Yang et al [2003]

An important new proxy series, which is one of only 8 in Mann and Jones [2003] and one of only 11 low-frequency proxies in Moberg et al [2003], is the Chinese composite of Yang et al [2003]. Unlike the Hockey Team, Prof. Yang promptly provided the underlying data set upon request. Here are some early thoughts.

The Yang article is here. The Yang composite is a weighted average of 9 series, which are all scaled and then plotted below relative to 20th century mean (I’ve done this so that the impact on hockey-stick-ness of each series can be seen more readily.) The 10th cell is the archived version of the series at WDCP. I’ve been able to closely replicate the archived version from the underlying series using weights provided by Yang.

Figure 1. Yang composite, together with underlying series.

The close connection between the Yang Composite and the Dunde series, which is highly weighted (as are the East China documentary and Guliya series), is obvious.

The Great Ghost Lake, Jiaming Lake and Jinchuan sediments all show high MWP values; they are low-weighted and have little impact on the final result. The original articles are in Science in China D and I have been unable to track them down (I’d like pdf’s if anyone can locate them.) The Japanese tree ring series is a dC13 series as I recall [I have this article; I’ll check and update]. One wonders whether this series has been calibrated correctly and whether it should be reversed.
[Note – it does appear to have been used in the reversed orientation.]

The S Tibet tree ring series is obviously discontinuous; I have been unable to track down the original reference. (Wu, X. D., and Z. Y. Lin, Climatic change during the last 2000 years in Tibet, Proceedings of Symposium on Climatic Change, 18–25, Science
Press, Beijing, in Chinese 1981)

The E China documentary series is inaccessible to me. Yang et al say:

The reconstruction of winter temperature from Eastern China are based on 5 proxy observations of changes in distribution of temperature sensitive biota and other climate indices, such as dates of flowering of shrubs, freezing of rivers, or displacements of the boundary of the farming zone [Zhang, 1996]. This record represents a noticeable improvement over the temperature reconstructions of Zhu [1973] and Wang and Wang [1989].

Meritorious as this series may be, it would be nice if the underlying proxy data were made more generally available, as Zhang, P. Y., 1996, Climate change in China during historical times, Scientific and Technological Press, Jinan, 435-436, is in Chinese and, to my knowledge, none of the underlying data is available.

Junipers in the western U.S. are strongly associated with moisture (see Paul Knapp and others) and negligibly associated with temperature. An article on Dulan junipers has been published in the western literature, but I haven’t parsed through the links yet. But it seems plausible to me that these junipers may reflect precipitation rather than temperature. (I once saw a very pretty picture on the Internet about 12 monbths ago of Chinese junipers (I think Dulan) at the edge of a desert, but I haven’t been able to re-locate the picture). Yang et al. say that the junipers are correlated to autumn temperature based on Kang et al [1997] a Chinese-language publication which is inaccessible to me. There is a later western publication, which seems to be by the same group of authors: Zhang et al, 2003.A 2,326-year tree-ring record of climate variability on the northeastern Qinghai-Tibetan Plateau, GRL, 30(14), 1739, doi:10.1029/2003GL017425, which I’ve posted up here.. They state (and this seems consistent with known behavior of similar vegetation in the U.S.:

We find that the annual growth rings mainly reflect variations in regional spring precipitation.

The Dunde version is one of 4 different versions that I’ve seen floating around. In one version, the 13th century was "warm"; in another, it was "cold". Thompson has never alerted anyone to the differences. Perhaps some day, Thompson will deign to archive all the sample information so that these differing versions can be reconciled.

The Dunde and Guliya series are dO18 series, with higher (less negative) dO18 anomalies interpreted as being warmer temperature. The problem is that the relationship between dO18 anomalies and temperature on an annual basis is reversed in monsoon regimes. The most negative dO18 occurs in the summer rather than the winter. Thompson claims that this relationship is reversed on a centennial basis. It seems pretty obvious that more negative dO18 could equally be explained by increased summer precipitation (the "amount effect" in ice core terminology). When you see the Dunde and Guliya dO18 series up against the Dulan juniper series and then recalls that dO18 in the Himalayas immediately reflects the amount of summer precipitation (with any link to temperature being a "tele-connection") , one wonders if the Yang Composite isn’t really a precipitation reconstruction.

Zhang [1994] reported from taxation records that some temperature-sensitive crops in the 13th century had expanded well to the north of present ranges. Obviously this information is at odds with Thompson’s interpretation of Dunde and Guliya ice core dO18. You will understand why I view the Yang Composite as more or less a re-packaging of the Dunde series – which seems to be the active ingredient in it.


  1. Greg F
    Posted Sep 11, 2005 at 6:45 AM | Permalink


    Could you post the data sets that were provided by Yang?

  2. Steve McIntyre
    Posted Sep 11, 2005 at 6:52 AM | Permalink

    The data is at http://data.climateaudit.org/data/Yangbao.data.txt. You should be able to open it in Excel. I’ve attached a script here http://data.climateaudit.org/scripts/collation.yang.txt to collate this version into a regularized time series and to produce the graph shown. This is just a working script and I haven’t tidied it. I cut and paste a lot and was working quickly. But anyone interested can follow it.

    PS: that wasn’t exactly Hockey Team hang time, was it?

  3. Greg F
    Posted Sep 11, 2005 at 7:33 AM | Permalink

    Thanks Steve. Some syntax errors I found.

    Data FileR script
    East China.STDEast.China.STD
    Japan treering.STDJapan.treering.STD
    Dulan treeringDulan.treering
    south.Tibet. treeringsouth.Tibet..treering

  4. Greg F
    Posted Sep 11, 2005 at 7:43 AM | Permalink

    Oh well the HTML table didn’t work.

    that wasn’t exactly Hockey Team hang time, was it?

    LOL … No it wasn’t. I think the hockey team prefers the slap shot and always thinks they should be on a power play. There has to be some statistical significance to switching sports. Getting Yang’s data has to be a slam dunk and your persistence speaks well of your hang time.

  5. Douglas Hoyt
    Posted Sep 11, 2005 at 11:23 AM | Permalink

    It would be interesting to see a 12×12 table of the r2 values for these 11 proxies and the composite.

  6. Steve McIntyre
    Posted Sep 11, 2005 at 11:36 AM | Permalink

    Your wish is my command. I don’t know how this will come out. I’ll post up an Excel-readable *.txt file if it’s too bad.

    WDCP guliya dunde dulan stibet echin ggl jiaming jinchuan japan
    WDCP 1.00 0.29 0.44 0.64 0.48 0.49 0.43 0.35 0.49 -0.55
    guliya 0.29 1.00 0.04 0.09 0.16 0.01 0.10 -0.21 -0.13 0.01
    dunde 0.44 0.04 1.00 0.24 -0.04 0.16 0.00 0.06 -0.01 -0.24
    dulan 0.64 0.09 0.24 1.00 0.32 0.25 0.11 0.14 0.31 -0.26
    stibet 0.48 0.16 -0.04 0.32 1.00 -0.11 0.09 0.21 0.09 -0.09
    echin 0.49 0.01 0.16 0.25 -0.11 1.00 0.29 -0.24 0.47 -0.12
    ggl 0.43 0.10 0.00 0.11 0.09 0.29 1.00 0.04 -0.01 0.02
    jiaming 0.35 -0.21 0.06 0.14 0.21 -0.24 0.04 1.00 0.03 -0.38
    jinchuan 0.49 -0.13 -0.01 0.31 0.09 0.47 -0.01 0.03 1.00 -0.12
    japan -0.55 0.01 -0.24 -0.26 -0.09 -0.12 0.02 -0.38 -0.12 1.00

  7. Chas
    Posted Sep 11, 2005 at 1:15 PM | Permalink

    Do they need to be all filtered similarly, for the r to be helpfull?

  8. Brooks Hurd
    Posted Sep 11, 2005 at 5:01 PM | Permalink

    When I look at the Yang et al graphs they indicate to me that we are nearing the end of a warming cycle. The recent up tick in temperatures looks very different on the Yang et al graphs than it does on the Mann, Jones, or Crowley graphs. If the M. J. and C graphs were plotted on the same y axis scales as the “weighted” and “complete” China graphs, the differences, particularly the blade of the M, J, and C hockey sticks, would be even more apparant.

    If I were shown the 5 graphs in Figure 4 of Yang et al, and I were told that the 5 graphs plotted the results of 5 different instruments measuring the same parameter on the same system at the same time, I would suggest that the technicians re-run the tests. My impression is that the data comes from two distinct systems. The M, J and C data seem to agree with each other AND the “weighted” and “complete” China data seem to agree with each other.

  9. TCO
    Posted Sep 11, 2005 at 5:04 PM | Permalink

    Actually the “feel” that I get from the different graphs (even those with LIA and MWP) is that this recent temp increase is faster than previous ones (and maybe different in character therefore).

  10. Brooks Hurd
    Posted Sep 12, 2005 at 9:39 AM | Permalink


    I would agree that the M, J, and C graphs show this, however the two Yang et al graphs show a similar, or steeper slope after the LIA. There are even more such temperature increases on the 2,000 year Yang et al graphs.

    On the 1000 – 2000 graph comparisions, consider how the amplitudes and slopes would compare when the 5 graphs are re-plotted with the same y axes.

    I have another comment on the five 1000 – 2000 graphs. I can almost make out the MWP and LIA on the M, J, and C graphs. These periods stand out more clearly on the Yang et al graphs. I think that the reason is that the M, J, and C graphs show much more “noise” in the 1000 – 1900 period than do the Yang at al graphs. This “noise” in the M, J, and C graphs de-emphasizes both the MWP and LIA.

    This “noise” seems to vanish from the M, J, and C graphs after 1900. This emphasizes the blade of the hockey stick. I believe that this is what M&M have been pointing out for several years.

  11. TCO
    Posted Sep 12, 2005 at 9:47 AM | Permalink

    I agree. If you draw a solid line through the noisy reconstruction period and then append the well-defined 1900s data, it gives a false view that recent jumps are way out of character from previous history. IOW, that climate is generally quite stable, but now we’ve suddenly got this curious recent jump. Mannites will when pushed agree that one should not draw a conclusion from their work that climate has been very stable (iow read into the work a conclusion on variability) from 0000-1900. However, they will then go on the other hand and say that the recent rise is unprecedented. That smacks a bit of some tricky pool to me…

  12. TCO
    Posted Sep 12, 2005 at 9:50 AM | Permalink

    There’s an excellent point from one of the AGW advocates, that the MBH style of argument (by their graph or the use it has gotten in the process) can actually hurt the AGW movement. If climate really does have a lot of natural variability, we could have a drop or constant temp over next couple decades even though the AGW forcing function over time continues to be a danger. If you really think of the climate as non-variable, you might be inclined to think that, that means we’ve reached a plateu. However, it would just be a case of strong (for the moment) natural variability swamping the still dangerous AGW trend.

  13. Steve McIntyre
    Posted Sep 12, 2005 at 11:41 AM | Permalink

    TCO, re #12: this is the position in Crowley and North in their 1990 textbook. The temperature increases projected in the climate models in the 1970s and 1980s had not materialized. So they were using natural variability (cold-side) to explain the non-warming then experienced.

    Re #11: the problem with red noise series is that if you data mine random red noise series for series with 20th century trends and then average them, you get a hockey stick. The common pattern wears off and you simply get an average of noise and the mean has very low amplitude. One of the interesting feature of multiproxy reconstructions is the attenuation of variance, which can be explained by this type of selection bias, but not through the so-called regression bias, advocated by von Storch.

  14. TCO
    Posted Sep 12, 2005 at 12:19 PM | Permalink

    I liked my explanation better. 😉

  15. Paul McGinnie
    Posted Sep 12, 2005 at 5:00 PM | Permalink

    From #2 above it appears you have a scripts and a data directory (neither of which is browsable).
    I can’t see how to access any other contents than those pointed to in #2 above, but I would guess
    that there are many data and R-script goodies of interest to the numerate amateur. Would it be
    possible to make the directories browsable, or to add instructions on how to get there?

    My apologies if this is a time wasting request – I have tried to find directions elsewhere
    before bothering you – but failed.

  16. Posted Feb 17, 2006 at 6:22 AM | Permalink

    I am very glad that someone is interested in our paper. I don’t want to make comments about these discussions because it is not correct in many ways and meaningless. However, sometimes I feel I should give simple ideas. About these discussions, I’d like to express my ideas. I have more than 10 years of working experience for historical temperature reconstruction and thus I should have priority to say how well the data used are in our paper. These data in China are chosen based on both resolution and representative of temperature change. In our paper, we gave three temperature composites in China during the last two millennia using different ways. The three composites show good agreement between one and another (please see the original paper), indicating good confidence of our reconstructions. Specifically, I don’t agree with Steve. About Dulan tree-ring width chronology, new width data from nearby region (Shidalong) representing winter temperature change are consistent with Dulan series usded in our paper in trend variations, giving strong evidence that Dulan chronology is an indicator of tempeature change at least. About Dunde ice-core chronology, we adopted the series with 50a resolution, which has the most ice-core samples (more than 7000) than other series. Concerning the more details about the ice cores, I have no right to say whether the data is how well or not. Although there are some different ideas about temperature-sensitivity of the ice-core data, there are extensive agreement between Dunde data and the temperature series from parts of China. Therefore it is no question that the Dunde-ice core represent temperature change on long-term timescales. The other series are also good indicator of temperature change. If one still has doubt about the data we used, please examine our paper more carefully and revisit the original references. Thank you very much for your attention!

  17. Steve McIntyre
    Posted Feb 17, 2006 at 7:45 AM | Permalink

    Dr Yang, thank you for these comments. As I mentioned in my post, I’ve had trouble locating original references for some of the series. I’ll follow up on some of these offline. For others, as I mentioned in my post, Dr Yang promptly and cordially provided me with the time series used in his calculation upon my original request.

    As is well-known, I’ve been quite troubled by potential problems arising from the use of tree ring chronologies in sites which are both cold and very arid – such as the bristlecone pines in the U.S.
    I’ve also been troubled by the difficult issues involved in testing for statistical significance between trending series and how to test for “spurious regression”.

    The implied history of the various proxies here is obviously quite different. While Dr Yang refers to the fact that he obtained quite similar results under three different weighting schemes, I don’t think that these 3 weighting schemes in any sense exhausted robustness issues – as mere inspection of the graphs of the underlying series shows.

    On Dulan, I remain concerned about using junipers as a temperature proxy. Dr Yang, you mentioned a new data set from Shidalong – what type of trees are these? Is there some ecological information on the site? Has it been published yet?

    On Dunde, I’ve posted up here about the problem of inconsistent versions of this series which have been distributed by Thompson. It is no fault of Dr Yang that Thompson failed to archive source data and then distributed different grey versions. However, the different Dunde versions need to be reconciled from original data. I have attempted to get Science to enforce their data policies and require Thompson to provide this data – see here. Since the Dunde versions are quite different, this means that estimates of their correlation to temperature need to be carefully examined depending on which version was used.

    As I mentioned in a post on Dunde here, I was unable to verify Thompson’s claimed correlations in a cross-validation calculation.

    I appreciate the cordial tone of your post. Cheers, Steve Mc

  18. Bao Yang
    Posted Apr 28, 2006 at 11:02 AM | Permalink

    Dear Prof. Steve McIntyre,

    I am sorry to bother you. I need your help. I saw a paper below.

    D’Arrigo, R., Jacoby, G. C., Frank, D. C., Pederson, N., Cook, E., Buckley, B., N
    achin, B., Mijiddorj, R., and Dugarjav, C. (2001). 1738 Years of Mongolian Temper
    ature Variability Inferred from a Tree-Ring Width Chronology of Siberian Pine. Ge
    ophysical Research Letters 28, 543-546.

    I can not find the related data of 1738 years tree ring data. Would you please te
    ll me how to get the data?

    Also, if possible, would you please tell me about the data of this paper?

    Loren D. Meeker and Paul A. Mayewski: 1400-year record of atmospheric circulation
    over the North Atlantic and Asia, the Holocene, 2002, 12, 3, 257-266.

    Thank you very much in advance!

    Best regards,
    Bao yang

  19. Steve McIntyre
    Posted Apr 28, 2006 at 11:37 AM | Permalink

    Dr Yang, I appreciate the imposing titles, but it’s just Mr and here it’s just Steve.

    A digital version of the Jacoby chronology has not, to my knowledge, been published. Two vesions are shown below, the first of which says it was scanned.

    Measurement data has been archived back to AD900 only, with the authors saying that the reconstruction is unreliable prior to that (despite publishing).

    I’ll see what I can locate on Meeker-Majewski for you.

    Cheers, Steve McIntyre

  20. Bao Yang
    Posted Apr 28, 2006 at 8:01 PM | Permalink

    Dear Steve, thank you very much for your reply and help. I hope the 2000-year record of atmospheric circulation is available. All the best to you!

    Bao Yang

  21. Bao Yang
    Posted May 17, 2006 at 12:56 AM | Permalink

    Dear Steve, do you have any information about 1400-year record of atmospheric circulation? Would you tell me about your email address? I am sorry I have lost your address. I can send you a paper which you may be interested in.

    regards, Bao Yang

  22. Armand MacMurray
    Posted May 17, 2006 at 1:30 AM | Permalink

    Dr. Yang, Steve’s address is smcintyre 25 AT yahoo.ca (it’s published here on the site, but isn’t quite as obvious as perhaps it could be).

  23. John A
    Posted May 17, 2006 at 1:32 AM | Permalink

    Dr Yang,

    Steve’s e-mail is smcintyre25 AT yahoo.ca

  24. Steve McIntyre
    Posted May 22, 2006 at 8:43 AM | Permalink

    Dr YAng, you asked:

    Also, if possible, would you please tell me about the data of this paper?

    Loren D. Meeker and Paul A. Mayewski: 1400-year record of atmospheric circulation
    over the North Atlantic and Asia, the Holocene, 2002, 12, 3, 257-266.

    Thank you very much in advance!

    I don’t guarantee that the following is correct as I’ve not used this data myself. It’s my best guess as to where the data comes from. You’ll have to plot up the series and compare to tell for sure. (It’s too bad that authors don’t regularly provide exact data citations.)

One Trackback

  1. By Yang et al. [2003] « Climate Audit on Nov 25, 2011 at 1:45 PM

    […] posted up once earlier on Yang et al [2003] here , raising some questions about some of the proxies. The Yang composite is NOT independent of the […]

%d bloggers like this: