages, thanks for that. On that grounds, I agree there is little point in continuing. You don't understand hypothesis testing, you don't understand what the null hypothesis is, you don't understand that size of effect is not related to size of p-value and that separate analysis is required to confirm that, you don't understand that long term persistence seriously undermines what you claim and you don't understand how to critique a paper when the data and equations are in front of you.
Oh dear. Getting details right in science is important. Pointing out elementary errors in statistics might be tedious, but it prevents arriving at the wrong results, as you have.
That's not evidence, that's hand-waving. Sorry. The evidence is in the obs and the analysis. And that shows there is a small component to orbital cycles, and the rest is natural variability. Yes, CO2 is a greenhouse gas but the evidence has no requirement for CO2 to be invoked to explain temperature. Of course you can force it in, but that isn't how science works.
You have yet to show any evidence beyond wishful thinking. That may be vast, but it isn't evidence.
No worries Spence_UK. As I said I don't feel like having a tedious blog argument, especially the particulalry dreary sort where one's text is salami-sliced for misinterpretation.
I've cited a bunch recent published papers that illustrate the growing understanding of the relationships between earth orbital properties and glacial-interglacial transitions; we could look at another dozen. Sadly you're not interested in this science. That's O.K. Spence...no one's forcing you to learn the science on this fascinating subject!
Koutsoyiannis and his group haven't ever published on glacial transitions and Milankovitch cycles. It's easy to do a Web of Science search and establish that simple fact. Don't know what you gain by asserting porkies! When they actually publish the stuff they presented as a poster a couple of years ago in the Athens of the North, then we can see exactly what they did and whether it's of any interest. However they don't seem that keen in doing this so far. Maybe they don't consider their analysis sufficiently robust - who can say? You love their poster - it supports your point of view apparently..what could be more satisfactory?!
Great to see your pleasant debating style, chris. Thanks for your contributions, Spence. I can see who Wunsch would be more likely to agree with.
Thanks Jeremy...I also agree that if we really wished to know who "Wunsch would be more likely to agree with" it's pretty easy to answer by looking at his published science. After all in his most recent paper on the subject with Peter Huybers Obliquity pacing of the late Pleistocene glacial terminations; Nature 434, 491-494 (2005) he indicates that the data are entirely consistent with orbital pacing by Milankovitch forcing. For example he (and Hybers) state:
"The simplest inference consistent with the test results is that the ice sheets terminated every second or third obliquity cycle at times of high obliquity, similar to the original proposal by Milankovitch(12)."
That could hardly be more clear.
Of course it's not about who Dr. Wunsch might or might not agree with, but what the evidence shows...
Chris,
Anyone can dump a bunch of references, but that doesn't help if you can't explain the contents. I asked how the papers addressed the issues raised in Koutsoyiannis but you seemed unable to answer - just continually citing papers.
The first step in science is understanding the papers content. If you do this - whether in the peer reviewed literature or presented at a respected conference or even on a preprint server like arXiv - that's where doing science really begins.
I'm not really interested in who can cite the most papers, it's a pointless exercise, what is more interesting is getting a complete understanding. One of the most interesting aspects of climate science is not what the papers say, but the blindingly obvious things they often miss out. Milankovitch cycles are a prime example of this - it's not what the papers say that is interesting, it's what they leave out that is the really interesting stuff.
This comes from many, many years of reading the peer reviewed literature. (In my experience, even in good fields, 50% of the peer reviewed literature is dross, and around a third of it is seriously flawed or false. Those who consider the peer reviewed literature to be a gold standard are naive. You need to understand the papers first and foremost)
Great summary of peer reviewed reality Spence - it sure rings true to me. Even in good fields, as you say. And as for climate science ?