Dulan Response Coefficients

A standard technique of dendroclimatologists is to calculate coefficients between ring width chronologies and monthly temperature and precipitation for 12-18 months relevant to the annual growth. twq has reiterated to us that Gou et al 2007, which is a few minutes off the press, has claimed high correlations to temperature of a site in the upper reaches of the Yellow River that is within several hundred miles of Dulan. Today I report on response functions reported by other authors in the area, which consistently report positive correlations with spring precipitation and negative correlations with summer temperature – what you’d expect from precipitation limited growth in arid and semi-arid regions like Dulan. I reiterate that I’m not trying to adjudicate between specialists here – I’m only observing that the results of Gou et al 2007 are inconsistent with authors reporting on Dulan junipers.

Zhang et al 2003

Here are the response coefficients for Zhang et al 2003. This is a very important study, since 55 of the cores were used in Kang et al 1997 (which is the source of the Dulan chronology used in Yang et al 2002 and the multiproxy studies.) Black- temperature; white – precipitation. The authors interpret these results as justifying a reconstruction of spring precipitation. Recently a study involving Cook of chronologies in the Tienshan use similar correlations to reconstruct the Palmer Drought Index, which is a positive function of precipitation and negative function of summer temperature – an approach applied by Cook in the United States. (Cook’s precipitation work appears to me to be much the best work of anything done by the Team.) There is a weak positive correlation to temperature in the preceding fall, so he ring width is obviously a complicated integral.

Original Caption: Figure 2. Response function coefficients showing the relationships between radial growth of Sabina przewalskii Kom. and monthly mean air temperature and total monthly precipitation for the regional climatic data set for the period 1953—2000. Lag 1 represents the tree-ring growth in the previous year. Black bars stand for temperature, white bars for precipitation, and x for significance at the 0.05 level as tested by bootstrap method.

Shao et al 2004

Here is a similar plot from Shao et al, also showing positive correlations.


Sheppard et al 2004

Here is the response coefficients from Sheppard et al 2004, again showing a similar pattern to Zhang et al 2003 and Shao et al. – positive correlations to spring precipitation, negative correlations to spring temperature.

Original Caption: Fig. 4 Monthly correlations using ring-width indices from the combined juniper chronology and meteorological data from Dulan (1955—1988) for precipitation (black bars) and temperature (gray bars). The critical correlation value for n=30 and alpha = 0.05 is ⯰.30 (Rohlf and Sokal 1981)

Gou et al 2005
Gou et al 2005 is an interesting article about 4 sites on the north slope of the Qilianshan, which contains many insightful comments about high-elevation trees and which I’ll try to come back to on another occasion. Their table of response coefficients shows a positive correlation to spring precipitation and negative correlation to summer temperatuyre.

Gou et al 2006
Here is a table showing response coefficients from Gou et al 2006, from 3 sites in the upper headwaters of the Yellow River, near the site of Gou et al 2007. Again we see the characteristic pattern of positive correlations to spring precipitation, negative correlations to spring temperature. Gou et al used the negative correlations to temperature to reconstruct temperature inverting the chronology.


Gou et al 2007

Now we get to twq’s favorite: Gou et al 2007. This is a study of a single site in the upper headwaters of the Yellow River. The site chronology is taken from the lower border as this is said to have a higher correlation to temperature (they say that these relationships are “complicated”) and report strong positive correlations of this site to local temperatures. Unfortunately they did not archive their measurement and the meteorological station data is either not archived or stale. For an article that is essentially a data paper, it’s too bad that they don’t actually archive any data.

Original Caption: Figure 3. Correlation coefficients between the Standard (STD), Residual (RES) and Arstan (ARS) chronologies index and the observed minimum temperatures recorded at Xinhai meteorological station (1960—2001). Ave. in the Figure 3 means average minimum temperature of previous October to current April. The dashed lines indicate the 95% and the 99% significance levels.

The consistent positive relationships between temperature and ring width chronologies are a very different pattern than we observed with the Dulan junipers. There are a couple of possibilities: this site is a magic thermometer; the correlation, while seemingly significant, is nonetheless “spurious” – a topic discussed on many occasions on this blog, the type example being the “99% significant” correlation between alcoholism and Church of England marriages reported by Yule in the 1920s. Or maybe there’s something about the location of the site that makes it a better but non-magic thermometer. Regardless, the positive correlations at site HBL don’t alter the results of Zhang et al 2003, which are the ones that apply to the south-facing Dulan junipers used in Yang et al 2002.

While the correlations at site HBL are intriguing, at cliamteaudit, we don’t accept the dendroclimatological procedure of simply cherry-picking one study. Any use or reliance on Gou et al 2007 needs to include a proper accounting and reconciliation of all the other studies in the area, something that Gou et al 2007 made no attempt to do.

Let me conclude with a quote from Gou et al 2005 – an interesting article as I noted above and one which, unlike Gou et al 2007, did not have any members of the Team (Jacoby, Cook). Gou et al 2005 concluded with the following observation:

trees growing at high elevation show a lower sensitivity to climate. This conclusion is of fundamental importance for tree-ring research in arid and semi-arid regions. It is important to understand the relationship between climate change and the growth of the trees in order to develop an appropriate ecological model of plant environmental reactions and to establish a valid basis on which to reconstruct long-term climate change over wide geographical areas…. Understanding the ecology model between trees and climate will allow us to examine differences in the long-term climatic response that may be related to changes in climate.

Everyone at climateaudit endorses the view that understanding the ecological relationship between trees and climate is a prerequisite for using ring width chronologies in temperature reconstructions. Maybe Gou could send a memo to the Team.


  1. MrPete
    Posted Apr 16, 2007 at 6:27 PM | Permalink

    Ummm… do ANY of these studies explain how the sites were selected?

    What is there to suggest that this is not a perfect example of the Texas Sharpshooter fallacy? Is that the meaning of “we’ve moved on”… that we see a pattern in one study, and, using much the same data cherry-pick a new study to zero in on the “higher correlation” data we’ve found???

    I want to believe the following oversimplification misses something essential… yet it sure seems to fit what I read above:

    Studies with large number of samples shows positive correlation for precipitation and minor neg corr for temp.
    A subsequent study with three sites strengthens both the positive and negative correlations
    Finally, a single-site study zeros in on just the “negative” correlation to demonstrate the ability to extract temperature.

    Sure sounds like covering the barn with shots, then walking up close to put a bullseye where you find the most “interesting”/”useful” data…

    I’m beginning to think this field needs to consider the rigors of double blind placebo-controlled analysis. Just think…

    Data for the various sites is stored in ‘hashed’ (randomized) filenames. Only the system knows the real sources.
    A random set of sites is served up for analysis. The researcher has no opportunity to select for “better signal”.
    Some of the data sets are actually ‘placebos’, i.e. computer-generated noise that shares similar characteristics with the real data.
    After the scientists’ analysis process has been perfected, the data sets and scripts are archived in ‘escrow’.
    Only then, a third party ‘unlocks’ the keys. Placebos are removed from the data sets. Results are re-run without modification to data or scripts, and true significance determined.
    If changes are needed, new randomized/placebo-laced data sets are generated.
    No research is accepted for publication unless it has been run through a proper double-blind / placebo-based study.

    Why are we seeing such elementary errors of scientific process at such an advanced professional level of work? Any student in a good high school science class could easily see through these issues.

  2. twq
    Posted Apr 17, 2007 at 1:32 AM | Permalink

    RE: Today I report on response functions reported by other authors in the area, which consistently report positive correlations with spring precipitation and negative correlations with summer temperature – what you’d expect from precipitation limited growth in arid and semi-arid regions like Dulan.

    It is ture for this correlation between spring precipitation with ring widths. But it is also clear that temperature (especially prior autumn and winter temperature) has effect on the radial growth of dulan ring widths because the annually frozen soil layer may exert influence by way of delaying or speeding the thawing time in the coming spring.

  3. Don Keiller
    Posted Apr 17, 2007 at 3:37 AM | Permalink

    Not being a dendroclimatologist, or indeed a statistian, I’m a bit puzzled by some of this data. Maybe tqw could explain. Take Fig 2., for example, which shows “Response function coefficients showing the relationships between radial growth of Sabina przewalskii Kom. and monthly mean air temperature and total monthly precipitation” Looking at this figure May and June response coefficients are both the largest and significant. May has an approx 0.2 positive coefficient and a negative 0.3 temperature coefficient. A similar situation exists in June. So am I correct here- the radial growth rings are influenced positively by precipitation in these months and negatively by temperature? Now bearing in mind that temperature and precipitation have influences of similar magnitude, but opposite sign, how does one differentiate, say, between high precipitation and low temperature in one year and low precipitation high temperature in another?

  4. Willis Eschenbach
    Posted Apr 17, 2007 at 4:44 AM | Permalink

    This might be a dumb question, but I asked it before and didn’t get an answer.

    From the first figure above, we can see various correlations between temperature and ring width. For example, ring width is positively correlated with the previous October temperature, and negatively correlated with the current April temperature.

    However, looking at the Dulan GISS temperature dataset, I find that previous October temperatures are negatively correlated with current April temperatures (correlation = -0.38, p = 0.03).

    Since this correlation between temperatures is greater than the correlation between temperature and ring widths, is this adjusted for in the calculations, and if so, how? I mean, is the real temperature/ring width effect taking place last October, and the April temperature/ring width correlation merely an artifact caused by the natural October/April temperature correlation, or vice versa, or some mix of the two, or how?

    I suspect that this is not adjusted for, and should be, but I can’t figure out how to remove the inherent temperature correlation from the data to expose the true effect of the temperature on ring width. All comments gratefully accepted, I fear this conundrum is above my pay grade …


  5. Paul Linsay
    Posted Apr 17, 2007 at 7:10 AM | Permalink

    #1, MrPete,

    You’re absolutely right about the need for double blind in “climate science”. I’ve thought for a long time that it’s a scandal that the modelers are the same people who analyze the surface temperature record. The opportunities for self deception are immense, especially in the hands of strong AGW proponents like Hansen. Once the surface record is properly analyzed the whole AGW edifice is going to come crashing down because the “hindcasts” will be hindcasts of junk and the models will be shown to be fudged. As I said long ago on this site, that’s going to be a huge fight, much larger than the hockeystick. Good luck SteveM. [Sorry, a bit off topic for this thread, maybe it should be on one of the temperature threads.]

  6. Don Keiller
    Posted Apr 17, 2007 at 8:58 AM | Permalink

    And of course there will have been no effect of increased CO2 to further muddy the waters…

    Well all Dendros should read this paper very carefully (Soule, P.T. and Knapp, P.A. (2006). Radial growth rate increases in naturally occurring ponderosa pine trees: a late-20th century CO2 fertilization effect? Volume 171, Number 2, pp. 379-390(12)) Basically they find a post-1950 radial growth enhancement that was and I quote “more pronounced during drought years compared with wet years, and the greatest response occurred at the most stressed site.”

  7. jae
    Posted Apr 17, 2007 at 9:53 AM | Permalink

    Steve M: I’m confused (again) by Fig. 3. Why are the correlations with MINIMUM temperature? Doesn’t this say that the trees are growing better when it’s colder?

  8. Steve McIntyre
    Posted Apr 17, 2007 at 10:06 AM | Permalink

    The theory is that colder wonters cause the ground to freeze up more and thus it delays spring growing. However, Luckman and Wilson 2003 do a study in B.C. in which they reconstruct maximum temperature.

    From a statistical point of view, the studies seem statistically opportunistic and one would like to see the hypothesis whatever it is demonstrated on out-of-sample data.

  9. jae
    Posted Apr 17, 2007 at 10:08 AM | Permalink

    It’s too bad the authors you cited didn’t provide the correlations between growth and the Palmer Drought Index. It looks like they would be very high. One can really see the combined effects of high temperatures and low precipitation during the peak growing period April-June.

  10. twq
    Posted Apr 17, 2007 at 10:25 AM | Permalink

    RE #9,
    Jae, How do you know if they did it or not? Give it a try. Then you will find who (you or they) are bad. I am fed up with your remarks. I think you are so crude and unpolite.

  11. Steve McIntyre
    Posted Apr 17, 2007 at 10:34 AM | Permalink

    twq, given that the authors did not archive their chronologies and used meteorological station information that is not archived at GHCN, a third party reader can hardly be criticized for being unable to test PDSI. One of the reasons for archiving data is to enable precisely this sort of analysis. Suggestions that a third party go to Dulan and take samples are very unhelpful in this respect.

  12. jae
    Posted Apr 17, 2007 at 10:56 AM | Permalink

    10, twq: I think there is a misunderstanding here. I am not criticizing anyone and I regret that you think I’m being rude. I just think such an analysis would be interesting.

  13. twq
    Posted Apr 17, 2007 at 12:20 PM | Permalink

    11, 12, thank you for the explanation. Thus it is sure that we have a misunderstanding here. I regret and apologize for my misunderstanding. To my knowledge, PDSI has its caveats in some way, especially in the special region where very few meteorological and other observations are available. I remember someone used Qilianshan ring widths for moisure index reconstruction (P/T), which is published in Chinese early. So it is sure the dendros have made such reconstructions by combining both temperature and precipitation. But like these reconstructions listed here, it has similar correlation.

  14. Willis Eschenbach
    Posted Apr 17, 2007 at 10:31 PM | Permalink


    Well, I’ve stumbled across a very interesting oddity. I love science, there’s always so much to learn.

    The dendrochronologists assume that certain tree ring widths are generally related to temperature, under the assumption that increasing ring widths signify higher overall temperatures, and vice versa. Now we know that there’s a host of problems with that involving linearity, but let’s assume for the moment that ring widths are linearly correlated with temperature. Let’s also assume that there are no other factors in play (precipitation, soil moisture, etc.). Where does this set of assumptions lead us?

    The first figure in this thread gives the monthly correlations of temperature and ring width for the Zhang et al. study. These are:

    Prev. Sep, 0.04
    Prev. Oct, 0.17
    Prev. Nov, 0.14
    Prev. Dec, 0.07
    Jan, 0.02
    Feb, -0.06
    Mar, 0.16
    Apr, -0.11
    May, -0.27
    Jun, -0.19
    Jul, 0.04
    Aug, -0.07
    Sep, -0.03

    So, I figured I’d see if I could reconstruct what the tree ring widths looked like from the correlations to the Dulan temperature. Much to my surprise, I found out that there is no unique inversion from correlation to ring widths. Here are four synthetic ring widths, each with identical correlations to the Dulan month by month temperature, along with the Dulan temperature itself:

    As you can see, the four synthetic ring width series are all quite different the Dulan temperature. In addition, the correlations with the quarterly (D-J-F, M-A-M, J-J-A, and S-O-N) and annual Dulan temperatures are very poor. None of the correlations are significant. Finally, the trends are very different.

    Now, if we can’t compute the ring widths from the correlations, and we assume that the temperature is related to the ring widths because of the correlations, doesn’t that mean that we can’t compute the temperature from the ring widths? And since the trends can be quite different and still give the same correlations, doesn’t this mean that there is not necessarily a relation between temperature trends and tree ring width trends?

    For my next excursion into dendro, I’m going to see if this is true of the Wilson study as well …

    OK, six hours later, went to work (I’m currently building a handicapped access ramp on a commercial building), back again. Here’s Wilson et al. data, compared to the 1950-1990 instrumental record. Wilson used principal components, I’ve looked at the PC1 correlations. It has a number of very good correlations with various months, and it only has one negative correlation. The correlations also extend over a longer period, 21 months instead of 12 as in the Zhang et al. data. Here’s the correlations.

    Prev. Jan, 0.09
    Prev. Feb, 0.24
    Prev. Mar, 0.31
    Prev. Apr, 0.36
    Prev. May, 0.34
    Prev. Jun, 0.14
    Prev. Jul, 0.02
    Prev. Aug, 0.16
    Prev. Sep, 0.22
    Prev. Oct, 0.16
    Prev. Nov, 0.05
    Prev. Dec, -0.04
    Jan, 0.08
    Feb, 0.16
    Mar, 0.36
    Apr, 0.26
    May, 0.44
    Jun, 0.32
    Jul, 0.21
    Aug, 0.28
    Sep, 0.30

    So I expected that the synthetic PC1s would be very close to each other. Once again, I was surprised … here’s four synthetic PC1s that all have the same correlation with the monthly data, month by month:

    As you can see, not only are the PC1s different, but their trends are also different.


    1) Very different tree ring width patterns can give identical correlations with a given set of monthly temperatures.

    2) The fact that tree ring widths are correlated with monthly temperatures does not mean that tree ring width trends are correlated with temperature trends.

    3) For any given set of monthly RW/temperature correlations, there exists a family of individual different RW curves which will give the same correlations with the monthly temperatures (within instrumental accuracy).

    … am I crazy for thinking that this makes it very hard to put confidence intervals on a historical tree-ring based reconstruction, and that it makes the calculated trends very suspect?


  15. Steve McIntyre
    Posted Apr 17, 2007 at 10:45 PM | Permalink

    Willis, that’s interesting, but your description of what you did to construct your synthetic series is opaque (code would help or something else that’s specific).

  16. Willis Eschenbach
    Posted Apr 17, 2007 at 11:39 PM | Permalink

    Sorry for the opacity, Steve M. I used industrial strength trial and error, also known as the “Solver” function in Excel, to minimize the squared error between the correlations of a given set of numbers with Dulan temperature, and the original correlations with Dulan temperature given in the Zhang paper. Solver changes the given set of numbers until the conditions are met (each month’s correlation is the same as the ).

    Using different starting series in this process generated different final series, all of which have the identical desired month-by-month correlations.

    I found the differences in the synthetic ring width series quite surprising, particularly the differences in the trends, despite having identical correlations.


  17. Posted Apr 17, 2007 at 11:40 PM | Permalink

    Hi Willis, Are not the various reconstructions with the same correlation just an illustration of the confidence interval that should surround any particular reconstruction (or inferred parameter value such as trend)? Or am I missing something? Cheers

  18. Nicholas
    Posted Apr 18, 2007 at 1:27 AM | Permalink

    David, I think it goes deeper than that.

    What Willis is effectively saying is that if you cored 4 stands of trees, and averaged the ring widths, and got each of these 4 fairly different curves, each of them could equally well explain the instrumental temperature for that reason, under this linear relationship hypothesis.

    Therefore I think we need to decide whether this hypothesis is at all reasonable and/or has any predictive power, when what we’re in effect saying is that as far as we can tell, all of those growth outcomes would be acceptable to be interpreted as the actual temperature.

    If you do think that is reasonable, then yes, I guess your confidence interval has to encompass all of those possibilities – and indeed, probably any possibilities that have HIGHER correlations as well (since that’s more desirable, right? And if you had a stand of trees with even higher correlations, you’d use it for the reconstruction, right?)

    Sorry if this is semi-incoherent.. but I guess what I’m saying is.. this is only specifying the minimum confidence interval, not the maximum. The maximum could be a lot larger, because nobody ever said the correlations have to be exactly these to be valid, only that these are good enough correlations to be confident of a good reconstruction. I bet if Wills does more testing he will find that there are a very large set of “proxies” that would correlate as well or better with temperature, and yet all of them would be “valid” reconstructions under this hypothesis.

  19. Nicholas
    Posted Apr 18, 2007 at 1:31 AM | Permalink

    Addendum : I guess you can ignore any synthetic proxies which are physically impossible (or at least improbable), but you have to consider the effects of averaging/PC analysis too. However, even those make me doubt the predictive power of this data under this hypothesis. At the very least, those correlations seem like very weak evidence that the proxy is actually representing temperature after these revelations.

  20. Willis Eschenbach
    Posted Apr 18, 2007 at 1:48 AM | Permalink

    What surprised me was the difference in the trends, not the difference in amount, but the difference in sign. Look at the blue and the red line in my first graph above, one goes up, one goes down. Look at the

    Next, remember that the correlation is not changed by a linear transform such as a multiplication … but the trend is changed.

    Which means that regarding Dulan, we can produce a series that has any trend we want, either positive or negative, but that still has the same identical correlation. The same is true of Wilson’s data. Although there are no trends with a negative slope in the graph above, I have produced these as well.

    Weird, huh?


  21. Posted Apr 18, 2007 at 2:19 AM | Permalink

    In addition to the response of tree ring widths to moisture and temperature, the response to increased CO2 levels in (semi-)arid conditions is of importance too. As mentioned by Don in #6, tree ring width growth is enhanced with increased CO2, especially in drought conditions. This is in part due to the fact that higher CO2 levels lead to less stomata density, which in turn reduces evaporation loss and thus enhances water use efficiency. But this is also species dependent.

    This further complicates the tree ring width correlation with temperature and moisture…

  22. twq
    Posted Apr 18, 2007 at 3:35 AM | Permalink

    RE: 14,

    I suggest you write a paper detailing these analysis and submit it into climate journal. Thus it would be better for climatologists. Here it is not clear and maybe cause misunderstanding. I can not understand what you did. Indeed, I just trust these allegations published in speciallized Journals rather than discussions here or there.

  23. Willis Eschenbach
    Posted Apr 18, 2007 at 4:07 AM | Permalink

    twq, thank you for your encouragement. I have written articles and had them published by scientific journals in the past. However, I prefer to preview and discuss them here first, to avoid obvious errors and mistaken inferences.

    As I mentioned above, this finding was a surprise to me, and I don’t really understand it.

    What I showed above is that we could have four (or more) different stands of trees. We could core them, and average them, and find that the averages have different ring width patterns, and different trends. However, all of them could have the exact same correlation with the Dulan temperature record.

    What are the implications of this? Quite frankly, I don’t know … which is why I have posted my research, and invited comments on the phenomenon.

    My very best to you,


  24. Posted Apr 18, 2007 at 5:04 AM | Permalink

    Interesting post, Willis! Tried with Matlab (matrices in bad condition, use pinv!), solutions seem to be very sensitive to initial guess.
    Example here.

  25. KevinUK
    Posted Apr 18, 2007 at 5:20 AM | Permalink


    I think what you’ve shown is that, I’ve long suspected but you have now proved, that tree ring width growth is useless as a proxy for temperature. To use the metaphor from the recent video link provided by Lubos, this ‘pillar of the AGW temple’ is now severely cracked/damaged. I think Steve M and others like yourself have done more than enough to back up my statement. Its now time to follow the HT’s lead and ‘move on’ (as Steve M is already doing by auditing aspect sof the other ‘pillars’ like Jones UHI correction etc) and leave the mess that is dendro paleoclimatology behind.


  26. Dave Dardinger
    Posted Apr 18, 2007 at 7:39 AM | Permalink

    re: #23 Willis,

    I’m wondering; while the correlations may be identical regardless of which of these synthetic series you pick, is the same true of the various statistical checks which can be performed? After all, one of the big hits on Mann is failure to test his results with r2, etc.

  27. Willis Eschenbach
    Posted Apr 18, 2007 at 11:46 AM | Permalink

    Dave, because the monthly correlations are the same, the R^2 are the same, but only for the period covered by the monthly correlations (e.g., previous September to current September).

    But their correlation to anything outside that period is different. This includes the correlation to the current year (January – December) or meteorological year (previous December – November) if those periods are not covered by the monthly correlations.


  28. bender
    Posted Apr 18, 2007 at 9:15 PM | Permalink

    Re #4

    Since this correlation between temperatures is greater than the correlation between temperature and ring widths, is this adjusted for in the calculations, and if so, how?


  29. bender
    Posted Apr 18, 2007 at 9:18 PM | Permalink

    I suggest you write a paper detailing these analysis and submit it into climate journal.

    Meanwhile the Dendro Truth Squad could write a group letter to IPCC explaining how the science is uncertaiin, imprecise, fallible, ….

  30. Willis Eschenbach
    Posted Apr 18, 2007 at 10:21 PM | Permalink

    Bender, I appreciate your elegant and terse answer to my question, viz:

    Since this correlation between temperatures is greater than the correlation between temperature and ring widths, is this adjusted for in the calculations, and if so, how?

    to which you replied ..


    Clear and to the point. Moving right along, I had also asked:

    I suspect that this is not adjusted for, and should be, but I can’t figure out how to remove the inherent temperature correlation from the data to expose the true effect of the temperature on ring width. All comments gratefully accepted, I fear this conundrum is above my pay grade …

    Should this be adjusted for? If so, can it be adjusted for? And if so … how?

    Thanks as always for your mathematical insight,


  31. bender
    Posted Apr 19, 2007 at 3:09 AM | Permalink

    Should this be adjusted for?
    If so, can it be adjusted for?
    And if so … how?

    1. Yes IMO.
    2. To some degree, maybe. Depends how “naturally orthogonal” the inputs and outputs are. Temp and precip inputs can be indepedent in the long-run but either (+) or (-) correlated in any short-run. Also, P and T interact synergistically (as in the PDSI), so even if the inputs are “naturally orthogonal” the outputs may not be.
    3. If you are asking for a recipe for statistical adjustment, I do not believe one exists that will IMO do the job required. I believe the only thing that will satisfy the hardest skeptics are controlled physiological experiments. When it comes to dendrostatisticology – a subject I am learning to loathe – cautious interpretation is the order of the day.

  32. Brandon Shollenberger
    Posted Jul 16, 2012 at 4:56 PM | Permalink

    I know this is an old post, but there don’t seem to be posts for off-topic discussions anymore, so I’m posting it here for posterity’s sake.

    While looking at a recent temperature reconstruction during a discussion at The Blackboard, I noticed a number of proxies were being used which I remembered as being problematic. I decided to review the proxies I recognized to refresh my memory, and one proxy I looked at was the Dulan series. Since I was in communication with an author (Bo Christiansen) from the paper regarding another issue already, I thought I should ask him about this series. He forwarded my question to the other author, Fredrik Ljungqvist, who said (in part):

    From what I have learnt, the signal can either be interpreted as temperature or precipitation but it is most likely both and they cannot be easily separated. This region of China/Tibet follows, on decadal and longer time-scales, a clear warm-wet and cold-dry pattern. Warm periods are wet and cold periods are dry.

    This means, at least in theory, the Dulan series could be used as both a temperature and a precipitation proxy. That would be an interesting twist on the issue.

    However, I am unconvinced on the matter for two reasons. First, as I told him, the fact warm-wet/cold-dry periods are correlated does not prevent temperature and precipitation from varying independently of each other. Second, he kindly sent me a link to a recent paper that seems to contradict the position.

    That paper examines the relationship between temperature and precipitation in China during five warm periods, and it is interesting. The problem is in two of the four periods the Dulan series covers, the paper shows dryness for the series. In fact, about half the time, series in that area appear to be dry while warm.

    Both authors have responded promptly and courteously, and I don’t fault them for having used the series whatever the “right” answer may be. Enough people have used the Dulan series as a temperature series that it’d be easy enough to think it is one. And perhaps I’ve just missed something. Either way, I’ve sent another e-mail to Ljungqvist asking about that, so I’ll see.

    • Steve McIntyre
      Posted Jul 16, 2012 at 5:21 PM | Permalink

      Brandon, a good reference for the importance of not conflating precipitation and temperature proxies is Mann et al EOS 2003 – their attack on Soon and Baliunas 2003. There’s a lot of irony and hypocrisy in the article, since Mann et al 1998 actually used 11 instrumental precipitation series as temperature “proxies” while Soon and Baliunas distinguished precipitation proxies from temperature proxies. Wigley even conceded that they had a point on previpitation, but told his collaborators that the important thing was to deny any “credit” to Soon and Baliunas. So there was a lot of projection in the attack.

      But check it out as an authority for the idea that precipitation proxies should not be used as temperature proxies. –

      • Brandon Shollenberger
        Posted Jul 17, 2012 at 1:17 AM | Permalink

        Steve, I don’t remember what that paper says offhand (I’ll try to check tomorrow), but I can’t see any inherent problem with using a proxy as both a precipitation and temperature proxy. I don’t think a proxy necessarily needs to be one or the other. One could easily create hypothetical situations where using a series as both would be viable. For example, one could imagine there was a region where temperature and precipitation were perfectly correlated in a linear function.

        It would probably take quite a bit of work to show any particular proxy series could be used like that, and that work doesn’t seem to exist for the Dulan series, so I don’t think it’s appropriate to use the series as a temperature proxy. Still, I wouldn’t rule out the possibility, especially not with work by Mann and cohorts as my basis!

  33. Skiphil
    Posted Jul 16, 2012 at 6:30 PM | Permalink

    Some “interesting” remarks on this thread at RC, about the recent Esper et al (2012) and dendro/proxy reconstructions more generally. Michael Mann sounds as usual, but some of the exchange is worth looking at:

    Rob Wilson dialogue with Michael Mann, Gavin Schmidt, and Jim Bouldin at Real Climate

    • Kenneth Fritsch
      Posted Jul 16, 2012 at 7:38 PM | Permalink

      “Some “interesting” remarks on this thread at RC, about the recent Esper et al (2012) and dendro/proxy reconstructions more generally. Michael Mann sounds as usual, but some of the exchange is worth looking at:”

      I read the exchanges and I see not much except the same old same old. Rob Wilson talks about the divergence problem as no big deal as he often does. Mike Mann is a little emotional about references to a couple of his papers. I do, however, think that Gavin Schmidt captures the importance of the paper and that is that it claims that MXD captures the orbital cooling in a small area of the globe and whereas TRW proxies did not. MXD is thus claimed to capture low frequency temperature changes better than TRW.

      Rob Wilson also correctly notes that the cooling claimed to be measured in the paper has nothing to do with global cooling. I hope the more thoughtful skeptics clearly see this point and do not suddenly start accepting reconstructions out of hand because of what they think it might portend.

      What I hope to find next is the data that were used in this paper.

      Steve – some of it is at Esper’s publications page.

%d bloggers like this: