Krummholz and the Yamal Chronology

Just before Climategate, we were reflecting on the apparent non-“robustness” of the Yamal chronology to inclusion of Khadyta data. Briffa’s response was that he could still “get” a HS by adding in Yadoyayakha and Porzayakha ring width data that hadn’t been used in Briffa (2000). As noted in other posts, Khadyta isn’t the only site with a divergence problem. Divergence, if anything, is the rule, rather than the exception. Briffa didn’t discuss why there were such differences between the sites – something that I’ll consider briefly today.

First here is a plot showing the ten YAD core chronologies (the chronologies, as usual, divide ring width by a negative exponential to adjust for aging.) YAD061, the “most influential tree in the world”, surges off the page, closing at an astonishing 8 units.

Next is a plot showing 15 Khadyta River core chronologies that go to 1990. None of these core chronologies are rocketing off the top of the page – an entirely different impression to the YAD chronologies.

Statistically, the inconsistency between the two sites creates all sorts of problems. The idea behind using chronologies as a proxy for temperature is that ring widths, after an allowance for aging, are a linear function of temperature. However, something else needs to be invoked in order to explain the huge difference between nearby sites. How can one “thermometer” read 8-sigma, while the other is below normal?

This is a difficult and interesting statistical question – one that isn’t even posed in Team literature.

A relatively recent study, Devi et al 2008 (needless to say, not cited by Oxburgh), appears to me to shed considerable light on the conundrum. here’s their Figure 8, showing ring widths for Siberian larch that switched from a creeping (krummholz) form to an upright form in the 20th century. The generic shape of these diagrams is highly reminiscent of the YAD core plots at issue.

Devi et al 2008 Figure 8.

Devi et al described their observations as follows:

Vertical stems started to appear after 1900, with most of them appearing in the 1920s and 1930s (Figs 7 and 8). The change from creeping to vertical growth led to significant growth enhancements of the multi-stemmed trees, as indicated by the increases in the ring widths of the horizontal stems of multi-stemmed trees (Fig. 8). Approximately one to two decades after the stems had started to grow vertically, ring widths increased 2- to 10-fold, which were much greater than the increases in ring width of single-stemmed trees during the same period.

2- to 10-fold increases in ring width! The same sort of thing that we’re seeing at the YAD site. Look back at core YAD061. It grows at relatively low rates for its first 100 years or so and then surges to 8 times the growth rate around 1950 – in this case, a little later than the growth surge in the Polar Urals cores shown here.

Now look at core YAD041 – its fluctuations are highly correlated to YAD061, but it experienced a growth surge early in the 20th century and has dramatically tailed off in the second half of the 20th century.

The shapes of these curves is very similar to the Polar Urals krummholz-to-upright curves. It sure seems plausible to me that the huge “late-life” growth surge in important YAD cores might be due to the same effect.

Reference: NADEZHDA DEVI, FRANK HAGEDORNw, PAVEL MOISEEV, HARALD BUGMANN, STEPAN SHIYATOV, VALERIE MAZEPA and AND REA SRIGLING, 2008. Expanding forests and changing growth forms of Siberian larch at the Polar Urals treeline during the 20th century, Global Change Biology, 14, 1581–1591


  1. Skip Smith
    Posted Apr 28, 2010 at 4:17 PM | Permalink

    Steve, forgive my ignorance as I’m relatively new to this debate, but has anyone besides the original researchers ever located and photographed YAD061?

  2. Posted Apr 28, 2010 at 4:35 PM | Permalink

    It seems plausible that a change in growth pattern is accompanied by a change in the rate of ring width growth.

    I wonder how this influence could be removed through some rational adjustment so that perhaps the temperature proxy would produce a valid indication of temperature change.

    At the least, as pointed out by Skip Smith, it would be useful to know for sure how these trees were growing, so we don’t have to wonder if they changed to vertical growth when their ring width growth rate changed.

    I wonder why the scientists who collect and analyze these samples don’t record such information at the time of collection, so they can later figure out what influenced the trees’ growth rates.

    • Shallow Climate
      Posted Apr 28, 2010 at 6:37 PM | Permalink

      Your “I wonder why…” musing: Yup, the devil, as usual, is in the details, in the metadata. It seems to me–and others have proffered this here before–that “in the good old days” (maybe not all that long ago) dendro science was such a scientific backwater that nobody paid any attention to it, so you could be sloppy about stuff like your concern here and who cares? Ditto, of course, the CRU data-keeping. If it weren’t for people like SM, this slop would be regarded today as Received Truth.

  3. Brooks Hurd
    Posted Apr 28, 2010 at 4:37 PM | Permalink

    Excellent find, Steve.

    Briffa did discover an interesting phenomenon of Yamal larch growth. Based on the near by cores you showed here, he seems to have misinterpreted the reason for the significant change in YAD061 ring widths, in spite of the fact that YAD061 had the right shape at the right time.

  4. justbeau
    Posted Apr 28, 2010 at 4:44 PM | Permalink

    In his papers, does Briffa display pictures of the trees or describe their micro-habitats? Is Briffa mindful whether a tree was vertical or multi-stemmed?

  5. Bruce
    Posted Apr 28, 2010 at 5:20 PM | Permalink

    Dear Steve,

    I would urge you to read the (Swiss Forest Service) Jan Esper (et al.) paper in Science, published in the year 2002, on intrinsic mathemetical problems of detrending in tree-ring studies (by which reduced tree ring widths in trees of greater ages are accounted for in dendro-chronology). Jan employs special methods to ‘rescue’ for example, the intrinsic suppression of the detection events like the Little Ice Age in detrended tree-ring data.
    With respect to the detrending, it is important to remember that this is a proceedure for chronology building, NOT longer-term climate history reconstruction. It is mathematically INTRINSIC to detrending that long-term, low-frequency signals of the latter type are suppressed. Perhaps not by accident then, Mann and Briffa, using methods not really suited to their palaeo-climatological purposes, come to incorrect conclusions.
    My impression from the phrasing of comments of Mann and Briffa re. Esper’s work (although Jan’s name is not mentioned, it is clear from context that Esper’s work is regarded as a ‘challenge’) in the leaked CRU emails is that they actually do not fully apprehend the detrending problem (in short, they are probably not corrupt, simply not the best in their field). The late Jean Grove of Cambridge U., author of two books on the Little Ice Age (and who gave her final lecture here at Durham), DID see the significance of Jan Esper’s work. I have made comments on this in official book reviews for QSR, but think it proper to also regard these sorts of (web) sites as appropo. places to make comments, as you do good work.

    Bruce M. W. Albert, Ph.D., PDRA, Durham U., UK (obviously not intending to get anything published in the journal The Holocene, which under present circumstances would not be regarded as any honour)

    Steve: i”m well aware of Esper’s work and have discussed it in the past.

  6. Craig Loehle
    Posted Apr 28, 2010 at 5:30 PM | Permalink

    I seem to remmeber from before climate gate (BC?) that it was pointed out that when you use RCS or similar to correct for age effects, which assumes exponential decline of growth with age, you will accentuate any late-in-life growth spurt like at YAD or in the bristlecones with strip bark. A perfect storm.

    • Rattus Norvegicus
      Posted Apr 28, 2010 at 11:46 PM | Permalink

      Read Briffa’s writing on modern sample bias.

  7. EdeF
    Posted Apr 28, 2010 at 5:33 PM | Permalink

    Interestingly, I found the same article last night while looking for
    Yamal instrumented temperature data, but didn’t have time to read the
    whole article. This is really important stuff, I think you have put your finger on the massive growth spurts due to krummholz growth. This is the type of metadata that needs to be talked about in the proxies studies.

  8. timheyes
    Posted Apr 28, 2010 at 6:00 PM | Permalink

    Apologies if I’m just being stupid (e.g. mixing up anomalies and temperatures) but the graphs don’t seem to correlate well with the temperature data. The peaks in temperature at 1880 and 1940 don’t seem to correlate with tree ring width.

    or the anomaly either…

    I couldn’t find temperature records for the yamal area alone.

  9. dearieme
    Posted Apr 28, 2010 at 6:00 PM | Permalink

    Well, obviously the change in growth mode is a consequence of Global Warming, innit? Simples.

    Steve: actually it is. The krummholz form was a way of surviving the Little Ice Age. The warmer 20th century resulted in the change to upright. The direction of the effect is correct – the only trouble is that the method becomes very sensitive to the sampling proportion of krummholz changes.

    • Pat Frank
      Posted Apr 28, 2010 at 9:17 PM | Permalink

      From reading the sites, the Krummholz effect seems to be from wind not cold. Cold alone produces upright but badly stunted trees that reproduce vegetatively. So could it be that the recovery from of Krummholz flattening is due to less wind stress, rather than higher temperatures?

      Could a connection to warmer alpine/arctic temperatures be through a smaller latitudinal thermal gradient, and therefore less severe winds?

  10. Steve McIntyre
    Posted Apr 28, 2010 at 6:28 PM | Permalink

    Another point to note – the krummholz effect contributes to the Divergence Problem. If you look at the plots in Devi et al shown above, the krummholz effect begins in mid-century and then tails off in a negative exponential of a typical sort. If there are enough of these so that the site average is affected (and it wouldn’t take too many with 8 sigma pulses), you would get a divergence problem.

    • geo
      Posted Apr 29, 2010 at 7:09 AM | Permalink

      Hmm. Surely that is deserving of some more research? Because what I think you may have just implied is that the “divergence problem” could be an artifact of cherry-picking their cores for temporary growth spurts at the right moment in history (a form of confirmation bias) that then peters out when the temporary cause is removed.

      Or perhaps “temporary” isn’t quite the right world. Threshold transition?

    • geo
      Posted Apr 29, 2010 at 10:02 AM | Permalink

      Is there any evidence of an inverse correlation between the amplitude of the Divergence Problem vs sample size of a given chronology? That would be a fingerprint for what you’re pointing at, right?

      Steve: I’m not saying that this is THE answer to the divergence problem. That would be a big undertaking that I don’t have time to do. I’m merely observing that these cores would contribute to an apparent divergence effect.

    • Clark
      Posted Apr 29, 2010 at 11:33 AM | Permalink

      Nice. A single observation can explain both the YAD061 8-sigma explosion and the divergence problem – simply a matter of when relative to the 1990’s a tree switched to single-axis growth.

      The perfect test of this hypothesis would be a species in the region that did not have the two growth types. Is there such?

  11. Benjamin
    Posted Apr 28, 2010 at 7:11 PM | Permalink

    Same thing happens in northern canada

    Click to access 032604ar.pdf

    Seems to have been common knowledge in dendroclimatology for quite a while that you cannot use tree rings just like that….

  12. Bermie
    Posted Apr 28, 2010 at 7:24 PM | Permalink

    The silence from the rest of the dendro community on Briffa’s argument is deafening.

  13. Marine_Shale
    Posted Apr 28, 2010 at 7:39 PM | Permalink

    Very interesting stuff,

    It could be quite instructive to revisit the plots relating to the bristlecone and foxtail pines,collected in Krummholz zones and used in previous “Team” studies.

    The Sheep Mountain data for instance may reflect the confounding influences of both strip bark and the Krummholz effect.

    I will have to go back and look at the divergence issues with that data as well, as I never did think that teleconnection explained anything.



  14. AnonyMoose
    Posted Apr 28, 2010 at 7:59 PM | Permalink

    If Briffa had to add more data to get a HS, particularly specific data, then it sure don’e look so robust.

  15. Robert
    Posted Apr 28, 2010 at 8:07 PM | Permalink

    Paging Dr. Wilson…

  16. phinniethewoo
    Posted Apr 28, 2010 at 8:20 PM | Permalink


    not the strongest suit of The Team.

  17. Pat Frank
    Posted Apr 28, 2010 at 8:32 PM | Permalink

    A picture of Krummholz spruce in Nova Scotia Canada.

    They’re all very scrubby. How does one get useful cores from such trees?

    • EdeF
      Posted Apr 30, 2010 at 2:10 AM | Permalink

      We call that kindling.

  18. Posted Apr 28, 2010 at 9:23 PM | Permalink

    As Dendro Ed Cook states in the CRU emails:

    …the results of this study will show that we can probably say a fair bit about 100 year variability was like with any certainty (i.e. we know with certainty that we know f***-all).

    With a statement like that, dendroclimatology seems pretty useless for paleo reconstructions.

  19. Posted Apr 28, 2010 at 9:25 PM | Permalink

    Ok, the quote got messed up, had some greater than and less than signs in it. Here’s another try:

    the results of this study will show that we can probably say a fair bit about <100 year extra-tropical NH temperature variability (at least as far as we believe the proxy estimates), but honestly know fu**-all about what the >100 year variability was like with any certainty (i.e. we know with certainty that we know fu**-all).

  20. Arn Riewe
    Posted Apr 28, 2010 at 9:30 PM | Permalink

    Steve – I’m sure you’ll get a contact from Briffa tomorrow saying this is an interesting observation and that he’d like to explore the ramifications with you.

    OK, not serious, but the boys could do themselves a lot of good if they’d get off the wagon and intellectually discuss this with the interested community.

  21. Geoff Sherrington
    Posted Apr 28, 2010 at 9:50 PM | Permalink

    You’ve mentioned this before and I read the papers then, so I’ve had rumination time.

    The type of effect described (prostrate or upright growth)is capable of being induced by plant breeders by playing around with minute traces of organic compounds under the general name of “auxins” such as gibberellic acid and relations, which can inhibit or promote growth. I’ve used gib myself and still have some in the fridge.

    IIRC, I have mentioned gravitropism/auxins in a past post. One demonstration of auxins is that they are involved in ensuring that plant roots usually head downwards rather than upwards. Here we have the possibility of the upside-down version of a similar mechanism, not to be confused with upside down like Tiljander graphs.

    Other auxins like the gibberellins that increase growth rate are used above ground, for example, to increase grape yields for vintners. They are also used to bring forward the growth cycle, so that flowers for breeding can be opened earlier for crossing, by periods of typically a month. Gibberellic acid discovery dates back to 1898 when Shotaro Hori demonstrated that it was caused by a fungus, now known as Gibberella fujikuroi.

    However, while such plausible mechanisms for changes of growth forms are available and known, the question of whether they are applicable here remains open. It is noted that the effect you demostrate is positionally related, to a good degree, to selected river valleys. Who knows what active chemicals, including fertilizers and agents altering pH, have been added to some valleys and not others, by accident or design, over the past century? Matters like this should have been studied to exhaustion before papers were published to claim that growth rings reflect emperature. That they were not is simply a reflection of the immaturity of the science behind the topic.

    For reasons like this I continue to hammer the multitude of uncertainties in calibration of tree rings and temperature.

    • Follow the Money
      Posted Apr 30, 2010 at 7:19 PM | Permalink

      What about growth effects from radiation, say from atmospheric atomic and hydrogen bomb testing, especially in the 1950’s and early 1960’s? Any public studies on that?

      The Yamal Peninsula is downwind from the island Novaya Zemalya, where the Soviets tested their bombs.

      Interestingly, bristlecones, esp. the Sierra ones, are close to American above ground test sites in Nevada.

      • Geoff Sherrington
        Posted Apr 30, 2010 at 7:43 PM | Permalink

        Hard to think of a nuclear fallout effect. The concentrations are minute (but easily measured because the instruments are so sensitive). I’ve been involved in some work where seeds were irradiated with X-rays and gammas, but this was to try to produce strange mutants. Nothing much happened, but other workers have reported effects.

  22. geo
    Posted Apr 28, 2010 at 11:03 PM | Permalink

    I am soooo heartily sick of “robust”. When I first heard it used in this kind of context it was the early 1990s and it was a perfectly useful and intuitively understandable concept. Twenty years later it has become a caricature of itself that deserves instant suspicion when seen.

  23. Posted Apr 28, 2010 at 11:07 PM | Permalink

    It is as if growing kids start playing in the yard, and someone inferred that all the sun exposure must have caused the growth.

  24. a reader
    Posted Apr 29, 2010 at 7:01 AM | Permalink

    The Devi paper was listed under the references to the Briffa and Melvin paper in response to CA, “Examining the validity of the published RCS Yamal tree-ring chronology”, so they certainly had knowledge of this issue.

  25. Posted Apr 29, 2010 at 8:02 AM | Permalink

    I really wish someone would do a cartoon of a thermometer tree.

    Things that aren’t temperature:
    – bicycles
    – chickens
    – wallabies
    – Persian rugs
    – Spaghetti noodles
    – Spaghetti sauce
    – trees

    But, but if you mash them together with selective scaling or data elimination processes, you can get good correlation to temp.

    You can paint the Mona Lisa from hard edged multi-color polygons by amplification and elimination of preferred features in nearly the same way.

    Seeing so much difference from tree to tree is one thing but does anyone have data resampled over and over from the same tree? If you were collecting data from a tree and you were an evil scientist with intent to manipulate data, a good spot to take your core would be from a bulge in the tree stem. Bulges would be covered up over time as new layers evened the tree out so they are recent growth enhancement totally visible on the outside of the tree. Since trees are cored 1 or 2 times, it’s not difficult to see where this would cause problems – perhaps even as severe as Steve’s graphs above.

    Does anyone have any information on which area of the tree should be cored or methods to address this ‘bulging’ problem?

  26. Steve McIntyre
    Posted Apr 29, 2010 at 8:40 AM | Permalink

    Jeff, this sort of bulging is the nub of the strip bark problem. Ironically the post on which the CLimategate dossier was first reported (unnoticed at the time) was on a thread discussing this precise problem.

    Pete Holzmann took cores on one Almagre tree about 5 cm apart and had huge differences in results – this is reported in an Oct 2007 post.

    If there are methods to avoid this problem, they are not routinely observed. Some strip bark chronologies have telltale plots (in the measurement details) indicating bulging – but no one in the field cares.

  27. Tom P
    Posted Apr 29, 2010 at 9:01 AM | Permalink

    “…something else needs to be invoked in order to explain the huge difference between nearby sites.”

    How about looking at the temperature variations for the region before suggesting alternatives?

    The correlation between Yamal (YAMALAD) and temperature in the late 20th century is reasonable, thought adding in Khadtyta (YAMAL_KHAD) reduces the agreement. It’s the Khadyta cores that appear to be failing as a temperature proxy, not Yamal.

    “It sure seems plausible to me that the huge “late-life” growth surge in important YAD cores might be due to the same effect.”

    There appears to be a straightforward reason for the surges in the Yamal ring widths – the temperature during the Yamal growth season has itself oscillated pretty wildly in the last one hundred years.

    Steve: Tom, it is entirely possible and even likely that the Khadyta cores are “failing” as a temperature proxy – that’s the divergence problem. Clearly temperatures have been rising in Yamal in the past 50 years However, there’s no a priori reason why the YAD cores are magic treemometers. Temperature alone cannot explain 8-sigma effects without a huge nonlinearity like that from krummholz. You are also not attending to the statistical issues of making averages when you have 8-sigma excursions from krummholz effects – stipulating that they exist. The krummholz effects will wear off, simply postponing the divergence problem.

    • bender
      Posted May 4, 2010 at 6:18 PM | Permalink

      Tom P with the brilliant insights. It’s a miracle he came back.

  28. Josh Keeler
    Posted Apr 29, 2010 at 9:22 AM | Permalink

    it would seem to me that the only way to accurately determine growth per year would be to ascertain the total volume of new growth per year, not just a tiny sliver or core. The only way to realistically calibrate would be to observe all species from seedling to death, and compare a record of all local conditions to the record in the rings. Once this has been done, if there is any correlation between rings and temperature that is closer than the correlation between rings and any other local condition such as soil or water, or sensitivity to the time of the year that rainfall occurs, or land use changes around the tree, etc., it could be determined that the trees could be reliable proxies to the extent that they all correlate. Until then, it is conjecture based on personal bias as far as I’m concerned, knowledge of conditions during certain years can be used to reasonably infer the cause of varying ring widths when compared year by year, but the varying ring widths are much less reasonable to use to infer variations in any on of multiple conditions that can affect those widths.

    Steve: this is a far more general issue than being discussed here.

    • Josh Keeler
      Posted Apr 29, 2010 at 1:46 PM | Permalink

      sorry, didn’t mean to go off topic. I was just struck by the comment above by Jeff about bulging and different data from the same tree, and my thoughts kind of got ahead of myself… I’ll try to keep confined to the actual issue discussed in the future, or refrain from posting.

      Steve; you’re welcome to post. it’s just that I prefer threads to stick to the very narrow issues under discussion – otherwise every thread becomes the same.

  29. Tom P
    Posted Apr 29, 2010 at 11:38 AM | Permalink

    “Temperature alone cannot explain 8-sigma effects without a huge nonlinearity like that from krummholz.”

    Steve, you have erected a straw larch. That 8-sigma refers to YAD06, and nobody would claim that temperature is going to be the sole input into the growth of any one tree. But the Yamal series as a whole has a correlation between the growth-season temperature and chronology above the published prior threshold for acceptability. Including Khadyta drops the correlation below that value.

    You are hypothesising, without any specific evidence for Yamal, that krummholz effects are significant here. But temperature does move with much of the variation in the growth. As for attending to the statistics, why did you not feel it worthwhile to present either the decent correlation values for Yamal, or the poor values obtained by including Khadyta, in any of your previous posts?

    Steve: Tom, there is circumstantial evidence for krummholz effect at YAD. At Polar Urals, when the krummholz effect began, there was a huge increase in ring widths disproportionate to temperature effects (as seen in upright trees). The form of key YAD plots is sufficiently similar to the corresponding Urals plots to warrant the hypothesis that these particular cores were affected by krummholz. Given the unfortunate lack of metadata from the original sampling, this remains just speculation, but you shouldn’t take the a priori position that I’m wrong about this. (ON another occasion, I surmised that some foxtail cores were strip bark from similar plots. Andrea Lloyd checked her field notes and they were.)

    As to invoking temperature correlations at this stage, you really demonstrate very little understanding of statistical methods. The implicit model of a chronology and reconstruction is that Yamal larch have a linear response to temperature. If YAD does, but Khadyta doesn’t, then that shows that there is something else at work and your model is ill-specified. If some cores at YAD have 8-sigma residuals, then that is a statistical problem that requires attention. It is your prerogative to believe that the YAD trees come from a magic grove, but personally I don’t.

    • Brooks Hurd
      Posted Apr 29, 2010 at 7:22 PM | Permalink

      Please list the natural events which have an 8 sigma change in such a short period of time. There are a few that I can think of such as volcanic eruptions, major seismic events, and super novas. Temperature clearly is not such an impact.

      How would describe the use of a data series such as YAD061 in a temperature proxy reconstruction? Personally, I would not describe this as an example of good science.

  30. Original Mike
    Posted Apr 29, 2010 at 1:42 PM | Permalink

    The phrase “lack of metadata” obscures what’s going on here; they don’t know or won’t tell what trees they sampled, right? As a professional scientist, I’d be embarrassed if this was my work.

    Was this data taken before the availability of GPS?

    • Ben
      Posted Apr 29, 2010 at 2:53 PM | Permalink

      Exactly correct, when dealing with statistical methods that is when metadata is especially needed. The first thing I learned in my first data mining course was that the sampling was the most important part of the data mining experience. If you don’t document what you did to grab your training data correctly, the training will not come to correct values and thus garbage in, garbage out.

      I think we are dealing with many different issues along with our variable of temperature, from the (possible) krummholz effect as seen here to just simply other unexplained phenomena that have been included in the proxies perhaps simply because they (seem) to be good temperature proxies…just in my opionion a bad mix for trying to model anything, especially something as complex as the summation of weather across the globe (AKA climate.)

      Steve, you are doing good work on finding these small tell-tale signs of problems in the data, keep up the good work, some of us just read your site and don’t post much like me…

    • Sean
      Posted Apr 30, 2010 at 9:34 AM | Permalink

      Interestingly (given Steve’s previous analogies to corporate/financial scandals) this is precisely the charge being leveled at Goldman Sachs by the SEC.

      Both sides agree that the actual list of mortgage bonds was fully disclosed to the buyers and buyers could have analyzed each of the individual bonds and the mix of bonds to see if they wanted to buy it.

      But the SEC alleges that they buyers were not told how the bonds were selected. Out of all the possible bonds that could have been used in the deal, the SEC alleges that Paulson (who was on the other side of the deal) carefully selected the ones he thought were most likely to default. Paulson and Goldman deny this.

      The SEC’s specific legal claim is Goldman should have disclosed to the buyers that Paulson was Paulson was involved in picking the bonds, and not to do so was a material omission of a crucial fact that the investor would have needed in order to properly evaluate the investment. Goldman denies that it would have been material, and they deny that Paulson was primarily responsible for picking the bonds.

      Whatever the merits of the SEC’s case, it’s interesting to see the US government taking the position that the selection process itself (independent from the results of the process) is material in evaluating the items selected.

      Steve: Yes, I’ve pointed out on many occasions the seriousness of disclosure in business transactions, as it’s what gives securities authorities leverage when deals go south. Even Oxburgh acknowledged that selection was potentially fraught with problems, but the studies where this was an issue (Mann and Jones 2003; Osborn and Briffa 2006; Jones et al 1998 etc…) were carefully de-selected from their remit. The UEA-Oxburgh sham selection process wouldn’t stand up for a minute in the real world.

      • Sean
        Posted May 4, 2010 at 10:42 AM | Permalink

        Incidentally, if anyone is interested in pursuing this analogy, here is the relevant text fromt he SEC complaint:

        “In sum, [Goldman Sachs] arranged a transaction at Paulson’s request in which Paulson heavily influenced the selection of the portfolio to suit its economic interests, but failed to disclose to investors, as part of the description of the portfolio selection process contained in the marketing materials used to promote the transaction, Paulson’s role in the portfolio selection process or its adverse economic interests.”

        And here is the link to the SEC complaint:

        Click to access comp21489.pdf

    • AMac
      Posted Apr 30, 2010 at 1:36 PM | Permalink

      Re: Original Mike (Apr 29 13:42),

      One of the earlier threads in which similar issues arose was last November’s “Miracles and Strip Bark Standardization.” A dendrologist engaged in back and forth in the comments (likely a dendrologist, as “CB” was pseudonymous). My too-long comment provides an entry to that part of that thread. It seems some of the issues discussed there may be relevant here, as well.

      • Original Mike
        Posted Apr 30, 2010 at 3:31 PM | Permalink

        I’m new here (sought this blog out after reading Montford’s new book). Thanks much for pointing that thread out for me, AMac. Only part way through it, but it’s fascinating stuff. I’ve got a prior interest because I recently performed CT scans on two bristlecone pine logs a colleague brought me. Thanks again.

  31. Tom P
    Posted Apr 29, 2010 at 2:27 PM | Permalink

    “The implicit model of a chronology and reconstruction is that Yamal larch have a linear response to temperature. If YAD does, but Khadyta doesn’t, then that shows that there is something else at work.”

    There’s much more these analyses than a simple linear relationship between growth and temperature with the same constant of proportionality for a single species, as you well know. The same species in a different, less stressed environment, might show very little response to temperature. Your second sentence therefore does not follow logically from your first.

    And what evidence do you have for a non-linear response for the Yamal series? My plot above does not show any obvious sign of such a relationship.

    Steve: There is no a priori reason why the YAD cores came from a “stressed” environment while the Khadyta cores didn’t. If the Khadyta cores are not “stressed”, then you have to throw out pretty much everything in the field. Aside from that, Climategate documents confirm that many subfossil cores came from the Khadyta river. Not even Briffa suggested that Khadyta cores were somehow inappropriate.

    Tom, as a matter of interest, do you have any experience doing any statistical analyses or is this sort of a tutorial for you? The YAD cores have a very non-linear response to temperature. You can’t just look at the chronology average – you have to look at the data set.

    • Dave Andrews
      Posted Apr 29, 2010 at 2:52 PM | Permalink

      “The same species in a different, less stressed environment, might show very little response to temperature.”

      But then surely you are into identifying the ‘ more stressed environments’ and how could you possibly do this and surmise anything about tree ring response to temperature over the centuries given that landscapes and climate are constantly changing?

    • Philometre
      Posted Apr 29, 2010 at 6:02 PM | Permalink

      Tom: I thought the purpose of statistical test was to verify a result with more than an unic serie … in order to validate a theory.
      Maybe others series were stressed by environnment. But your “good serie” too, giving a false result of perfect linear correlation with real temperatures.

      The issue is that we don’t have large number of series matching your ….. desires.

    • Tom C
      Posted Apr 29, 2010 at 10:22 PM | Permalink

      Tom P

      Actually, there are very solid thermodynamic reasons to be very suspicious about growth spurts of tree rings. The rate of ring growth will be either limited by the reaction kinetics of the underlying photosynthesis reactions or by the mass transfer of nutrients from the air and soil. Both processes have Arrhenius type dependence on temperature and the coefficients are well known, tending to cluster around typical values. A temperature increase of a degree or two cannot lead to extraordinary increases in growth rate since neither the kinetics nor the mass transfer can possibly have increased proportionally. The only things that could lead to huge changes are in the tree morphology or a drastic environmental condition such as availabililty of water, etc.

  32. Tom P
    Posted Apr 29, 2010 at 3:03 PM | Permalink

    “The YAD cores have a very non-linear response to temperature.”

    Rather than just repeat yourself, could you please provide some evidence?

    • Marine_Shale
      Posted Apr 29, 2010 at 6:01 PM | Permalink

      Tom P

      You want evidence of non-linear response to temperature of individual trees at the Yamal site?
      Do a graph using only YAD01,YAD031,YAD041 and YAD121 and see what you get.
      As Steve said look at the data set, not just the average.

      I may be missing something here Tom, and if so I am more than willing to learn, I hope you are as well.

    • Dave Dardinger
      Posted Apr 29, 2010 at 6:10 PM | Permalink

      Re: Tom P (Apr 29 15:03),

      Rather than just repeat yourself, could you please provide some evidence?

      Rather than just trolling, could you please look at the graphs at top and explain how they could represent a linear response to temperature? Indeed, how could they be any sort of response to temperature?

      • steven Mosher
        Posted Apr 30, 2010 at 3:00 AM | Permalink

        As I understand it Tom is arguing is this. He is looking at the average of a bunch of trees. Sure, a couple of them show growth spurts that are unprecidented. Sure they could only be explained by mechanical changes in the trees. But if you AVERAGE those trees with other trees, the average magically takes on the emergent property of treemometerhoodness.

        I think before on the YAMAL threads we had a botanist on to tell us that no way was an 8 sigma excursion cause by temperature… but nobody thought about Toms averaging trick.. its like an ensemble of GCMS

        • mpaul
          Posted Apr 30, 2010 at 2:17 PM | Permalink

          Except I suspect that some series are stationary and others (notably the morphological switch hitters) are non-stationary. I’d be cautious about averaging in non-stationary SP series and regressing the result.

    • Skip Smith
      Posted Apr 30, 2010 at 1:04 AM | Permalink

      Tom P, what kind of evidence would you like to see? A linear response to rising temperatures would plot out a roughly straight line, which even a cursory examination of the plots reveals isn’t the case.

      So, it’s either (1) a non-linear response, (2) a linear response but temperatures are not rising, or (3) there is no response at all.

      In any case, it’s hard to argue that trees act as thermometers and show evidence of increasing temperatures, unless the response to temperature is non-linear.

  33. Posted Apr 29, 2010 at 4:37 PM | Permalink

    This strikes me as very wierd:

    “The idea behind using chronologies as a proxy for temperature is that ring widths, after an allowance for aging, are a linear function of temperature.”

    You don’t have to be an expert in tree growth to observe that trees are influenced by the following factors:

    a) the CO2 level of the air (up to some maximum that is dependent on other factors)
    b) the water supply to the trees during growth season
    c) the soil
    d) the wind conditions (which also affects heat and water effects)
    e) temperature
    f) % sunshine hours received vs. how cloudy it is or how much shadow the tree has from sourroundings
    (other trees, rocks etc)
    g) attack on the tree from insects (might depend on the particular year)
    h) attack on the tree from fungi or other microbiologocal things (might depend on the particular year)
    i) age of the tree

    Many of these factors are dependent on eachother, which make focusing on a single factor very unlikely to prcduce meaningfull result. I think the first step would be to draw a causal graph and
    use Judea Pearl’s ( (and others) causual theory to determine if the effect of the factors that have been observed are sufficient to reveal an effect of temperature.
    If not, then the measurements need to extended and we can say nothing before that is done.

    The assumption of a linear relationship strikes me as extremely poor.

    Thorsten Ottosen, Department of Computer Science, Aalborg University, Denmark

    • Posted Apr 29, 2010 at 4:49 PM | Permalink

      One additional comment: many biological/chemical processes are logarithmic rather than linear as a function of temparature. Are there any papers showing that tree growth is actually linear?


    • Posted May 2, 2010 at 3:05 PM | Permalink

      I’ll try to review the potential impact of the factors I mention above. In the end the conclusion must be that it is /very/ difficult to interpret tree rings as data for temperature without kwowledge about other factors.

      a) Effect of water

      If the summer is very hot, growth may be limited as the tree lacks water. Therefore this will lead to a large underestimate of the true temperature. Only trees that are not likely to suffer from lack of water can thus be trusted.

      b) Effect of wind

      Larch is tree that has very poor wind resistance. So if the tree in question has been subject to a lot of wind during some period of it’s life, those years it’s growth cannot be compared with a later or previous period where little wind was present, or where other tres became big enough to provide shelter from the wind.

      c) Effect of sunshine.
      The photosynthesis could be larger in a colder climate with more sunshine hours than in a warmer climate with fewer sunshine ours.

      d) Effect of age

      It seems that age affects the tree-rings in indirect ways as well as direct ways. In particular, in a forrest the trees will be too close for the trees to devolop large branches horizontally. This implies only the topmost part of the tree will have green parts as the trees fight for the light. The effect is that the stems become elongated, and slimmer than they would normally have been.

      So the tree might actually be growing as fast as usual (with no direct age impact). The width of the tree rings will be determined by the green parts of the tree relative to the amount of stem (vertical and hirizontal). The (green parts)/(amount of stem) ratio may of course be dependent on age, but also on other factors like wind, soil, and neighbouring trees.

      So for a particular year, a particular tree can have its own particular (green parts)/(amount of stem) ratio. How may one possible reconstruct those?

      In turn, this seems to imply that any approach that adjust for age by some exponential factor would need to use a different expoential function per tree. Is that what is being done?

      It seems like a formidable task to interpret the tree-rings correctly. I would much more like to see comparisons of tree-rings of trees of the same age, under the same conditions (water,soil,wind etc), but growing at different periods in time, than I would see comparisons of tree rings on the same tree.


  34. Steve McIntyre
    Posted Apr 30, 2010 at 7:44 AM | Permalink

    Steve mosh,

    The same sort of thing occurs in MBH. Mann’s bristlecones (like YAD061) way overshoot as treemometers – calibrated in the way implied by MBH, they imply that prior to the 19th century, Califoria was colder than the Ice Age. They are averaged with a bunch of series that are in effect white noise so that the average “works”.

    Jean S’ remarkable post on Mann’s “Milankowitch” adjustment (recently summarized excellently by Bishop Hill) links Mann’s weird PC1 adjustment in MBH99 not to any physical effect known to the rest of the world but to accomplish in the MBH99 extension precisely what you observe here – without the adjustment, the MBH style calculation overshot because there were fewer white noise series to get the 20th-19th difference to work. By “adjusting” the PC1 in the AD1000 network, Mann got the difference to “work”. (People interested in this aspect of MBH99, please comment on the relevant thread).

    • steven Mosher
      Posted Apr 30, 2010 at 9:19 AM | Permalink

      When your data looks too good to be true, sometimes its too bad to be true.

      Or think of it this way. The real divergence is how these tree diverge from what is know about the magnitude of reponse to changes in temperature.

      Take any other sample of cores in the whole of denro. does any other tree in the whole lot respond to a temperature change of say 1C with a 6 sig spurt. excluding bristlecones? any other tree in the whole of dendro.

      Rank all cores by the magnitude of the biggest spurt in their time series. What’s that ranking look like and where do these cores fall.

      opening bets: the species that are most susceptible to having “mechanical issues” strip bark and Krummholz will be at the top of the stack.

  35. Craig Loehle
    Posted Apr 30, 2010 at 7:46 AM | Permalink

    Is it possible to get a 6 or 8 sigma growth spurt in trees by other then release from krumholz form? Yes, by release from competition. If a suppressed tree has its neighbors die it might start to grow rapidly. Not likely from the pictures at Yamal. Not due to temperature.

    • Hu McCulloch
      Posted Apr 30, 2010 at 8:29 AM | Permalink

      If a suppressed tree has its neighbors die it might start to grow rapidly. Not likely from the pictures at Yamal.

      What about erosion? If the river takes out a tree’s neighbors, the survivor (conveniently accessible by boat) would surge in growth.

      It has been noted on the Yamal photo thread that most of the trees are on the insides of the meanders, rather than the outsides where the meanders grow. However, such meanders eventually get cut off and become oxbow lakes. The new channel will cut right through the thickest growth, creating a growth surge in the old growth on the new shoreline.

      In a region where there are few, if any, roads, it was cost-effective to collect the specimens by boat. But unfortunately this may have created some biases in itself.

      Just a thought — Craig’s our forestry expert.

  36. Jack Maloney
    Posted Apr 30, 2010 at 10:22 AM | Permalink

    Interesting to compare lead stories on leading AGW web sites vs. leading ‘skeptic’ sites:

    From RealClimate (“Climate science from climate scientists”)

    Claude Allegre: The Climate Imposter
    Second CRU inquiry reports
    The Economist does not disappoint
    Krugman weighs in
    Climate scientist bashing

    From Climate Audit:

    Krummholz and the Yamal Chronology
    Yamal Aerial Photo
    Yamal and the Decline
    Yamal and the Vaganov Network

    Thanks, Steve, for covering the science!

    Steve: In fairness, I’ve also been pretty preoccupied with Climategate and will post more on it.

    • Chris M.
      Posted Apr 30, 2010 at 11:20 AM | Permalink

      Not to stray OT, but this is precisely the sort of thing that prompted me to seek other sources of information when I was a proponent of AGW. I think anyone of serious integrity would likely view websites of that sort with the same disdain.

  37. deadwood
    Posted Apr 30, 2010 at 12:16 PM | Permalink

    It is really quite astonishing that something as obvious as a morphology change would be missed let alone ignored by Briffa.

    What is even more amazing is that other experts in his field would not catch this before publication. It’s not like the literature is silent about the issue. Pal review perhaps?

  38. Michael D
    Posted Apr 30, 2010 at 1:03 PM | Permalink

    The depth of winter snow pack and the height of the Larch may play a role in the growth spurts. Trees that extend above the depth of the snow pack tend to melt out around the bole in the spring, much quicker than something totally covered by snow pack. This sould allow the soil around the tree to warm up sooner than those buried under the snow pack. So this could be increasing the growing season for trees that are taller, and thus a growth spurt.

  39. Craig Loehle
    Posted Apr 30, 2010 at 1:30 PM | Permalink

    The conversion from shrubby (covered by snow all winter) to tree form can result from a threshold response to winter dessication and/or freezing. If you go over that temperature threshold, the trees are not killed back in winter if sticking out of snow, and begin to grow like trees (see Payette refs on prior threads). So in a sense, this transition could be a temperature response, but of a very on/off type. By the way, once you release a tree from this type of suppression (given that the species can respond–not all can) you can get accelerating growth as the tree grows larger, just as it did as a sapling, sort of an exponential increase (each year larger than previous) for some decades, the opposite to the expected age effect that RCS (etc.) factor out. If you assume exponentially declining growth but have exponentially accelerating growth–8 sigma (after standardization)!

  40. sbd
    Posted May 1, 2010 at 3:18 PM | Permalink

    Not sure if this has been posted on here or not, but the AG in Virginia is investigating Mann and what he did with the state Grant Money he was receiving.

    Here is a link to the university along with the documents demands requested for the investigation.


  41. Kiminori Itoh
    Posted May 10, 2010 at 11:55 PM | Permalink

    Thank you very much. The information is very informative as usual. I will include it in my (Japanese) book which I am preparing now.
    Thank you again.

2 Trackbacks

  1. By Top Posts — on Apr 29, 2010 at 7:22 PM

    […] Krummholz and the Yamal Chronology Just before Climategate, we were reflecting on the apparent non-”robustness” of the Yamal chronology to […] […]

  2. […] *4 Krummhilz, a term from the early 20th century. German; Krumm: bent or twisted, holz: wood. Trees growing in twisted, perhaps largely horizontal directions do not grow at the same rates as to trees whose growth is largely vertical and single stem. Link: […]

%d bloggers like this: