A Cook's Tour

Cook et al. [2004] is a “reconstruction of past drought across North America from a network of climatically sensitive tree-ring data”. It uses 835 sites in North American regions overlapping Mann’s PC network. I thought that it would be interesting to compare the two networks. Of course nothing is straightforward when you’re dealing with the Hockey Team.

Cook et al. [2004] was published in Science in October – see reference below. There is an Internet presentation at WDCP here.

I had originally been drawn to this article by Mann’s reference to it in realclimate as support for his bizarre claim that you only needed to look at an RE statistic for statistical verification. This was a pretty strange article to cite as proof as Cook et al. [2004] specifically report on R2 (and other verification) statistics, including a long listing here . In the gridcell containing the bristlecone pines (35.5N, 117.5W), the verification statistics were high – R2-calibration 0.761, R2 (Verification) 0.634 , RE 0.558, CE 0.550.

I thought that it would be interesting to see how much overlap there was (if any) between the Cook sites (with their successful drought reconstructions) and the sites used in Mann’s PC calculations. Readers may be interested in how this sort of process works.

On Jan. 31, I wrote to Connie Woodhouse of NOAA, one of the co-authors (who had been one of the panel organizers who had accepted my AGU presentation), saying that I was interested in looking at the overlap with Mann’s list. She said that I would have to get the information from Ed Cook and mentioned:.

I’d imagine there is not much overlap at all with Mann et al. since we targetting primarily lower-elevation moisture sensitive species, and Mann et al. were intererested in the temperature signal which is typically (but not always) strongest in a different set of species.

I tried Cook again, but, consistent with current Hockey Team policy, he once again did not acknowledge my email. I wrote back to Connie Woodhouse complaining about this, but she said:

As co-author, it is not my responsibility to acquire these data.

I then tried Science, who had published the article, asking Richard Kerr, a staff writer who had interviewed me after publication of our GRL article. He replied that it was not his area, referring me to Steward Wills, the online editor, but mentioned in passing:

Requirements for archiving data seem to focus on the biological and genomic, or at least those areas with prominent public archives. Climatic and geophysical data archives exist, but Science doesn’t seem to notice.

I re-iterated my requests to Wills, who answered:

Dick Kerr is correct that our policy on deposition of data to public repositories has tended to focus on the life sciences. However, one of our published conditions of acceptance is that "any reasonable request for materials, methods, or data necessary to verify the conclusions of the experiments reported must be honored," which would seem to cover the situation you discuss herein.

Wills passed the request on to Brooks Hanson on Feb. 15, which was the last that I heard of it from them. I also included a request for Thompson’s unarchived data.

Meanwhile, I noticed that the results had also been published at the NOAA website, so I wrote to Bruce Bauer of WDCP, who has been exceptionally cooperative in the past. On March 11, Bauer replied:

Ed Cook sent me the full list of 835 sites used in the 2004 Science PDSI paper:
http://www.ncdc.noaa.gov/paleo/text/pdpubs.html. 602 of these are in the West as mapped in the supplementary materials and our website – looks like all those west of 93W. This is the full starting set – some are unfortunately not publicly archived, and others failed the screening he did at each grid point.

Needless to say, the file does not include any ITRDB codes, so matching the Cook sites in this list to ITRDB sites (and then to Mann or Jacoby) is a fairly painstaking process. I’m pretty efficient at this since I’ve spent the time to make an R database, collating WDCP data on site codes, locations, types, start, end, latitude, longitude, altitude, author, etc. I can do these comparisons in a semi-mechanized way. But there are always spelling mistakes and other defects in the lists, so in the end, you have to check each one individually. I’ve done a first cut concordance, matching 560 Cook sites to ITRDB codes, which I will post up if anyone is interested.

I’ve done some quick cross-comparisons of the lists to check the distinctness of the Cook sites to other lists:

  • all 11 Jacoby Northern Treeline sites appear to be in the Cook drought reconstruction, including the Gaspé site.
  • 44 of the 70 sites in Mann’s AD1400 North American network were in the drought reconstruction. This included a number of bristlecone and foxtail pine sites, although some bristlecone sites were not included.
  • over 120 of the 212 sites in Mann’s AD1750 network were in the drought reconstruction
  • Another very interesting point, about which I’ll discuss more later, is that 422 of these chronologies go past 1990 and 365 go past 1995. I’ve looked quickly at tree ring chronologies currently archived at WDCP and there are hundreds that go well into the 1990s. Thus, the claim that it is necessary for temperature reconstructions to rely on series ending in the 1970s looks pretty bogus. The problem for the Hockey Team seems to be that tree ring widths correlate better with drought than temperature.

    Cook et al. show the following interesting image, using the dreaded term "Medieval Warm Period" to refer to the period AD900-1300.
    Cook Sciecne I’ll be returning to this topic a little later.

    Cook, E.R., Woodhouse, C.A., Eakin, C.M., Meko, D.M., and Stahle, D.W. 2004. "Long-Term Aridity Changes in the Western United States". Science, Vol. 306, No. 5698, pp. 1015-1018, 5 November 2004.

    Cook, E.R., Meko, D.M., Stahle, D.W. and Cleaveland, M.K. 1999. "Drought reconstructions for the continental United States." Journal of Climate, 12:1145-1162.

    Woodhouse, C. and Overpeck, J.T. "2000 Years of Drought Variability in the Central United States." Bulletin of the American Meteorological Society, Vol. 79, No. 12, pp. 2693-2714.


    1. Michael Ballantine
      Posted Mar 16, 2005 at 11:20 PM | Permalink

      Sounds like Cook, Mann or both of them picked the wrong sites. What happens to their respective reports if all overlapping sites are discarded?
      And one small nit to pick, the last graph wipes out most of the last sentance. At least in Safari.

    2. Murray Duffin
      Posted Jul 15, 2005 at 8:02 AM | Permalink

      Steve, when do you expect to return to this topic. It seems like it could be a quite critical indicator. in the graphs above I’m surprised that the dark curve in the upper, long term graph doesn’t show a higher peak near 1940, corresponding to the detailed lower curve. Is this some artifact of the way smoothing was done? I am betting that if you can push the dendrochronology out 15 years, to about 1995, it will agree with the curve where you tacked the satellite record onto Moburg. Murray

      Steve: Nice of you to ask. It’s rather fallen off my radar screen. I really feel overwhelmed right now with topics. I also need to finish some journal articles. butI appreciate the reminder and thought. steve

    3. Murray Duffin
      Posted Jul 20, 2005 at 4:44 PM | Permalink

      Have you looked into the possibility of getting a couple of college students to do detail projects for you? It is clear that you need some way to augment what you can do yourself. Every time you turn over another rock a whole lot of anomalies crawl out. I surely admire your one man effort and productivity, but you now clearly need help to pursue the several interesting trails you have identified. I wish I could help, but I know almost nothing of statistics. Murray

    4. TCO
      Posted Aug 26, 2005 at 9:50 PM | Permalink

      1. How does this post (or the article or research) show that drought correlates better with trees than temp?

      2. You state that much data does run past 1980. Do you no longer feel additional feild studies are warranted?

      3. To what extent has careful observation (or even direct experimentation) on trees been able to determine the multiple regression style response equation to sun, temp, rain, etc. etc? It would seem possible to do this type of experiment and get some really good intutions to baseline what mean and SD of tree ring size or density or whatever is expected. Is this baseline setting work well done and agreed on?

    5. Steve McIntyre
      Posted Aug 26, 2005 at 10:05 PM | Permalink

      1) As I recall, about half the trees in the MBH dataset as temperature indicators also occur in this dataset as drought indicators. Lots of tree rings correlate better with drought than temperature. The question was what they were doing in the MBH network in the first place.

      2) The post-1980 data hasn’t been integrated and much of it isn’t archived. Jacoby has done nearly 100 North American series since 1990 and (I’m pretty sure has) archived only one. I don’t think that Esper has archived anything. Hughes has not archived any bristlecone updates. I feel certain that, if the Sheep Mountain bristlecone samples taken in 2002, were showing off the charts results, then we’d have heard about it.

      3) I think that botanists are a long way away from being able to predict tree rings or even make a good stochastic model. However, there’s solid evidence that conifers have an upside-down U (quadratic) response to temperature in botanical studies; it’s only dendrochronologists that try to distort this into a linear relationship to temperature.

    6. TCO
      Posted Aug 26, 2005 at 10:20 PM | Permalink

      1. Meaning based on situations where the drought or temperature was independantly measured? So what were the differences in correlation coefficient or rsq?

      2. One way to deal with this is to urge additional field studies. One is to try to get the data unlocked. What’s your vote?

      3. So is this kind of baseline work done to allow the proxy work itself? What is it’s quality. If I would think one could do some pretty decent measurements and even controlled experiments here.

      4. So the response is U-shaped, do the proxy guys refer to this and then use a “U-transform” to extract the implication? Do they take efforts to deconfound the effects of other variables (rain etc.), perhaps by referring to studies done on rain sensative species?

    7. Steve Sadlov
      Posted Nov 14, 2006 at 10:33 AM | Permalink


    One Trackback

    1. By The Trouet Ocean Proxies « Climate Audit on Sep 11, 2010 at 10:34 AM

      […] of an RE statistic – an untrue claim, since Cook claimed verification RE and CE as well. Also here in connection with replication issues, but I haven’t considered recently. I’ve […]

    %d bloggers like this: