Updated Polar Urals Data

I have obviously been very critical of the 11th century portion of the Polar Urals dataset used in Briffa et al. [1995], without which Jones et al [1998] does not have a cold 11th century (and which is a staple of other studies). It turns out that Schweingruber collected new subfossil samples for this site in 1999, which are dated back to 778 and have replication in the 11th century. If my hypothesis were correct, then the new information should diverge significantly in the 11th century.

A curiosity first. The data set that I’ve worked with – which ties in to what Briffa used – is russ021. It has archived measurements, but (anomalously) no archived chronology. The new samples ahve an archived chronology at russ176, but no archived measurements in the russ sequence. However, there is a duplicate archive of Schweingruber’s data and the series "POLURULA" corresponds to what should be russ176.

First, I show below "grass plots" for the Briffa data set collected in 1991 and second the 1999 dataset. The third and fourth panels show respectively a Shiyatov dataset collected in 1968 and spruce (rather than larch) at Polar Urals. I’ve shown pictures from time to time of medieval subfossil stumps located above modern treeline. The russ176w series looks like it must have been taken above modern treeline – the last sample is dated in 1892. The relative strength of medieval growth can be discerned quite easily here – with some trunks obviously much exceeding modern trees in girth.

So what is the effect on RCS "chronologies". First, I am making no adjustment for altitude, although the (presumbale) higher altitudes of russ176w would make this effect even greater. First, for review, I show (my) emulations of the ring width and MXD chronologies for russ021 (the Briffa version). If you recall, 1032 was supposed to be the "coldest" year of the millennium (from the MXD version) and the early 11th century notably cold.

Figure 2. RCS RW and MXD chronologies for russ021 (similar to Briffa et al 1995)

Next I show the new chronologies incorporating russ176 data with russ021 data. The 11th (and 15th)centuries are higher for both RW and MXD.

Figure 3. RCS RW and MXD chronologies for russ021 and russ176 combined

Finally here is the difference between the two versions: combined – Briffa 1995.

Figure 4. RCS RW and MXD chronologies – difference between russ021 and russ176 combined and estimated Briffa 1995.

These differences do not appear to make any difference to the 19th and 20th century values and thus to the temperature calibration. They do make a big difference to the 11th century values – perfectly consistent with my hypothesis that the Polar Urals material in the 11th century portion of russ021 was misdated. (The misdating is not corrected here – it’s just compensated for by replication. The true effect is probably even greater) .

I find it impossible to believe that Briffa and/or Schweingruber did not check the impact of the russ176 samples on their Polar Urals reconstruction. If this were a mining promotion, when you did new sampling pertaining to an important study and the samples, you would have to report the new results. In this case, Schweingruber at least archived the data, although the data was inexplicably not transferred to the main WDCP archive.

If the results showed an even colder 10th and 11th century, what do you want to bet that we’d have heard about the new samples? As it stands, if I re-run Jones et al[1998] just with the new Polar Urals data, it will have a warm 11th century – guaranteed. The Hockey Team are really something.


  1. TCO
    Posted Sep 26, 2005 at 2:26 PM | Permalink

    1. Perhaps there is a survivorship or other selection bias that explains why samples from older trees are girthier.

    2. Wonder what a tree from long ago that died long ago (per the grass chart) looks like. Is it a stump? Standing (seems hard to believe)?

    3. Sigh…hang in there. If you are so disgusted, why don’t you get ross to publish some more papers. He writes, you do the science (just don’t tell the jounrals). You have about 10 gazillion analyses in this blog. Lot more meat than at RC.

  2. John Davis
    Posted Sep 26, 2005 at 3:09 PM | Permalink

    The question is whether the tree rings mean anything in any case. Inverted “U” curves anyone? It seems to me that elsewhere on this site the argument is that growth isn’t proportional to anything much, so what, in this case, are you proving or disproving?

  3. TCO
    Posted Sep 26, 2005 at 3:11 PM | Permalink

    I don’t think that ALL of the possible critical arguments are mutually compatible.

  4. Steve McIntyre
    Posted Sep 26, 2005 at 3:32 PM | Permalink

    RE #1.2. Look here for a standing medieval foxtail pine. Also in the post about Constance Millar’s site. I’ve seen some pictures from the Urals and will try to locate them.

    Assuming that I’ve not erred here (and I’m working fast and going straight online), there are two possibilities: either they never checked the impact of the update on the Polar Urals study, which I don’t believe, and which would mean that they are incompetent. Or they checked the impact and didn’t report it, which is worse. This is just like the CENSORED bristlecone directory.

    Further, they’ve kept using the old results, while presumably knowing of the new results. Jones is a co-author of the Polar Urals study amd of Jones and Mann [2004]. In the latter, he used the non-updated Polar Urals study, which he presumably knew to be inaccurate in the most controversial period. Crowley presumably used old Polar Urals. Esper presumably used old Polar Urals. What a bunch of …… [hockey players].

  5. Steve McIntyre
    Posted Sep 26, 2005 at 3:34 PM | Permalink

    Re #2: I’ll try to collect some inverted U-diagrams that I’ve seen. It will be a handy reference.

  6. Douglas Hoyt
    Posted Sep 26, 2005 at 4:09 PM | Permalink

    Dead wood tends to dry out and contract over time. I wonder, if you have a long dead tree, then wouldn’t its tree rings tend to be narrower compared to a live tree of the same age? And if these tree rings are narrower, wouldn’t it be likely that one would interpret it as being colder? It might explain why trees from the MWP would show it colder than it really is. Is this an issue? Is the effect considered and corrected for? Just asking.

  7. fFreddy
    Posted Sep 26, 2005 at 4:34 PM | Permalink

    Steve, the link in #4 is pointing to this page.

  8. John A
    Posted Sep 27, 2005 at 1:20 AM | Permalink

    As it stands, if I re-run Jones et al[1998] just with the new Polar Urals data, it will have a warm 11th century – guaranteed.

    I’d like to see that re-run (with a comparison to the original).

  9. TCO
    Posted Sep 27, 2005 at 7:16 AM | Permalink


    Could we get some feature so that one can see the date or date/time of last post to a thread? Or so that one can go back and check for any threads that have had recent posts?

    It’s not just for unreasonable things like me going back through last year. Even with normal usage, one misses stuff, misses replies.

    If you do that, I will offer to moderate my usage (e.g. switch to once/week posting)

  10. Mats Holmstrom
    Posted Sep 27, 2005 at 10:25 AM | Permalink

    Regarding the question how tree rings correlate with temperature, I found an interesting paper by McCarroll et al.: Multi-proxy dendroclimatology: a pilot study in northern Finland. The Holocene, 13, 6, pp829-838, 2003.

    Ten potential proxy measures of past climate were recovered from Scots pine (Pinus sylvestris) at three sites on a latitudinal transect close to the pine limit in northern Finland (earlywood, latewood and annual ring width; earlywood, latewood and maximum density; stable carbon isotope ratio; height increment; needle production; pollen deposition). Cambium dynamics were also monitored. The aim was to determine how climate influences each potential proxy and to decide which proxies are potentially useful for reconstructing climate. Height increment, needle and pollen production are strongly influenced by the temperature of the previous July, which is when the bud forms, but needle and particularly flower development may also be influenced by spring frosts. Maximum and latewood densities provide proxies of net photosynthesis. 13C is controlled mainly by summer sunshine, reflecting the influence of photon flux on photosynthetic rate, and moisture stress which reduces stomatal conductance. By combining proxies, the strength of climate correlations is increased and the range of extractable parameters extended. The multiproxy approach provides a powerful means of extracting climatic information from long tree-ring chronologies.

    Note that Multiproxy here means that they try to combine different proxies from the same trees to better predict temperature.

    For me it is striking in the paper (Fig. 3) how bad the different measures correlate with temperature. Also, I realized something obvious I have not seen mentioned: The best correlation is with July temperature, thus all tree related proxies at best only provide information on the summer temperature, there is no information on winter temperatures, e.g., the Thames froze over during the little ice age. Are the instrumental temperatures used in the multiproxy studies only for summer months?

  11. Steve McIntyre
    Posted Sep 27, 2005 at 10:45 AM | Permalink

    Usually they are calibrated against summer months. Some of the recent Russian studies at Taimyr focus on periods as short as 10 days in late June/early July. The assumption is that summer temperatures correlate with annual temperatures – which is true to some extent.

    Sometimes they data mine – for example, Hughes and Diaz [1994], a still-cited multiproxy report, used a “winter” reconstruction in a case where the tree rings did not correlate with summer temperature, but a correlation to winter was found. Given the lack of any similar studies, one questions whether this might be simply a staistical artifact.

    In Norht America series, correlations with precipitation are stronger than correlations with temperature and hot summers usually shut down growth.

  12. John G. Bell
    Posted Sep 27, 2005 at 10:54 AM | Permalink

    I read the article differently. To me he seemed to say that the tree related proxies he had gathered matched July sunlight/cloud cover best. I agree about the bad correlation of the individual proxies with temperature, RW and MXD among them.

  13. TCO
    Posted Sep 27, 2005 at 11:02 AM | Permalink

    If precipiation and temperature have some predicatable covariance, it may still be possible to extract temperature signal from precipitation proxies.

    I thought everyone understood and agreed that the tree rings reflect summer conditions. However, it would seem reasonable (and easy) to look at the last 150 years and see how much covariance there is between hot summers and hot years. You may still be able to extract hot year inferences from a hot summer proxy.

    I have been singing the song for a while that we could combine measurements (MxD and RW) and get better proxies. Bothers me when the people who jump up and down about 3 equations and 2 unknowns, don’t consider this. This article takes that (obvious) concept forward. I would also think that one might be able to use other techniques to sharpen the proxy relationship, though. For instance: Perhaps use the trees at the northern extent for a temp proxy and the trees in the middle of the range for a moisture proxy and (by having the different data sets) deconvolute temperature signal from precip signal. This seems so obvious, I’m sure others are doing this, but we’re just not talking about it since this sight is so critical.

  14. Mats Holmstrom
    Posted Sep 27, 2005 at 11:12 AM | Permalink

    Re #12. John, yes you are correct. Sunshine gave the best correlations. What I meant was that of the different monthly temperatures, July gave the best correlation.

    Re #11. Yes, it is reasonable that July temperature is correlated with annual temperature, but it introduces another indirect step if one wants to study climate.

  15. John G. Bell
    Posted Sep 27, 2005 at 11:19 AM | Permalink

    I went over it again and see you are right. Sorry for the noise. Interesting article. A tighter fit between late wood density and mean dialy sunshine June – August than mean July temperature makes me wonder about MXDs and RWs being a better proxy for sunshine. Keep a salt shaker around. I often jump the gun.

  16. Steve McIntyre
    Posted Sep 27, 2005 at 4:01 PM | Permalink

    This is an interesting article. It’s nice that someone’s actually trying to validate the proxies through collection of data, rather than data mining.

    You have to be careful in going from these trees to bristlecones, which are at much lower latitudes, about 40 degrees; much greater altitudes and in very arid sites (they compete with sagebrush). Many American tree ring sites are categorically precipitation proxies.

    The observations about sunshine are interesting. I’ve mentioned that Jacoby was only able to get a “signal” from south-facing sites. It struck me that a north-facing site should be a “temperature” proxy and the south-facing site might be integrating cloudiness.

    I noticed this issue in particular with respect to the very low ring widths in northern Canada in the years following 1816. There’s a surprising amount of information from old Hudson’s Bay company records. My recollection (and I’m not 100% sure) is that the periods of very low ring widths seem to be characterized more by cloudiness; while the very cold years (1816) did not have as low a ring width as expected.

    It would be nice to a chart of ring width against sunshine (although I presume that if it was really high correlation) they would have reported it.

    The dC13 is interesting – for the bristlecones, it’s associated strongly with precipitation rather than sunshine. Since ring growth is always limited by something, I wonder if dC13 is responding to the limiting factor and could measure sunshine in one place and precipitation in another.

  17. TCO
    Posted Sep 27, 2005 at 4:08 PM | Permalink

    We (you 🙂 ) should invite the authors over here to the discussion…

  18. Mats Holmstrom
    Posted Sep 27, 2005 at 9:20 PM | Permalink

    (Oops, I posted this under the wrong thread the first time. Anyway, here it comes again:)

    It is interesting to note that the pine trees studied in the McCarroll paper is probably very similar to Briffa’s Tornetrask pines. These regions of northern Sweden and Finland are quite similar.

    The connection between tree growth and sunshine is interesting, especially since it seems that the solar radiation at Earth’s surface varies quite a lot (R.T. Pinker et al., Science, 308, p850, 2005). In that paper they state that At high latitudes, plant growth is light-limited, and a decrease in solar radiation can affect net primary productivity. Consistent with high latitude tree growth being a proxy for sunshine.

  19. beng
    Posted Sep 28, 2005 at 8:45 AM | Permalink

    Steve, you prb’ly know this, but 1816 was the “year without a summer” in eastern NA & Europe. Hvy volcanic dust from the eruption of Tambora in 1815.



  20. Larry Huldén
    Posted Sep 28, 2005 at 10:56 AM | Permalink

    We can not assume that summer mean temperature correlate with annual mean temperature. At least in Scandinavia summer mean temperature has not increased during the last 280 years but the winter mean temperature increased by at least 1-2 centigrade (I don’t have the figures at hand just now).

  21. Murray Duffin
    Posted Sep 28, 2005 at 1:25 PM | Permalink

    I noted in a post some days back that northern tree rings correlate with summer conditions, and northern temperature change has been a winter phenomenon. The heat wave in Europe in 2003 produced reduced plant growth, ie narrow tree ring corresponding with hot, not cold. The summer of 2004 in North America was the mildest (coolest) in about 100 years, but 2004 on average was a top decile warm year. Contrary to 3) above, it seems that all of the critical arguments support the conclusion that tree rings are not a proxy for temperature on any scale. Murray

  22. TCO
    Posted Sep 28, 2005 at 4:57 PM | Permalink

    well if you can have a proxy that magically responds to the global but not local trends, maybe you can do the same with year versus summer. 😉

  23. Steve McIntyre
    Posted Sep 28, 2005 at 10:06 PM | Permalink

    You are perhaps thinking of Gaspe cedars and bristlecones.

  24. Chas
    Posted Sep 29, 2005 at 1:38 PM | Permalink

    Re: July temperature and annual mean
    Out of curiosity I had a look at this within the CET series:
    July is part of the annual mean, so to avoid having a part-whole correlation , I compared July to the average of the other 11 months: July temperature variation explains only 10% of the variation of the rest of the year average. Not a lot, and this might be spurious.

    PS Armagh Observatory has strung together its various thermometer records (1796 to 2002):

    Click to access 445.pdf

2 Trackbacks

  1. […] information for the medieval period for the Polar Urals site used in MBH and Jones et al 1998 (see here (Aug 2005) for my first mention of the Polar Urals update). In Feb 2006, D’Arrigo (Wilson) et […]

  2. […] data set is what I’ve called the “Polar Urals Update”. On Sep 26, 2005 at CA here, I did a standard RCS calculation on the combined larch data from russ021 and russ176, observing […]

%d bloggers like this: