delayed.oscillator

Climate, paleoclimate, huevos rancheros, and general asymmetry

Yamal V: … but they pull me back in …

with 17 comments

Via Jeff at his blog The Air Vent, in the midst of a post that includes a nontrivial quantity of ALL CAPS and multiple exclamation points (no <blink> tag?), the following statement (since that time stated in a more reasonable tone here):

For some reason EVERY RCS CORRECTION Briffa can conceive of refuses to turn upward to fit the ACTUAL data. This lack of flexibility in the RCS curves is what creates the HS.

I’m not sure why he doesn’t take my post on growth curves at Yamal seriously, but I’ve already explained why this is probably incorrect.  The problem is an intermingling of the signal and the noise.  As I showed previously, when I constructed an RCS curve using only the recent, living trees from Yamal, because these trees were growing roughly at the same time, and at that during a period of increasing temperatures, estimating a growth curve from them potentially intermingles the climate signal with the geometric growth curve.

One way to look at this is to consider only those trees that ceased growing prior to 1900.  I’ve extracted these from the full Yamal set and estimated the regional curve standardization using a 67% spline.  Here it is:

non20c

Let me be as clear as I can be, there is no sign that I can detect that it is old trees that increase their growth at Yamal (even if identified, this phenomenon would require some hypothesis as to the cause),   At Yamal, a portion of the old trees are those that were growing together during a period of climate warming.  If you examine the raw ring width, there are a few fossil series that have rapid increases toward the end.  If Jeff’s hypothesis were correct, we’d expect these to be the oldest, right?  In fact,  the seven subfossil samples I identified as having rapidly increasing growth in their later years, six had a wide range of ages from 90 to 180 years (this comes with the caveat that we don’t know the exact pith age).

Furthermore, and as my earlier posts have shown, and as I’ll show again below, using a 67% spline (so that the later part of the regional growth curve bends slightly upward) you still get an increase in the chronology values during the 20th century.

Finally, what of the claim that combining the mean-detrended series demonstrates that the RCS method is invalid?  One way we can test this is by first aligning all the ring widths by age (as was done here), observing where the curve  of the juvenile grow trend flattens in the majority of trees (eyeballing it at around 175 to 200 years), throwing away the ring widths for the time when every tree was between 1 and 199 years of age (you get essentially the same result if you use 175 years), then realigning by time (year A.D.), and removing the mean.  Lets look at the truncated growth curves in time and the mean of these series:

meancrn

What immediately jumps out at you is that the mean for the most recent century and half is noticeably higher than  the earlier part of the millennium.  This doesn’t a priori indicate warmer temperature, but as I explained to Jeff here, once again there is the potential for this approach to remove climate signal in the process of detrending.  For the most part, the mean following truncation is lower for the non-living trees at Yamal, but for the living 17 trees, the mean is similar or in some cases higher after truncation. Why? At least partially because after truncation the mean of the recent, living trees is of a large portion of the growth occurring during 20th century warming.

So, what do the chronologies look like if we use three different types of detrending?  Let’s try constructing an RCS chronology using a 67% spline, negative exponential, and a generalized negative exponential curve:

cmpcrn

The top shows the regional curve fit to all the raw Yamal data, the bottom panel zooms in on the part of the chronology since A.D. 1700 (heavy lines are 20 year low pass Butterworth filtered values).  The regional curve fit that I think Jeff would endorse (67% spline) to fit his ‘U shape’ hypothesis actually results in slightly higher chronology values following the Little Ice Age, and slightly higher values at the end of the chronology in the mid 1990s.

My point is not to indicate a ‘correct’ method here — that goes well beyond ‘Blog Science’.  Rather my point is this: detrending and standardization is one of the most challenging tasks in accurately estimating past low frequency climate variability from tree rings.  Divergence is a serious challenge worthy of further study.  What I don’t understand, however, is why methods or analyses (mean detrending, identifying growth curves from a small number of simultaneously growing trees) that can be shown to have the potential biases in specific instances that I’ve demonstrated here and in a previous post can still touted by those with  a visceral dislike of paleoclimatologists as proof that another method is incorrect.  If this debate was a collegial one that might be one thing, but there is nothing collegial — or scientific — about the language and tone from the other side. Too bad.  To endorse once again Rob Wilson’s comment from Climate Audit:

“In fact, the fatal flaw in this blog and what keeps it from being a useful tool for the palaeoclimatic and other communities is its persistent and totally unnecessary negative tone and attitude, and the assumption that our intention is faulty and biased, which keeps real discourse from taking place.”

UPDATE: Minor grammar corrections

About these ads

Written by delayedoscillator

November 3, 2009 at 2:07 am

17 Responses

Subscribe to comments with RSS.

  1. […] Comments Layman Lurker on UAH Temperature Anomaly V…Yamal V: … but… on Hockeystickization Revisi…Jeff Id on Tax^2SteveAUS on Tax^2Jeff Id on UAH […]

  2. DO,
    I had glanced at Jeff ID and wondered if he was confusing the effects of age on growth of trees in general with the climatic conditions for a particular set of aging trees (i.e. the living trees). Thanks for an illuminating post.

    Deep Climate

    November 3, 2009 at 6:26 am

  3. Thanks for calculating the RCS curve for pre-1900 trees only. This was something I suggested to Jeff Id after reading his recent post. This curve shows that the exponential decay curve is a reasonable fit for very old trees that died before 1900.

    I think there are still some legitimate critiques of Briffa 2000 and 2008(9?):

    1. Briffa should have made clear in his papers that the post 1990 reconstruction was based on very few trees, and so should be “treated with caution”, as he explained in his recent web post.

    2. It is impossible to tell if the increased growth rates post 1900 were due in part or in whole to increased CO2 concentration. The 1900 to 1990 growth rates do not appear to track 20th century temperatures very well.

    3. Two of the four Yamal region data sets (including the one substituted by Steve McIntyre) show little or no accelerated growth in the 20th century, while two show substantial increases. Since all four sample sites experienced the same regional temperature trends, it is difficult for me to believe that temperature increases are the only (or even the dominant) cause for accelerated 20th century growth in only two of the four data sets. [snip – motive, loaded language ]

    stevefitzpatrick

    November 3, 2009 at 8:15 am

    • Hi Steve,

      Thanks for your inquiries. A few comments on your points:

      [1] Briffa et al. Phil. Trans. R. Soc. B (2008), Figure 3, shows sample size for each chronology.

      [2] There is no definitive evidence yet of which I am aware that links increased tree growth in mature trees in natural settings to CO2. If you’re interested, start here with these:

      http://www.pnas.org/content/94/16/8350.abstract

      http://www3.interscience.wiley.com/journal/118832066/abstract

      http://www3.interscience.wiley.com/journal/121436013/abstract

      but the literature is vast, including lots of papers that hypothesize (but don’t prove) ring width growth trends due to CO2 ‘fertilization’.

      In my quick analysis here, it looks to me that the Yamal chronology does track recent temperature trends.

      [3] See my post here, which includes identification of apparent divergence in the Khadyta River site.

      delayedoscillator

      November 3, 2009 at 11:48 am

    • Steve,

      Your idea is a good test to see if the increase in growth by age of the trees is a legitimate phenomenon or is it created by unusual recent climate or something else. I really did plan to spend the time on it. DO’s version is a little different than the one I would do.

      Later tonight I want to repeat the analysis DO shows, then do it only on trees greater than 250 years old as these are the ones which seem to hook back upward.

      Also, I will include all the recent chronologies used by Briffa’s latest posts because we know that the Yamal data prior to temperature measurement had a very small selection of older trees. It will be interesting to see if DO’s contentions hold I don’t know for certain which way it will come out.

      Jeff Id

      November 3, 2009 at 1:37 pm

      • Hi Jeff,

        Looking forward to seeing what you come up with. As I’ve indicated here and before in other posts, I suspect you’ll have to account for the fact that age and recentness may covary (some of the oldest trees are the most recent, if I recall correctly).

        I’m curious however — why are you interested in continuing to invest in this analysis?– your posts make rather strong statements about what you consider to be science, so I’m surprised you’re interested in investing your time and energy in something you have such clear disdain for (which just so happens to be one of my areas of research)?

        delayedoscillator

        November 3, 2009 at 1:56 pm

  4. DO,

    One thing I learned many years ago is that no matter how deeply you believe something you can always be wrong.

    [snip – no more of this please, there are plenty of other places to continue to fight that battle, but not here]

    So in short, I’m curious about the methods of reasonable scientists on this topic. When we see that the original yamal with its huge blade is a result of RCS and an assumed rather than proven growth curve – that’s a problem and it’s frustrating.

    Also, I’m interested in different proxies as the dendro stuff is not terribly convincing at this point. I am open to being convinced but there are some very difficult questions to answer. Linearity, divergence and correlated overlayed signals. These are not problems that can be answered with a hand waive.

    Can you explain why your exponential fit didn’t replicate the original Yamal in the last graph?

    The last plot has a fairly similar look to the results I got including all of Briffa’s data from his sensitivity test and the Schweingruber set SteveM chose.

    Jeff Id

    November 3, 2009 at 3:50 pm

    • Hi Jeff,

      I’m confused that this keeps coming up (let alone that you think it is proven!), but to emphasize the point of this and other post here, the ‘hockey stick’ in the Yamal data set is not a result of RCS.

      I don’t think I’m dismissing anything with a hand wave, although blogging has to take a back seat to actually working on some of these issues.

      Part of the difference (but only part) is that you can’t really see 1996 in my plot (it is the rightmost tick on the x axis). The differences in my emulation and the published Yamal chronology seems to be due to the precise detrending techniques used (they are pretty full described and references in the Briffa reply I linked to in the previous post). So, my emulation is just that and my test of the different regional curve fits is, as you say, a sensitivity test. With respect to that, in more absolute terms, my chronology falls slightly below the published chronology, while McIntyre’s RCS code actually, if I recall correctly, puts the ‘blade’ slightly above the published version.

      delayedoscillator

      November 3, 2009 at 10:12 pm

  5. DO,

    With regard to Briffa’s disclosure of the Yamal data set size for the 20th century, this was not (apparently) done in 2000, and 8 years seems to me a long time to not make this disclosure, especially considering the very frequent use of the Briffa 2000 reconstruction in other publications and in the IPCC analysis. I applaud Briffa for finally disclosing all the data that went into the 2000 reconstruction, but I wish he had done this years ago.

    With regard to the “divergence problem”, recognition of this problem leads (I think) to hand selection of those tree core samples which do not exhibit “divergence”. Any such selection criteria can never (of course) be applied to cores from before the instrument temperature record, and is therefore fraught with problems. Since the whole object of Briffa’s Yamal study is to compare recent growth rates to samples from before the instrument temperature record, any difference in selection criteria pre/post the availability of the instrument record automatically (in my mind) calls the validity of the entire exercise into question. I think it is fair to ask how dendro-climatologists know there was not a similar divergence of growth from temperature among many or even most trees during different periods before the instrument record. Without a clear and verified explanation of the cause(s) of the current divergence, which shows that divergence could never have happened before the instrument record, I think the selection method should be uniform for all samples. If the selection criteria used lead to substantially different rates of sample rejection pre and post the instrument record, then the legitimacy of the selection criteria and the results of any study based on those criteria would seem to me doubtful.

    It was not clear to me in Briffa’s recent web post on Yamal what selection criteria were used, and more importantly, if those criteria were applied uniformly across all sample ages. Since the Russians collected the samples for a different type of analysis, I wonder if the selection criteria they used were suitable for an RCS study. Again, Briffa does not appear to address this.

    [snip – politics, motive]

    stevefitzpatrick

    November 3, 2009 at 10:05 pm

    • Hi Steve,

      [1] Core counts were available in H+S2002, and if I understand correctly the data had been made available by request in 2004 (maybe Deep Climate can point me to where I read this). But in any case, I’m really not interested in second guessing what did or did not go into a manuscript nearly a decade ago. As I’m certain you know, Briffa 2000 is a large review of tree rings from around the world, including temperature and precipitation-sensitive chronologies.

      [2] The type of selection you are describing doesn’t occur. For a typical site, cores are collected and crossdated and dated cores enter into the master site chronology. H+S are somewhat different because they were dealing with subfossil sites and the characteristics of their particular standardization method, but again they had selection method that was not related to comparing cores to the meteorological records.

      Let me emphasize again that we don’t pick cores based on comparison to the meteorological record. Identification of divergence happens at the site level.

      [3] Finally, you state ‘I think it is fair to ask how dendroclimatologists know there was not a similar divergence of growth from temperature among many or even most trees during different periods before the instrument record’ — and you’re right — As I’ve stated elsewhere on this blog, the concern about divergence is as you describe it — we need to understand why divergence occurs and how to detect it (is it unique to the anthropogenic era? when is it just a consequence of detrending?), and for this reason this topic is one of ongoing research in the field.

      delayedoscillator

      November 4, 2009 at 12:49 pm

  6. DO,

    “[1] Core counts were available in H+S2002, and if I understand correctly the data had been made available by request in 2004″

    I think this is not a fair representation of what happened. Steve McIntyre has made very clear that at no time did Briffa explain that the H$S cores were exactly the same as used for Briffa 2000. Had McIntyre been aware of this, then he would for certain have asked H&S for the data. Briffa in his recent post (specifically addressing McIntyre’s request for data), stated that he was unable to divulge the requested data because it was not his to give away. Perhaps this true, but if, so he should have made that clear to McIntyre, made clear that it was the self-same data as H&S used, and sent McIntyre along to contact H&S.

    [Your description is at odd with others, see http://scienceblogs.com/deltoid/2009/10/mcintyre_had_the_data_all_alon.php. I opened this door, kinda, so you had every right to walk through it, but no more on this, as it takes us away from the science. You have other venues available to you if you want to discuss this]

    [2] H&S “had selection method that was not related to comparing cores to the meteorological records”

    OK, but that is really only part of the issue. If the selection method excludes “diverging” living trees, then the effect may be very much the same as selecting based on rainfall or temperature records. Since “divergence” is obviously not understood, and since it is obviously a very important effect, I do not see how a level of confidence can be stated for a study like Briffa 2000. [snip]

    [3] With regard to your removing any and all comments related to motives: [SNIP! – nope, told you no motives here. If you want to impugn the motives of scientists, you have other venues available to you, I’m well aware. This will not be one of them]

    stevefitzpatrick

    November 4, 2009 at 8:44 pm

    • To your point [2] — As I’ve explained, trees in a chronology are not excluded based on divergence — the level of assessment is whether a chronology (site) diverges. I can’t be more clear than that. The rest of your comment appears to be boilerplate from other sites. I’m happy to have you contribute, but a contribution has to be more than cut-and-paste from elsewhere.

      delayedoscillator

      November 5, 2009 at 1:05 pm

  7. DO,

    “a contribution has to be more than cut-and-paste from elsewhere”

    I cut and pasted nothing, and I have no idea why you imagine that I did. I do not know what you mean by “boilerplate from other sites”.

    With regard to exclusion of entire data sets because they are divergent: I do not see any justification for choosing to exclude a complete data set from the same (close) geographical region, since it seems very likely that all the trees in the region experienced similar historical climate changes. To make any exclusion of divergent data sets rational, some local variable or variables would have to be identified and quantified with confidence for each and every exclusion. But were it possible to identify and quantify what drives site divergence, the effect(s) of the identified variable(s) could then just be incorporated into the analysis…. the divergent set would no longer be “divergent”, and would not need to be excluded.

    But even if the above could be done, you would also need to be able to determine accurate value(s) for the same variable(s) for the whole of the historical record. So unless the values for the variable(s) driving divergence can be determined from the cores themselves, it appears to me impossible to uniformly treat all the data.

    Selection/exclusion of entire data sets based on whether or not there is agreement between data sets is something I have almost never encountered outside of dendro-climatology. In fields where I have worked (chemistry, instrumentation, nano-technology), any exclusion of data is usually (if not always) based on a specific and clearly documented problem. In cases where a cause for clearly divergent data can’t be identified, the divergent data is always retained and noted, and sufficient additional analyses are done using the same conditions to verify that excluded data truly represents an “extreme outlier” for which there is no known explanation. I have not read many published dendro-climatology studies (just 3 or 4), but I do not remember reading how many (if any) additional data sets were excluded from these studies based on lack of agreement with the data sets that were in fact used. At an absolute minimum, I believe dendro authors should include in any publication the number of data sets that were initially considered for the study, and the number that were excluded due to divergence from the data sets that were used.

    Considering the exclusion of entire site data sets based only on divergence from other data sets, how can the uncertainty in the results of a study be determined?

    If you think the above does not constitute a reasonable contribution, then I do not understand the purpose of your blog, and will not comment further.

    stevefitzpatrick

    November 5, 2009 at 9:47 pm

    • Hi Steve,

      Thanks — there is a lot in your comment, so apologies if I miss something in responding.

      [1] My observations about identification of divergence were only mean to clarify that it is done at the site, not core, level, in general (there has been work with responders/non responders as a way of looking at whether recent divergence is unique or whether microsite differences might account for divergence). Now, once a diverging site has been identified, what to do (which I think your current comments go to)? I’m more ambivalent about this, since temperature reconstructions are not, per se, my specific interest. Were I doing it, though, my inclination would be to assess the sensitivity of the final temperature reconstruction to the inclusion of each possible set of subpopulations (again, most likely grouped as sites), by doing an ensemble of reconstructions and using this to bootstrap the uncertainties associated with the subpopulations (of which divergence might be only one source, albeit probably the most significant). After all, reconstruction are estimates with uncertainty, and divergence is going to be a major source of uncertainty if it exists. Again, I can only speak hypothetically here, since I don’t do large-scale temperature reconstructions.

      Note that you wouldn’t necessarily expect all the trees in the region to diverge, depending on the cause, even if it was climatic, for a number of reasons (but site environment differences come to mind) — some ‘microsite’ characteristics might be more susceptible or closer to some sort of optimum temperature/growth threshold.

      On one final note, I appreciate you being so upfront when you say ‘I have not read many published dendroclimatology studies (just 3 or 4)’ — I’m quite pleased (as you can see in other threads) to answer questions from genuinely interested amateurs, but I do hope that people asking questions have a genuine interest in learning more or discussing issues that they are actually curious about, as opposed to taking this opportunity to be confrontational with a dendrochronologist. I don’t mean to pick on you here, Steve, because I do appreciate you being upfront with this statement, but hopefully you (and others that come here from blogs antagonistic to me and my colleagues) will understand if I only want to spend time and energy with authentic inquiry. I wouldn’t expect to be well received if I were to arrive combatively on a, for instance, nanotechnology (again, just to riff off your comment) blog. Maybe a chemistry blog, though :)

      delayedoscillator

      November 5, 2009 at 10:38 pm

  8. […] data it amplifies the older data creating an artificial uptick at the end of the reconstruciton. DO has made the following statement at his blog: Let me be as clear as I can be, there is no sign that I can detect that it is old trees that […]

  9. McIntyre has done what he calls an Esper-style inhomogeneity test on Yamal data.

    But as far as I can tell, Esper tests two populations over the same period of time, while McIntyre is comparing the live tree cores with the older sub-fossil “population”.

    I can’t understand why the relationship between age and growth should be the same for times of very differing climatic change.

    http://www.climateaudit.org/?p=7720

    Deep Climate

    November 13, 2009 at 2:12 am

    • Hi DC,

      I believe you are correct. The Esper et al. 2003 paper looks at two different species or two different sites (among other issues). I’ve already demonstrated to Jeff, for instance, why comparing 12 trees that grew at the same time during a period of changing climate vs. the full living and subfossil data over more than millennium doesn’t show what he wants it too.

      delayedoscillator

      November 13, 2009 at 1:50 pm


Leave a Reply

Please log in using one of these methods to post your comment:

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.

%d bloggers like this: