Wilson, Pisaric and Gaspé

Two respected dendrochronologists wrote in criticizing my recent post averaging new white spruce chronology contributions to the ITRDB data bank. [Update: see post here with further thoughts on the location of this site. Next post in category] Rob Wilson wrote, using an uncharacteristic Gavin-esque sigh, as follows:

Sigh —
Location, location, location!
or even
ecology, ecology, ecology!
you are confusing the issue again.

Your little “spiel” above is meaningless without some site information.
Where are the spruce sites located (?Map)?
How close are these sites to the high elevation or high latitude tree-line?
For what purposes were these sites originally sampled?

For readers of this blog, PLEASE understand that one cannot randomly sample trees from any location and expect there to be a valid climate signal (temperature or precipitation).

In fairness, Rob, I don’t think that there are any readers of this blog that expect that one can randomly sample trees from any location and expect there to be a valid climate signal. I’d say that the concern of this blog is more whether it is valid to choose sites ex post and expect there to be a valid climate signal.

Mike Pisaric of Carleton University in Canada writes:

As Rob points out in 3 and 28, Steve’s analysis is flawed! You can easily search out for yourself where the Meko sites are located using Google Earth. They are a long way from latitudinal tree line or the mountains!

Both are concerned that “misinformation” not be spread through inappropriate selection of sites. Rob says:

Criticism is fine, but misinformation is counter productive.

I endorse this comment and I do not wish to contribute to any spread of misinformation. I am happy to correct or amend any incorrect statements. As Mike Pisaric suggested, I located the Melo sites on Google Earth, as shown below together with the Gaspé site, used in Jacoby and d’Arrigo 1989, 1992 (their northern treeline reconstruction used in many multiproxy studies); MBH98 as Ross and I discussed and, as far as I can decode in the absence of a listing of sites, even in D’Arrigo et al 2006.

Figure 1 showing Meko location and Gaspé location

Now here is a closer view of the Gaspé peninsula. I’ve used the coordinates for the Gaspé site (to the nearest minute) from ITRDB series cana036 (St Anne River). The site itself looks like it is in a river valley, which I dare say will prove to be the St Anne River. As far as I can tell, this site is nowhere near latitudinal or altitudinal treeline. I’ve endeavoured to obtain a more precise location for this site, so that I could commission re-sampling, but Jacoby said that the sampling was done before GPS and he was unable to provide directions to the site.

I’m grateful that both Rob Wilson and Mike Pisaric are taking such keen interest in rooting out inappropriately selected sites from multi-site averages. I’m sure that Mike Pisaric will agree that the use of the St Anne series as a temperature proxy, given its distance from latitudinal and altitudinal tree line, makes it “flawed!” and will communicate this opinion to all of the authors of Mann etal 1998 and D’Arrigo et al 2006 with the same alacrity with which he reproached my use of the Meko sites.

Likewise, I’m sure that Rob Wilson will write to the authors of D’Arrigo et al 2006 advising them that their ” little ‘spiel’ is meaningless without some site information” and ask “where are the sites located (?Map)?” (and, Rob, a big red dot in Alaska for multiple sites doesn’t count as a location map). If Rob gets that information, maybe he can pass it along to me.


Figure 2. Gaspé peninsula


  1. Posted Mar 22, 2007 at 3:33 PM | Permalink

    Rob Wilson has started a dialogue that he may not wish to continue. As a matter of blog etiquette, one should not insult a blog host in word or manner, unless he is willing to back up his superiority complex. Rob, Gavin, and others suffer from the misapprehension that graduate training confers superior reasoning and methodology within a chosen field. To the contrary, it is quite easy to be a well-credentialed boob.

    Steve is quite up front with his methods, his data, his assumptions, and anything else relating to his commentary on the very public topic of climate change. This is an open-source worldwide party, that anyone with a good argument and data to back it up, can join.

  2. bernie
    Posted Mar 22, 2007 at 4:15 PM | Permalink

    I think Rob to date has been moreorless a gentleman in his postings to date. He has articulated a set of conditions for including dendoclimate data that seem to make sense. Steve has now pointed out that it is difficult to verify that these standards have been adhered to with respect to proxies that have been included in highly publicized articles and have been relied on to draw far reaching conclusions. As Steve put it, an understandably a touch acerbically, where is the site data that confirms adherence to these standards?

    Let’s hope the replies are constructive and move us forward.

  3. Posted Mar 22, 2007 at 4:20 PM | Permalink

    Dear Steve, it is good that the dignified Gentlemen care. You’re no longer not even wrong – at least you became wrong. 😉 I suspect that the right calculation whether the use of a series as temperature proxies is “fine” or “flawed!” is some linear combination of its distances from tree lines as well as of the ability of the result to justify cataclysmic global warming.

    As a classic whom you like to e-mail with wrote, each of us must choose a linear combination of being right and honest and being effective. Of course he hoped that it means being both right and honest and effective. 😉

  4. John Lang
    Posted Mar 22, 2007 at 7:02 PM | Permalink

    The Meko location in particular would have been subject to episodes of forest fires. The close-up Google image shows this location hasn’t been burnt recently but in these northern areas, the fires are just left to burn and there are very few mature trees.

  5. Don Youngblut
    Posted Mar 22, 2007 at 7:36 PM | Permalink

    In the absence of more detailed site specific information I don’t think you can necessarily rule out that the St.Anne site is far from treeline. The range of co-ordinates given (48N, 65W) are in close proximity to treeline areas on the Gaspe such as the Parc de la Gaspesie (Mont Albert for example which has a treeline hike which overlooks the Ste-Anne River).

  6. bender
    Posted Mar 22, 2007 at 8:43 PM | Permalink

    Rob Wilson’s point is simply that averaging chronologies across vastly different ecological circumstances is meaningless. And he’s right.

    The dominant feature of those averaged chronologies are the spruce budworm outbreaks of the 1940s and 1970s in eastern North America. That’s one of the reasons you don’t see a hockey stick. Granted: Ste. Anne River is not near all that close to treeline; but central Gaspé is generally a fairly high elevation region. Summers used to be so cold there that spruce budworm outbreaks were rare and short-lived. Time will tell if that is still true.

  7. bender
    Posted Mar 22, 2007 at 8:49 PM | Permalink

    Oh, and the “Meko” PIGL chronologies were originally collected by Stockton (1973) – from sites sensitive to changes in the flow of the Peace-Athabasca River complex. Again, Wilson’s good point about location & ecology.

  8. Pat Frank
    Posted Mar 22, 2007 at 11:14 PM | Permalink

    Rob’s objections, and Mike Pisaric’s, are tempests in a teapot. I’d still like to know how they expect to derive quantitative measures of temperature from trees qualitatively judged to be temperature-limited. You can make any argument you like about latitudinal or elevational tree lines, and historical aridity vs. temperature, but such arguments are all still qualitative judgments. There is no quantitative theory to extract temperature from tree rings.

    Until that problem is solved, arguing about location in the context of an extracted so-called temperature is so much twaddle. It’s like homeopathists arguing whether it’s better to potentiate their concoctions by tapping their bottles on a leather pad or on the heel of one’s palm.

    Even ice core data, or other dO18 measures, which do have a quantitative theory, have serious confounding environmental influences that must be taken into account and that can subvert the most otherwise careful work.

    Tree ring temperatures have all the same environmental confounds but no quantitative theory. There remains no way to extract a true measure of local temperature from a tree ring. Tree ring temperature chronologies take their validity from portrayal of a false precision, and from nowhere else. It’s all show and no scientific substance. There was a time when all tree ring workers did first rate work. I have in mind Harold Fritts and his like. But in their galloping need to participate in the high-profile AGW bandwagon, the paleotemperature honchos have betrayed their field.

  9. Steve McIntyre
    Posted Mar 22, 2007 at 11:41 PM | Permalink

    #6. Hi, bender, Nice to hear from you. The Ste Anne River chronology in question is cedar (I think that you knew that, but may have forgotten). A question: does cedar grow at treeline in these areas or would it be more likely to grow in a river valley? LArson of Guelph, who’s studied cedars for over 20 years, said to Ross and I that cedars like cool moist weather. He’s also written some interesting articles observing strip bark formation in long-lived cedars.

    Larson’s group has done interesting work on the cedars of the Niagara Escarpment (which will not be news to bender). There’s a site with 1000-year old cedars from which you can see the CN Tower in Toronto. I should do some posts on these, because they are intersting.

  10. Rob Wilson
    Posted Mar 23, 2007 at 3:24 AM | Permalink

    OK – last time.
    But this really will be it because it will generate many more replies and I cannot keep coming back. I have to stop at some point.

    Let’s take an idealistic view on trying to reconstruct the temperature history for a particular region. And when I say region, I mean a region that expresses a relatively homogeneous climate signal. There are two strategies:
    1. Sample a network of sites
    2. Sample a single site

    The latter scenario is arguably problematic as a single site may also express local site specific ecological effects and may not be too representative of the larger region. When Brian Luckman and I did the Icefields study, we tried to ensure by comparison with other independent proxies that this ‘single site’ portrayed the larger regional signal. I think we did this successfully. The fact that Steve has not ‘audited’ it may mean he is generally happy with this work. Of course, I could be dangling a carrot at him and may regret mentioning it.

    So – a much better approach, if there is enough time and resources is to sample a network of sites. As a basic rule of thumb one would go up to high elevation tree-line (or latitudinal tree-line depending on the location of the region) and sample as many sites within the region as possible.

    Willis Eschenbach asked me what makes a good site. There is no easy answer. The first step is to go to the right environment – in this case upper tree-line. However, there are many choices at this stage and quite often this might be the first time that a researcher has been to a particular region, so what works at some other location may not be relevant here. Some things to consider – (1) sampling on north or south facing slopes; (2) sampling trees on ecologically extreme locations (i.e. talus/scree slopes) or tree stands with deep soil profiles; (3) etc etc. However, we are speaking in idealistic terms, and quite often when you climb up a mountain, you may find only one relevant stand of trees which might not quite fit into some theoretical ideal – whatever that ideal is.

    So again – depending on time and resources, it would be best to sample as many sites as possible within a region. Also it would be useful to sample a couple of different sites with different ecologies in close proximity of each other. Then, when all the samples have been measured and the analysis is later made, it might be possible to objectively decide which are the better sites and knowing the ecological information, one could come up with a better strategy (with regards to an idealised site) for future sampling expeditions in that region.

    So imagine the situation that 20 ‘sites’ have been sampled within this idealised climatically homogenous region, and that for each ‘site’, I sampled two stands with varying ecologies – so 40 chronologies in all. Lets say that the dominant response of all the trees is with June/July temperatures. All the chronologies correlate with this season, but some sites will correlate more strongly with JJ temperatures than others. This is a very typical situation. From > 10 years experience, I can honestly say that I have never yet found a consistent pattern between those sites that correlate more strongly with climate compared to those that correlate weakly when taking into account the ecology of the sites. This is quite frustrating, but perhaps highlights the need of sampling networks.

    So now we come to what Steve calls “ex-post picking“.
    Let us not forget that all 40 sites crossdate with each other and this common signal across the region suggests a common forcing upon growth that can be assumed to be climate related. There are three options:
    1. All the series can be averaged together to derive a regional mean time-series.
    2. The 40 chronologies can be screened and only those that correlate most strongly with JJ temperatures should be averaged together.
    3. Utilise principal component analysis to derive the main modes of common variance, with PC1 reflecting the dominant response of the trees in the region.

    From my experience, there is often very little difference between the resultant reconstructions from option 1 to 3 although option 1 will generally end up with an R2 value lower than options 2 and 3.

    ex-post picking ” is an important step in deriving a reconstruction and essentially is a process to optimise the final calibrated r2. We are not talking about random time-series and forcing a calibration fit due to overfitting. We are talking about trees that have been sampled in environments where we would expect temperature to be the dominant limiting factor on growth and we then use statistical procedures to optimise this signal.

    With respect to Gaspé – the only way to assess the signal at this single site is for someone to go back to this region and sample a network of sites. This would be invaluable. As Steve saw at the AGU, I have been looking at divergence issues in NH temperature reconstructions and there is a distinct lack of recent data from eastern Canada. Perhaps if Steve needs a rest from sitting at his computer, he might consider going on some fieldwork to this region. The result would be interesting.

    Lastly, a personal observation with respect to:

    If you want the next generation of studies to be better than the last generation of studies, then you should annotate and archive a broader spectrum of relevant information for each sample. Take a geologist with you on your next sampling expedition and ask him to take notes and map things; he’ll know what to do. OR take a geology course and find out how to do proper maps. I can’t specify all the things that need to be done. It’s just that I’ve seen a lot of geological reports in my life; they are expected to be useful 15 years later or even 50 years later and usually are.

    I had to chuckle at this. My BSc was in Geology (1992) and I worked as a geologist for a building company from 1994-1997. I hated the job as the work was controlled by speed and money (i.e. cost-cutting). There was also absolutely NO data exchange between different building companies. I do not think the Geology world is as perfect as Steve would have us believe.

    From my experience in science, people do what they do because they have a passion. Monetary gain is simply not an issue. On the whole, I have also generally not had a problem with acquiring data from colleagues. But that might simply reflect my strategy in life of being nice to others (even Steve). I’d rather have critical friends than critical enemies. Buy hey – I am probably being naive.

    And on that note (with distant sounds of laughter from Jan Esper), I am finally signing off.

  11. Posted Mar 23, 2007 at 4:35 AM | Permalink

    The 40 chronologies can be screened and only those that correlate most strongly with JJ temperatures should be averaged together.
    We are not talking about random time-series and forcing a calibration fit due to overfitting. We are talking about trees that have been sampled in environments where we would expect temperature to be the dominant limiting factor on growth and we then use statistical procedures to optimise this signal.

    You can check whether it is overfit or no: divide data to two parts, repeat the procedure and verify that the order of most correlated series does not change. (Monte Carlo with H0: series are independent red noise will probably help to find significant order changes ?)

    with distant sounds of laughter from Jan Esper

    ..all I hear and see is laughter.. 🙂

  12. MarkW
    Posted Mar 23, 2007 at 4:43 AM | Permalink


    “I mean a region that expresses a relatively homogeneous climate signal.”

    These are about as common as Easter Bunnies.

  13. bernie
    Posted Mar 23, 2007 at 5:07 AM | Permalink

    I appreciate the thoughtful, patient and good humoured responses. Help me out here since I am in serious learning mode. If you essentially control for all in situ variables except for temperature and precipitation how do you partial out temperature when there is an interaction between precipitation and temperature? Are you saying that the trees sampled are not responsive to precipitation but only to temperature? Given the complexities, is a sample size of 40 sufficient for the level of precision illustrated in conclusions of many of your colleagues?

    One final point, you clearly recognize that the temperature estimates are tricky and as a result the error terms may be larger than any one would like. If so, how come the “bet the farm” conclusions as to no MAW and LIA.

  14. Jean S
    Posted Mar 23, 2007 at 5:32 AM | Permalink

    This is quite frustrating, but perhaps highlights the need of sampling networks.

    Or more advanced statistical methods. I would vote for the latter. Why is it so hard for ‘climate scientists’ to consult people who actually know statistics/math? IMO, a positive example of such co-operation is seen in

    J. Weckström, A. Korhola, P. Erästö, and L. Holmström. Temperature Patterns over the Past Eight Centuries in Northern Fennoscandia Inferred from Sedimentary Diatoms. Quaternary Research, 66:78-86, 2006.

  15. Steve McIntyre
    Posted Mar 23, 2007 at 6:38 AM | Permalink

    I had to chuckle at this. My BSc was in Geology (1992) and I worked as a geologist for a building company from 1994-1997.

    Touché. I had to chuckle at this comeback as well. Geologists in my experience have been good-humored and thoughtful people; so maybe your training was not as wasted as you think.

  16. Willis Eschenbach
    Posted Mar 23, 2007 at 6:57 AM | Permalink

    Rob, thanks for your patience and your replies. I understand the time constraints that all of us are operating under, and I regret that yours are such that they limit your participation here.

    You are right that your post raises more questions than it answers, but that is the nature of science. I will formulate them, although you won’t be here to answer them, in the next few days. When and if you return, I’d love to hear your responses … if not, they’ll be there for others to answer.

    My appreciation for your contribution,


  17. Spence_UK
    Posted Mar 23, 2007 at 7:06 AM | Permalink

    ” ex-post picking ” is an important step in deriving a reconstruction and essentially is a process to optimise the final calibrated r2. We are not talking about random time-series and forcing a calibration fit due to overfitting.

    While site selection remains subjective, these two aspects are inseparable. Even if all of the groups working on this are well-behaved and do not “close the loop” on their study around site selection, if you have (for example) five groups working on this, all of whom end up with different site selections, and only one group pass significance tests, that is the one that gets published – but the statistical significance is inherently weakened. And in the real world, experimenters will close the loop on site selection, perhaps even subconciously. The bottom line is that statistical significance tests with ex-post picking are probably meaningless until new out-of-sample data (that the original experimenters were not even aware of) comes along – but if the site selection remains subjective even that could be prevented.

    Of course quantification of the site selection process may be sub-optimal, but I’m afraid that is what you have to live with if you want your statistical significance tests to be meaningful.

    From my experience in science, people do what they do because they have a passion.

    I have no doubt that Blondlot did what he did because of a passion for science when he discovered n-rays. I’m sure Fedyakin did what he did because of a passion when he discovered polywater. I’m sure Witztum and Rips did what they did because of a passion when they studied the Torah codes. I’m sure Benveniste did what he did with a passion when he found water had a memory.

    What all of these examples have in common is that poor experiment design resulted in scope for experimenter bias to influence the results (albeit subconciously). Similar weaknesses appear in all of the recent attempted temperature reconstruction that I have seen, and Rob’s comment above actually reinforces the concerns I have.

    Until such time an experiment is derived which is not open to subconcious experimental bias, I cannot consider the recent attempts to estimate historical global average temperature as scientific. Just MHO.

  18. bernie
    Posted Mar 23, 2007 at 7:18 AM | Permalink

    So does this mean we declare victory and leave the dendroclimate guys to their own devices and focus on more contemporary, instrument derived temperature records? Are their models that flawed?

  19. Steve McIntyre
    Posted Mar 23, 2007 at 7:28 AM | Permalink

    I realize that Rob is not necessarily going to respond, but there are a couple of other points that I’d like to make (besides acknowledging the irony of the geology comeback). Not all paleoclimate people do things the same way. Rob says:

    ” ex-post picking ” is an important step in deriving a reconstruction and essentially is a process to optimise the final calibrated r2. We are not talking about random time-series and forcing a calibration fit due to overfitting. We are talking about trees that have been sampled in environments where we would expect temperature to be the dominant limiting factor on growth and we then use statistical procedures to optimise this signal.

    Obviously the MBH controversy has been the one that I’ve been most prominently involved in and, while Rob’s own networks may represent the picking style described above, that is very much not the case with the MBH network -as Jean S observed as well. MBH has proxies that are unabashedly precipitation proxies (he even has instrumental precipitation series in his network). I didn’t notice any dendrochronologists standing up and saying this or even defending me against some pretty vicious and unfounded criticism. I say this so that readers don’t get the impression that because Rob’s networks may be well-chosen, that this somehow vindicates the MBH network and Rob and the other field dendrochronologists have remained as silent as lambs on this.

    Obviously Rob and I are at an impasse on ex post picking. There’s a simple little article that I should write illustrating the perils in connection with Jacoby picking practice – along the lines of things that David Stockwell and I have both posted on from time to time. Jacoby’s 1989 network picked the 10 “most temperature sensitive” sites from the 36 that he sampled and he has refused to disclose information on the other 26 sites. If you cherry pick the 10 most HS-shaped series from 36 red noise series generated with persistence characteristic of the 10 remaining series, the HS-ness of the Jacoby composite is about at the median HS-ness of the red noise composites. So there is not a whole lot of statistical significance to any HS resulting from this type of ex-post picking.

    Rob can keep saying that that’s the right way to do it until he’s blue in the face, and maybe Jan Esper and Keith Briffa agree, but I’m honestly baffled by his conviction on this (as are doubtless Jean S, UC and other statistically minded readers). I would defy them to present this methodology to (say) the American Statistical Association. I don’t see any harm in exchanges that probe these assumptions and I think that it’s too bad that Jan Esper disapproves of such discussion.

    As to going to Gaspe to re-sample, I was quite prepared to commission a re-sampling program as early as 2004. I’ve been on enough geological sites to know that there’s no value added to the geological knowledge by my being there, but I like the outings and I like the people. The tree ring sites look like pretty nice places to visit so I’d be quite happy to make an outing to Gaspe during the summer along with a qualified person. Bender and I have vaguely chatted about such a possibility.

    However, I’ve sought information from both Jacoby and Cook about the location of the cana036 site so that the sampling could replicate prior sampling. Both of them have failed to provide directions to the site where the sampling took place. Rob (and I’ll send this offline as well), perhaps you could provide the exact location from which the cedars were sampled and maybe bender and I will go on a trip there. Given that this site was used in D’Arrigo et al 2006, I presume that,as a trained geologist, you know where the site is in the way that geologist does (and not just 48N; 64W).

  20. Dave Dardinger
    Posted Mar 23, 2007 at 7:35 AM | Permalink

    Since Rob in #10 above says he won’t return, I’ll just address a problem he deftly avoided to the general audience.

    The problem is that it isn’t primarily the existence of a temperature signal in the data which is an issue, but how large it is and whether it varies with time. This means we need a series of sites which we can statistically compare to see whether or not time A is warmer than time B. If we just pick a few sites which we find ex post to correlate well with temperature at time A, how are we sure that they correlate well with temperature at time B when B doesn’t have an instrumental record to prove it? It might be that there are factors not being considered which make the selected series work well in the instrumental period but not when the climate was somewhat different at time B.

    Having all the sites available to examine let more robust statistics be calculated. It may be true that the particular sites which would be selected ex post are the best temperature proxies, but it obviously wouldn’t hurt to have the others available to check this out.

    Finally we need to see if the basic hockey stick generation method applies. We know that because there’s been a general rising temperature during the instrumental period, that’s the primary signal that any analysis is going to choose. Also, as Rob pointed out, all the sites in an area are going to have a common signal since they will presumably crossdate. This means that when you’re sitting down to select ex post you’re going to see various series which have correlation due to local climate signals and also have a rising signal. If you make this your selection criteria, then either you’ll have succeeded in finding a true temperature responder or you’ll have a number of semi-random series which have a hockey blade but not much signal outside the instrumental era. When you average them, they will necessarily wash out signals outside the instrumental era and create a stick. If you start with a mindset that says the present warming is unique this won’t be a problem. But if you want to match past work which showed a MWP and a LIA then the failure to find it should set off an alarm.

    I don’t think Rob realizes how the mindset of the investigator can trip up the analysis, but it can. And the way to avoid it is to either create an objective ante selection criteria or to archive all the data and metadata for future analysis. This will let others with different biases see just how robust the final analysis is.

  21. Posted Mar 23, 2007 at 7:53 AM | Permalink

    When you average them, they will necessarily wash out signals outside the instrumental era and create a stick.

    ..and then you add variance adjustment procedure + ad hoc smoothing, and we have a very nice hockey stick.

  22. Steve McIntyre
    Posted Mar 23, 2007 at 7:59 AM | Permalink

    ONe other point about these cedars. The Gaspe cedar has never been reported in the literature in the sense that its ecological particulars and location have been provided. It has merely appeared as a time series. So while it may have great ecology, ecology, ecology and great location, location, location,… without actual knowledge of the ecology and location, it is mere speculation.

    The only passing reference to the Gaspe site by Ed Cook (who collected the data) is in Sheppard and Cook, 1988. Scientific Value of Trees in Old-Growth Natural Areas, Natural Areas Journal 8, 8. There he says:

    Northern white-cedar is another species used in eastern dendrochronology. Unlike eastern hemlock, little of the northern white cedar range has been sampled by dendrochronologists. This laboratory’s only northern white cedar chronology, form Gaspe, Canada, shows a peculiar “ramp” of increasing relative growth for the last 130 years (See Figure 3)….Any short-term departure from [ a long term mean of 1] such as shown in this chronology from 1850 to the present warrants special attention. Does that “ramp” of increasing indices truly indicate a climate change lasting 130 years? Or is it that some influence other than climate change is affecting northern white cedar growth in Gaspe Peninsula? Some hypotheses exist, based on tree physiology, about possible causes of such a trend, but since only one northern white cedar chronology exists, other basic questions remain. Do stands of northern white cedar from other areas show similar trends of increasing relative growth? Are their climate responses similar? Also do other species in the Gaspe Peninsula show a similar growth trend? Until other northern white cedar chronologies are analyzed, we are currently analyzing cores from two other sites in Maine and Michigan), these basic questions cannot be answered.

    Sensible comments and good questions – they are the sort of questions that reasonable people should be able to ask. Some time after the above comment was written – and before the “basic questions” were answered- , Jacoby used the Gaspe series in his treeline composite and the door got shut on discussion and analysis.

    At present, there are only 3 eastern white cedar chronologies in the ITRDB data base, Gaspe (cana036), Sag Pond ME and Duparquet, Quebec. Ed Cook collected Sag Pond, so I presume that this is the “Maine site” referred to above; but nothing further has bee reported on the Michigan site. (IF there is any info on Sag Pond ME – maybe a CA reader knows this location – , that might be a clue as to what Cook was sampling in Gaspe). The chronologies below obviously do not replicate the Gaspe pattern, but this negative answer to the “basic questions” asked by Cook did not affect the use of this data in multiproxy studies.

  23. Steve McIntyre
    Posted Mar 23, 2007 at 8:01 AM | Permalink

    BTW there is a fairly correlation 0.29 between the me022 and cana036 chronologies, despite their very different appearance.

  24. Peter Hartley
    Posted Mar 23, 2007 at 8:10 AM | Permalink

    I thought that a critical issue with some of the studies — particularly those pooling series from widely separated locations — is that the series were not chosen on the basis of correlation with locally measured temperatures but instead by correlation with a wider “global temperature field”. If the purpose of the exercise is to extend local instrumental records back in time, it seems to me that choosing series ex-post on the basis of high correlation with local measured temperatures is perfectly legitimate. The problem arises when the series are chosen ex-post to construct a “global temperature field”, the time series properties of which have been pre-defined by the researcher on subjective grounds, or perhaps computer model output. We even end up in the ludicrous situation where series are chosen because they have a high correlation with the “global temperature field” even though their correlation with actual local measured temperatures is very low.

  25. Steve McIntyre
    Posted Mar 23, 2007 at 8:34 AM | Permalink

    Google Earth wasn’t available when I tried to decode Gaspe informatino back in 2004. If you go to the stated lat-longs for the Ste Anne site cana036, within half a degree E-W and N-S of the coordinates reported to the nearest minute, the elevations are as low as 827 feet in the river valley and elsewhere about 900-1200 feet. Just browsing green areas of the Gape peninsula, it looks like the altitudinal tree line is easily over 2200 feet. I haven’t seen evidence elsewhere of white cedar being a treeline species, but haven’t examined this nuance in detail. Maybe a reader knows.

  26. Steve McIntyre
    Posted Mar 23, 2007 at 8:43 AM | Permalink

    #24. This is very much an issue with the Gaspe chronology, which has negligible correlation with gridcell temperatures, but is a type-case of a Jacoby-Mann “teleconnection”.

  27. Hans Erren
    Posted Mar 23, 2007 at 9:25 AM | Permalink

    Sag Pond, USGS Big Brook Lake (ME) Topo Map

    google map

  28. Steve Sadlov
    Posted Mar 23, 2007 at 9:33 AM | Permalink

    RE: #9 – Here out West, all you see near tree line are pines. There are no other conifers up that high. Our cedars are low to mid elevation trees. Same deal for firs, spruce and of course redwoods (with the slight exception of Sequoia Gigantum, which occur up to within 1500 feet of tree line at the upper most extreme, but still, not truly near it).

  29. Douglas Fox
    Posted Mar 23, 2007 at 9:35 AM | Permalink

    The Silvics of North American Trees (link below) says that “Northern white-cedar grows from near sea level to more than 600 m (2,000 ft), but within most of its range it is found between 150 and 600 m (500 and 2,000 ft).”

    Note that NWC grows on quite a variety of sites, from acid to alkaline, bogs to well-drained. The cores are easy to read. I’m a forester, not a dendrochronologist, but I’d suggest from my experience with hundreds of cores that one should get help from a dc expert. I expect that dc has ways of dealing with all kinds of variables, but the change in reaction wood (cell and growth differences in response to physical pressure from wind and weight)over time as the tree grows would seem to be difficult. (Not really meant as a criticism of the field of dc; just a warning to amateurs with increment borers.)


  30. Steve McIntyre
    Posted Mar 23, 2007 at 9:38 AM | Permalink

    Sag Pond is at about 1400 feet and would hardly be a “treeline” site. Hans, do you know of a topo map like the one you found for Sag Pond for the area containing the Gaspe site? 48 35N 65 55W

  31. Jim Edwards
    Posted Mar 23, 2007 at 10:12 AM | Permalink

    #25, 30

    Re: Treelines

    Are you sure when dendro people mention treeline they’re talking about THE Treeline. If you think about it, there are actually many treelines: one for black oaks, sugar pines, noble firs, sequoias, etc. Each species will have it’s own treeline, based on that species ability to survive climate extremes which are principally controlled by a combination of latitude and altitude, but secondarily affected by the microclimactic factors: North v South face, shade, etc.

    THE Treeline near the Arctic circle or at high elevation is merely the treeline of the most cold-tolerant species – not necessarily the treeline of the species being sampled.

    Douglas Fox’s [#29] comment implies the treeline for NWC is about 2000 ft at the lower latitudes of its range, and presumably much lower at higher latitudes – where it is colder.

  32. Steve Sadlov
    Posted Mar 23, 2007 at 10:12 AM | Permalink

    RE: #30 – Steve M, I navigated a bit around the US – Canadian border area near by Sag Pond. I saw peaks above 1800′ but no where did I manage to encounter tiberline. In my business travels, I always try to break free for hiking. I’ve been to Western Maine. There you have serious elevation as spurs of the White Mountains are there. At that point, timberline has to be at least 3 thou if not higher, and that’s more of an inland climate than Gaspe so being further south would not raise the treeline by anything more than 5 to 600 feet versus Gaspe. Do that math from there. (Someday I plan to hike in Gaspe, what a wonderful area based on what I’ve heard!)

  33. Bill F
    Posted Mar 23, 2007 at 10:30 AM | Permalink

    Rob, I think you pointed out in your post exactly why the use of tree rings for paleo climate studies is so problematic. We have no experimental data to tell us which trees are responding to temperature and which are responding to other factors such as precipitation, sun exposure, CO2 fertilization, soil depth, etc. So the only way you can decide if a given tree ring series is responding to temperature is to compare it to an instrumental record to see if it correlates. However, if you are sampling multiple sites over a larger region and comparing to a single instrumental record, the same features you mentions (n-s slope, elevation, distance to treeline, etc) will bias each tree’s signal in ways that may make it not match the instrumental record. That doesn’t mean those trees aren’t responding to temperature…and as you said, we have no way to know what they ARE responding to. So what happens is instead of fidning trees with a clear temperature driven signal, you are finding trees with the clearest signal. In the case of Juckes, he has stated in responses to comments, that he doesn’t care if the signal matches the local record or not. Which begs the question of how he can tell if it is a temperature signal or not. So you are stuck when you start picking trees for “signal” characteristics, because if you only pick trees that match a local record, there is no guarantee that the local record has any relation to a “global signal”. But if you pick trees that don’t match the local record, but do match a “global signal”, then you can’t ever be sure that they are really responding to temperature and not some other climate feature such as precipitation.

    Given all of that, how can you say for sure than any given tree ring series…not just the ones Steve picked at random…have ANY reliably proven ability to record a distinct temperature signal without automatically biasing the results by only picking the ones that show such a signal? You state that picking the ones with the best signal improves the r2 values, but doesn’t that automatically involve inserting a bias by making the assumption that the signal “should” be there? What if reality is that the is no relationship between the series and question and temperature. Throwing out the ones that don’t have a signal doesn’t change that fact…it just erooneously makes it look like there is a relationship.

  34. Hans Erren
    Posted Mar 23, 2007 at 10:35 AM | Permalink

    re 30:


  35. Ken Fritsch
    Posted Mar 23, 2007 at 10:51 AM | Permalink

    I think RW is a really nice guy who is very comfortable in his job/profession — and it shows. I think it helps explain his and other scientists’, in this field, seemingly different reactions to MMs, i.e. the Michael and Mac flavors. Be that as it may, I need to thank him for his explanation for rationalizing the “cherry picking” selection methods used for tree ring proxies (it gives much insight into the problem), and particularly, for inducing the responses of Dave Dardinger, Steve M, UC and Spence UK which in effect describe in easily understandable terms why RW’s approach, limited to what was described here, is wrong.

  36. MarkW
    Posted Mar 23, 2007 at 11:36 AM | Permalink

    Let me see if I have this straight.

    The way to find what the global temperature signal is, is to find trees whose growth rate matches the global temperature signal?

  37. Steve Sadlov
    Posted Mar 23, 2007 at 1:05 PM | Permalink

    RE: #36 – The ones which “match” the signal (whatever that is) during the past 150 or fewer years. Then, extrapolate back 1500 years …. 😉

  38. Peter Hartley
    Posted Mar 23, 2007 at 1:10 PM | Permalink

    Re#36 Yes, and it doesn’t matter if the growth rate does NOT match the local temperature signal.

  39. Steve Sadlov
    Posted Mar 23, 2007 at 7:40 PM | Permalink

    Altitudinal timberline on Mt. Washington, just a bit west of the Maine – NH border (therefore subject to less marine moderating influence than Gaspe) and not-all-that-much-further south in latitude, is 4500 feet.

  40. Doug Fox
    Posted Mar 23, 2007 at 8:20 PM | Permalink

    The Mt Katahdin, Maine treeline is at about 3500 feet. (Balsam fir, and I believe, Black spruce.) Kind of embarrasing, but I can’t recall how far up we lose northern white cedar.

  41. Nathan Schmidt
    Posted Mar 23, 2007 at 11:01 PM | Permalink

    re: 30, 34

    The NRCAN Toporama link provided seems to show contours, but no labels showing elevations/contour intervals. There’s an alternate source at NRCAN that appears to link to much the same data, but includes elevation labels:


    Hope that helps; compliments on the site.

  42. Douglas Hoyt
    Posted Mar 24, 2007 at 5:40 AM | Permalink

    Halfway between Katahdan and Washington, is Little Jackson with a treeline at 3000 ft. Just a few miles away, Mt. Abrams has its treeline at 4000 ft.

  43. John Baltutis
    Posted Mar 24, 2007 at 2:42 PM | Permalink

    Re: #10

    Rob Wilson opines:

    From my experience in science, people do what they do because they have a passion. Monetary gain is simply not an issue.

    Maybe not. However, this 1996 comment (still relevant today), although pointed to global warming science (in general), is an issue and applies to all associated scientific work funded via government grants:

    The most important explanation as to why so much theoretical work in the development of climate models has been done during the last ten years is that the development of models sustains funding and secures jobs at research institutions.
    ‘€”Professor Aksel Wiin-Nielsen, former Secretary-General, UN World Meteorological Organization (cited in The Skeptical Environmentalist, Lomborg, pg. 37).

  44. Doug Fox
    Posted Mar 24, 2007 at 6:58 PM | Permalink

    I got out Fritts’ 1976 book Tree Rings and Climate today. If anyone can point me to a few good, recent sources of current dendrochron theory so I can do some homework and be more helpful, I’d appreciate it. In response to a few recent posts, here’s some forestry conventional wisdom (NB: CW can be mistaken) and a concern:
    ‘€¢ The actual treeline is determined primarily by wind, not temperature.
    ‘€¢ By some definitions, treeline includes krummholz–very twisted, low growing trees, usually black spruce or balsam fir in the east. I would think it impossible to get anything from these tree rings. I suspect dc takes cores from near the top, in the dense growth but before the krummholz.
    ‘€¢ Density (trees per hectare) determines diameter growth; soil conditions determine maximum height growth.
    ‘€¢ On poor soils, trees fail to differentiate crown classes and self thin. All the trees grow up slowly at a fairly uniform height and diameter. On decent soils you might start with 2000 trees per acre and end up with 800 after 80 years as some trees grow ahead of others thereby getting more sun and shading their competition. Near treeline, this doesn’t happen.
    Concern with DC: On conifers, the compression side of the wood (from a lean or wind) will exhibit faster growth and the tension side correspondingly slower growth. The total volume growth is probably the same. The thicker growth takes place exactly where an engineer would put it (cf. Claus Mattheck’s work), according to the axiom of uniform stress distribution whereby engineers and trees spread physical stress as uniformly as possible. This abnormally high stem growth from lateral loading will change as the tree grows: taller height can mean more “wind sail,” increasing the reaction wood; lower branch death from shading (the typical process), will reduce the windsail. Therefore, the growth ring width is determined by crown size, tree height,typical wind speed,overall stem diameter, temperature, moisture, and competition–some of which we know change each year and all of which are subject to change over time.

  45. Willis Eschenbach
    Posted Mar 24, 2007 at 7:15 PM | Permalink

    From my experience, Rob is right, monetary gain (getting rich) is not an issue for most scientists. But he misses the point. Continued funding is a huge issue for scientists, one that shapes the nature and direction of their research, and which, if the scientist is not careful, may shape their results as well.


  46. bender
    Posted Mar 25, 2007 at 8:40 AM | Permalink

    Ok, Steve M, let’s go. Somewhere around the first week of July? Rather than theorize about the altitudinal limit at which cedar is not found, let’s go look for ourselves. We’ll do an elevational transect of multiple species, cedar and spruce and whatever else we find at treeline. Maybe compare strip-bark cedar and normal cedar. We’ll do our own temperature reconstruction, and suggest Rob Wilson as a reviewer on the manuscript.

  47. Doug Fox
    Posted Mar 26, 2007 at 4:41 AM | Permalink

    Re: #46
    In case bender is serious: I’ll apply for the increment borer technician position. Steve, you have my email–say the word and I’ll send you my CV offline (and I’ll start practicing hitting the pith on cedars).

  48. Don Keiller
    Posted Mar 26, 2007 at 9:05 AM | Permalink

    RE Rob Wilson’s original comment;
    Location, location, location! or even ecology, ecology, ecology!”
    As a plant physiologist (PhD, Cambridge), I have to agree.
    Quite apart from all the confounding variables (moisture etc.) that can impact on annual growth ring width, we need to add CO2 concentration.
    Partial pressure of CO2 falls with altitude and I have done enough A/Ci curves (assimilation v internal CO2) to know that at lower than normal (sea level) partial pressures of CO2 that a plant’s response to increased CO2 is more or less linear.
    THe only way I can think to deconvolute this confounding variable is to do a comparison with a latitudinal tree line at sea level. (I’m sure the dendrchonologists will have done this simple test!)
    If both respond in a HS type way then I would concede that there may be a temperature signal.
    If, as I suspect, they do not then the whole edifice of tree-ring temperature proxies falls.

  49. fFreddy
    Posted Mar 26, 2007 at 11:15 AM | Permalink

    Re #48, Don Keiller
    Do you fancy writing a paper on the subject ?

  50. crmanriq
    Posted Mar 26, 2007 at 12:04 PM | Permalink

    In the “there are no stupid questions, just stupid people” category… As a layman (neither mathematician nor climatologist, though I am in mining and really really know what it’s like to have to be able to repeat results), what comes to mind in reading about selection of trees and treelines is that it seems that if the selections of trees on the treeline is so important, wouldn’t you expect the treeline to shift as the controlling variables change? (ie. if temperature is rising with time, wouldn’t you expect to find the treeline moving to higher altitudes with time? or if CO2 is rising with time, wouldn’t you expect the treeline to move upward as well?)

    And if the treeline is shifting, then it would seem that a tree that you sample that is on the treeline now, probably wasn’t for most of its growth. Does this mean that the tree earlier in its growth was growing at a slower rate or a faster rate?

    One other question: How is changing population density controlled for in dendrochronology. As high-density tree populations self-thin, I would expect that as the population self-thinned, the surviving trees would have less local competition for light and soil and would be able to grow at a faster rate. Or is this something well documented and controlled for?

  51. Doug Fox
    Posted Mar 26, 2007 at 8:11 PM | Permalink

    If there is still interest in the Graybill and Idso stripped bark v. whole tree divergence in growth, let me know. I have comments, but you folks might be beyond that discussion now.

  52. John Baltutis
    Posted Mar 27, 2007 at 1:39 AM | Permalink

    Re: #51

    If you think it’s suitable, post away; it’s your stuff and only you can decide. No one here has moved on!

  53. Don Keiller
    Posted Mar 27, 2007 at 2:28 AM | Permalink

    RE #50 All these are highly pertinant questions that I, as a plant physiologist, would like to see answered before any conclusions regarding tree rings and temperature were made.
    I may be wrong, but I haven’t seen this kind of rigour in the dendrochonologists work to date.
    Any dendro’s out there, please correct me (and direct me to the relevant primary literature if I am mistaken)

  54. Don Keiller
    Posted Mar 27, 2007 at 2:33 AM | Permalink

    RE #49 I’d love to. Once I have finished rewriting several degree courses (hopefully this summer), I’ll get round to writing a grant application. Coded, of course, with a suitable title
    such as “Investigation of climate change by assessing photosynthetic responses of trees growing at their altitudinal and latitudinal limits”

  55. fFreddy
    Posted Mar 27, 2007 at 4:03 AM | Permalink

    Re #54, Don Keiller
    Excellent, that would be a useful paper to have on the record.

    You know, everyone keeps saying that science these days is all about “publish or perish”. Surely there must be an awful lot of “low hanging fruit” papers like this that could be written in the real scientific journals, pointing out the flaws in the climatology nonsense.

  56. fFreddy
    Posted Mar 27, 2007 at 4:29 AM | Permalink

    Re #48, Don Keiller

    Partial pressure of CO2 falls with altitude

    Dumb question :
    If I understand it right, a plant holds its stomata open to get the CO2 it needs for photosynthesis, and pays the price of losing water to the atmosphere.
    All other things being equal, what happens to the relative partial pressures of CO2 and water with altitude ?

  57. Doug Fox
    Posted Mar 27, 2007 at 6:39 PM | Permalink

    I’m feeling a bit like a charlatan with my posts, given that I need to do much more homework on the dendro issues. No clear answers in this post but a few thoughts…

    Notes on Graybill and Idso

    Why the difference in carbon dioxide fertilization response in whole tree v. strip bark trees? (SB trees grew faster in diameter)

    1. I agree with others that the strip bark phenomenon is likely caused by root death, and the root death is likely caused by physical injury by rocky substrate. (cf. wave regeneration in fir)

    2. Root death will cause cambial death above it.

    3. When tree roots are killed, typically C allocation goes to the remaining roots for added growth, at the expense of top growth, until the root equivalent is recovered. Even the stem is bypassed–stem growth suffers until the roots recover. This is true even in situations where water and nutrients are supposedly not limited. (Similarly, cut the top back on a tree and root growth with suffer until the crown has sufficiently recovered.) [Still, I wonder what wood happen if there were no place for new root growth to go’€”under a low or no soil condition.]

    4. Woundwood (the wood that grows after injury, typically growing over and eventually closing the wound) grows more rapidly than non-woundwood, but the total biomass increment is the same. Thicker wood around the wound is countered by less wood laid down elsewhere in the growth ring. Conventional wisdom regarding the mechanism includes two notions: first, less cambial tissue means the existing, limited number of cambial cells will get more carbohydrate and therefore grow faster. For example, tie a rope around a tree and the trunk above the rope will grow to be larger than the trunk below. Phloem flow is cut off and carbohydrates pool above the rope. Second, the tree responds to growth stresses by placing the maximum amount of growth at the physical stress points. For example, cut holes with a hole saw in a vertical line on a trunk, then bend the tree. The growth will occur as predicted by engineers based on “notch stress.” On high diameter trees with little bending stress, this is probably not an issue.

    5. Conclusion: On the one hand, I would predict that the stripped trees would grow slower in diameter because of their corresponding loss in roots (C allocation going to extending existing roots). On the other hand, I’d predict faster stem radial growth on the stripped trees due to photosynthate feeding fewer cells. A theory that would bring both together: root death occurred several years before the stem diameter increase, perhaps accompanied by a radial growth slowdown. When the roots recovered, the stem woundwood go back to its original allocation but fed these carbs to a smaller cambium.

    6. I need to peruse some study of carbon allocation foll0wing carbon dioxide fertilization to see if it makes a difference. All I remember from the literature is that CO2 fert favors production of secondary compounds (nonstructural) in some plants. BTW, carbohydrate partitioning/allocation results can be misleading depending on whether or not they include fine roots and whether the roots form mycorrhizae.

    7. To believe Graybill and Idso on reproductive allocation, I would need to see how their trees reacted to stress. Some trees actually reproduce heavier under stress.

  58. Bob Weber
    Posted Mar 27, 2007 at 7:06 PM | Permalink

    #53. You might take a look at these 2 references.
    K. Briffa brief at
    aric’s Global Climate Change Student Guide


  59. bender
    Posted Mar 27, 2007 at 7:24 PM | Permalink

    Re #47.
    #46 was serious. Therefore your offer of employment is accepted. Now we need to hear from the white hat, Mr M.

  60. Steve McIntyre
    Posted Mar 27, 2007 at 7:36 PM | Permalink

    #59. I’m in. Didn’t you get my email?

  61. Don Keiller
    Posted Mar 28, 2007 at 3:39 AM | Permalink

    re #58, yes I am aware of this. But it is all based on correlations. Correlations do not prove cause and effect.
    Im any case there is an increasing descrepency between tree ring width and “temperature” over the latter half of the 20th century, so the correlation breaks down.
    CO2 is undoubtedly a significant confounding factor -(eg Telewski et al (1999). Wood properties and ring width responses to long-term atmospheric CO2 enrichment in field-grown loblolly pine (Pinus taeda L.). Plant, Cell and Environment 22: 213-219.
    The fact of the matter is there is not enough experimental data available to underpin and justify the assumptions that are being made about the relationship of tree rings and temperature.

  62. Don Keiller
    Posted Mar 28, 2007 at 3:48 AM | Permalink

    re 56, you are right stomates open up at low [CO2]- and in doing so decrease the leaf’s water use efficiency as the water diffusion pathway is the same for CO2.
    The ultimate limiting factor, however, is not a stomatal limitation, but that of substrate availability- (in this case [CO2]) at the active site of RUBP carboxlyase.

  63. Doug Fox
    Posted Mar 28, 2007 at 8:11 AM | Permalink

    Another “if I could ask a dendrochronologist…” question:

    Trees grow most rapidly at the base of the live crown. The base moves higher over time due branch death generally attibuted to loss of hydraulic capacity on long old limbs and/or to a reduced photosynthesis to dynamic mass ratio (living tissue on large limbs is too great for the amount of light captured on shaded leaves and storm-damaged leave-bearing twigs).

    Fritts (1976)indicates that the first sentence is understood, but is this taken into account on these studies? I have dated branch death on some trees but I haven’t tried cedar yet. (It is much easier on dissected trees than with cores.)

    Is there a 12-step group for those of us who are not retired and need to spend more time away from this site?

  64. Steve McIntyre
    Posted Mar 28, 2007 at 8:15 AM | Permalink

    #63. You’ll have to ask TCO who has returned today after a year’s absence. Hi, TCO.

  65. Doug Fox
    Posted Mar 29, 2007 at 10:00 AM | Permalink

    “If I could ask a dendrochronologist” part III:

    Transverse sections of xylem generally reveal at least three zones: 1) active transport and storage of water and carbohydrates, 2) inactive water transport (plugged vessels or tracheids) but active carbohydrate storage, 3) heartwood (in heartwood forming trees), 4) either instead of or around heartwood, an inactive, dry, low nitrogen, protection wood formed in response to death of several lower branches. Within the first zone, at least, there is a moisture gradient with the latest ring containing the highest m.c.

    Particularly in regard to moisture content, does this affect dendrochronology and how so? (Would recent growth be inflated due to high moisture content?)

  66. DeWitt Payne
    Posted Mar 31, 2007 at 1:15 AM | Permalink

    Re: #56

    All other things being equal, what happens to the relative partial pressures of CO2 and water with altitude ?

    Assuming that CO2 is well-mixed, which I believe is the normal situation outside urban areas and other local sources, the ratio of partial pressure of CO2 to total pressure should remain constant. Water vapor pressure, OTOH, decreases faster than that because the vapor pressure of water is also determined by temperature and temperature decreases with altitude. In other words, water vapor condenses to the liquid (or the solid if it’s cold enough), with clouds being a classic example. So the ratio of CO2 to water vapor should increase with altitude.

One Trackback

  1. […] A data set that was almost as controversial in MBH98 as the Graybill bristlecones was the Gaspé cedar chronology used by Jacoby and d’Arrigo. An interesting new cedar chronology from Quebec has just appeared at NCDC, shown below. The third chronology shown below is an unreported update to the Gaspé series. I reported the unreported update in a 2005 post (see comments by Martin Wilmking, a young dendro interested in the divergence issue). Also see 2007 discussion here. […]

%d bloggers like this: