Mann 2008: the Luterbacher Mystery

Jeff Id has identified another intriguing mystery in the arduous problem of determining what Mann’s realdata was. Jeff observed that the version of the Luterbacher lutannt10 series in Mann’s infilled data version allproxy1209 was different than the version of lutannt10 in allproxyoriginal (Sep 5 version), illustrating this as below:

I’ve verified this information; below is my replot of lutannt10 versions:

I also compared versions of this series in the 6 versions of Mann’s data archived over the past week (3 infilled versions, 2 distinct; 3 ‘original’ versions, 2 distinct). Two different versions of lutant10 occur in the 4 distinct versions. The version labelled “1209” above occurs in the Sep 4 allproxy1209, Sep 5 allproxy1209 and WDCP proxy-infilled data sets, while the version labelled “original” occurs in the Sep 4 allproxyoriginal, the Sep 5 allproxyoriginal and the WDCP proxy-original data versions i.e. one version occurs in all infilled versions and an inconsistent version appears in all “original” versions.

The Luterbacher data is not itself archived and so there is, at present, no way to verify which is the “original” ‘original’ realdata. (I think that Hans Erren tried to obtain Luterbacher’s data a couple of years and was unsuccessful.)

Whatever the result, I guess that we’re going to see still one more version of either the infilled or original data and maybe even get the real realdata. At that point, we won’t just have the “original” data; we’ll have a Team of original data.


  1. Steve McIntyre
    Posted Sep 19, 2008 at 10:22 PM | Permalink

    For anyone who doubts the inconsistency of the versions, here’s a script to read and plot the two inconsistent versions presently at Mann’s website:

    loc= file.path(“”,”allproxy1209″,”lutannt10.ppd”)
    Data1=read.table(loc); Data1=Data1[4:nrow(Data1),];Data1=ts(Data1[,2],start=Data1[1,1])

    loc= file.path(“”,”allproxyoriginal”,”lutannt10.ppd”)
    Data2=read.table(loc); Data2=Data2[4:nrow(Data2),];Data2=ts(Data2[,2],start=Data2[1,1])

  2. Pat Frank
    Posted Sep 20, 2008 at 12:46 AM | Permalink

    Why would Mann scrap a proxy that has a steep 20th century rise, and substitute a different version that’s ~flat after about 1960? That doesn’t seem consistent with Team strategy. The new version looks much less noisy than the original version as well, and doesn’t seem to wiggle-match very well anywhere. I’m wondering whether they’re not different versions of the same proxy after all, but rather different proxies entirely. Maybe the mystery would be solved it the original Lutannt10 proxy was misidentified as such.

  3. Posted Sep 20, 2008 at 8:36 AM | Permalink

    My software picked out more than 50 series which seem to be different. I haven’t tried to look if they were just mixed around with each other yet. This would explain the elimination of the high peak in favor of another. The fourth graph at the link below shows the count per year of variation in the data.

    In my more recent (and a little crazy) post, I discovered that the data used to extend the proxies had an amazing up slope in the last 50 years. This up slope was pasted to the end of 90% of the proxies prior to ‘calibration’.

    Data was added to the 90% of the proxies based on a specifically selected 55 proxies.

  4. anna v
    Posted Sep 20, 2008 at 8:45 AM | Permalink

    This digging into the consistency of the data is a real contribution to the effort of clearing the dross from climatology related science.

    This is an aside, but I wonder whether you are aware of it

    It is the first link I found with a yahoo search, but the names are there:

    Seventy degrees is a lovely, comfortable temperature for most people. And the same turns out to be true for all sorts of tree species. In fact, scientists have found that trees actually have tricks they use to keep their leaves close to that perfect temperature.

    Brent Helliker thought maybe he could tease out something useful about today’s environment from the samples Richter collected. He discovered that the leaves on these trees did most of their work — that is, capturing solar energy — at 70 degrees Fahrenheit. And it didn’t matter whether the trees were growing in the hothouse of the tropics or in the frigid North.

    Now if this study is not proven wrong, it at least confuses any proxy for temperature use of tree rings particularly from forests. It seems that the trees control their temperatures and collectively create a microclimate as far as temperature goes, which anybody who has walked in woods winter or summer is aware of.

    I am amazed the reviewers of this recent revival of the proxy tree ring method did not raise this point. I would think it would stop the industry in its tracks. Is this point addressed in the publication?

  5. Steve McIntyre
    Posted Sep 20, 2008 at 9:32 AM | Permalink

    Jeff, I looked at the discrepant proxies and, as far as I can tell, they are all in the Luterbacher section.

  6. Posted Sep 20, 2008 at 9:37 AM | Permalink

    I did a quick analysis to search the 86 series of Luterbacher to see if they were just mixed up.

    They are not! My software didn’t find the same or a similar series in any of the other data. So my possible explanation of inconsistency in labeling is not the answer.

    This data is different.

    I expanded the search to all the 1357 data series from the scrapped set and found no matches for any of the 71 Luterbacher series in the 1209 used data series.

  7. Ernie
    Posted Sep 20, 2008 at 9:53 AM | Permalink

    Is it possible to get hold of the original data from archive of the scientists who did the actual proxy research, and bypass Mann altogether so we can see what the graphs should look like? That way we have a reference point to know when he is tinkering with the data.

    – Ernie.

  8. deadwood
    Posted Sep 20, 2008 at 10:28 AM | Permalink

    Is these what Mann refers to as “pseudo-proxies”? I just can’t believe he has substituted data in the way this looks. Surely his reviewers would have been familiar with the data and have caught such a problem.

    Be that as it may. Are there not procedures at PNAS to dispute his findings? To write and publish a rebuttal?

    Steve: PSeudoproxies are completely different. PNAS permits comments for only 3 months after publication. They might try to run the clock by stonewalling information as much as possible.

  9. Steve McIntyre
    Posted Sep 20, 2008 at 10:32 AM | Permalink

    #7. That was one of the things that I did in MM2003, where we found such interesting things as Mann’s unreported “adjustment” of the Gaspe series to “get” it into the troublesome AD1400 network. This was the ONLY such extension frontwards of a series in the entire archive. From an accounting perspective, this had alarm bells all over it, but no one in climate science cared. I collected quite a few series, but could only get a portion of them.

    In the new data, many underlying series are at WDCP, but many are not. To my knowledge, the Briffa MXD data (called “SChweingruber” here) is available nowhere. I’ve sent an FOI to CRU in England but have received no response yet. They generally refuse FOI requests on one pretext or another but we keep trying and get occasional snippets.

    I haven’t been able to obtain Luterbacher data yet. A reader emailed me this morning saying it was KNMI, but I am trying to clarify that as I can’t get it yet.

    There are grey versions of data from Bradley’s associates e.g. speleothem data and that will be very hard to get.

  10. Steve McIntyre
    Posted Sep 20, 2008 at 10:37 AM | Permalink

    #6. Jeff, there’s one other possibility. In the WDCP infilled data set, there are 1355 series with the same core of 1209 series, but the other 146 series are 2×71 Luterbacher series (said to be summer, winter) and 4 log versions of the disturbed Finnish sediments. Why don’t you check and see if they’ve labelled a winter or summer Luterbacher series as annual?

  11. Hans Erren
    Posted Sep 20, 2008 at 1:36 PM | Permalink

    I discussed the summer temperature by Luterbacher at al. in
    originally published as
    Jürg Luterbacher, Daniel Dietrich, Elena Xoplaki, Martin Grosjean and Heinz Wanner, European Seasonal and Annual Temperature Variability, Trends, and Extremes Since 1500, Science, Vol 303, Issue 5663, 1499-1503 , 5 March 2004

    for which the results were put online as late as june 2006

    Luterbacher European seasonal average temperature reconstruction results can be downloaded here

    here is a comparison for luterbacher summer vs Chuine wine harvest dates

  12. Steve McIntyre
    Posted Sep 20, 2008 at 1:48 PM | Permalink

    Hans, I’m famliar with that data reference. Mann uses gridded versions, for 71 different gridcells. I can’t locate original archives for the gridded data.

  13. Posted Sep 20, 2008 at 2:14 PM | Permalink

    Ok, I have spent enough time on this. It sucks being a newbie. When you say gridcell, do you mean that the Luterbacher data was in individual gridcells? Mann used some averaging or RegEM between gridcells?

    I thought he did that for measured temperature only.

    I just ran the average of the ‘original’ deleted 89 luterbacher series and it matches the average of the new Luterbacher within .15 standard deviation at the worst point. It is a near perfect fit.

    I have chased this around my software for hours now. I kept asking how could they be the same average?

    Some enlightenment if you don’t mind.

  14. Steve McIntyre
    Posted Sep 20, 2008 at 2:18 PM | Permalink

    In temperature datasets, they divide the world into 5 degree gridcells with centers at say 62.5N, 12.5E. If you look at the information in the table SD1 in Mann’s SI, the Luterbacher series are all associated with different gridcells. However, we have no original Luter data to compare it to, so it’s hard to say what he’s using.

  15. Posted Sep 20, 2008 at 2:47 PM | Permalink

    So, the fun continues.

    I don’t know how you caught that. The Luterbacher data is on even .5 degree coordinates while the rest of the data isn’t.

    Very nice. There has to be some kind of interpolation going on.

    Well the average of the 89 series of the original original set is the nearly an exact match to the average for the 71 series from the final original set.

  16. Steve McIntyre
    Posted Sep 20, 2008 at 3:04 PM | Permalink

    the luter data is in 5 degree gridcells, not 0.5 deg gridcells. Luterbacher et al 2004 uses Mannian style methods to calculate gridded data, but it includes instrumental info, that’s why the correlations are so high. IT’s totally inappropriate for saying that 484 series are significant.

  17. Sam Urbinto
    Posted Sep 22, 2008 at 11:38 AM | Permalink

    AFAIK, the grid starting at 45N 65E would have a center of 47.5N 67.5E

    Steve: Yes. CRU gridcells are centered at the 2.5, 7.5 degree marks. GISS gridcells are different, but CRU are more common.

  18. Sam Urbinto
    Posted Sep 22, 2008 at 12:12 PM | Permalink

    Given the 5×5 cells at 0 are 5,408 km, 30 are 154,000 km, 60 are 267,000 km and 90 are 309,000 km, that would, combined with the coverage of stations in various areas, certainly add a large degree of complexity to the grid calcuations and importance.

    Where do the GISS gridcells center? IIRC you’d mentioned they do some kind of size equalization or something like that. Weighting or whatever. I forget.

%d bloggers like this: