Taimyr is one of the sites in Esper et al [2002]. It does not make a material contribution to any hockey stick-ness in Esper. The authors have some very interesting things to say about AGW which I’ll post up in a day or two. For now, I want to post up the following graph, which is of great methodological interest to me. Using one site age curve is sweeping the field in dendroclimatic studies. While the various authors genuflect towards the need for homogeneity, there is precious little evidence in their articles of operational procedures to ensure it.
A big and obvious difference and potential source of bias is that some of the newer reconstructions are being done by stitching together cores from live trees and cores from dead trees. Here’s a remarkable graphic from Naurzbaev et al [2002]
Original Caption: Figure 2 The two “Åregional’ age curves (1) obtained for cores and stem wood samples taken at a height of 0.5–1.0 m above ground and (2) as derived from discs of dead trees where the lower part of the stem is preserved. An offset in the curves is required to account for the sampling height bias and to produce a single valid standardization reference curve.
The take-home message to me from this graphic is that the usual negative-exponential age-standardization curve will penalize subfossil samples in their early portion and make the chronologies seem colder than they really are.
Reference: Mukhtar M. Naurzbaev, Eugene A. Vaganov, Olga V. Sidorova and Fritz H. Schweingruber, 2002. Summer temperatures in eastern Taimyr inferred from a 2427-year late-Holocene tree-ring chronology and earlier floating series, The Holocene 12(6), 727–736
Climate Audit 
22 Comments
Steve,
I realize that this is work in process and that you are exploring things, looking for problems (not nescessarily making an accusation…yet). However, it is still darn hard to even tell what you are driving at, what hypothesis you are formulating (for examination).
1.
Huh? What is this in nitty gritty? Is it better to have a single curve or multiple ones? I thought the actual dendroCHRONOLOGY was not in much debate?
2.
Homogeneity is a big concept. What are you talking about here (avoiding confounding factors, accounting for them with a multiple regression)? Are you just talking about factors related to chronology (previous sentence) or the whole moisture, CO2 etc litany? I just don’t get the thought thread.
3.
So what is the possible problem here? Does the wood dry out? Is your concern age based, juveline spurt and all that (in which case, you can have age issues regardless of dead/alive, just depending on ring age within the core) or is it actually a problem of live and dead wood?
3.
Why is it remarkable? Explain to me as if I were a plodding linear thinker. 😉
4.
a. What is the “usual negative..blabla”.
b. Why does it expecially penalize subfossil samples?
c. What about the curve suggests this?
d. Does “subfossil” mean “dead” (graph caption implies that stem is still rooted, even for dead samples)?
(now parsing the figure caption, not your work…but maybe you can answer questions)
5. The two “regional’ age curves (1) obtained for cores and stem wood samples taken at a height of 0.5–1.0 m above ground and (2) as derived from discs of dead trees where the lower part of the stem is preserved. An offset in the curves is required to account for the sampling height bias and to produce a single valid standardization reference curve.
Re:
” Why is it remarkable? Explain to me as if I were a plodding linear thinker ”
Well perhaps it’s remarkable because Taimyr is above the treeline? Ain’t no trees in Taimyr:
http://sciborg.uwaterloo.ca/~twdedwar/reprints-pdf/2000-qr-wolfe.pdf
What I had in mind was this: if you are standardizing all your subfossil cores using the curve in the Top Panel, when you should be using the curve in the Bottom Panel,you are going to screw up all your comparisons. If you do that for every site, every site chronology done this way will be screwed up and will be biased in one amount or another towards a high 20th century. It would affect every single one of Esper’s chronoogies. I’m amazed that Schweingruber is a listed author on both papers. I wonder how much input he actually had in the papers, since this is an important difference between them.
I don’t know what you mean (mechanically) by “standardize” a core.
Are you concerned that cores are being misdated?
Are you concerned that dead and live wood have differet widths (different populations) based on expansion/contraction or based on difference among the two in age or in what types of samples survive (which)? IOW that one should include liveness/deadness as another confounding factor that needs to be solved for in the multiple regression?
Interesting difference between dead & living trees. Some kind of sampling bias? Wildly guessing, perhaps the dead trees started out “suppressed” by shading, less precip, colder temps, etc (and gradually released). Maybe the “germination date” of the dead samples are on average older than the living ones, and conditions during that older period harsher.
However, if the above curves are the basis of an eventual temp vs RW calibration curve, wouldn’t they just apply the appropriate dead or living curve to the various samples?
Re #5: in this case, they probably did. I’m going to post on it. But Teimyr doesn’t show a 20th century hockey stick. It’s the potential impact of this effect on other sites, where the effect was not discussed or addressed: e.g. Polar Urals, Tornetrask, Gotland etc. etc. I’m pretty sure that Esper did his own RCS calculations on Taimyr, so it might apply to Esper’s (but not Naurzbaev’s version of Taimyr.)
This is one more reason why I’d like to see Esper’s component series (as other people). If there’s a difference between Naurzbaev and Esper on this site, I’d like to know about it.
Allow me to pin this down so a non-Nobel-Prize-math-contest-three-years-younger-beating-person (;)) is sure of the inference:
1. Is your point that “liveness” is a seperate confounding factor that needs to be accounted for?
2. Are you just worried that it may be an issue or feel strongly that it is one and is not being dealt with?
3. Do you think that “age compensation” takes care of the possible confounding factor of “liveness”.
4. What do you think is causing the difference in ring width in the two populations (speculate freely please). Is it sample selection/survival? Age (therefore maybe covered in (3)? Wood expansion/contraction? Insects? Other?
At present, just worried about (2). There’s a lot levered on these tree ring standardizations and I’m just paying attention to things as I notice them. I’ll probably come back to this issue.
I don’t know whether the popluations are different. Unfortunately the original scientists don’t mark enough information about each sample to be able to extract the relevant information – for example, I’d like to know the altitudes. Based on this Figure, it seems like you need to know where the core was taken for each sample i.e. was there a stump. I’ll post up some pictures on Taimyr which are interesting and may answer the question.
While I’m thought of as a nihilist on paleoclimate, I like most articles written by non-Hockey Team authors. They describe interesting details, do not spend a lot of time on poorly done statistics and are not in a rush to get to an inflammatory conclusion.
Still don’t know what you mean by “standardization”. Is this the chronology? Or is it the procedure of deconvoluting confounding factors?
TCO, I believe “standardization” is the process of modeling the “normal” growth of the trees (w/o e.g. temp variations) on the theory that the ring width/density/whatever depends in part on the age of the tree, and then normalizing the raw numbers by applying an inverse function.
Thus, environmental effects are inferred from departures from the normalized numbers, not from the raw numbers. In recent posts, Steve seems to be showing that typically, a negative exponential curve is used to model the “normal” growth; at first glance, this looks like it would work OK for curve 1 in Fig 2 at the top.
However, the initial positive slope of curve 2 in that figure certainly doesn’t fit such a model curve. Were they modeled by such a curve, the early years of trees in that population would (artifactually!) seem to have severely “environmentally” depressed ring widths after normalization, leading to an erroneous “cool” bias for early years of a temperature reconstruction.
Steve, please correct me if I’m wrong; like most of your readers, I haven’t bothered to read the actual papers!
Thanks armand. So it has nothing to do with the wiggle matching of actual chronology. And is just the process of deconfounding for age dependancy of ring size?
I guess I’m still a bit confused about how to read the graph above, though. The “age in years” has two different scales. That is the age of the trees themselves? ARe they all from the same epoch? Or is that a chronology? Why the offset?
My informed guess on the Age in Years: Naively, I would expect it to be “the age of the tree at the year when the tree ring was laid down”. However, I think what it really represents in that graph is more like “count of years from the earliest ring visible on the sample to the ring of interest.” If the ring from year 1 of the tree’s life is visible on the sample, then both definitions are the same; however, as hinted at in Steve’s description and the quotes, typical samples from living trees don’t included the earliest years of the tree’s life (core taken from above ground level, so years until tree grew that high are missed; core didn’t intersect center of tree, so years near core missed; etc). For dead trees OTOH, the best samples may be full slices from ground level, which would tend to include early rings missing from the live sampling.
To make the plot, for both “living” and “dead” series, for each “ordinal sample year” they average the ring widths for that “age” year from all the trees in the series (e.g. all 100th year rings from “living” samples) and plot that average width for that sample year. Thus, the extent of the x axis shows the sampled lifetime of the trees.
The “offset” exists (presumably) because of the missing early years in the “living” samples (of course, the number of missing years needs to be pretty uniform for this to work). They inferred the alignment by matching the curves.
great. that explains the offset, the lack of juvenile growth for living trees and the top and bottom scales.
I guess this study is also all of trees of the same epoch? It’s still interesting to wonder how do you deconvolute for the age effect versus the climate effect (for instance on your “calibration runs” to determine “age effect” won’t there also be long term climate trends affecting the overall curve? Especially over last 200 years, where we know climate has warmed.
Still think it would be good to have some better calibrations done. Either by very careful use of locations where temp was recorded for a long time instrumentally (or some other very accurate proxy) or via greenhouse measurements.
I still find it somewhat disconcerting that people are using healthy tree growth as the predictor of our doom.
My Bayesian level estimate is that GW is happening at a 3:1 likelihood (nothing like a scientific consensus that only evil babykilling creationist dubyas would disagree with…but still…more likely than not). I’m not so sure that it will not taper off or that the real effects will be that bad. I mean, I would love it if VA was more like NC let alone like FL. I love alligators…
I’m hoping for cooling — I think Florida would be much nicer if it were about 15 F cooler!
Back to the science, I would hope that the trees were from different epochs, not from the same one, so the per-ring-width averagings would each be over a variety of different conditions. I definitely agree on the need for a lot more good quantitative work at the boundary between tree physiology and ecology.
“I’m not so sure that it will not taper off ”
Considering that the thermal effects that govern the behavior of the process are logarithmic and will effectively reach 0 at some point, it cant do anything but taper off
Depends on the forcing function (for instance rate of carbon introduction) and also if we change into a strongly positive feedback regime in the system. Of course, in the pedantic sense that we will not have infinite temperature, the effect will level off somewhere. But of concern would be a radical move up.
Addition of CO2 the atmosphere = Warming is the process that I’m speaking of. That s the process that is logarithmic. There is an upper limit based upon atmospheric density, solar output etc etc etc.
re: #19
Please have a look at these two links
http://home.casema.nl/errenwijlens/co2/howmuch.htm
http://www.grida.no/climate/ipcc_tar/wg1/519.htm
The absorption of CO2 is indeed logarithmic (as Arrhenius observed), and is therefore a constant for every CO2 doubling. The ‘hot’ SRES scenarios however, assume an exponential growth, leading to an increasing temperature trend.
Hans (currently from USA)
ref:
Arrhenius, S, 1901, Ueber die Wàƒ⣲meabsorption durch Kohlensàƒ⣵re, Annalen der Physik Bd 4. 1901, p690-705.
http://home.casema.nl/errenwijlens/co2/arrhenius1901/index.html
Absorption of CO2 into what is logarathmic. Trees, water, soda Pop?
Again you have sompletely and totally mis-understood what I have said.
ET, I think you are “spouting words”, but don’t feel like bothering more with this.