the Air Vent

Because the world needs another opinion

Baby Dendro’s

Posted by Jeff Id on July 30, 2009

How is dendroclimatology taught to the college kids?

This is from an on-line text for undergrad climatology students HERE.  It has some goodies in it which I found  interesting.  They discuss the mythical thermometer tree, statistical processing of data and assert the validity of the methods used.

It’s good to know that parents who put up the 15K/yr are getting their money’s worth.  I bolded the stuff which I take issue with.

=======================

Dendroclimatology

The study of the annual growth of trees and the consequent assembling of long, continuous chronologies for use in dating wood is called dendrochronology. The study of the relationships between annual tree growth and climate is called dendroclimatology. Dendroclimatology offers a high resolution (annual) form of palaeoclimate reconstruction for most of the Holocene.

The annual growth of a tree is the net result of many complex and interrelated biochemical processes. Trees interact directly with the microenvironment of the leaf and the root surfaces. The fact that there exists a relationship between these extremely localised conditions and larger scale climatic parameters offers the potential for extracting some measure of the overall influence of climate on growth from year to year. Growth may be affected by many aspects of the microclimate: sunshine, precipitation, temperature, wind speed and humidity (Bradley, 1985; Fritts, 1976). Besides these, there are other non-climatic factors that may exert an influence, such as competition, defoliators and soil nutrient characteristics.

There are several subfields of dendroclimatology associated with the processing and interpretation of different tree-growth variables. Such variables include tree-ring width (the most commonly exploited information source, e.g. Briffa & Schweingruber, 1992), densitometric parameters (Schweingruber et al., 1978) and chemical or isotopic variables (e.g. Epstein et al., 1976).

A cross section of most temperate forest tree trunksNot all trees are suitable for dendroclimatology. Many trees in tropical areas where growth is not seasonal do not reliably reflect varying conditions and are unsuitable as proxy indicators of palaeoclimates. will reveal an alternation of lighter and darker bands, each of which is usually continuous around the tree circumference. These are seasonal growth increments produced by meristematic tissues in the cambium of the tree. Each seasonal increment consists of a couplet of earlywood (a light growth band from the early part of the growing season) and denser latewood (a dark band produced towards the end of the growing season), and collectively they make up the tree ring. The mean width of the tree ring is a function of many variables, including the tree species, tree age, soil nutrient availability, and a whole host of climatic factors (Bradley, 1985). The problem facing the dendroclimatologist is to extract whatever climatic signalThe climate signal may be regarded as the response due to an identified forcing factor, as opposed to inherent random climatic variations (the noise) is available in the tree-ring data from the remaining background “noise” (Fritts, 1976).

Whenever tree growth is limited directly or indirectly by some climate variable, and that limitation can be quantified and dated, dendroclimatology can be used to reconstruct some information about past environmental conditions. Only for trees growing near the extremities of their ecological amplitudeThe range of habitats that a tree can grow and reproduce within is termed the ecological amplitude., where they may be subject to considerable climatic stresses, is it likely that climate will be a limiting factor (Fritts, 1971). Commonly two types of climatic stress are recognised, moisture stress and temperature stress. Trees growing in semi-arid regions are frequently limited by the availability of water, and dendroclimatic indicators primarily reflect this variable. Trees growing near the latitudinal or altitudinal treeline are mainly under growth limitations imposed by temperature; hence dendroclimatic indicators in such trees contain strong temperature signals.

Furthermore, climatic conditions prior to the growth period may precondition physiological processes within the tree and hence strongly influence subsequent growth (Bradley, 1985). Consequently, strong serial correlation or autocorrelation may establish itself in the tree-ring record. A specific tree ring will contain information not just about the climate conditions of the growth year but information about the months and years preceding it.

Several assumptions underlie the production of quantitative dendroclimatic reconstructions. First, the physical and biological processes which link toady’s environment with today’s variations in tree growth must have been in operation in the past. This is the principle of uniformitarianism. Second, the climate conditions which produce anomalies in tree-growth patterns in the past must have their analogue during the calibration period. Third, climate is continuous over areas adjacent to the domain of the tree-ring network, enabling the development of a statistical transfer function relating growth in the network to climate variability inside and outside of it. Finally, it is assumed that the systematic relationship between climate as a limiting factor and the biological response can be approximated as a linear mathematical expression. Fritts (1976) provides a more exhaustive review of the assumptions involved in the use of dendroclimatology.

The general approach taken in dendroclimatic reconstruction is:

1) to collect (sample) data from a set of trees (within a tree population) which have been selected on the basis that climate (e.g. temperature, moisture) should be a limiting factor (Bradley, 1985);

2) to assemble the data into a composite site chronology by cross-dating the individual series after the removal of age effects by standardizationYounger trees generally produce wider rings, and this confounding influence of age on growth must be removed before analysis. Standardisation is achieved by fitting a mathematical function to the data which characterizes the low frequency variations in the data. Division of the original series by the resulting set of values then removes the growth trend.. This master chronology increases the (climate) signal to (non-climate) noise ratio (Fritts, 1971);

3) to build up a network of site chronologies for a region;

4) to identify statistical relationships between the chronology times series and the instrumental climate data for the recent period – the calibration period (Fritts, 1976);

5) to use these relationships to reconstruct climatic information from the earlier period covered by the tree-ring data, and;

6) finally, to test, or verify, the resulting reconstruction against independent data.

Bradley (1985) gives a full account of the methods (1 to 6 above) of palaeoclimate reconstruction from tree-ring analysis. This approach may be applied to all the climate-dependent tree-growth variables, specifically tree-ring width, but also wood density and isotopic measurements. The latewood of a tree ring is much denser than the earlywood and interannual variations contain a strong climatic signal (Schweingruber et al., 1978). Density variations are particularly valuable in dendroclimatology because they to not change significantly with tree age, and the process of standardisation (removal of growth function) can therefore be avoided.

The use of isotopic measurements in dendroclimatology also avoids the need for a standardisation process. The basic premise of isotope dendroclimatology is that since 18O/16O and D/H (deuterium/hydrogen) variations in meteoric (atmospheric) waters are a function of temperature (see also section 3.3.2.1), tree growth which records such isotope variations should preserve a record of past temperature fluctuations (Epstein et al., 1976). Unfortunately, isotope fractionation effects within the tree, which are themselves temperature dependent, will create problems associated with this technique (Libby, 1972).

===============================

I suppose I shouldn’t be surprised that the standard sort, throw-away and rescale techniques are taught as standard starting from an early age.  It’s too bad that they don’t practice these techniques on fabricated data too see if they actually work!  Of course then a bunch of PhD’s would be looking for jobs.

16 Responses to “Baby Dendro’s”

  1. Layman Lurker said

    Jeff, I’m sure you have read over some of Dr. Koutsoyiannis’s work on long term persistance, and how such phenomenon (the “Hurst phenomenon”) – if misunderstood – can be viewed as “signal”. This particular paper is interesting in that it actually uses a Graybill proxy series from Utah to demonstrate the Hurst phenomenon:

    http://www.itia.ntua.gr/getfile/511/1/documents/2002HSJHurst.pdf

  2. TAG said

    Trees growing near the latitudinal or altitudinal treeline are mainly under growth limitations imposed by temperature; hence dendroclimatic indicators in such trees contain strong temperature signals

    And it teh climate changes and the tree line moves??

  3. Jeff Id said

    Lurker,

    I haven’t read the paper. It will give me something to do tonight.

  4. Kenneth Fritsch said

    Jeff, I think the explanation that you excerpted here of what the dendro do, or at least attempt to do, rings true with this layperson.

    My beef with the dendros, as witnessed by some papers and postings at CA by Rob Wilson, is that they do not appear to appreciate the dangers of overfitting their models when they look for trees (ring widths and densities) to use for reconstructions. They apparently know how to get a good fit in the calibration and verification periods by trial and error selection of tree candidates and then picking the favorable model variables of time periods (months from 2 to 6 as I recall), choice of maximum, minimum and mean temperatures and a mix of tree ring widths and densities measurements. They appear to me to think that that emperical fitting can be used as a legitimate test for torturing a temperature signal from the trees. Their seems little a prior input into their models. I think they would gain more understanding and do it more efficiently if they attempted to aggressively find flaws in their models.

    If they took the necessary precautions dealing with the statistical aspects of these processes, I think that they would still have some major problems showing the uniformitarianism of their model assumptions.

    They, of course, have the “divergence” problem staring them in the face. They try to get around it by using classifications of trees and their locations that are and are not subject to the problem.

  5. Jeff Id said

    #4 I think you’re right about it ringing true. However, I’m lucky enough to have a few guys like yourself around who have seen the problems.

    How about that amazing assumption list?

  6. JAE said

    IMHO, here is a big bugaboo in much of dendroclimatology:

    “Trees growing near the latitudinal or altitudinal treeline are mainly under growth limitations imposed by temperature; hence dendroclimatic indicators in such trees contain strong temperature signals.”

    There is absolutely no reason to assume that moisture stresses are not just as important at treeline as they are at lower elevations and arid areas. Moisture availability must be established clearly before one starts attributing growth variability to temperature. And I don’t know just how one does this with any degree of certainty. Most treeline sites are characterized by very porous soils which dry out very rapidly. Treeline sites in summer are virtual deserts, when you think about it. There may be some trees that show a clear temperature effect, but I’ll bet there are very few. Tree rings are very good for studying droughts, but of dubious value in studying temperature changes. Just my opinion….

  7. Ryan O said

    My biggest issue is the lack of appreciation for the multivariate nature of tree ring growth. It is not possible to separate effects by mere correlation, yet that is what is done.

  8. Jeff Id said

    There are too many issues to list for me.

    Besides the clear mixing of multiple signals, which Mannian PCA was supposed to magically sort, there is a clear amplification factor on calibration vs historic data. I’ve spent a lot of time over the last two weeks thinking on how to correct the signal magnification before and after correlation sorting. I believe there is a way to do it but haven’t got my head around it yet.

    Non-linearity is just as big an issue, as is signal separation and the assumption of a continued long term relationship to whatever caused the correlation with temp.

    Moisture has to be the primary factor as JAE says but is growth linear with moisture — HEll no.

    Some of the samples they choose come from damaged trees where the rings aren’t complete. I don’t have a clue to the thinking for that but it seems crazy to me.

  9. JAE said

    “Moisture has to be the primary factor as JAE says but is growth linear with moisture — HEll no.”

    No. And neither is temperature. It is an upside-down U-shaped quadratic. How the hell do these “dendroclimatologists” presume to sort out these problems?

    Answer: IGNORE THEM.

    BAD SCIENCE, ALL THE WAY AROUND.

  10. Jeff Id said

    It’s so bad that it’s hard to imagine it get’s published. Can you imagine how easy it would be to write a dendro temp paper?

  11. Rob R said

    If you did write a paper who would read it except a few dendros, and how many people could you convince? My bulldust gauge runs fairly high when examining most stuff written in connection with temperature and tree rings.

  12. woodNfish said

    JAE: There may be some trees that show a clear temperature effect, but I’ll bet there are very few.

    I am unaware of any research that proves trees make good proxies for temperature. Everytime I read about tree proxies I think they are pulling this crap out of their asses.

  13. Kenneth Fritsch said

    I think, as a layperson in this discussion, that the dendros have the notion (and are convinced) that if they can show calibration and verification periods are in reasonable agreement, after selecting for trees and the “appropriate” variables that are selected after the fact, that they have demonstrated a legitimate reconstruction model. They will then proceed to also show that in the reconstructed part of the series that the model “responded” to (most of the) extreme events such as volcano eruptions.

    I have seen some otherwise very intelligent people use this same rationale in picking investment strategies. I truly think that they do not appreciate how easy it is to overfit a model. They also, I think, like the dendros do, see the verification period as akin to out-of-sample testing.

    And that out-of-sample testing is a very available way for the dendros to go back to previously sampled and modeled trees used in reconstructions, like Mr. Pete and Steve M did not far from a Starbucks, and determine whether their models work truly out-of-sample. The observation that I see little of that work being persued by dendros, or at least reported, and in light of the divergence problems, is telling for this layperson.

  14. Page48 said

    “Trees growing in semi-arid regions are frequently limited by the availability of water, and dendroclimatic indicators primarily reflect this variable.”

    This is a true statement, as far as it goes, but note that this is a direct reversal of the assumption made by Mann and crew for the ’98 Hockey Stick. Bye, bye bristlecones.

    “Trees growing near the latitudinal or altitudinal treeline are mainly under growth limitations imposed by temperature; hence dendroclimatic indicators in such trees contain strong temperature signals.”

    I don’t even think this is a true statement. Maybe some do – maybe some don’t.

    As for the rest – Science by assumption, I guess. I’m assuming that the temperature “signal” is strictly a statistical phenomenon, not something actually measured by experiment.

    It makes sense to me that tree growth might respond positively to GW if the overall global warmth extends the growing season in a given area, but it has never made a lot of sense to me that trees would be exquisitely sensitive to minute variations in the daily ambient temperature. What’s the benefit to the tree, from an evolutionary standpoint?

  15. crosspatch said

    In addition to climate, the trees at the edge are under nutritional stress. Soils are thin, pH can be extreme, not much organic material, all sorts of stresses. What might be interpreted as a “warm spell” could be a change in pH of rain from a volcano (maybe hundreds of miles) upwind that allowed the tree to absorb more nutrients. Could mean an elk or other large animal died in the tree’s root zone and fertilized it for a good long time or that grass grew bringing grazing animals that fertilized the area. And in the end what are they proving anyway; that growing conditions vary on the timescale of centuries? I think that should be expected.

    A dendro study I would like to see is on the wood being exposed by receding glaciers in the European alps that show the valleys currently glaciated having been forested some 7,000 to 5,000 years ago (according to 14C dating of samples). Was the change to glaciation abrupt or gradual?

  16. Billy said

    Whenever I see these discussions about dendroclimatology it always seems like this field is crying out for a multi-year study where they grow various types of trees in various controlled environments — basically a bunch of mini biospheres — where they have control over all inputs: temp, moisture, sunlight, fertilization, etc. Then after say 5 or 10 years they cut the trees down and look at the rings to see what kind of correlations can be made with the original inputs.

    Maybe such an experiment would be too expensive for a field as small as dendroclimatology, but given how influential the field is in paleoclimatology, it seems like it would be worth it to get to the bottom of whether or not they truly are a good temperature proxy. For my money, I’m betting the only temp correlation you get is the one where everything but temperature was held constant. You start varying any of those other inputs in any significant way from year to year and I bet any correlation to temperature disappears.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

 
Follow

Get every new post delivered to your Inbox.

Join 140 other followers

%d bloggers like this: