The Updated Gaspé Series

We have focused considerable attention on the role of the Gaspé series in MBH98. While we alluded previously in MM05 (E&E) to information in our possession that the updated (1991) version of the series differed from the 1982 version and to our frustration that the updated version remains unarchived (see Jacoby #1) , we are now able for the first time to show the updated Gaspé site chronology and perhaps add additional flavor to our concerns over the impact of the use of the obsolete version.
Updated Gaspé Version

Figure 1. Gaspé Site Chronologies. Purple (Ed Gas) – the 1982 version used in MBH98; Black (Gas) – the unreported and unarchived 1991 version.

In MM05 (E&E), we showed that the hockey-stick shaped Gaspé series was one of two influential outliers (together with the North American PC1 based on bristlecone pines) that caused the difference between MBH and MM-type results in the early 15th century. We showed that the unreported extrapolation of this series in MBH98 had a material impact on the reconstruction and that this extrapolation was concealed (perhaps unintentionally) by a misrepresentation of the start date of this series.

The impact of this series is compounded, because, in MBH99, the Gaspé series contributes to the Northern Treeline composite used for the MBH99 "adjustment" of the North American PC1. In the first part of the 15th century, the Gaspé series is the only contributor to Mann’s composite (including portions where there is only 1 or 2 trees) and undoubtedly contributes disproportionately to the MBH99 conclusion of "similarity" between the two series.

In passing, and I’ll post this on the Bristlecone Pine Adjustment theme as well, Jacoby begins his temperature reconstruction only in 1601, while Mann’s composite of the Jacoby versions (used for "adjustment") begins in 1400. But what bugged me more than the use of 1-2 trees is the withholding of the updated version. In the updated (1991) Gasp” chronology, there is obviously no hockey stick shape at all. However, Jacoby et al. are "mission oriented" and decided that they would not archive the new data, even though it had been funded with public money. Their argument is presumably that there’s something wrong with the new data – let them argue that and show it.


  1. Posted Feb 12, 2005 at 12:39 AM | Permalink

    This is really astonishing news.

  2. Posted Mar 10, 2005 at 2:13 PM | Permalink

    How about modern ice core isotopes. Deuterium ratio is supposed to be directly linked to temperature. How about the comparison with the records of nearby McMurdo weather station. A staggering 15% correlation.

    This is Talos Dome on Antarctica data 1200 – 1996. Almost suitable as cue for pool?

    Stenni, B, et al 2002, Eight centuries of volcanic signal and
    climate change at Talos Dome (East Antarctica),
    Journal of Geophysical Research,
    vol, 107, no, D9, 10,1029/2000JD000317,

  3. TCO
    Posted Aug 26, 2005 at 10:28 PM | Permalink

    This sucks. Any explanation more? How did you get the data?

  4. Martin Wilmking
    Posted Sep 25, 2005 at 8:42 AM | Permalink

    Well, that is a nice diagram fitting well our recently published results on diverging growth trends (Geophysical Research Letters) in circumpolar boreal forests. is this gaspe update available now at the international tree ring data bank?

    the blue line might well be individual trees which show growth reduction because of moisture stress or other factors. should be relatively easy to check.

    how many trees are in each chronology? it might well be that averaging all trees at that site is not the proper method to extract climate information for neither the “old” series, pink, nor the “new” series, blue.

  5. TCO
    Posted Sep 25, 2005 at 10:26 AM | Permalink

    how did you get the data.

    Please stop publishing anything meaningful in EE. Put your stuff in the real journals. There are enough of them that you can still be in the regular abstracted, looked at by the field, journals; even if you are blackballed a bit by the hockeyteam from Nature or even GRL (at times).

  6. Steve McIntyre
    Posted Sep 25, 2005 at 11:59 AM | Permalink

    I felt obliged to give EE a second article. They asked me to write the 2003 article and took a lot of (undeserved) heat. I figured that we owed them a followup article, even though there was an exposure price. We’re square now.

  7. Steve McIntyre
    Posted Sep 25, 2005 at 12:46 PM | Permalink

    Hello, Martin, thanks for dropping in. I had a brief post on Wilmking et al here

    The Jacoby data is not archived. (He’s archived virtually no North American data taken after 1991.) It was a bit accidental that I came into possession of this graphic. I requested a copy of the measurement data unsuccessfully. When I tried to find out further information about the new version, they got all shirty and said that I should look at the old data as it was more "temperature sensitive". If you go to the right bar category -Jacoby, you will see some of my frustrating efforts with Jacoby.

    Ross McKitrick and I met with cedar specialists at the University of Guelph (Larson, Kelly) who were very dubious (to say the least) about whether cedars could be used as a temperature proxy. They’ve studied hundreds of cedars for over a decade and said that they like cool moist conditions – an upside-down U.

    As to checking site conditions, I asked JAcoby for the location of the Gaspe site. He said that they did not know whether the original trees were sampled as it was done pre-GPS (unconvincing to someone who’s worked with geologists) and wouldn’t say where the second sample was taken place. I was going to get someone to check the site. Pretty frustrating. Cheers, Steve

  8. Martin Wilmking
    Posted Sep 25, 2005 at 1:17 PM | Permalink

    Thanks for getting back to this, but what data is actually behind that graph? any information on how many trees contributed to it? as i can see from the very limited information on the graph, those two curves are not necessarily contradictory. they might be from the same site, same species, but different individuals reacting differently to similar climatic conditions. you mentioned, cedars seems to like cool moist conditions, the blue curve might just show that. its not so cool and moist. of course that leaves the pink line, and i agree it would need some more investigation. possible that those individuals can take advantage of microsite conditions (moist sites).

    to answer a question like that it would be necessary to how many trees contributed to each curve. by the way is this data detrended? or is the y axis mm?

  9. Steve McIntyre
    Posted Sep 25, 2005 at 1:51 PM | Permalink

    You’ll have to ask Gordon Jacoby. Their answers to me were very unsatisfactory. There are 29 cores from (by memory) about 15-20 trees in the original 1982 series (archived as cana036.)

    You’ll probably have more luck than me. (They completely stonewall me.)

    At the time, I was interested in the series primarily because of its extraordinary impact on MBH98 results. It was the ONLY series (out of over 400) which Mann extrapolated at the beginning. He did this to get it into his 15th century reconstruction, where it brought down the 15th century result significantly. I was criticla of the extrapolation, the quality of the 15h century portion (1 tree for many years), the invalidity of the cedars as temperature proxy, the failure to disclose the extrapolation, the misrepresentation of the start date of the series in the article so that no one (without crosschecking as I did against original data) would have known about the unique extrapolation.

    This series and the bristlecones are what drive MBH98 in the 15th century reconstruction.

    The y-axis is dimensionless units in a usual chronology (1 instead of 1000). I’m 99.99% sure that the pink series was created using ARSTAN with negative exponential options. I’ve replicated this calculation as a check on some of the software that I’ve written in R to do site chronologies (I’ve experimented a lot on this)- actually there are some interesting numerical results.

    The convergence of negative exponentials is sensitive to parameters in the numerical methodology. Using ARSTAN, 16 of 29 series converge (with 12 modeling as mean and one a negatively sloped line). If you change the numerical convergence parameters, I could get 27 of 29 to converge. There is a really odd error in ARSTAN whereby the parameters for the first series are always wrong. I have no idea why.

%d bloggers like this: