Positive and Negative Responders

Just when you think that you’ve heard of all the possible problems with tree rings, the newest issue comes from “positive” and “negative” responders to temperature within the same site. These issues are discussed in a number of articles by various post-docs associated with Jacoby and D’Arrigo with the latter as co-authors. So in fairness to them, they are not just ignoring the problem of the failure of tree rings to respond to post-1980 temperature – although the issue is dealt with more candidly in specialist articles than in the general literature. Also in fairness to Jacoby who I’ve criticized for failing to archive information (and will continue to criticize), unlike Mann, he’s collected a lot of information. When I twitted him for supplying data to Mann for use in Mann and Jones [2003] and then refusing to supply to me, he said that he didn’t supply it to Mann and that he had no idea how Mann got the data and seemed to have antipathy towards all data re-processors. However as I’ve said before, exploration geologists also go into remote spots of the world and it would never occur to them that this labor made them owners of the data. But it’s the compoanies that manage the geologists. I’ve got a bigger beef with NSF being co-opted by dilatory archivers than even the dilatory archivers. But on to positive and negative responders.

Driscoll et al [2005] report of Alaskan sites:

During standardization, it became clear that the chronologies from Fish Trap Lake (FT) and Portage Lake (PO) contained two subpopulations within each site and hence they were separated into four chronologies, two of which responded positively to increased temperatures in the latter half of the 20th century and two of which appeared to experience an overall decline after the 1950s (Figure 2)….

Growth of the divergent subpopulations correlate well prior to 1950 and diverge thereafter, supporting the idea that contemporary warming has introduced unprecedented stresses on some trees and thus impeded growth. Furthermore, strong positive correlations with August precipitation among negative responders suggest that temperature induced drought stress may be relieved in these populations by higher levels of precipitation late in the growing season. [14] The interplay between climate variables (e.g. temperature, precipitation), highly localized non-climate variables (e.g. competition, depth to permafrost) and individual tree growth rates and allometry is not well understood in the context of recent warming at high latitude tree-line sites. It appears that unprecedented climatic stresses are triggering diverse growth responses between and within study sites that may greatly complicate dendroclimatic reconstructions of past climate conditions. Such intervening variables, if undetected, may seriously threaten the accuracy and validity of dendroclimatic reconstructions of past climate conditions.

Wilmking et al [2005] summarize two earlier studies as follows:

In a recent study of treeline sites in northern Alaska, [Wilmking et al., 2004], where pronounced warming has taken place in recent decades [Hansen et al., 1999], we systematically sampled 1558 white spruce (Picea glauca) at thirteen sites in the Brooks Range and the Alaska Range. Opposing types of tree growth response (positive or negative) to temperature were demonstrated, with both types of trees occurring within a given sample site. In such cases, some individual tree-ring series contributed to a statistically significant relationship of the chronology with a particular predictive function of climate, while others degraded it. Even though these opposing growth responses were present in all sampled sites, their relative proportion varied between sites following patterns of regional and local moisture availability [Wilmking and Juday, 2005].

Then they proceed to generalize the phenomenon to many sites, including Tornetrask (TK) and Polar Urals (PU):

Here we present evidence that this phenomenon is not a regional abnormality, but is operating in several dominant tree species in forests across the circumpolar North (Figure 1).All trees with a significant positive correlation to site-specific mean monthly temperatures were grouped into “‹Å“”‹Å“responder chronology A”. All tree ring series with a statistically significant negative correlation to a site-specific mean monthly temperature (mostly July; an indication of drought stress) were grouped into what we call “‹Å“”‹Å“responder chronology B”. Trees could only be members of one responder chronology, and possible overlaps occurred only very infrequently. Trees with no significant correlation with the main climate indices were excluded from further analysis (for number of trees in each group; see Table 1).

As you see from Table 1, there are a lot of opposite responses in all the sites considered except Labrador (where there was a cooling trend):

They go on to conclude:

Without accounting for these opposite responses and temperature thresholds, climate reconstructions based on ring width will miscalibrate past climate .… Our findings suggest that the observed divergent response to climate at circumpolar treeline, overlapping the warming of recent decades, could be important for a significant proportion of the circumpolar forests and their dominant tree species.

Driscoll, W. W., G. C. Wiles, R. D. D’Arrigo, and M. Wilmking (2005), Divergent tree growth response to recent climatic warming, Lake Clark National Park and Preserve, Alaska, Geophys. Res. Lett., 32, L20703, doi:10.1029/2005GL024258. url

Wilmking, M., R. D’Arrigo, G. C. Jacoby, and G. P. Juday (2005), Increased temperature sensitivity and divergent growth trends in circumpolar boreal forests, Geophys. Res. Lett., 32, L15715, doi:10.1029/2005GL023331

Wilmking, M., and G. P. Juday (2005), Longitudinal variation of radial growth at Alaska’s northern treeline”¢’‚¬?Recent changes and possible scenarios for the 21st century, Global Planet. Change, doi:10.1016/j.gloplacha.2004.10.017,

Wilmking, M., G. P. Juday, V. A. Barber, and H. S. Zald (2004), Recent climate warming forces contrasting growth responses of white spruce at treeline in Alaska through temperature thresholds, Global Change Biol., 10, 1724– 1736.


  1. John A
    Posted Feb 12, 2006 at 4:23 PM | Permalink

    Growth of the divergent subpopulations correlate well prior to 1950 and diverge thereafter, supporting the idea that contemporary warming has introduced unprecedented stresses on some trees and thus impeded growth.

    I admit it. When I see stresses on trees, I try to help them, calm them down. Stress can be managed through proper relaxation and imaging therapy.

    To be serious, it’s not difficult to see why there will be two populations, one positive and one negative, within a population of trees: it’s called competition for light, water and nutrients.

    As those increase, one population will out-compete another, causing the “losers” to grow less than the “winners”. It’s not the climate that’s stressing them, its other trees and plants.

    If that’s unprecendented, then its been unprecedented for the last 4 billion years.

  2. Louis Hissink
    Posted Feb 12, 2006 at 4:36 PM | Permalink

    Re #1


    that seems to ring pretty true.

  3. Douglas Hoyt
    Posted Feb 12, 2006 at 4:55 PM | Permalink

    Do they report the correlation for all tree ring widths to temperature for the case where no trees are excluded? Just looking at the numbers in Table 1, it would seem that the correlation must be very close to zero.

    Perhaps all the paleoclimate studies should just drop all tree ring data.

  4. Steve McIntyre
    Posted Feb 12, 2006 at 5:00 PM | Permalink

    #3. I haven’t parsed through any data sets from this viewpoint, but tree ring widths are very non-normal, more like log normal. Maybe when you take averages, the mean is very biased; the big responders probably make an inordinate contribution to any correlation.

    The interaction of log-normality and positive and negative responders would be interesting to look at.

  5. john lichtenstein
    Posted Feb 12, 2006 at 5:26 PM | Permalink

    Why don’t people focus on tree-rings from sites where temperature ought to be the limiting factor because other inputs are saturated or constant? Maybe rain forest (water saturated, no dry lake beds) above the flood line (no silt deposits).

  6. per
    Posted Feb 12, 2006 at 5:39 PM | Permalink

    Have I got this right ?

    For a population of trees, there is an average response/ measurement. Some trees are above average, some are below average, and some are in the middle.

    Is there something I am missing here ?

    Not to put too fine a point on it, you could take some of the data itemised above to suggest that there is no relationship between tree growth and temperature.


  7. Dave Dardinger
    Posted Feb 12, 2006 at 5:40 PM | Permalink

    So what these researchers have discovered is basically a way to rectify (as in what a diode does to a current) tree-ring response. This whole idea, of course, was discussed here a while back. Apparently what we were discussing is close to reality. But did a similar thing happen at the time of the MWP? If so, has an adjustment been made to only count strong responders? I seem to recall from Steve’s hair charts (or whatever they are called) that there were some trees with much faster growth than others.

    The thing that bothers me is that unless the total group of trees are archieved and available, it won’t be possible to test various ideas as to what caused some trees to respond more to whatever than others. Throwing out a third or more of the trees sounds like a crime in the making if the raw data isn’t first archived.

  8. per
    Posted Feb 12, 2006 at 6:11 PM | Permalink

    Re: #7
    omigod. I can see this legitimising a whole new industry.

    We only archive the records of the trees that are positive responders, because we are not interested in the rest. We can therefore show a perfect correlation to temperature. I believe in a few good men…


  9. Steve McIntyre
    Posted Feb 12, 2006 at 6:27 PM | Permalink

    I just looked at the data set for Sol Dav,Mongolia (mong003.rwl). The gridcell temperature data starts in 1936. There are 36 cores ending after 1990. I don’t know how they got 33, but I’m surprised when stuff matches. I did correlaitons of ring width as is, and the delta from an RCS standardization. There were 19 positive responders on mean width and 21 with RCS standardization. The correlations in mean width ranged from -0.4 to +0.52 and fairly evenly in bins; there was exactly the same range after standardization.

    This stuff looks goofier and goofier all the time. It’s like falling down the rabbit-hole.

  10. John A
    Posted Feb 12, 2006 at 6:33 PM | Permalink

    Re #8

    Really? You reckon?

    How about this? There are two populations of trees within any sample which react differently to any climatic parameter, some thriving, the others either no correlation or "stressed".

    Come to think of it, if you analysed stock prices during the twentieth century, you might find that some stocks react positively to temperature and some are "stressed".

    I’m beginning to wonder if tree ring widths are like climate science’s version of Rorschach Tests: what you see in them depends on your preconceptions.

  11. Louis Hissink
    Posted Feb 12, 2006 at 6:35 PM | Permalink

    Re #8


    didn’t Stephen Jay Gould remark some time ago that data which produced nix results inevitably got put in the circular filing cabinet, or away in some other dark recess to be forgotten? Who publishes a paper on data that yield nothing?

    I just realised I am guilty of the practice because I never report negative results, only samples which reported a positive result. (Implicit in the reporting is that the rest are negative anyway).

    But a negative result is not necessarily a true negative result if incorrect to sloppy procedure occurred. And of course human error which happens all the time.

  12. john lichtenstein
    Posted Feb 12, 2006 at 6:43 PM | Permalink

    JohnA that stock market study would be great material. Maybe we could start an index of stocks that go up with tree growth.

  13. Douglas Hoyt
    Posted Feb 12, 2006 at 7:24 PM | Permalink

    Another thing that should concern people is the stability of the correlations. Do some positive responders later become neutral or even negative responders as time passes? And, of course, vice versa. Dividing the time intervals in half and comparing the responses would answer this question. Of course, if there is migration, then it becomes hopeless to use tree rings for climate reconstructions.

  14. Paul Linsay
    Posted Feb 12, 2006 at 8:12 PM | Permalink

    From the table there are 337 trees in category A and 343 trees in category B, 50/50 within statistical errors (remarkably close to 50/50). The only sites that show even modest deviation are MC with a 1 sigma difference and LB with a 2 sigma difference. I’d doubt their claim that LB shows any real correlation, it could easily be a statistical fluctuation. The data looks just like what one would expect from a random walk. Without producing a temperature independent criterion for placing a tree in A or B, this study would indicate that there is no way to use tree rings from these species as thermometers.

  15. Steve McIntyre
    Posted Feb 12, 2006 at 10:42 PM | Permalink

    #*. Per, Here’s what Wilmking et al. say (confirming your fears) eta l = D’arrigo, Jacoby, Juday:

    Selection of cores that have a coherent and consistent relationship with the climate variable of interest [e.g. Jacoby et al 2000] can thus greatly improve the accuracy of the estimates.

    That’s all we need.

  16. Paul Linsay
    Posted Feb 12, 2006 at 10:47 PM | Permalink

    Re #15: You’re surprised?

    A bit off topic but, this just in … http://www.scrappleface.com/?p=2177

    Snow Blankets East Coast Causing Global Warming

    by Scott Ott
    (2006-02-12) “¢’‚¬? A thick blanket of snow that covered much of the northeastern United States this weekend may increase global warming by preventing the heat that radiates from earth’s molten core from escaping into the atmosphere, according to former vice president Al Gore.

    Mr. Gore, a noted global warming expert who also once ran for president, dismissed suggestions that the biggest snowstorm in New York City history diminishes his case that the planet is warming at an alarming rate.

    “First of all,” Mr. Gore said, “the reason for all of this snow is that greenhouse gasses trap reflected solar heat causing the polar ice caps to melt, increasing the volume of oceanic water that evaporates, then freezes to become snow. The warmer the planet gets, the more massive snow storms we’ll see.”

    The former vice president, former senator and founder of the red-hot Current TV Network, said the only solution is to remove the snow, not just from the ground, but from earth’s water cycle.

    “We must pack the snow into giant containers and launch it into outer space,” Mr. Gore said. “Every day that George Bush fails to do this, the threat to Mother Earth grows exponentially.”

  17. Steve McIntyre
    Posted Feb 12, 2006 at 10:52 PM | Permalink

    #12 – john, if you look at Reply to Huybers #2, there’s a graphic (and I ought to re-present this in its own right) in which I substitute tech stock prices for bristlecones and then do an MBH98 type analysis using white noise for the other proxies – and out-perform MBH 15th century proxies using Mann’s “preferred” RE statistic.

  18. Posted Feb 12, 2006 at 11:11 PM | Permalink

    #15: So let me get this straight. All we have to do is, know what the climate was like over the life of the trees, then pick the trees which responded most directly to the climate variations we’re interested in. Then we can use that tree ring data to reconstruct what the climate was like over the life of the trees?


  19. Steve McIntyre
    Posted Feb 13, 2006 at 12:25 AM | Permalink

    See Jacoby on sites here http://www.climateaudit.org/?p=29

  20. beng
    Posted Feb 13, 2006 at 7:51 AM | Permalink

    It appears that unprecedented climatic stresses are triggering diverse growth responses between and within study sites that may greatly complicate dendroclimatic reconstructions of past climate conditions.

    That’s baseless speculation. How has it been determined that the climate stresses are “unprecedented” or natural?

    The work of mostly one person here is showing that these dendro studies are simply chosen (prb’ly just by eyeball) selections of the “proper” proxy series w/a few non-trenders tossed in, processed by a few lines of code, spliced at the end w/Jones proprietary surface record, and presto, a “temperature” graph. IOW, a sham.

  21. fFreddy
    Posted Feb 13, 2006 at 8:30 AM | Permalink

    When I see stresses on trees, …

    Maybe they should talk to the humans …
    &lt/prince charles&gt

  22. fFreddy
    Posted Feb 13, 2006 at 8:40 AM | Permalink

    Seems to me that “survivor bias” is going to be relevant somewhere here. All these tree cores are from trees that have survived for however many centuries until today: what data are we losing from those trees that lived a couple of centuries but then died in 1900 ?

  23. Posted Feb 13, 2006 at 8:42 AM | Permalink


    Well, the “bristlecone” series and some others do seem to have something unusual happening post-1900 or so.

    I’ve only seen data back to about 1400 (the MBH98 data) but a sudden change of that magnitude does not seem to be very common, especially before 1850.

    I assume dendrochronologists are sure that the ring width and density of wood never changes over the life of a tree (thus ruling out that the outer rings are different from the inner rings due to the outer rings being newer). If so, then the question remains what caused the unusual change (positive or negative) in growth of late. It could be stresses, or in fact it could be fertilization or some such.

    Take these graphs of the raw MBH98 data (three sets), for example. Ca535 (blue) looks like a random walk pretty much. Ca535 (red) has a little less predictable variance, but doesn’t do much unusual until about 1925, where it looks like it should come back down but stays high instead, then goes up. The tree(s) is/are growing larger rings than at any time since 1400, if the data is correct. Ca534 (green) spikes even earlier and harder. This is not particularly typical of the data, but out of the hundreds of data sets, it seems that a few do that (perhaps 5-10, I haven’t looked at every single one). Others have a weaker, but detectable, upwards swing in the 20th century.

    So, I think in the case of these few proxies, “unprecedented” (at least in terms of the last 600 years) is valid. But what’s causing it does not seem obvious to me, nor it seems to a number of scientists. It seems to me warming could explain it, but I believe at least in some cases it’s been ruled out. The CO2 transpiration theory seems reasonable but I’m not sure how it can be proven.

  24. jae
    Posted Feb 13, 2006 at 10:31 AM | Permalink

    #23 There are literally hundreds of studies that show CO2 has a fertilizing effect. See http://www.co2science.org/scripts/Template/MainPage.jsp?MerchantCode=CO2ScienceB2C&Page=Index.

    #1 Good grief! Talk about bias! It’s obvious that Driscoll has already decided to blame whatever he sees on AGW. I cannot believe these guys call themselves scientists. You can’t just ignore data that don’t show what you want it to show. Do these guys adjust ring width for tree age? I’m pretty sure ring width is strongly negatively correlated to age; and density is positively correlated.

  25. Steve Sadlov
    Posted Feb 13, 2006 at 10:35 AM | Permalink

    RE: John Lichtenstein’s comments about selecting proxies with most growth inducing factors held relatively constant.

    A Marine rain forest in the mid 50s of latitude would be interesting to assess. Such an area should in general have a quite moderate climate and only the big picture of global climate ought to result in any trend of more than a couple or three years. Even the odd drought due to a further-north-than-normal blockage from the oceanic semi permanent high pressure center tends to not affect areas north of the lower 50s in N latitude as much as areas further south. For example when Washington state is having a drought, northern BC may be near normal.

    So indeed, I would imagine that in Northern BC, and SE Alaska, temperature is the limiting factor for growth.

  26. Steve Sadlov
    Posted Feb 13, 2006 at 10:39 AM | Permalink

    RE: Al Gore Jr’s flawed understanding of heat flow from the earth. Crustal heat flow in the Eastern US is quite low. Contrast that with areas such as the thinned crust in the Great Basin or a mid oceanic ridge. Beyond that, while snow may affect the heat flow at the crust – atmosphere interface it is only a transient effect. Anyone running the PDEs will see that over time, there is no net effect. All such a thing would do is create a slight oscillation, but overall, the flow is unaffected. Boundary conditions 101 ….

  27. Dave Dardinger
    Posted Feb 13, 2006 at 10:42 AM | Permalink

    re#24 (and othera)

    Yes the proxy people do adjust for tree age. You’ll see it in anumber of Steve’s presentations. I’d have thought he’d have chimed in by now, but that should be clear if you read enough of the old posts. OTOH, it’s not clear (to me at least) that this is always done properly. It seems that there’s a subjective aspect to how the growth curves are adjusted based on the shape of the curve and this could result in a subtle bias in how the long-term trend comes out. The errors may all cancel out, but it should be kept in the back of ones mind that the curve-fitting was done and so one should ask if this might might impart a bias in a particular situation. And, of course, the raw data should be archived to allow the adjustments to be double checked in case of questions.

  28. Dave Dardinger
    Posted Feb 13, 2006 at 10:45 AM | Permalink

    Steve S,

    That Gore “quote” was a satire, not something he really said. You just need to look at the source, “Scrappleface” to be sure of it.

    I admit I was startled when I read the first couple of sentences, but that’s how a good satire wants you to react.

  29. beng
    Posted Feb 13, 2006 at 11:16 AM | Permalink

    Thanks for the comments, Nicholas.

    My point about the Bristlecone/Foxtail/arid-mountain-site trees is that their unusual 20th-century growth-spurt is rare in the totality of tree-ring proxies (I think that’s been demonstrated satisfactorily on this site). Look at the “Upper & Lower Bristlecone site” thread here, and it’s seen that Bristlecones on the very same mountain but lower show no such growth-spurt. So how can prb’ly non-temperature differences (unless there’s a big temp-trend difference between trees prb’ly within easy sight of each other :)) between very-closely located groups of the same tree-species be used to infer GLOBAL TEMPERATURE rise? It’s not science — it’s patently ridiculous.

    An extreme “test” of the growth-spurt question is: Even if most trees were showing a recent growth-spurt (and they’re not), how much of that chould be attributed to temp rise? It could also be CO2 fertilization, increased precip, nitrogen/dust deposition, even lower temperatures if they have a upside-down quadratic temp-growth relation!

    From a general aspect, I think the RW/density of tree rings is far too problematical to assume linear relations w/temp, especially from precip-limited arid sites. If anything could be salvaged from tree-rings, I’d think it’d be from cold, wet sites, like the Cook et al. (1993?) Tasmanian study. These seem the least precip-limited & most cold-limited tree sites. Some of the other proxy-types have simpler, more direct indication of past temperature, like boreholes and cold-limited treelines. But those’re just my laymen’s thoughts.

  30. Steve McIntyre
    Posted Feb 13, 2006 at 11:21 AM | Permalink

    #24. They adjust for age and this is a big topic with them. The usual method is a negative exponential for ring width and a straight line (slightly declining) for density,

  31. Steve McIntyre
    Posted Feb 13, 2006 at 11:23 AM | Permalink

    #29 – see my recent post on Briffa for a more detaild summary of evidence showing late 20th century decline on a wide scale

  32. Steve Sadlov
    Posted Feb 13, 2006 at 5:46 PM | Permalink

    While the Gore quote is satire and I took it hook line and sinker, nonetheless, it is satire that cuts close to the bone. It is not far off of the real thinking of “Earth In Balance!”

  33. Willis Eschenbach
    Posted Feb 13, 2006 at 9:28 PM | Permalink

    The cool thing with trees is, they’re either thermometer (which works normally for “positive responders” and works inverted for “negative responders”, or, they’re a rain gauge … or both …

    Cut down a tree, discover the next big drought
    Tree rings used to predict droughts
    Current methods may be inaccurate

    Feb. 11, 2006. 01:00 AM

    Engineers need to design and build dams big enough so that the reservoir will still have water during bad droughts. Otherwise, irrigation systems, municipal water supplies and even hydroelectric generation could all be in trouble.

    Yet, how can engineers determine whether a really bad drought is likely once every 50 years, every century or every 400 years? How do they avoid underbuilding or overbuilding the dams and reservoirs?

    Devices that continually measure the flow of water in Canada’s rivers, called stream gauges, can convey a misleading picture because there’s seldom more than 40 or 50 years of recorded data.

    So, more and more engineers are turning to substitute means to measure past stream flow. The most useful is turning out to be tree growth, recorded for posterity in the width of each year’s tree rings.

    Good rainfall normally means healthy tree growth and rings that are wider than normal, while drought curbs growth, leaving narrower rings.

    The usefulness of this approach is shown by findings just published in the Canadian Journal of Civil Engineering, put out by the National Research Council of Canada.

    Using readings from ancient tree rings, researchers David Bonin and Donald Burn showed that a particularly severe drought was likely in Alberta’s Athabaska River basin every 100 years rather than every 400 years as suggested by the short-term records from stream gauges.

    Now, all we need to do is determine which category a tree is in, positive, negative, or rain … then deal with the inverted quadratic response, and we’re there!

    Easy money …


  34. Steve McIntyre
    Posted Feb 13, 2006 at 10:28 PM | Permalink

    Here are some interesting comments from a review article in Botanical Review about 1952 (I’ve got a copy and can’t determine the exact citation for now:

    Giddings (1941) in Alaska was troubled by finding more than one chronology. A spruce chronology on a gravel terrace differed from surrounding trees and gave a radical departure from the usual climatic reaction, being due perhaps to a combined rainfall and temperature effect. In contrast a different chronology from timberline gave foog agreement with summer temperature. Trend analysis between June temperature and tree growth showed 21 parallel and 8 opposite. Mean June temperature and a group of trees from NOatak had 17 parallel and 11 opposite trends. In comparison June precipitation and Norton Bay trees gave 14 parallel and 14 opposite trends. …LAter Giddings (1951) reported on two grops of spruce trees 50 miles apart and 80 miles from the weather station. The two groups showed 31 parallel and nine opposite trends. Compared with growing season temperture, June-July. each group gave 20 parallel and 13 opposite trends.

    Sounds like positive and negative responders.

  35. Posted Feb 14, 2006 at 5:19 AM | Permalink


    What I’m saying is, something unprecedented in the last 600 years happened to those trees. But you’re right that, this is only a few tree samples and we don’t actually know what caused it. I agree that making the leap from some trees going through an unusual stage in the 20th century to saying that’s proof of warming specifically is silly, especially since as you pointed out, if it were warming we’d expect nearby trees to be similarly affected.

    Yes, I’d say careful selection of which tree sites to sample WITHOUT looking at the actual data itself is important. Pick those you EXPECT will show the correct relationship, then analyze the data. Don’t pick and choose those data sets which suit your purpose.

    As I’ve mentioned before, looking at the raw data, the MXD density data seems to contain more sensible information than the ring widths – certainly a higher signal/noise ratio. Whether that translates into them being a good representation of temperature I don’t know. But someone on this thread earlier commented that there was a greater correlation found between MXD and temperature than between ring width and temperature. Having compared both to O18 and similar non-dendro proxies, I think there might be something in it.

    Steve – fascinating. That makes some kind of sense. I guess some individual trees just grow differently from others? Now if only we could work out what causes it and how to test for it…

  36. TCO
    Posted Aug 5, 2006 at 9:02 AM | Permalink

    1. Is there a bimodal distribution (two humps)? How does one differentiate (statistically) “positive and negative responders” from expected variation within a group?
    2. I like the looking over time concept (see if groups stay the same).
    3. Do the papers identify or posit a physical rationale for why some are positive and others not? Could one use whatever “tag” that they come up with, to do (or at least design) a follow-on study that would look at new trees, with the “tag” and see if posited relation to +/- is really characteristic of the stands in general?
    4. JohnA’s comments seem overly strong and simplistic. Maybe if he talked about how competition would vary with temp, it would be more interesting. Or if he identified some way to measure competetiveness (perhaps location of +/- next to each other) that would be more helpful.
    5. Let’s say that the researchers are right and that one of the effects of higher temps is this kind of divergence into two populations. How could we use or recognize this b ehavior in the past to make sure that MWP is properly assessed? Would we look for a widened standard deviation as a signal of this kind of warming?

  37. John Creighton
    Posted Aug 5, 2006 at 9:49 AM | Permalink

    I think if Tree ring people could solve these problems and actually get good temperature extrapolation from trees alone, they would be winning fields awards. The mathematical challenges seem overwhelming to me from a theoretical, numerical and computational perspective.

    I think that any good attempt would not be trying to relate mean temperature to mean tree growth but distributions in annual temperature to distributions in tree growth. However, it is not event that easy because there are so many factors affecting tree growth, thus we need joint distributions. We need to know how these distributions propagate with time (survivor bias). There will be many nonlinear effects and the optimization will be extremely computationally challenging. There could be an infinite amount of local maximum and local minimum. There could be even times where it is mathematically impossible to deduce the temperature because all of the needed information about the warm periods is lost.

  38. TCO
    Posted Aug 5, 2006 at 9:56 AM | Permalink

    Things won’t get better until mathematicians and statisticians and physicists and such invade the tree ring field and start doing science there. And it won’t be an immediate change. will take a while. Steve is at the pointy end of the spear, penetrating the soft underbelly.

  39. Steve McIntyre
    Posted Aug 5, 2006 at 10:05 AM | Permalink

    #37. Here’s an interesting result of the House Energy and Commerce committee hearings. As a result of people seeing my presentation, I’ve now received my very first invitations to speak at a university – not by climate scientists, but by a statistician and an econometrician. Both at very good universities with terrific statistics deparments. Both invitations came from very junior people. I’m hopefully of interesting some young statisticians in these issues – I’m going to try to introduce them to tree rings as a statistical problem.

    I’m surprised that so few have seized on the opportunities. In the mining business, competitors would follow up on anything that received as much publicity as this. Given how slow I’ve been in following up with reports on the other Hockey Team studies, I’m amazed that some young statistician hasn’t stepped in.

  40. TCO
    Posted Aug 5, 2006 at 10:13 AM | Permalink

    You have done a poor job of proselytzing within the academic channels. Only one real publication (GRL05) And no, the replies to comments and EE papers (not even abstracted by science citations) do not count! Nor do blog posts. WRT invitations to universities, sure it is surprising that you don’t get invites, but why not be proactive and ask for some.

  41. TCO
    Posted Aug 5, 2006 at 10:14 AM | Permalink

    Congrats on the stats invite and have fun.

  42. Steve McIntyre
    Posted Aug 5, 2006 at 10:26 AM | Permalink

    TCO, I’ll publish some more stuff when I’m ready. I’ve had my reasons for doing things the way I have, and, for all the disparaging comments that I get about how I could have been more effective if I’d done things differently, I’m not dissatisfied with the impact.

    The NAS and Wegman reports really change the situation for me. Any papers that I submit now will have more impact than they would have 6 months ago. I suspect that I’ve probably not even lost much time, since the review processes at journals are going to be more expedited and less adversarial for anything submitted now than they would have been pre-reports.

    If I’ve not submitted anything else 6 months from now, then bring the topic up again, but for now, please give it a rest.

  43. KevinUK
    Posted Aug 5, 2006 at 11:19 AM | Permalink


    Steve (unlike Mann et al) has learned from his experiences. When he’s ready to publish he will publish. He won’t do this before he thinks it is all right. You can see how busy he is, so lay off the constant taunts as they are likely to if anything have the opposite effect you would like i.e. Steve will end up wasting his time answering you posts when he would otherwise be spending the time researching and documenting his analysis/finding.


  44. TCO
    Posted Aug 5, 2006 at 11:55 AM | Permalink

    I totally disagree if for only the reason that being forced to write papers will make his sharpen his points and make them more fair. I guess it’s his caboose. But I’m not JUST giving him advice on how to prosecute more effectively*, I’m sticking up for SCIENCE. Not advocacy.

    *and papers WOULD do so…to be honest, when I don’t see them, but see the little snippet gibe type analyses, it makes me think Steve’s “case” may not be that strong, given he won’t leave his dojo.

  45. TCO
    Posted Aug 5, 2006 at 11:57 AM | Permalink

    I guess I can give it a rest, since I’ve made the point.

  46. bender
    Posted Aug 5, 2006 at 1:55 PM | Permalink

    One lesson to learn from this whole process is that the “publish or perish” ethic can lead to a mad race toward premature conclusions. A little pressure may be a good thing. But alot of pressure can be bad. Like the U-shaped response curve of a tree to its environment, you don’t want to go over the top and have that positive response go negative.

  47. TCO
    Posted Aug 5, 2006 at 2:01 PM | Permalink

    The problem is not people publishing, when they shouldn’t. It’s not being honest about limitations. Read the classic by Katsoff.

    Posted Aug 5, 2006 at 2:16 PM | Permalink

    TCO, in the light of the obviously few papers Steve has submitted to the scientific journals put up against what he has achieved till now, I certainly find it a bit unfair to keep taunting him. I´m shure that Steve must be a bit overwhelmed by his succes till now and need to let things settle for a while. I for one am sure that Steve knows himself what comes next and when!

    Hans Kelp

  49. Pat Frank
    Posted Aug 5, 2006 at 2:25 PM | Permalink

    #37 — The problem of finding a temperature signal in tree rings has been discussed extensively here in the past. The central problem is that no one wants to do the hard, long-term greenhouse studies correlating tree rings with physical conditions by doing experiments that vary one parameter at a time.

    But just now, an interesting possibility occurred to me. Maybe Dave Dardinger could comment on the idea. That is, the C-12/C-13 kinetic isotope effects of metabolic processes are well known — or should be, in any case. The ratio of C-12 to C-13 in the carbon framework of each tree should follow the isotope effect. There is a temperature component (Boltzman’s kT) in the exponential term relating the reaction rates.

    If all the metabolic reaction rates producing, e.g., cellulose were known along with their isotope effects, the ratio of C-12 to C-13 in the product cellulose ought to allow one to back-calculate the temperature at which the cellulose was synthesized by the tree.

    Mass spectral analysis of the tree rings, counting the C-12 and C-13 atoms, should then allow an explicit year-by-year temperature reconstruction that ought to be free of the quadratic effect illustrated at the top of the “Survivorship Bias” thread, and also solve the divergence problem.

    The temperature of the growth wood of the tree should be very close to the external temperature, but even if somewhat removed it should be possible to come up with a scaling function that produces the temperature just external to the bark, relative to the temperature deduced at the growth wood.

    Dave, what say you?

  50. Steve McIntyre
    Posted Aug 5, 2006 at 2:38 PM | Permalink

    Since you ask, what about dC13 in bristlecones at Sheep Mountain – wouldn’t that be interesting – read about it here

  51. KevinUK
    Posted Aug 5, 2006 at 2:43 PM | Permalink


    To placate TCO and perhaps some others, might I suggest that, as has now become standard practice on open source software development projects that you perhaps put a ‘roadmap’ on your blog. Perhaps it may also be a good idea to start another thread to canvas visitors for ideas on what should be on the roadmap and what the priority of tasks should be? Even if you don’t want to canvas opinion, I still think we would like to know what your future plans/thinking are.


  52. bender
    Posted Aug 5, 2006 at 2:50 PM | Permalink

    Re #49
    This will be a discussion worth listening to. (Tree ringers know a fair bit about tree ring chemistry. Would be interesting to know if they’ve thought of this angle yet.)

  53. bender
    Posted Aug 5, 2006 at 2:53 PM | Permalink

    Bibliography of Dendrochronology provides this for a search on “C13”:
    Libby, L.M., Pandolfi, L.J. 1973. Measured temperature coefficients of C13/C12 and O18/O16 in European oaks. In: Proceedings of the International CLIMAP Conference. Norwich, Norfolk, United Kingdom: 21-39.

  54. Dave Dardinger
    Posted Aug 5, 2006 at 2:57 PM | Permalink

    Pat, I think someone, probably Bender, posted a link to something like that here recently concerning disentangling isotope signals from evaporation from those of rainfall.

    One problem I see is that most of the isotope separation would come at the step where CO2 from the air is captured. I’m not sure of the exact details, but I know that grasses tend to capture energy from the sun during the day and then used the captured energy to capture CO2 at night when it’s cooler and they can open their stomata to let more air in without losing too much water in the process. That’s one big aspect of the CO2 fertilization effect, for instance; it lets non-grassy plants keep their stomata more closed thus saving water loss.

    But that brings up one thing to be considered with trees. Since they aren’t grasses, they must open their stomata while they’re photosynthesing, and this means that to be effective they must be able to afford to lose the water which escapes while they’re capturing CO2. So it’s precisely when it’s warmest, and thus they’re most able to produce sugars for growth that they’re most likely to be vulnerable to water stress. If it’s either not THAT warm so not too much water is being lost or they have ample water availability they will grow lots, but if its really hot and they don’t have enough water then the opportunity is lost which would show up as smaller rings than there might have been.

    But as far as the isotope composition of what growth there was, that might be tricky. If you had dry ground but a wet spring and lots of growth in early summer then less growth during the hotter weather then it’d look isotopically like it was a cooler year than if you had lots of carryover moisture in the soil and this allowed plenty of growth in the hottest part of the summer even though the actual average temperature was higher. I suppose if you could actually measure isotope composition from various parts of the year this would help untangle things but it sounds tricky.

  55. James Lane
    Posted Aug 5, 2006 at 3:25 PM | Permalink


    I guess I can give it a rest, since I⢢¯ve made the point.

    That would be great, because you’ve made the point maybe 100 times.

    I’m sure that Steve has publications in the pipeline, and it would be improper for him to presage them on the blog.

    That said, there has been an explosion of quality thinking on CA in the last couple of months – I don’t know if that’s due to the profile of the NAS panel or the vindication in the Wegman report (or neither) but it’s very entertaining and worthwhile.

  56. Steve McIntyre
    Posted Aug 5, 2006 at 4:03 PM | Permalink

    #51. Kevin, can you point me to a template for what you have in mind?

  57. john lichtenstein
    Posted Aug 5, 2006 at 6:37 PM | Permalink

    SteveM you might want to touch base with Wegman about his August 9 jsm presentation. Now that you have shown everyone the directions to Flawed Study Mountain don’t be surprised to find rival prospectors showing up. You might want to think about work you would like to see other people doing. For example it would be good to have a real EM expert looking at the paleo RegEM stuff. I remember a post here a while ago that someone thought it was impossible to have verification statistics when using EM. Oh boy.

  58. Pat Frank
    Posted Aug 5, 2006 at 6:51 PM | Permalink

    #50 — Thanks, Steve. Tang 1999 looks very interesting, and it may be that the data are already in hand to test whether a carbon kinetic isotope temperature could be calculated from tree ring ratios. It’s peculiar to me that there is a specific stromatal diffusion rate difference between 12-C and 13-C, but it means that the eventual isotopic ratios in the tree ring organic matrix reflects more than just the metabolic rate differences.

    #54, Dave, thanks for your comments. I suspect that seasonal climatic variations ought to factor out in a kinetic rate ratio, because both carbon isotopes of CO2 would be influenced identically, leaving nothing but a temperature signal. I don’t have time just now to look into the idea and it’s been a very long time since I did any physical organic chemistry, but maybe it would eventually be worth checking out. It certainly has the promise of producing results faster than generation-long greenhouse growth parameterization studies. 🙂

  59. Nordic
    Posted Dec 12, 2006 at 1:23 AM | Permalink

    RE #34: I don’t know if this has been posted on CA before, but some of Giddings publications (and many other interesting papers can be found here:http://www.treeringsociety.org/TRBTRR/TRBTRR.htm

    Just a teaser: The 1947 Mackenzie River delta ends with this sentence: “Whaterver the cause, we can look to these northernmost American trees for proof of borderline stability for at least five centuries”

    It is too bad their chronologies did not extend back further – eyeballing the graph seems seem to get interesting in the 1400s, not to mention at the right-hand side of the graph in the ’30s.

3 Trackbacks

  1. By Wilson et al 2007 « Climate Audit on Feb 24, 2011 at 5:53 PM

    […] 2005,, also see ppt here, coauthored by Wilmking, which was previously discussed in CA in the post Positive and Negative Responders and are type examples for Wilmking’s positive and negative responders. (Also see previous […]

  2. […] widths going down while temperatures go up. Previous discussions included: Wilmking in Alaska and Positive and Negative Responders. Other related posts include Upside Down Quadratic, Twisted Tree Heartrot Hill , bender on Gaspé […]

  3. By Survivorship Bias « Climate Audit on Nov 12, 2011 at 10:13 AM

    […] Down Quadratic Twisted Tree Heartrot Hill Wilmking in Alaska Positive and Negative Responders bender on […]

%d bloggers like this: