Tasmanian Tree Rings

Update: The rwl data at WDCP is a mixture of measurement formats with one set of measurements in units 1/10 of the other set. This accounts for the bimodal distribution. These graphs need to be redone. End Update 2006] .

Thinking about some of these tree ring sites reminds me of some stuff I did on Cook’s Tasmania while I was learning the tree ring data, while I was first looking at Mann. Cook is on the Hockey Team, on the same line as Jacoby. Cook’s Tasmania reconstruction is a staple of nearly every multiproxy study: Mann, Jones, Crowley, etc. Here are some graphics that I did a couple of years ago, which I’ve never quite figured out what to do with. Here’s a histogram of all ring widths at Lake Johnston (there are over 170,000 measurements in this figure). A simple histogram is usually the first thing that a statistician would do, but I’ve never seen a simple histogram like this in any tree ring study. Yet it says volumes.

Figure 1. Histogram of All Ring Widths at Lake Johnston, Tasmania

Most histograms that I’ve drawn have a unimodal distribution – I think that they look like a type of gamma-distribution. But this one — a very long series back to –2136 — has a unmistakable bimodal distribution. It was a fluke that I noticed it, since it’s a very small bump in the default histogram plot, but it obviously sharpens up when by specifying the breaks as I’ve done here. It looks like a mixture of two gamma distributions, since the "main" mode is also more like a gamma distribution than a normal distribution. (I have no idea why tree rings chronologies assume normal distributions when they never are.)

The left mode is not an artifact since the dataset has about 170,000 individual measurements and the left mode has about 30,000 measurements by itself — so it’s statistically there. Another plot that I do on these sites( which again I’ve never seen elsewhere) is simply plotting the cumulative ring width "rooting" each tree in its assigned start date. Here’s what this looks like for the Tasmania site: you certainly don’t get the sense that the modern trees are exceptional. There are a couple of super-giant trees in the Greco-Roman period.

Figure 2. Lake Johnston, Tasmania – "Rooted" cumulative ring width by tree.

The scale of this is squashed by accommodating the super-giant trees, so I zoomed in a bit, with very interesting results as shown below. You can see that the slow-growing trees are concentrated in the millennium just ended.

Figure 3. Lake Johnston, Tasmania – "Rooted" cumulative ring width by tree – zoom.

I illustrated this by plotting first the distribution of "narrow-ring" trees and then the distribution of "wide-ring" trees as below. As you can see, the narrow-ring trees are "modern" phenomenon. They seem to start about half way through the first millennium. Interestingly, the earliest layers in the Quelccaya glacier (the recession of which is comlained about a lot) are dated as starting about AD450 – around the same time as the commencement of the "narrow ring" trees here. At the time I wondered if there was some connection.

Figure 4. Lake Johnston, Tasmania -Distribution of "narrow-ring" ring widths

Figure 5 . Lake Johnston, Tasmania – Distribution of "narrow-ring" ring widths

There were some observations in the article articles about this stand of huon pine being rather a surprise, since it was found quite a bit above the elevation where huon pine presently grows [I’ll add in the exact quote later].

Since this was a Hockey Team study, they concluded that the 20th century was unique in some or another [again I’ll add in the quote]. [I’ll add in Cook’s temperature reconstruction so re-vist this if you’re interested].


  1. Armand MacMurray
    Posted Sep 5, 2005 at 6:53 PM | Permalink

    A quick question about units from a tree-ring novice. It was my impression that typical ring widths are around 1-10mm, yet the titles of Figs 4 and 5 imply sizes perhaps 100x wider. Could the titles of Fig 4 and 5 (and some of the other numbers shown) be missing a decimal point, or be in units other than mm?

  2. Steve McIntyre
    Posted Sep 5, 2005 at 7:43 PM | Permalink

    I presume that the units are 0.01 mm here. These are slow growing trees usually.

  3. Ed Snack
    Posted Sep 5, 2005 at 8:05 PM | Permalink

    As I recall this study also used potentially inappropriate local temperature records for calibration, including Low Head where there is a non-temperature artifact in the recent record. Worth a quick check on the calibration side as well maybe ?

    But intriguing looking data, and will follow with interest any explanations.

  4. Paul Gosling
    Posted Sep 6, 2005 at 3:19 AM | Permalink


    Looking at figures 4 and 5. Are these percent of total for that time period, or percent of all rings. If the former then it is a very small proportion of the sample. Should we base much emphasis on them?

  5. Steve McIntyre
    Posted Sep 6, 2005 at 4:21 AM | Permalink

    If you look at Figure 3, you see that “narrow ring” trees make up the vast majority of the modern samples.

    The graphs are a histogram of “narrow rings” and “wide rings” by 50-year bins – sorry not to make that clearer. I was just experimenting with quantifying in some way the idea that the narrow-ring “regime” seemed to be modern and some better representation might spring to mind.

    I sent this information to Cook before I became notorious but received no reply.

  6. Louis Hissink
    Posted Sep 6, 2005 at 4:28 AM | Permalink


    What you have in your original histogram are two lognormal distributions and any stats analyses should be performed on the logarthims of the values rather than the values themselves. And that is something geologists are very, very conversant with.

    There is a lot of literature on this, texts and peer reviewed papers, so I won’t get too detailed.

    Lognormal because in earth science we cannot never have readings which are

  7. Louis Hissink
    Posted Sep 6, 2005 at 4:33 AM | Permalink

    hmm, post got truncated, but in geoscience there are no ‘negative’ values – so all distributions are skewed and best approximated by the lognormal distribution.

    If most of the climate data are so characterised, then the whole house of cards that is AGW should fall.

  8. Paul Gosling
    Posted Sep 6, 2005 at 6:40 AM | Permalink

    Does anyone else think that the end (last 500 year) sections of figures 4 and 5 have a hockey stick shape, or am I just imagining it?

  9. beng
    Posted Sep 6, 2005 at 10:22 AM | Permalink

    Assuming the Tasmanian series deals w/essentially the same tree species or mix of species over the entire record length, the species seems to have two growth “modes” — wide & narrow-ring (krummholtz vs upright forms perhaps?). For whatever reasons, the narrow-ring mode has become increasing prevalent since ~1000, almost exclusive now. My first guess (less growth = “worse” conditions) would be harsher conditions in the last 1000 yrs.

    If the above assumption is wrong, it could also be the prevalence of narrow-ring species for the last ~1000 yrs & wide-ring species earlier.

  10. Steve McIntyre
    Posted Sep 6, 2005 at 10:37 AM | Permalink

    Unfortunately the publication of these sites is typically about 4 pages long with minimal ecological data.

  11. TCO
    Posted Sep 6, 2005 at 2:45 PM | Permalink


    1. The histogram result is quite interesting and shows value of a basic high school analysis of data.

    2. The “hair” curve just looks cool!!

    3. I think you ought to publish some of this stuff. Both to share the immediate insights (give to the community) as well as to encourage others to look for more insights in their data. I think this should be done even if it has no relation to your agenda of discrediting temp reconstructions.

    4. These are interesting features. what do they mean? What do you suspect/wonder? Even if you don’t know, publish anyway, so people can start looking at these features and thinking about what this means (either from a methodology artificat arena or some sort of bimodal expressions in plants).

  12. Steve McIntyre
    Posted Sep 6, 2005 at 2:58 PM | Permalink

    I’ve done lots of these hair diagrams. I’ve called them “grass” diagrams. I’ve actually got the makings of a methodology for doing tree ring statistics in a proper way, using proper statistics instead of the ad hoc dendro methodologies which lack statistical diagnostics.

  13. Posted Sep 6, 2005 at 3:10 PM | Permalink

    Could evolution be a cause, after all the big assumption in all proxy analysis is that the genotype is identical throughout the series.

    Tasmania is an island and only the new individuals show the trend.

  14. TCO
    Posted Sep 6, 2005 at 6:34 PM | Permalink

    Where have you published your grass charts?

  15. TCO
    Posted Sep 6, 2005 at 6:37 PM | Permalink

    What are you preparing a magnum opus, Steve? Publish as you go with LPU (least publishable unit heee heee). Serious…sorta…don’t have to have it all figured out. Just start sharing some insights. Then rest of the community will jump in and help. That’s science. Unless, you are worried about losing credit? But you still have a lot to show before others join in.

    Are you just so wrapped around the fighting global warming thing, that you can’t publish basic science?

  16. TCO
    Posted Sep 6, 2005 at 6:41 PM | Permalink

    oh and shouldn’t narrow rings mean cold temps recently?

  17. Ross McNaughton
    Posted Sep 6, 2005 at 7:54 PM | Permalink

    Can I point you to some comments by the late John Daly on Cook and his use of the Huon Pines:
    John Daly – Huon Pine

    I will also point out these comments on the species from the Tasmanian department of primary industry:

    Although extremely slow growing, the tree may attain heights of over 40 m. Growth rates average a mere 1mm per year, but can vary from 0.3 mm to 2 mm, depending on conditions. Huon pine can reproduce both vegetatively (from fallen individuals) and by seed. Seed dispersal is largely limited to the area downstream from riverine stands.

    The Huon pine can reach prodigious ages, often in excess of 2000 years, making it among the longest-lived organisms on Earth. Only the bristle-cone pine of North America exceeds it in age. International headlines were made with the discovery of a stand of Huon pines on Mt Read that was widely quoted as being in excess of 10 000 years of age. All the individuals in this population are genetically identical, and are all males. The stand arose from one or a small number of individuals, and has maintained itself by vegetative reproduction. It is important to remember that no individual tree in the Mt Read stand is 10 000 years old — rather, the stand itself has been in existence for that long.

    So depending on which trees were sampled, genetically they may have been identical which could account for the Thick and Narrow ring widths.
    Some other notes on Huon Pines:

    Cultivation details
    Requires a light, freely draining soil[164] in a sheltered position with protection from cold winds[166]. Requires high rainfall and humidity if it is to succeed[200]. This species tolerates shade and probably requires it in drier areas if the tree is to survive[82, 200].

    So some other factors which significantly affect their growth are exposure to cold winds, rainfall, humidity and shade. Can they distinguish growth based on a cool wet summer as opposed to a dry warm summer in a growth ring of 1mm? So are they good temperature proxies well if you are able to account for wind exposure, rainfall, hunidity and shade over their 2000 year lifespan then sure they are!!

  18. Steve McIntyre
    Posted Sep 6, 2005 at 10:18 PM | Permalink

    Re #8: Paul, no but the percentage of coldwater diatoms offshore Oman has a hockey stick shape, so even though it measures water coldness, it becomes the most important Moberg proxy for global warming. It’s a quite a system that these guys have going.

  19. beng
    Posted Sep 7, 2005 at 7:06 AM | Permalink

    Looks like Cook’s study is essentially on Huon pines at harsh, treeline conditions. A good guess would be that these are temperature-limited in their growth.

    John Daly’s link (always well ahead of his time) shows Cook’s Fig.1 graph portrays the 1880-2000 period — quite convenient. Steve_M’s graphics for the whole 4000 yr period tells a much different story — most of that entire period has had greater growth than at present. The late 18th century shows the lowest growth! Another case of graphic fruit-picking?

  20. beng
    Posted Sep 7, 2005 at 7:09 AM | Permalink

    Jeesh, can’t get a simple post right. I meant the late 19th century had the lowest growth, right where Cook’s Fig.1 starts.

  21. TCO
    Posted Sep 7, 2005 at 7:11 AM | Permalink

    I just love the hair graph. It’s worth publishing almost for the pure graphical value of it.

    P.s. Anyone ever see that graph (famous graphic) showing the size of Napolean’s army as it moved in/out of Russia. It is part of some famous coffee table book of cool graphs.

  22. Posted Sep 7, 2005 at 3:04 PM | Permalink

    Charles Joseph Minard “Carte figurative des pertes successives en hommes de l’Armee Franàƒ⦡ise dans la campagne de Russie 1812-1813”.


  23. Evan
    Posted Sep 7, 2005 at 4:25 PM | Permalink

    Time to quit lurking and contribute something…
    I think the slow-growing trees are a mistake. Perhaps cm units mis-labelled as mm, or maybe an old column-format file that was read as, say, “F7.3” instead of “F7.2”. Either way, 30,000 errors with a single keystroke! Multiply those rings by 10 and they should match the others. A handfull of errors in the other direction might account for your giants.

  24. Steve McIntyre
    Posted Sep 7, 2005 at 7:40 PM | Permalink

    Re #23: I’ve seen giants turn up when I’ve done plots of other sites, so they are probably real. There seems to be a type of “alpha tree” phenomenon where they keep strong even as they are very old. They really screw up the RCS adjustment where a standard aging adjustment is done, because the alpha trees just keep on going.

    Could there be a lot of errors in the underlying dataset? I would be the last person to deny this possibility. A lot of the ITRDB chronologies end in the 22nd and 23rd century because of a computer programming error. But in this case, I think it’s real. I sent these diagrams to Cook some time ago but he did not reply. (These are the same people who won’t reveal the location of the Gaspe cedar site.) It’s worth asking, but I won’t be able to get an answer.

  25. Steve McIntyre
    Posted Sep 8, 2005 at 10:37 PM | Permalink

    Re #17 – Mount Read and Lake Johnston are the same site.

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

    Publish, publish, publish! Sheesh, we finally have a chance for you to publish something that is not just a Republican/Exxon plot to discount GW. You’ve actually added a bit of knowledge here, vice just exposing a fraud.

  27. Douglas Hoyt
    Posted Feb 27, 2006 at 2:47 PM | Permalink

    Re #23, I would tend to agree with Evan. The slow growers were probably measured by one group using centimeters. The other trees were probably measured by another group using millimeters. The two groups were probably combined and no one noticed the units problem. Perhaps that is why Cook did not answer.

    In any case, it looks like the data as it is would give a downward pointing hockey stick. Doesn’t one of Mann’s steps invert downward sticks into upward sticks? Maybe that is why they are happy with the data as it is.

  28. Steve McIntyre
    Posted Feb 27, 2006 at 3:29 PM | Permalink

    Doug, it’s probably worth resolving. Cook’s standardization methods are probably on a tree-by-tree basis (not RCS) so that a gross error like this would not actually make it difference to his chronology – which is interesting in itself.

  29. Steve McIntyre
    Posted Mar 21, 2006 at 10:02 AM | Permalink

    I was looking at some Mongolian data and noticed that the mean ring width for the Jacoby series ws about 10 times greater than the mean ring width for Schweingruber series. I asked for clarification from a specialist who told me that later dendro versions tend to be in 1/1000 mm while earlier ones tend to be in 1/100 mm. The ITRDB data bank has no metadata describing the units, which creates a problem even for specialists.

    For Cook’s Tasmania data, I’ll bet that the modern data was done earlier and is in 1/100 mm, while the subfossil data is in 1/1000 mm. However, the two differing units are jumbled together in the ITRDB archive. As I noted before, this would not make any difference to STD standardization, but would make a difference to RCS versions applied to this data.

    I notified Cook some time ago about the bimodal distribution, but he did not reply and the matter does not seem to have been corrected.

  30. John A
    Posted Mar 21, 2006 at 3:17 PM | Permalink

    Do they really measure tree ring widths to 1/100 or 1/1000 of a millimeter?

  31. Steve McIntyre
    Posted Apr 11, 2006 at 9:58 PM | Permalink

    #29. Although Rob Wilson thought it unlikely, the Tasmania data is definitely jumbled. Some of the individual trees have 999 tags and some of them have -9999 tags. So the data can be disentangled. But the data has to be extracted in an ad hoc way to do so.

  32. ET SidViscous
    Posted Apr 11, 2006 at 10:07 PM | Permalink

    “Do they really measure tree ring widths to 1/100 or 1/1000 of a millimeter? ”

    Didn’t see this before.

    What is the variation around the entire circumference? I’ve yet to see a tree ring that was a perfect circle, and I don’t mean the band, (Real music Steve, not L’il Kim type stuff).

    I can’t see where the error is going to be less than .1mm, and that only if you measure at 720 angular points or so and determine an average.

    Hell I’d rather calculate Global Mean Temperature, lass grunt work.

  33. TCO
    Posted Apr 12, 2006 at 5:15 AM | Permalink

    Sid, that’s one reason for advocating double-coring trees. (Although one could probably make a DOE-type argument that you’re just as well off, sampling more trees).


Get every new post delivered to your Inbox.

Join 3,633 other followers

%d bloggers like this: