As I’ve mentioned before, the Von Storch and Zorita Comment was accepted by GRL and will appear soon. Von Storch has been very courteous to us both in public and private and it has been a pleasant exercise replying to their Comment (as opposed to wading through the bile of Ammann and Wahl.) They are somewhat in competition with us for the peculiar honor of breaking the hockey stick and readers have to bear that in mind (that goes both ways.)
Here is a preprint of our Reply (all rights reserved to American Geophysical Union- not to be reproduced without permission. )
The Can-Opener
I remember a joke from about 40 years ago about a physicist, an engineer and an economist stranded on a desert island with cans of food, but no can-opener. The physicist proposed the following solution: I’ll take the lens from my eyeglasses, focus the sun’s rays on the can and burn a hole to open the can. The engineer thought he had a better idea: use brute force, just take a rock and bang the can until it opened. The economist said that he had a solution that trumped them both – "assume that we have a can-opener."
If you look closely, the Von Storch and Zorita Comment assumes they have a can-opener.
They agree that the MBH98 principal components method is flawed (a welcome confirmation!); but argue that the flawed method doesn’t "matter". As always, the devil is in the details. They try to show this using "pseudoproxies" constructed from their ECHO-G model and endowed with correlations to gridcell temperatures of 0.3-0.7 (based on Jones and Mann, 2004). Sounds reasonable enough on the surface, but there’s a big problem: MBH98 proxies in the AD1400 aren’t nearly this accurate.
Here’s a histogram from our article showing the correlations to gridcell temperature, state precipitation and CO2 levels for the 70 proxy series in the critical AD1400 North American tree ring network (including the bristlecones). None of the 70 series has a correlation to gridcell temperature exceeding 0.2 and more are negatively correlated to gridcell temperature than are positively correlated. The vertical red line marks the minimum gridcell correlation used in VZ pseudoproxies (0.3). The correlations to state precipitation are much higher with many series have a correlations to gridcell precipitation exceeding 0.3. A number of series (all bristlecone/foxtail) have high correlations to CO2 levels. You can see why I am reluctant to say that there is a "right way" to extract a "temperature signal" from this particular dog’s breakfast of proxies.
Figure 1 Adapted from McIntyre and McKitrick [2005c]. Reply to Von Storch and Zorita.
So when von Storch and Zorita say: let’s assume that the pseudoproxies have a correlation of 0.3-0.7 to gridcell temperature, they might as well be saying: "let’s assume we have a can-opener".
We have one other serious issue with how they implemented the MBH98 PC methodology. In our article, we pointed out that the data mining was accentuated by two aspects: the short-segment centering AND using singular value decomposition on the de-centered data matrix rather than the covariance or correlation matrix (we noted that the latter would have substantially mitigated the effect, even with short-centering.) Unfortunately, von Storch and Zorita’s description of methodology indicates that they followed the short-centering part of the recipe, but used the correlation matrix instead of the decentered data matrix in their calculations. This would seriously attenuate their results.
The "Bad Apple" Effect
I was very intrigued by the following comment in von Storch and Zorita – they attributed the Artificial Hockey Stick effect in our simulations to the fact that we used "red noise" where there was nothing to "steer" the selection of the eigenvectors. I’ve found the image of "steering" the selection of eigenvectors to be a very useful concept and to be very helpful in showing the effect of a few "bad apples" on the selection of eigenvectors in the MBH98 method. While the Artificial Hockey Stick effect certainly existed in a red noise situation, it is far more intense with a few "bad apples" (such as bristlecones with nonclimatic trends). We were more or less aware of this before, but did not illustrate it in our first article. We used the Reply to show this more clearly.
In MM05 (GRL), we had already pointed out that MBH98 PC1 was made up entirely of bristlecones. We experimented with "pruning" the number of bristlecones in the network, reducing them sequentially from 16 to 0, looking at the "hockey stick"-ness of the PC1 and the steering of eigenvector selection.
Figure 2 below ( in our GRL SI) shows the loading of each site (squared) in the PC1, with the 16 bristlecone sites at the extreme left (Sheep Mountain in the first spot). The first panel on the left shows that the 16 bristlecone sites have loadings well in excess of those of the other sites (they contribute over 90% of the variance). The second panel shows the situation with only 3 bristlecone sites under the MBH98 method: the weights previously allocated to all 16 bristlecone sites are now nearly all allocated to only 3 sites. The third panel steps forward to only 2 sites: again 2 bristlecone sites are heavily loaded and dominate the PC1. The fourth panel shows the loadings for the PC2 where only one bristlecone site dominates the PC2. With no bristlecones in the 5th panel, outlier weights are much better controlled.
Figure 2. EOF1 loadings for North American AD1400 netowrk.
The next figure shows the effect of the pruning on the "hockey stick" index of the PC1 and PC2 using the MBH98 methodology. Down to 2-3 bristlecone sites, the hockey stick index of the PC1 is still extremely high. Even with 1 bristlecone site, the hockey stick index of the PC2 is higher than with the attenuated method of the VZ simulations. It is remarkable that 1 site can "steer" the selection of eigenvectors so strongly.
Figure 3. Top – Hockey Stick Index of NOAMER PC1 and PC2. Bottom – share of bristlecones in EOF1 and EOF1+EOF2.
Finally, I show here another figure from our SI. It would have been nice to show everything in our article, but GRL has limited space. The figure illustrates the PC series resulting from the pruning process. The top panel shows the MBH98 PC1 using 16 bristlecone sites. The second panel shows the PC1 with only three bristlecone sites and then with only 2 bristlecone sites. These are relatively unchanged. The fourth panel shows the PC2 with only 1 bristlecone site; the final two panels shows what happens with no bristlecones: no hockey stick.
Preisendorfer’s Rule N Again
We’ve seen commentary from realclimate about principal components – the notorious Dumb and Dumber Guide to the latest Hockey Stick Controversy , where Ammann and Schmidt try to show that Preisendorfer’s Rule N requires the use of 4 PCs. I’ve discussed Preisendorfer’s Rule N and the Dumb and Dumber Guide elsewhere. My main issue with Preisendorfer’s Rule N is that, at most, it is a necessary condition for significance; it is not a sufficient condition for significance. This can be seen by merely reading the original texts (or by considering the Tech Stock PC for an amusing counter-example.) Here’s one more demonstration: the PC1 with 2 bristlecone sites and the PC2 with one bristlecone site are "significant" under Preisendofer’s Rule N. The MBH98 data mining method can literally find one series out of over 50 and endow it with weights to ensure that it gets into the regression roster. Once it’s in the regression roster, it can imprint the final NH temperature reconstruction, which is why Mann and Ammann are so dead set on getting the bristlecones in the regression in the module.
Verification Statistics
Von Storch and Zorita do not report any verification statistics. We point out that (1) an accurate simulation of MBH98 using pseudoproxies needs to produce a high RE statistic and negligible R2 statistic (2) that we’ve accomplished this by network simulations using one simulated PC1 from MM05 (GRL) together with white noise series (as described in our Reply to Huybers, posted up a few weeks ago.)
Anyway, it was pleasant doing this Reply. I thought that James Saiers edited the article expertly. We took umbrage at a few passing comments; Saiers ably kept the hockey match under control and we ended up with a better reply than we started with. (This was a constructive process and I regret the current spat with Famiglietti.) There was never anything presented by von Storch and Zorita that worried me for a second as to the substance of our claims. I think that we answered their comment thoroughly and, in addition, used the occasion to amplify our earlier article.
Climate Audit 
119 Comments
1. I only see the von storch paper, not yours.
2. Your graphic would be clearer (in implication) if you put some lines where Von storch had their signinficance (.3-.7). You are forcing the reader to go back and forth vice showing the meaning clearly within a grapihc.
3. As usual, very dense post. Will need to wade through it.
Steve: Re #1. Fixed, I meant to post our Reply. Re #2. A red line has been added to the graphic – where else could you get such service?
Steve – You have shown how bristlecone tree rings tell us almost nothing about temperature. Is there some way to extract a CO2 chronology out of them. We know the CO2 levels for the past 250 or more years. Do you think some statistically valid results could be obtained that would allow us to extend this backward a couple hundred more years? It would be fun to show it wasn’t a flat line up until the 1850s with tree rings. But if it was, so be it.
Nice work, Steve.
I have seen you state before that the MBH98 methodology extracts a hockey-stick even with random noise data. Therefore, why, without the Bristlecone pines, is there no indication of a hockey-stick?
As usual an interesting and well argued piece. Surprised you used the metaphor of a few “bad apples” though; surely this demonstrates the value of a “few good men”.
SC
Re #4
I don’t think there’s anything remotely hockey stick shaped in the remaining data series for the algorithm to find. It’s done its job and found the hockey-stick series, and then Steve has discarded them. So there’s nothing left.
Re #4: we’ve said that the MBH98 method will obtain hockey stick shaped series from networks of persistent red noise networks nearly all the time. This has been demonstrated and confirmed. The bend depends on the persistence. The bend with white noise networks is small.
We’ve said that the method is severely biased, not that it will always produce hockey sticks. For example, it doesn’t produce hockey sticks from the Stahle/SWM network in MBH98 due to the nature of the series in the network. Surely the power of the data mining method is demonstrated in the examples here. In this example, we’re not dealing with random persistent data, but the actual tree ring network. Without the bristlecones, you’re getting some kind of consistent growth "signal" which does not have a strong hockey stick shape.
Steve
One way you might try illustrating the data mining “power” of the MBH98 procedure would be to train against say the decline from the Medevial Optimum to the Little Ice Age. Would certain proxies then become the canaries showing colder conditions?
Actually this seems familiar. Did you already do something of this sort? The disadvantage of such a data-rich site as this is it’s hard to remember everything which is here.
Re #2: there have been studies to deduce CO2 levels from looking at stomata in trees. I don’t know whether it works or not.
I’m also not necessarily arguing that CO2 fertilization is the only factor at work in 20th century growth of these trees. They grow in very nutrient-deficient soils. Perhaps it’s an effect of increased phosphorus or increased nitrate fertilization at high altitudes (which could have more impact on the more nutrient deficient soils of the bristlecones than elsewhere.) The non-robustness of MBH results to bristlecones should be setting alarm bells off, especially when these series turn up in Crowley and Lowery [2000], Mann and Jones [2003] and Jones and Mann [2004] – supposedly “independent” studies.
Re #8. I did this analysis about 5 months ago, so the point could have been mentioned in passing before. But I’ve never previously shown these illustrations and the commentary is new.
John
Re: 2
I would agree that we have data for CO2 from ice cores and tree stomata. As with other proxy derived data, there are questions as to the correlation of these with atmospheric CO2 levels.
Superficially, it would seem that ice cores would simply store CO2 and other gases. Thus on the superficial level, measurement of ice core gases should provide a close approximation of the atmospheric concentration of these gase at the time the water froze. Transporting any sample to a laboratory for analysis may effect the result. This is most likely the case with deep ice cores.
My area of expertise is trace gas analysis. I know of numerous examples where the sample changed en route to the instrument. The resulting data was erroneous, but was nevertheless believed by the analyst until someone pointed out the sample handling problems. Zbiegniew Jawarowski has pointed out problems of ice core sample handling in several papers as well as a summary written for the US Senate Committee on Commerce, Science, and Transportation, March 2004. This summary is the only elctronically retrievable document at the present time. Prof. Jawarowski proposes that decompression of the deep ice core samples allows CO2 and other gases to escape from the ice prior to analysis. This results in a lower CO2 measurement than was originally in the deep ice.
Jawarowski has received the same reception from the hockey stick team as have Steve and Ross. There are two assumptions underpinning belief in AGW. One is the rapid increase and historic high temperature that Mann et al show in their papers. The other is that atmospheric CO2 levels were significantly lower in pre-industrial times than they are today. M, M, and J are questioning the methods which lead to these basic AGW assumptions. If M, M, and J are correct, it does not follow that AGW does not exist. These questions do cast doubt on the efficacy of climate change mediation measures proposed by the IPCC and enshrined in the Kyoto Treaty. A logical response to these questions from the climate science community would be multi-disciplinary re-evlauation of their basic asusmption and methods. Sadly, this is not their response.
Re #11
Given that the present concensus is that Kyoto is essentially dead and that the best tack now is to do more research and to plan for coping with whatever global warming does occur, shouldn’t such a “back to the basics” plan be one of the most important things to do?
Trying to keep a lid on pent-up skeptical responses to the ‘concensus’ is only going to make the ultimate result worse from the POV of AGW alarmists. While there was a realistic chance that an advanced Kyotoist global regime could be established and basically take control of whatever it wished, it made some tactical sense to supress dissident opinion, but now the cat’s out of the bag and trying ever more draconean measures to assure conformity is only going to make the final outcome worse. The ragged retreat of the Mannean contingent will turn into a rout, and there’s a risk that whatever truth there is in their position will be dismissed as part of the earlier cover-up.
Re 11. Jaworowski’s piece on unreliable ice core measurements is here . Jaworowski has indeed received a similar treatment as M&M. It is really astonishing for me as a journalist to see scientists blabbering about peer review and checking facts and accusing Jaworowski of only publishing garbage without even bothering to ask him for his publication list.
Re #13, As a long-time reader and almost non-poster, it seems to me that those who are ‘blabbering about peer-review, checking facts and making accusations’ are revealed by their methods to be pseudo-scientists, rather than real ones. They are of a particularly expert brand, but the resort to obscurantism and dismissive insult is classic; indicating a highly skilled charlatanry. At the risk of an off-topic ejection, I’d observe that the current political brouhaha over “Intelligent Design” is not the only religious abuse of science going on. The other involves the secular religion of expiational environmentalism. The latter may be even more corrosive.
re: #11
Jaworowski relies on the outliers in the nineteenth century when analysis was crude and not corrected for the urban CO2 dome.
Below is some more background on the stomata debate.
http://www.sciencemag.org/cgi/content/full/286/5446/1815a
http://www.pnas.org/cgi/content/full/99/19/12011
Ok, I read the paper. Very nice pub, Steve. I think you win the match for content and for clarity (!!) of argument. Keep it up.
Is this the final approved draft? A preprint? How did it go in the review stages (did paper get rewritten or criticized a lot or are you learning to slam these things out at high level first time round)?
General question: How do reviewers review and then accept two papers which differ so strongly? If you are right and VZ “assumed the canopener” why publish at all?
and here is a complete dissertation on the topic!
Coupling between atmospheric CO2 and temperature during the onset of the Little Ice Age , Thomas Bastiaan van Hoof, 2004, Proefschrift Universiteit Utrecht
http://igitur-archive.library.uu.nl/dissertations/2004-1214-121238/
Re #16: hopefully, they simply want to let the reader decide. Peer reviewers aren’t in a position to tell who’s right or wrong in these things, but merely whether it’s worth publishing.
Ok. Had a more general question. If I take a series with hockey stick index of 10 and average it with 9 flat lines (0 index), I would get a resultant simple average series with stickishness of 1. If I do the same thing, using Mann’s methods, what stickishness results? (This may be easier to understand than the red noise…may also not be as apt as using red noise, but give me the answer first, then the caveat.)
Pretty please.
Steve, it looks like you added the red line I wanted, but it would be much more dramatic to show the range .3-.7 and have it hashed in or whatever. You need to communicate (in the chart) that the entire damn historgram is outside their range. You should also annotate the figure text to show what the hell is going on…
😉
RE #19: but afterwards, they re-scale the “average” (it’s not exactly an average) so that its variance in the calibration period matches the variance of the temperature index. So the fitted HSI looks just the same regardless of the other series. Look at Reply to Huybers #2
I don’t understand. Did you answer my question? If we do the experiment that I discussed, with Mannish methods, is the answer 1.0 or different? I don’t care about the intermediate steps. I want to know about the final figure. what it’s number will be.
It will be the same with or without the averaging of white noise.
Re #20 – not a bad idea, I’ll have to figure out how to hash the figure. It wasn’t an option at GRL.
You could have done it just fine in GRL. Just do it with gray-scale. It’s not the redness, it’s the showing the whole story on one chart. I thought you were a bizness weenie?
So if I (using Mannian methods) mix 9 straight lines and a stick of size 10, I get a resultant stick of 1.0 or of 10? Just want to be clear.
10 maybe 9.8
Give people the CEO charts. The killer analysis.
Re: 11
Escape of gases from drill cores is not the only problem Jaworowski has identified.
There is a lot of mixing of annual signals because of:
1) Surface melting in summer mixing annual layers;
2) Severe and very deep cracking of the ice from the surface down so that water from newer layers mixes with much older layers;
3) At a certain depth, the pressure of ice creates enough heat to cause melting and mixing over huge areas and of considerable depth. Antarctica is underlain in places by an enormous under-ice lake which is one of the largest freshwater lakes in the world.
Any signal can therefore be expected to be very coarse in resolution.
How far site selection can control for these problems is unclear. Ice caps are moving, dynamic masses which may fracture and heal giving the misleading impression that they have always been intact (unfissured or unmelted), for example.
There is really no way of knowing the history of a particular vertical sample of ice, traversing millennia of time, so far as melting and fissuring is concerned.
In addition, the hydrocarbons used to lubricate the drill and drill string can contaminate the sample.
There is another problem which Jaworowski does not mention, as I recall, which is that the mathematical equation which is applied to the core data to compensate for the ice flowing faster as the depth of the ice increases is purely an assumption. It has not been validated. Indeed, it is hard to see how it could be. This severely distorts the dating of deeper layers by an unknown amount.
The problem of outgassing from the sample is similar to the problem of outgassing experienced in the early sampling of methane hydrates. Early cores brought to the surface from the ocean bottom contained very little hydrate leading to the conclusion the resource was too diffuse to be economic. Coring with a drill capable of maintaining seabed pressure all the way to the surface and the use of a pressurized examination chamber led to massive hydrate cores being recovered, indicating an economic resource. Clearly, the methane was outgassing from the early cores as they were drawn up through the water column and the pressure decreased. Same problem with ice cores, except that the water is frozen.
re #28: in a business situation, I wasn’t drawing graphs myself. You had artists doing it. However, I did spend a lot of attention on presentation graphics. Story-lines in the graphics have to be thought out carefully. A lot of sales in exploration finance are based on commentaries on a good graphic so you have to work hard on them and the conceptual design is hard to delegate in a micro-business.
That’s what intrigued me about the IPCC hockey stick: I knew what was involved in the selection of the presentation graphic and the importance of a good sales graphic. So it’s a little ironic that I’m not paying enough attention right now to these issues.
Steve,
I’m sure you’re not going to do this, but the Hockey Stick Team hung their hats on a good graphic, that, for all intents and purposes, was just a pretty picture…as fictional as any of Crichton’s stories. But a graphic, based on fact, can be very very compelling (http://www.edwardtufte.com/tufte/posters – but I’m sure you’ve seen this before, too).
(PS – I’m a architect/designer. I’d be happy to help with the pretty pictures)
Jaworowski et al’s criticisims of the CO2 “consensus” seem all-too familiar. Maybe there’s a CO2 hockey-stick team too, hiding inside ivory fortresses?
Re #19,26 &ff. TCO you asked a different question than Steve answered. If you take a series with a hockey stick index of 10 and do a simple average with 9 flat lines the result will have a small hockey stick index, roughly 1 (where 0 is flat). I think what Steve had in mind was, if you take the same combination and put it into the Mannomatic (i.e. the MBH98 PC algorithm), what will come out? The answer is, likely, a first principal component with a hockey stick index of 9.8, and (rough guess) the package will say it accounts for 90% of the explained variance. The other PCs will be flat lines with tiny assigned explained variances. But, of course, this will be a misrepresentation of the data you plugged in.
The trouble with your example is that MBH methodology doesn’t involve taking a simple average across proxy indicators. If they’d just used a simple mean they wouldn’t have run into the biasing error (but then neither would they have had a hockey stick).
Ross. I wanted to know what the Mannomatic would give me. I understand what a simple average would give me. I want to contrast the two. I suggest that the thought experiment (could be a real, numerical, one) is more illustrative of the biasing in the Mannomatic than use of “red noise” since autocorrelation is a seperate issue from biased selection of the Mannomatic. It seems like a good idea to isolate and define criticisms without drawing seperate contentious issues in.
One thing that has occurred to me since (overnight) is the possible argument by Mann, that the method should preferentially weight for the one hockey stick (in thought experiment given) since what he is doing is an inherent calibration of proxies with the experimental data (i.e a totally flat line is a lousy proxy since we know from instruments that temps went up this century.) Do you think he would make that argument and how would you respond? (I imagine with comments about the global field versus the local? anything else?)
Re 29:
In addition to all the problems mentioned in 29 in regard to ice cores, the chemistry in ice can be quite active for a number of years so that the chemical concentrations captured in ice differ from that of the atmosphere above. This is a particular problem for unstable compounds such as methane and their concentration is probably underestimated because it is partly destroyed. It means that the difference between modern concentrations and past concentrations are probably not as great as commonly assumed. Here is a quote from a press release on the subject.
“What we now know is that the near-surface atmospheric zone called the mixed layer (from the surface upward to between 20 to 200 meters) is a highly oxidizing environment at the South Pole,” says Doug Davis, one of the lead researchers and a professor in the Georgia Tech School of Earth and Atmospheric Sciences. “Equally exciting, we are beginning to see evidence that a lot of this oxidizing chemistry is also occurring down in the snowpack. Thus, once things get buried in the snow, there continues to be active chemistry — including oxidation — that could further modify chemical species before they are trapped in the ice in their final chemical forms.”
This finding suggests that glacio-chemists — who study climate change based on an analysis of trace chemicals trapped in polar ice — have to be far more careful in their interpretation of Antarctic ice cores, says Davis, whose research team is funded by the National Science Foundation. Changes in some chemical species buried may continue for another five to 10 years after they are trapped in the snowpack. Davis expects that scientists will soon focus more attention on this topic.”
From http://gtresearchnews.gatech.edu/newsrelease/SPOLE.html
Hans, Re: 17 & 15
I perused the van Hoof thesis which you linked. I could find only the following as a reference to CO2 analysis of ice cores:
T. B. van Hoof does not address the points made by Jawarowski that there are problems with ice core sample retrival, contamination, storage before analysis, and the analyses.
During the past 25 years, I have dealt with the problems of sampling and analyzing gases in the semiconductor industry. The primary trace gases for which we analyse semiconductor gases are atmospheric components. Getting a sample to the instrument without contamination is a major problem.
The basic points that Jawarowsky brings up in his articles are not unique to ice core sampling. The specific details of sample contamination mechanisms, loss of sample gas, and chemical reactions of sample gas are characteristic of ice core gas analysis, however these are analogous to sampling problems faced by anyone performing trace gas analysis. These are cross disciplinary problems faced by anyone doing trace gas analyses. The basic question underlying any trace gas analysis is: Did the sampling method effect the sample?
T. B. van Hoof has not answered this question regarding ice core gas analyses.
TCO: your example of one hockey stick tree series out of ten brings up another problem with PC analysis. As Ross said, you will get a PC1 of 9.8 or so and this PC1 will “explain” a large fraction of the variance. This is fraudulent not only because one series is doing all the explaining, but also because it makes no sense at a basic level. PC is supposed to help you find one variable from among a group of DIFFERENT variables that explains the variance of the dependent variable. So if you were just comparing the measures from ice cores versus tree rings to explain temperature, then PC might find that tree ring data was explaining more of the variation of temperature than ice cores. But Mann et al. are using different tree ring series as if they are all separate variables. Overall, tree ring data do not correlate with temp. Steve and others have shown that. But if Mann can get just one or a few tree ring series to correlate with temp, then his version of PC claims success. This crazy way of using the analysis will always produce the desired result if you have, say, more than twenty series to choose from.
Re #34. TCO, in our E&E05 paper, pp 79-80 there’s a discussion of a 2-panel diagram (Fig 3) that is much like our von Storch reply, and illustrates your point. I thought it was one of the highlights of that paper, but it takes some effort to understand and I think rather few people have actually read E&E05 (judging by, for instance, the GRL comments and referee reports arising). You know you’ve understood the graph when you snort your coffee all over it in surprise.
It’s hard to imagine that Mann would attempt such a crazy counter-argument that preferential hockey stick weighting is actually a good thing since, even though there’s no local temperature signal it boosts the fit to his temperature PCs, ergo it’s extracting a “global” signal from the data. But he did argue this, and yes, we responded by making fun of the global/local handwaving. See E&E05, Sct. 5.3, pp 90-91. The wording there was edited down from more sarcastic earlier drafts, but I think still conveys the fun we were having at that point.
More substantively, the other response is this. In any regression model you can arbitrarily improve the fit by lifting the hood and making a mechanical adjustment to your data, with a view to bending it into a pattern also found in the dependent series. But if you do so, you should say so, and also you have to revise your significance criterion. Since random numbers would also give a good “fit”, once subject to the same mechanical adjustment, the benchmark for your data has to be revised. Data has to do better than random numbers, to be judged significant. The test is always whether data+model outperforms noise+model. Our GRL paper showed that MBH98 tested data+model(1) against noise+model(2), so it wasn’t a proper test. When we used noise+model(1) the RE benchmark shifted and no longer yielded a conclusion of significance. So even if we allow Mann to insert the bristlecones and preferentially weight them, the results are insignificant. And, of course, there’s the insurmountable problems of using data known on other grounds to be lousy temperature proxies; the Mannomatic PCs are not actually PCs according to the algebraic definition so no one really knows what his “network series” actually are, etc.
Ross,
-you shouldn’t have published in EE (since as you say, few read that journal).
-one should not have to dig so hard to understand such a simple point. That it is the scientific literature is irrelevant.
My criticism of the communication stands.
Well TCO, Should Mendel have not published in some obscure Journal and settled for just talking his gardening up in the Abby lunchroom? Sometimes you just have to publish where they’ll take you.
I have not heard that yet to be the case. If so, yes, a small, controversial, journal is better than none. Doesn’t make sense to me, given the sucsess at GRL and the number of other options. I warrant that clearer communication would make the papers more easily accepted, even at journals where there is some degree of political headwind. My point stands.
Earth to TCO: MM03 in E&E got the ball rolling. MM05 in E&E provided additional information for those interested (and who need not be E&E subscribers). If people choose not to learn from it, that is their loss, not MM’s nor E&E’s.
None of that changes my point that MM should try to publish in better journals. A well-written paper (like that last comment reply) helps. Also, doing some looking at the rest of the options. And I still haven’t heard MM say that they did a good effort to publish in the regular literature before picking the skeptics (barely ever abstracted or held by libraries) journal.
I trust that you are suitable amazed that they managed to consider publishing in GRL without having had in advance the benefit of your expertise on where to publish.
No.
TCO: well we tried Nature… 8 months of chasing moving goalposts before they snatched the ball away and said they weren’t playing anymore. When we started out, and even up to early this year, if we had tried the famous climate journals we’d be 4-6 months at each before getting predictable condescending referee reports and dreadfully polite rejection letters, and nothing would be in print yet. So E&E is half a loaf for you and quit complaining about it. Now, after all that’s happened, I think we’d get more of a hearing at well-known journals. But it’s very time consuming (and aggravating) dealing with journals that you know likely won’t publish your paper anyway, and since climate publications count for nearly zero on my cv it’s hard for me to justify long shots at famous outlets. I’m not saying it wouldn’t happen if the occasion arose, but I have no regrets about our strategy up to now.
re # 36
Go for it Brooks, dig the source data out!
As for CO2, I am pretty convinced that current levels are unprecedented for the last several million years.
In the Eocene, Cretaceous etc., there seem to have been high CO2 levels. Crowley and North’s Paleoclimatology text attributes the warmth of the Eocene and periods like that to high CO2 levels. They seem to attribute the long-term (over millions of years) decline in temperature into the Pleistocene to lowered CO2 levels; I don’t recall seeing an explanation of why CO2 levels are supposed to have declined. I haven’t researched the topic as to why CO2 levels are supposed to have declined from the Eocene to the Pleistocene.
How they fiddled the CO2 figures
From Prof. Zbigniew Jaworowski
“An ad hoc assumption, not supported by any factual evidence[3, 9], solved the problem: the average age of air was arbitrary decreed to be exactly 83 years younger than the ice in which it was trapped. The “corrected” ice data were then smoothly aligned with the Mauna Loa record (Figure 1 B), and reproduced in countless publications as a famous “Siple curve”. Only thirteen years later, in 1993, glaciologists attempted to prove experimentally the “age assumption”[10], but they failed[9].
….The basis of most of the IPCC conclusions on anthropogenic causes and on projections of climatic change is the assumption of low level of CO2 in the pre-industrial atmosphere. This assumption, based on glaciological studies, is false.”
http://www.warwickhughes.com/icecore/
Re #47. Crikey, even you Hans 😉
Re Prof Z.J., I’m amazed anyone still quotes his study – it’s been thoroughly trashed! For starters, can anyone show me who, if anyone, from the Senate actually saw, or has seen, his ‘presentation’? Or indeed, if it ever made it anywhere near the Senate? Once someone has done that to general satisfaction we can go through the rest of it…
Peter… what a great comment! “This guy’s study has been trashed…” and “nobody saw his presentation.” What, exactly, does this mean? Nobody seeing a presentation doesn’t invalidate the presentation or the contents thereof. Exactly what in Prof. Z.J. study is wrong?
There have been several posts in this thread regarding the ice core sampling. Your comment does nothing to advance our understanding of the issues involved. Brookes in #36 made some very valid and interesting points. Do you dispute them? How would you address his posts? It appears that most tree ring data, as presently collected and analyzed, are not good temperature proxies. Why not make sure that the other data being used to support AGW is also valid and correctly collected an analyzed?
It appears that even questioning data an analysis is taken as a personal ad hominem attack. People don’t want to defend their science so they attack the skeptics.
Paul was his ‘presentation’ presented? No, afaic, it was not.
Don’t you find it strange that a presentation wasn’t presented? Funny isn’t is how impresive ‘Statement written for the Hearing before the US Senate Committee on Commerce, Science, and Transportation March 19, 2004’ sounds. A lot better than ‘Statement written for the Hearing before the US Senate Committee on Commerce, Science, and Transportation March 19, 2004, err, cough, but probably not actually seen by anyone there and it probably got nowhere near the Senate to be honest’ don’t you think? Perhaps you don’t think that?
That is the first of a series of errors and misleading statements in his ‘presentation to the Senate’ – do the google to find out more, clue: boomjums. Frankly, defending the indefensible is pretty darn difficult. Go on, have a go 🙂
I don’t actually see the word “presentation” or “presentation to the Senate” in the title. Maybe my eyes fail me but it appears to say “Statement written for the Hearing…”
Statements are often submitted to Committee Hearings as part of the documentary record. It does not imply that the paper was “presented”.
Whatever are you talking about?
Now let’s see if you can produce any of the “series of errors and misleading statements” or whether this, like all your previous attempts at fact-checking, is a work of your imagination as well.
I read the paper. What’s wrong with it, Peter?
‘John’ do the search. Go on, you’re clearly smart enough to be able to follow my clue, surely?
Still, even you have come close to acknowledging that the Senate committe referred to never saw his paper. Indeed, I doubt very much it’s part of the documentary record – but, go on again , prove me wrong…
Peter… YOU MISS THE POINT! (again)
It really doesn’t matter if the Senate has seen, read, understood (not likely), eaten the paper. The Senate is a group of 50 useless people. I really don’t care what the Senate does with it.
What is WRONG in the paper? How is Prof. Z.J. not right in his paper? All of this handwaving about the Senate not having it is mis-direction and an attempt to not answer the question.
So…are you going to tell us how Prof. Z.J.’s research has been trashed, or do we just get to go through the weekend in an extreme state of suspense?
RE #48
Steve I seem to recall, though its a long time ago since I read it. That it was due to the rise of the Himalayas, lots of limestone to dissolve, which locked the CO2 into ocean sediments. Though I guess its only a theory.
I read the paper. Would be interested in a thoughtful criticism.
Re #50 etc. on Jarowoski,
I think Peter is hinting at this: http://www.someareboojums.org/blog/?p=7
Which, depending on your POV, is either a comprehensive debunking or a vicious attack similar to those on M&M.
Even if the report for the Senate by Jaworowski were correct, it is based largely on an earlier paper (1991) if I recall, which criticises work which went on before that. Are the methods that Jaworowski criticises still used or has the science moved on? Most scientific fields, especially young ones like this do not stand still and I would be surprised if the same methodology is still being used as was used 20 years ago. Even if his criticisms stood then it does not therefore mean they apply to more recent/current work.
Good point, Paul. I’m not sure that the GWers have conceded the point even on the earlier criticism. Or have they? Do you? What did you think of the paper? Total trash? Or possibly on target?
Re #59, John Davis
I did like the guy’s second argument :
If only the hockey team would take the same view of M&M …
Re #59, the former :). I mentioned point 1. Why? Becuase it’s the first point…
And you’re right to use the word debunked, better than my trashed. As to a vicious attack, I’d say it’s pretty tame compared to here…
I still haven’t heard any discussion of the article. Pete, did you read it? Why do you think it poor work? I have a Ph.D. in the hard sciences and have read a lot of articles in many fields. Seemed like decent work to me. I could be missing something…so please elucidate.
Re #64. Why is it poor work? Because it seems to me the debunking is good work :).
TCO, do you think the ZJ paper good work? If so, why? Following from that what’s wrong, and I am open to this, with the debunking?
I didn’t see any debunking of the actual main points of the article (let me look at the debunking again, though).
Pete, I’m open to learning more…but I did read (in toto) the actual Z article reffed in this post and I didn’t see the flaws that I typically look for in hacks.
I just looked at the Boojum critique and it reminded me mostly of the supposed debunkings of Bjorn Lomborg which occured after his book came out. I read those critiques, Bjorn’s rebuttals and did various examinations of whether or not the critiques were warrented and decided that they weren’t, except for the items which Bjorn himself admitted were mistakes and which are still up on his website. I suspect the situation is similar here. Some of the criticism is correct, a lot is just plain ad hom and some, perhaps a lot, of the criticism is just wrong, though which is which can’t be determined just by reading ZJ, Boojum or both. Perhaps if Boojum were to do as thorough a job as Steve does here of explaining his his complaints it would be possible to decide. For now I’m more inclined to the Hack Job position.
reading “debunking” more thoroughly now. Have to wade through some extraneuous stuff (Dashiel Hammet!) to get to main arguments. Also, hard since he doesn’t even give a link for the actual testimony that he is criticising!
1-senate thing is irrelevant. next.
-2expert status: interesting I guess. would appreciate a fuller list of ze pubs and the ranking of journals and such. But fine.
3-liquid water claim: now we get a science/logic based criticism. Great. Would want see response and all (and if I had time go look at that pub and see if there were any other cited by Z.)
4. He notes that the 1992 paper (which we looked at vice the testimony) is pretty decent in objective. Says Z. doesn’t have fixes: so what!? Z does not need to know how to correct for an error to point it out. wrt 1994 article: only says that someone else “ripped it to shreds”. That’s pretty puffy speaking behavior. And he doesn’t cite the actual criticism. reading the O. criticism now…
–
Peter refuses to answer the question. Why?
O criticism:
a. Some extraneous stuff about how important CO2 is and all that, but nicely written (methods are either accurate or not…the importance is irrelevant.)
b. some horn-tooting when what really matters is proof that the methods are accurate (not assertion). For instance: “the project was conducted, and is succssessful” (imperfect qoute, control c not working).
c. no citations in “shredding” and no detailed factual arguments.
d. Basically, O is telling us that he is an expert and disagrees with the paper. But that’s not sufficient. He has to engage and actually get into the substance. Or else, it is just a “trust me” stance. That’s inadequate. Dick Feynman would frown.
e. good point about the gravitational enrichment made (but again with no reference).
f. paper closes with more fluttering and “trust me, I’m the expert”. He may very well be. J. may be a kook. But O. has not shown that. boo-hiss.
Another point on O. He makes the case that J. assumes that problems with data collection in the 60s still exist in the present (and also says that they don’t–field has improved.) But J. actually cites specific cores and studies from 80s that show problems in data collection. If O wants to debunk J, he needs to engage on those points…show that those studies are not cited much…or show that the cores were well handled (even though some melting occurred!). And if they were mishandled (but are nonrepresentative), O should CONCEDE that those cores were mishandled. To not do so is tendentious. The mark of an arguer, not a scientist. Dick Feynman would frown.
We are referred to the Indermuhle NATURE article as proof of how firn/air close-off occurs. But the article is a general reconstruction of the CO2 in the HOLOCENE (from 1999!). It referrs to refs 24, 25 for the method of firn/air closeoff dating. boo hiss!
that was 5 btw
6. So contamination with the drill fluid has no impact? There is no CO2 in solution? Want more facts here. Also, blithe statement that sample handling is “well worked out” is insufficient. That is the point in debate. Need to cite facts, studies, etc. not just assert victory.
7. This might be a good point. (aside from the ad hom sniping). Not clear why several day expansion would not be applicable to cores that are transferred from the field though.
8. Not clear to me how the shellfish records can bolster the ice core methods. Also, the citation is to a press release which does not support the point made.
Re #11 Brooks, do you know what is the accuracy of the instruments used to measure C02 levels, specificly the levels from the ice cores. The literature doesn’t mention this important detail.
9. graph in actually seems to support J! Especially after Gavin put in the correction. some bouncing around but keeps going down for 1000s of years. would expect instead to go down for a hundred years or so and then do some dithering or be flat. instead that thing drops!!
10. No real comment. Irrelevant to the issue of methods being accurate or not. Hard to read even as itself without seeing the original text, also. (Ok…found the link finally…kinda poor style not to make it more prominent and at beginning of article but ok).
11. See 10. Except that if the assumption was not ad hoc, then it should be backed up by a REFERENCE to a CALIBRATION or such thing.
12. “No they didn’t” is not much of an argument.
This is really more relevant to our disscussions here, since it is the article in question.
Re #69. Hey, I gotta eat and work sometime you know :), besides, didn’t I? TCO, re you recent posts. OK, lets say neither a ringing endorsment of ZJ or a crushing indictment? I can live with that given where you and I stand on these kinds of things 🙂
Dammit, Peter, in post #50 you described this as trashing Jaworowski. As soon as someone demonstrates he has read 60% of it, you roll over.
Have some self-respect, man. At least read the bloody thing yourself.
Chapter six from the van Hoof Dissertation:
Click to access c6.pdf
And he STILL didn’t answer the question…
#79, hey, fF, look, I have some respect for TCO – OK? ‘I can live with that’ – know what that means? I’m happy not to disagree with him, doesn’t mean I agree with him! Jese, slightest bloody chance and you’re in with the frapping hob nails…
#81 what question? Post #51? Look, I think the people who ‘do’ ice cores know what they’re doing. You’re not an ice core expert (?) neither am I, but one of us thinks they know better than the ice core experts…? despite lacking the training they have…?
Doh, ‘training he has…?’
Phil, Re: 75
I am not certain what instrument(s) were used for all the ice core CO2 analysis. There are many different techniques that may be used. FTIR, GC (several different detectors), MS, Laser CRDS, and others will quantify CO2. The CO2 lower detection levels (LDL) for these intruments range from fractional percent to below 100 ppt.
In reading Jawarowski’s paper (not the short summary), he mentions methane interference with a flame ionization detector. This would indicate that at least one team used a gas chromatograph with an FID and a methanizer. This apparatus, properly set up, can resolve CO2 down to low ppb levels. In lieu of a GC-FID, they may also have used an FID detector with a methanizer. These are readily available and can be used in the field, but their accuracy is at best 0.1 ppm.
A key aspect of Jawarowski’s paper which his detractors seem to ignore is the sample handling on route to the analytical instrument. As I pointed out above, this is critical to obtaining a good result. Analtical science has advanced since Jawarowski published his papers in the early 1990s. However, there were instruments available at that time which had sub 100 ppt accuracy.
Paul, (Re: 60) the issue is not analytical instrument development, but rather sampling technique. Anyone can use a highly sophisticated (and very expensive) instrument and obtain garbage data by using poor sampling technique. The issues pointed out by Jawarowski’s detractors concern various aspects of his paper, but do not get to the heart of the matter, which is the ice core retrieval, transport and storage prior to analysis. It should also be noted that the ice core measruements are quantitative, therefore, any change in the quantity of gas in an ice core will impact the results.
Peter, I read through v. Hoof’s thesis and I can find no mention of either the analytical instruments nor sampling methodology used by any of the people who have written on ice core CO2 analyses. The blog trash (perhaps I have the words reversed) discussed many issues, but ignored the issue of sample handling and anlaytical technique. The issue is trace gas analysis, not which Senate committee accepted a document into its records.
The bloggers seemed to not understand why you might lose CO2 (and other gases) when you retrieve a deep ice core and bring it to the surface. Here is a very brief explanation. Deep ice is most likely saturated with dissolved gas at the temperature and pressure of the location. When you bring such a core to the surface, the core is now in a supersaturated condition. Supersaturation is highly unstable, and the gases quickly come out of solution and in forming bubbles, fractures the ice. When the fractures occur, your primary gas diffusion mechanism transitions from Knudsen to Fickian. Knudsen diffusion is driven by the concentration gradient, whereas Fickian diffusion is driven by pressure difference. Once there are cracks in the ice, Fickian diffusion will cause some of the gas in the core to be lost to the ambient air. At the same time, some ambient air will diffuse into these same cracks. Therefore, the quality of any subsequent gas analysis may be very poor.
I would like to read the comments of someone with a decade or more experience in field gas analysis who can find problems with Jawarowski’s papers. What ZJ’s detractors have written indicates to me that they do not have this experience.
Brooks,
Van Hoof may not have elaborated on the core measurement technique. However what he does find is that the independent method of Stomatal Index agrees with the core measurements. IMHO Which reduces Jaworowki’s critique to marginal proportions.
Peter (Re: 83),
The people who “do” ice cores have many different reasons for studying the cores. These people may be world reknowned experts in their fields, but this does not make them into experts in trace gas sampling and analysis. I am not attempting to trash anyone by saying this. The problems are solvable.
re 87 IMHO the problems are exaggerated, like UHI in Europe. (sneakily trying to change topic)
Exaggerated in what way, Hans? In there not being much UHI or the ability to correct for it being easier than some say?
Gotcha! The latter, as is shown by Douglass et al, the disparity between sattelite and surface record is over the ocean, not over land.
Also my own investigation in long rural records in Europe show the same.
All red herrings to me.
Brooks (Re:85)
If the preanalysis retrieval, transport and storage of the ice cores isn’t well described in print you might look at the video record. Several TV science programs had footage of the cores being extracted, their storage, and the methods and equipment used to analyze them. I remember thinking “You can get away with that?” when I was watching one of them, Scientific American Frontiers?. That is, it seemed odd to me that the samples were exposed to the atmosphere and they wanted to analyze gass content.
I didn’t get the impression that gass exchange between the core and the atmosphere was thought to be a problem. I expected they’d have to extract their sample from the interior of the core and they had reason to think it would remain pristine under their methods of extraction and storage.
John (Re: 91),
If the the cores were exposed to air, then contamination might at least be a possibility. Most people do not consider atmospheric gas contamination to be a problem until have the problem is explained to them.
Most environmental samples do not suffer form this problem because the trace compounds like pesticides are not normal atmospheric constituents. These samples are routinely transported in plastic bags. Such bags can not be used if you are analyzing for atmospheric components because the bages particularly permeable to relatively small molecules like O2, N2, and CO2.
Hans,
“IMHO the problems are exaggerated”
Not if you are analyzing for atmospheric components like O2, N2, CO2, H2O, or Ar. Anytime you are analyzing a gas for the presence (qualitative) of an atmospheric components, contamination is a problem. When you perform quantative analyses for atmospheric components, even a tiny amount of atmospheric contamination will give you bad results.
Hans (Re: 86),
You are much more familiar with the stomata CO2 proxies than I am.
Can you explain how stomata are calibrated? What is the standard?
Re #85 Thanks Brooks, the accuracy is better than I expected.
Has the Siple data been moved 83 years?Figures 1A and 1B
The
data from shallow ice cores, such as those from Siple,
Antarctica[5, 6], are widely used as a proof of man-made increase
of CO2 content in the global atmosphere, notably by IPCC[7]. These
data show a clear inverse correlation between the decreasing CO2
concentrations, and the load-pressure increasing with depth
(Figure 1 A). The problem with Siple data (and with other
shallow cores) is that the CO2 concentration found in
pre-industrial ice from a depth of 68 meters (i.e. above the depth
of clathrate formation) was “too high”. This ice was
deposited in 1890 AD, and the CO2 concentration was 328 ppmv, not
about 290 ppmv, as needed by man-made warming hypothesis. The CO2
atmospheric concentration of about 328 ppmv was measured at Mauna
Loa, Hawaii as later as in 1973[8], i.e. 83 years after the
ice was deposited at Siple.
An
ad hoc assumption, not supported by any factual evidence[3, 9],
solved the problem: the average age of air was arbitrary decreed
to be exactly 83 years younger than the ice in which it was
trapped. The “corrected” ice data were then smoothly
aligned with the Mauna Loa record (Figure 1 B), and
reproduced in countless publications as a famous “Siple
curve”. Only thirteen years later, in 1993,
glaciologists attempted to prove experimentally the “age
assumption”[10], but they failed[9].
I put the html and link in for the figures but they didn’t show.
Pete, yes your comment is a fair statement of my views. I definitely don’t want to hitch my sail to the J wagon. I do realize that usually the skeptic is wrong and the expert right. I would rather that the responses were more detailed and willing to engage, so that I could explicitly see the issues with the J paper and weigh it based on science/logic not expertise.
At that same debunker site, there is a shorter (and better) discussion of the 92 article itself (not the testimony). that discussion makes some good points that I agree with: J has a laundry list of possible experimental concerns. J. does not well characterize the degree of error from these possible concerns. I know that any complicated issue, is one where one can try to find error or raise possible uncontrolled confounding factors. Still, I would feel better if the defense to at least some of them…was more along the lines of “here this possible confounder has been studied and proved irrelevant” rather than saying the J. has not proved it definitely to be at issue.
Re #90, Hans, which version of the sat record did Doulgas use? I suspect it was 5.1? Do the differences 5.1/5.2 make and difference?
One interesting thing in Jaworowski’s statement is the 1900th century measuremnts of CO2 levels. The mean value of all the measurements are close to the value of 1973. In IPCC only a fraction of the measurements have been taken as a standard for the preindustrial level. This “cherry picking” is biased towards a lower value. I have not digged in the literature to find out the reason for the selection. If someone knows it would be interesting to see it.
Glad you asked. In his statement, Jaworowski takes a figure from Fonselius et al. 1956 [1], and claims that “encircled values show a biased selection of data used to demonstrate that in 19th century atmosphere the CO2 level was 292 ppmv.” He cites Callendar 1958 [2] as the source for the “biased selection”, but this would appear to be an error, since Fonselius was not commenting on a paper not to be published for another two years. Rather, Fonselius et al. cite Callendar 1940 [3] and 1949 [4], saying that he “used the mean values of the longest and most accurate measurements published in the literature and thus obtained a series of values, showing a mean value of 294 ppm in the period 1865 to 1900, with no trend, and then an increase to about 322 ppm during the period from 1900 to 1935.” Fonselius et al. also remark that “in the diagram, the values used by Callendar and our mean values are encircled and we can see that our values fit in quite well.” For his statement, Jaworowski redrew the figure from Fonselius et al., removing the labels identifying from which studies the encircled values were taken. You can see both figures here.
In Callendar 1958, he elaborates on the criteria for exclusion of data:
In selecting data for study, Callendar set up clear a priori criteria for exclusion, and followed them faithfully. Other researchers, such as Fonselius, acknowledged the validity of his methods at the time (though you wouldn’t know it to read Jaworowski). Data taken since then, from independent sources using different methods, have confirmed the trend observed by Callendar.
Jaworowski deserves to be taken seriously if and only if, on the evidence, he appears to be more interested in discovering the truth than he is in scoring a predetermined debating point. I leave it to you to decide if Jaworowski has represented Callendar and Fonselius et al. fairly.
And to anyone who agrees with Jaworowski that the data in the figure were “cherry-picked” and that the pre-1900 data should have been averaged without first being winnowed for accuracy, I will propose the following question:
What physical mechanism do you think may have caused atmospheric CO2 levels to vary by 350 ppmv — an amount equivalent to 745 Gt of carbon, or 75% more than a typical year’s total global CO2 flux — from one year to the next before 1900, and then stopped doing it?
[1] Fonselius, S.; Koroleff, F. and WàÆà⣲me, Carbon Dioxide Variations in the Atmosphere, Tellus VIII (1956), 176-183
[2] Callendar, G.S., On the amount of carbon dioxide in the atmosphere. Tellus X (1958), 243-248
[3] Callendar, G.S., Variations of the amounts of carbon dioxide in different air currents, Quart. Journ. of the Met. Soc. 64 (1940), p. 395
[4] Callendar, G.S., Can carbon dioxide influence climate? Weather 4 (1949), p. 310
Amen
Re #101, yep, good work that, I’d missed that entry in your blog.
Thanks to Jim Easter for clarifying this aspect of Jaworowski’s statement.
I have discussed the topic with some people, but the answer has been that they don’t know if Jaworowski is right or wrong in this case.
” What physical mechanism do you think may have caused atmospheric CO2 levels to vary by 350 ppmv …”
When looking at the 19th century measurements I never expected that they represented varying C02-levels, but rather varying inaccurate measurements. So, it was still plausible to see the average of all measurements also. It is still interesting that the total measurements are gradually going down before the most accurate ones in the 20th century. It could have been because of improvements of the technology.
The most important thing is that some of the old measurements are taken in a wrong place or in a wrong way.
Jim,
Did Callendar have a method of evaluating the accuracy and precision of CO2 measurements he chose or rejected?
Hans (Re: 94 & 86),
Were you able to find out how the stomata were calibrated by the authors to indicate CO2 concentrations?
re# 106
Brooks, try reading the dissertation! But let me spoonfeed you:
Click to access c5.pdf
Indeed, he did. Much of the 1958 paper is given over to a discussion of the strengths and drawbacks of the methods used by the sources. For example, in regard to methods in general, Callendar said
In regard to the methods used by specific researchers, Callendar found it useful to give a few examples:
Throughout this paper, and the earlier papers cited by Fonselius et al., Callendar gives evidence of a very careful examination of his data sources, and of an honest and fair attempt to select among them on the basis of their probable accuracy. If anyone still feels that this selection was somehow biased, I’d suggest that the most productive criticism would be to come up with a better set of selection criteria, apply it to the data sources, and see if you get more reasonable results.
Jim
Thanks very much for the detailed answer as well as the reference.
On page 90 of the dissertation on the caption of Figure 5.1 is the following:
The stomata are calibrated against the ice core CO2 measurements. In other words, the one supports the other.
This is a calibration with one known point, that is the post indutrial CO2 levels and the post industrial stomata index. The fact that the stomata indices change in the older data records indicates that CO2 levels were changing, assuming that CO2 partial pressure is a major factor in the stomata index change. The problem is that you do not have a reliable point in the past to correlate stomata indices with CO2 partial pressure, unless you completely ignore the issues that ZJ brings up which would certainly effect the reliability of the ice core CO2 measurements.
You can of course accept that all the ice core CO2 measurements are correct. However, you are doing this based on faith unless you can show that the ice core retrival, sample handling, and anlaytical techniques are not effected by atmospheric contamination, loss of CO2 through fractures, interferences in the analytical instrument(s), or failure to properly calibrate the instrument with a tracable standard. These sorts of issues are common to any measurements of trace atmospheric components. There are some very knowledgable people at NIST who have an in depth understanding of the analysis of trace gases. If you do not wish to accept my explanations, you can certainly contact NIST concerning trace gas analysis and calibration.
You could argue that the major factor in stomata index change is the CO2 partial pressure. I could also argue that the ice core CO2 measurements are off-set as a result of sample retrieval and handling effects. I could consequently draw a caibration curve with almost any slope, negative or positive. Would my new curve be any more or less accurate than one based on accepting on faith that all the ice core CO2 analyses are correct?
Brooks,
Stop armwaving, do your homework, show _your_ version of the calibration curve.
That’s how science works.
Hans,
I am not hand waving. This is a circular argument. A proves B because B proves A.
When I calibrate an analytical instrument, I prefer to use a multipoint calibration with an NIST tracable standard blended by means of a blender which is calibrated for flow using a positive displacement cylinder. I calibrate the instrument over its full dynamic range. Lacking this, I at least do a zero and a span with a tracable standard. Of course in the case of zero and span calibrations, you must assume that the response is linear.
Here we have a system with no zero, thus a methodical approach would be a multipoint calibration through the full range of stomata indices and CO2 levels. The problem is that we do not have sufficient reliable data to do the calibration.
When I said that I could draw any calibration curve I desired through what is essentially a single point, I was trying to be humorous. If you only have a single point, or a couple of points on one end of the scale, you simply can not perform a calibration.
Hans asked me to clarify the stomata calibration a bit,
As you have noticed from my thesis, we generally use Mauna Loa data from 1950 onwards and compiled shallow ice core data prior to that period. I agree this is one of the weaknesses and we often hear this is a cicular argument. One of the alternatives we have is calibration against partial pressure by using altitudinal transects, however in the Netherlands this is not possible…. Furthermore, the wide range of climate over these altitudinal transects is so extreme that this might leave an imprint on the stomatal frequency. So, at the moment the only thing we can use to calibrate our PROXY is by using shallow ice core data. I have to mention that lets say the pre-industrial CO2 levels were actual higher then proposed by the ice core data, our calibration curves would have a steeper slope and therefore CO2 wiggles reconstructed with them will have a smaller amplitude. However, pre-industrial CO2 wiggles of 4 ppmv will not be generally picked up by a plant. So, at the moment we are still working (and trying to get funds!) to do low CO2 growth esperiments to make our calibration curves less dependent from ice core data. In the meantime we try to detect and replicate our proxy records using various different species from across the globe and apply different calibration techniques (including the altitudinal transects) to verify our records.
As a response to the comments about our firn densification smoothing study. We were just curious how big the effect of firn densification smoothing in ice cores actually was and if this alone could explain the difference in amplitude we observed. It seemed that the smoothing was bigger then expected and for a large part could explain this discrepancy.
The Co2 wiggle during the thirteenth century AD has been recorded now in two ice cores and two (maybe three) stomatal frequency records… I agree that more accurate calibation of the stomata records is needed, however that these trees pick up this signal at all (various sites : north-sweden, the netherlands, west-coast USA) is for me evidence that this perturbation should at least have been higher than the absolute maximum of 12 ppmv suggested by the ice cores, I think we provided some evidence that this is partly caused by an underestimation of firn densification based smoothing (note: on these small timescales!),…
furthermore… a study by Gerber et al (2003), did a climate model run with a higher variability of air-temperatures then suggested by the hockey-stick scenario, and found out that this would have generated CO2 wiggles of up to 20 ppmv…..conclusion: ice core records do not record such changes during the past millennium so air-temperature variability is constrained….Now that several studies start to question the low-variability of air temperature changes during the past millennium, maybe we should also re-reconsider CO2 variability…if you do not believe the stomata evidence…this model study provide independent evidence that under higher T variabilities CO2 fluctuations should have been higher
Tom van Hoof
btw here’s some of the references of chapters from my thesis
Wagner, F., Kouwenberg, L.L.R., Van Hoof, T.B. & Visscher, H. (2004). Reproducibility of Holocene atmospheric CO2 records based on stomatal frequency analysis. Quaternary Science Reviews, 23/18-19, 1947-1954
Van Hoof, T.B., Kaspers, K.A., Wagner,F., Van de Wal, R.S.W., Kürschner, W.M. and Visscher, H. (2005) Atmospheric CO2 during the 13th century AD, reconciliation of data from ice core measurements and stomatal frequency analysis. Tellus B, 57/4, 351-355. doi 10.1111/j1600-0889.2005.00154
Tom,
You’ve got plenty of greenhouses in the Netherlands…is it not possible to seal a few up from the atmosphere and run some CO2 sensitivity tests on something you do have a lot of: spagnum (sp?) moss or something else which is commonly found in the Netherlands?
Tom,
I forgot to ask: how sensitive are stomatal frequencies to temperature?
re #113. There have been quite a lot of FACE, CO2 enrichment studies done. What did they show as far as stomata goes? There may also have been some studies with lower CO2 concentrations as well, as John suggests. I suppose someone should go search the Idso’s site see if they have something on the subject.
Tom,
Thank you for an excellent summary.
I agree that stomata are giving an indication that CO2 levels have changed both up and down during the past thousands of years. I have a high degree of confidence in the Mauna Loa CO2 data since they are using a Siemens NDIR located at the collection site. With 4 samples per hour, and frquent calibrations, the people at the site would quickly be aware of any instrument malfunctions or drift.
Most of the other CO2 collection points send periodic grab samples to Scripps for analysis. These samples could be impacted by diffusional effects, depending on the length of time they are en route to the NDIR at Scripps. These effects are likely to be small compared to the potential impacts of contamination and gas sample loss in ice cores.
Regarding John’s post, I agree that it would not be very difficult to set up a low CO2 stomata study in a green house. It would be possible to do it in a smaller enclosure, however a larger space is easier to control.
Do you know how far the Moana Loa sampling station is from the most active lava vent of the most active volcano in the world? If its more than 5 miles I’d be very surprised.
I wonder how they calibrate for the volcano?
well we should have the funding in utrecht to built set ups for low CO2 in the near future….and they cost quite some money I can tell you! let alone hire people to do the experiments…..
there have been a lott of doubling experiments already going on for more then a decade right now, a lott of species seem to show a response under higher CO2 conditions….however, the response is species specific…..! e.g. my oak species do not go farther then 320 ppmv…..another problem which has to be taken into consideration with regards to reconstruction…therefore the multi-species approach…btw if anybody has access to leaf material from the FACE experiments……..we are always very much interested!
one general problem with most doubling studies I know is the usage of seedlings… they act different then adults..and you have to watch out to use them for calibration purposes…
about temperature….temperature is hardly an issue, some experiments with birch trees seem to give somewhat of a response but we are talking about temperature changes in multiple degrees celcius now..and the first response of a tree to temperature (regarding leaf development)lies in phenology, so the leaves start to develop earlier in the season if temperature rises…. the effect that can have an effect on stomatal frequency in general is water availability..this effects the amount of cell stretching in the epidermal cells..and so the density of the stomata…. therefore we use the index value between the number of stomata and the number of epidermal cells..this seems so far to be the most sensitive to CO2 changes
RE: trace gases in ice cores.
A new study has appeared which found new biota in ice which produce aromatic hydrocarbons.
Now what does that imply for the reliability of ice core measurements of trace carbon gases?
EXO LIFE
Life In Ice
L. Benning (Leeds) monitors sterile conditions in an ice coring tool. Credit: Kjell Ove Storvik/AMASE.for Astrobiology Magazine
Moffett Field CA (SPX) Oct 11, 2005
“We tested equipment that we are developing to look for life on Mars and discovered a rare and complex microbial community living in blue ice vents inside a frozen volcano,” remarked Hans E.F. Amundsen of Physics of Geological Processes (PGP) at the University of Oslo, Norway, and leader of the international AMASE team.
AMASE, the Arctic Mars Analog Svalbard Expedition, is designing devices and techniques to find life on Mars. Their test ground is Svalbard, (Norway) an area with a geology that is analogous to some Martian geology. “The instruments detected both living and fossilized organisms, which is the kind of evidence we’d be searching for on the Red Planet,” he continued.
Science leader of AMASE, Andrew Steele of the Carnegie Institution’s Geophysical Laboratory, explained that “ice-filled volcanic vents, such as these, are likely to occur on Mars and may be a potential habitat for life there.”
The carbonate rocks found within the approximately 1-million-year-old Sverrefjell volcano on Svalbard are similar to carbonate rosettes found in the Martian meteorite ALH84001 and may have been produced by common processes. The blue ice, trapped in the volcanic vents, may represent samples of water that formed identical carbonate deposits in the Sverrefjell volcano.
The scientists detected living and fossilized microbiota, in the ice and on the surfaces and cracks of other volcanic rocks, using their integrated life-detection strategy successfully tested by AMASE in 2004. “Our instrument, designed by scientists at the Jet Propulsion Lab (JPL), detected minute quantities of aromatic hydrocarbons from microorganisms and lichens present in the rocks and ice,” said Arthur Lonne Lane of JPL who made his 2nd voyage with the AMASE team.
Steele’s team from Carnegie deployed a suite of instruments to detect and characterize low levels of microbiota. “We performed several successful tests with a miniaturized instrument fitted with special protein microarray chips,” says Steele. “Our results showed that we were able to maintain sterile sampling procedures without introducing contamination from humans.”
Coring of the blue-ice vents and surface glacial ice involved developing a detailed procedure for sterilization of the ice-coring tool.
“The organisms found in ice are survivors! Small ecosystems in the ice have apparently adapted to extremely cold conditions,” says Liane Benning, University of Leeds. The ice and rock samples will be characterized further in labs at the Carnegie, the Smithsonian Institution, PGP, Penn State, and University of Leeds.
This summer’s AMASE expedition also involved interdisciplinary studies of the world’s northern-most thermal springs above sea level, rock weathering and pattern formation, and biota in glacial ice by the physicists, geologists, chemists, and biologists on the team. The AMASE group sampled sedimentary rocks that are roughly 780 million-year-old, which contain remarkable remains of microbial structures that still maintained morphologic structure.
“These rocks hold potential chemical markers of fossilized life. If there is similar evidence in ancient rocks on Mars, our equipment will be able to find it,” says Marilyn Fogel, biogeochemist and astrobiologist at Carnegie.
http://www.spacedaily.com/news/life-05zzzzzzp.html
Mike,
Any effect of trace hydrocarbons is instrument dependent.
The simplest method of CO2 analysis, which was alluded to by ZJ is an FID with a methanizer. The methanizer converts CO2 to CH4 and the FID measures carbon equivalents. This assumes that all carbonaceous gases are CH4. Thus a mole of C2H6 (ethane) would be detected as 2 moles of CH4, a mole of C3H8 would be detected as 3 moles of CH4, and so on.
Other more sophisticated analyzers would separate the components either by mass number or chromatography. A GC-FID/methanizer train would still count CH4 and CO2 as the same peak, however heavier hydrocarbons would have their own peaks. A mass spec would separate the components. NDIR, FTIR or laser based instruments would work as long as there were no interferences.