Thursday, June 23, 2011

Voldemort's Question

Updated June 25 & 26, 2011 -- see end of post

Are the Tiljander proxies calibratable to the instrumental temperature record, 1850-1995?

Reader Alex Harvey copied his submission to RealClimate.org as a comment to the just-prior post at this blog, "The Tiljander Data Series Appear Again, This Time in a Sea-Level Study." Some time later, it was allowed into RealClimate's "2000 Years of Sea Level" at position 22. The second of Harvey's two points concerned the use of Tiljander:
The study has also been criticised on various blogs for using “one of the multiproxy reconstructions that employed the four (actually three) uncalibratable [edit] Tiljander lakebed sediment data series” e.g. http://amac1.blogspot.com/2011/06/tiljander-data-series-appear-again-this.html.[edit].
RealClimate's moderators snipped the comment as shown.

Prof. Mann offered this inline commentary --
[Response: No. Just more of the usual deception from dishonest mud-slingers. More on that in short order. -Mike]
In short order, Update 2 appeared (link). Scroll down at this link to the RC post for its embedded links (Update: embedded links added).
People have asked whether the use of the Tiljander proxies in the Mann et al (2008) EIV surface temperature reconstructions matters for the conclusions of this or any related studies.
Wrong. "People" (me) have been asking a different question for over a year:

Are the Tiljander proxies calibratable to the instrumental temperature record, 1850-1995?

Prof. Mann continues:
The answer, as provided previously in the literature (see this reply to a comment in PNAS) is no.
A one-word answer to an undefined, postmodern question. Who knows what it means.
The impact of whether or not these proxies are used was demonstrated to be minimal for the Northern Hemisphere land+ocean EIV reconstruction featured in Mann et al (2008) [see Figure S7b of the Supplementary Information of that article, which compares the reconstruction both with and without 7 potential 'problem proxies', that include the Tiljander proxies; a similar comparison was also made in Figure S8 of the Supplementary Information for the followup article by Mann et al (2009)].
I have discussed the figures of Mann08's S.I. at this blog and elsewhere. Prof. Mann again invites the reader to puzzle out just how the linked versions of SI figures might support sweeping and vague claims. Not tonight.
The same holds for the specific global mean EIV temperature reconstruction used in the present study as shown in the graph below (interestingly, eliminating the proxies in question actually makes the reconstruction overall slightly cooler prior to AD 1000, which–as noted in the article–would actually bring the semi-empirical sea level estimate into closer agreement with the sea level reconstruction prior to AD 1000).
These 191 dense words were followed by a new spaghetti graph.
 Two variations of a smoothed "Mann et al (2008) global mean (land+ocean) temperature reconstruction" were presented, with (blue line) and without (red line) 7 proxy records (including the Tiljander data series).

The graph is virtually useless.
  • No uncertainty intervals are shown for either of the two curves.
  • There's no mention of which portions of the two traces pass the "validation test" described in Mann et al (2009), at the 95% or higher level.  Recall Gavin Schmidt's remarks on the closely-related Northern Hemisphere reconstruction: "...it's worth pointing out that validation for the no-dendro/no-Tilj is quite sensitive to the required significance, for EIV NH Land+Ocean it goes back to 1500 for 95%, but 1300 for 94% and 1100 AD for 90%" (link).
  • It's not clear whether the "without" red line excludes suspect hockeystick-shaped tree-ring datasets. Steve McIntyre claims that the red line includes bristlecones (Update:  The red and blue lines both include bristlecones, see below.)
A few years ago, Paul Graham posted his memorable essay What You Can't Say (link fixed 7/5/11). I don't want to make him mad, but I'll close this post with a simple question that is still waiting for a clear-cut answer:

Are the Tiljander proxies calibratable to the instrumental temperature record, 1850-1995?

Update, June 25, 2011

In the comments, MikeN suggested that I post the graph from Mann09 that shows the portion of the no-dendro/no-Tilj reconstruction that doesn't pass validation.

That graph is Figure S8 of the S.I. for Mann09. Reference --

ME Mann, Z Zhang, S Rutherford, RS Bradley, MK Hughes, D Shindell, C Ammann, G Faluvegi, & F Ni, "Global Signatures and Dynamical Origins of the Little Ice Age and Medieval Climate Anomaly" (PDF). Science 326: 1256-1260 (Nov. 2009) (S.I. at Science).
Legend:
Figure S8: Sensitivity of NH mean reconstruction to exclusion of selected proxy record. Reconstructions are shown based on “all proxy” network (red, with two standard error region shown in yellow) proxy network with all tree-ring records removed (blue), proxy network with a group of 7 long-term proxy with greater uncertainties and/or potential biases as discussed in ref. S1 (brown) and both tree-ring data and the group of 7 records removed (green; dashed before AD 1500 indicates reconstruction no longer passes validation).
As far as a non-postmodern approach to the question of whether the use of the Tiljander data series "matters": If one is to use this approach, the nearest thing to an apples-to-apples comparison is thus the Blue Line versus the Green Line. (Unfortunately, a two-standard-error region is not shown for either one.) However, it's very hard to see the dashed green line, made worse when it passes beneath the solid blue line. By increasing the magnification of the PDF, most of it can be discerned. In the following figure, I've overwritten the dashed green line in a drawing program, as best I could. The black vertical I added at 1500 AD is a reminder that the green line fails Mann09's validation test at all points prior to that year.

Blue Line: Proxy network, without tree-ring data (no-dendro/yes-Tilj).

Green Line: Proxy network, without tree-ring data and without 7 proxies with potential problems (no-dendro/no-Tilj).
Obviously, this is not the comparison that was offered by Prof. Mann in Update 2 at RealClimate. That graph, reproduced above, showed a Blue Line (with 7 proxies) and a Red Line (without 7 proxies) from Mann09 S.I. Fig S7b (I believe that was a typo, and that the intent was to refer to Fig. S8.) (Update 6/25/11: I was in error on this minor point.  Prof. Mann didn't specify whether the data in this graph came from Mann08 SI Fig. S7b or Mann09 SI Fig. S8.  It's from the latter.)

RealClimate Update 2 Blue (with 7 proxies) appears to be Mann09 SI Fig. S8 Red (yes-dendro/no-Tilj).

RealClimate Update 2 Red (without 7 proxies) appears to be Mann09 SI Fig. S8 Brown (yes-dendro/yes-Tilj).

Update, June 26, 2011

Apples to oranges.

In discussing Mann09 SI Fig. S8, I quoted Gavin Schmidt on the failure of the Green Line to pass validation at the 95% level prior to 1500 AD. It should be noted that the Blue Line is dashed prior to 700 AD. Presumably, that is where it also failed validation.

So, if this method of assessing whether Tiljander "matters" is to be used, we have to recognize three parts to a comparison of "yes-Tilj" to "no-Tilj":
* 1850 AD to 1500 AD -- Compare "validated" no-dendro/yes-Tilj to "validated" no-dendro/no-Tilj

* 1500 AD to 700 AD -- Compare "validated" no-dendro/yes-Tilj to "failed" no-dendro/no-Tilj

* 700 AD to 500 AD -- Compare "failed" no-dendro/yes-Tilj to "failed" no-dendro/no-Tilj
So the apples-to-apples exercise for Mann09 SI Fig. S8 involves evaluating two quite different and sometimes-failed reconstructions. The oranges-to-oranges comparison in RealClimate's Update 2 is not relevant to whether the use of the Tiljander data series "matters".

And certainly not relevant to whether these series are calibratable to the instrumental temperature record.

End of June 26, 2011 Update
End of June 25, 2011 Update
July 6, 2011: I inserted the phrase "Prof. Mann continues:" in the body of the post, for clarity.

38 comments:

  1. Voldemort is the evil wizard who bedevils Harry Potter. His power is such that decent folk avoid even thinking of him. In whispers, he is described only as "he who must not be named."

    One can imagine a Question that would evoke similar dread, and must not be named.

    ReplyDelete
  2. Amac, Mann's graph is the same one he posted in his updated SI (figure s7 b). By comparing the the versions of the EIV reconstructions in figure s6 b (no dendro vs full network) vs s7 b (no Tiljander vs full network) - in particular 1500 to 1850 - it seems obvious that Steve's comment is correct.

    ReplyDelete
  3. You might want to post Mann's graph with the green line that shows the portion that doesn't pass validation.

    ReplyDelete
  4. Nice one - a suitable candidate for one of Lucia's mugs?

    Clivere

    ReplyDelete
  5. Visually it is quite impressive how bad data is treated so differently in the different timeframes.

    A spurious and negative correlation with modern temperatures has been allowed through and strengthened the blade.

    On the other hand bad data (inverted from its correct orientation) has dampened historic variation.

    ReplyDelete
  6. Bart Verheggen has a post on climate science, and it got into a discussion of Tiljander.

    ReplyDelete
  7. I'm curious as to whether you are primarily interested in getting an answer to your specific question per se, or whether you rather see it as a doorway to some larger issue(s) regarding proxy analysis in general.

    I have the feeling that it is the latter. If it is, I think we could have a productive discussion that goes in some different directions than the usual.

    Note that the link to "What you can't say" appears to be broken.

    I'm a member of RealClimate, for whatever it's worth.

    ReplyDelete
  8. Jim Bouldin --

    I think the two questions you raise are related. If the current methods are working well, then further work in the same general direction should lead to further advances.

    My own sense is that Tiljander in Mann08 illustrates a set of problems with the "proxyhopper" approach. Some obvious (or "obvious," if you wish) and others subtle. If one can't recognize a problem, one can't think clearly about how to solve it.

    Becoming somewhat informed about clinical trial design has influenced my thinking about paleoclimate reconstructions. I think there are probably surprising opportunities for cross-discipline interactions, which would benefit the field. I don't read about that, but rather about things like CLIVAR, which strikes me as a big step in the wrong direction.

    Link to Graham's website fixed; thanks. Good essay.

    ReplyDelete
  9. "Becoming somewhat informed about clinical trial design has influenced my thinking about paleoclimate reconstructions."

    I strongly agree that cross-discipline interaction is typically beneficial, and in fact my own work right now is an example of just that, since I bring a biologist's perspective to the issue of tree ring-based climate analysis. Could you elaborate on your statement above, given that the two are very different in their basic natures?

    ReplyDelete
  10. I know this is not really related to your main point but one really has to wonder about the mindset of scientists who after removing a few proxies state casually that the reconstruction is worse because the verification statistics get worse. As if the removed proxies are the magic ingredient in divining an accurate temperature reconstruction. It is simply not scientific. They never deal with the issue of spurious correlations on these kinds of test.

    ReplyDelete
  11. Jim Bouldin --

    Re: clinical trials, I was very impressed by Dr Norman Walmark's remarks at a meeting I attended in 2009. I transcribed them here.

    His concerns are very different from the ones at issue here, but you can get the point. Surgeons following in Halstead's footsteps concerned themselves with the details of the surgical techniques of the radical mastectomy, and properly so. But in keeping their focus narrow, they lost sight of the most important questions for three-quarters of a century.

    In my opinion -- I could be wrong, paleoclimatology isn't my area -- a large gap has developed between the analytical approaches that are considered acceptable in your specialty, and those that are used in certain other fields. I'd guess that the dendro/paleo standard is more than a decade behind that required for clinical trial design.

    In some disputes, it seems to me that prominent people in your field take positions that have fallen out of accepted practice elsewhere. As examples, compare the current permissive dendro/paleo consensus views on "data selection" and "post hoc analysis" with the sophistication attending the intent-to-treat methodology (that's an 11-year-old article). I think dendros and paleos would benefit from recruiting some up-and-coming biostatisticians into your laboratories, and heeding their advice.

    In part 2 of this comment, I'll copy remarks originally left at Ron Broberg's blog "The Whiteboard", as part of a discussion about RL Smith's 2010 paper on reconstructions.

    --- continues ---

    ReplyDelete
  12. --- continuing from prior comment --

    --- begin Whiteboard comment ---
    AMac
    2010 October 14 at 8:31 am

    Well-written MS and an interesting analysis by you, Ron. I wish I knew more maths — enough to comment substantively on Smith’s variations on PC analysis of the tree-ring proxies. A lay reader can at least follow the general thrust of his approach, a seemingly simple task that is much more difficult when looking at BHM analyses.

    I’ll stray a bit off-topic to talk a bit about proxy selection, which Smith specifically excludes from this MS. Here’s a validation process that I would really like to see, performed by a disinterested and trustworthy statistician.

    1. Assemble candidate treerings.

    2. Predefine a first step for the selection process that does not include curve-matching to an recent instrumental record. For instance, it could be “trees within 50 meters of the current treeline” for montane sampling areas.

    3. Predefine a second step in the process that uses curve-matching to the instrumental record (say, 1890-1990) to select a subsample of the tree-rings that passed the first step.

    4. Predefine the conditions that are to be used for the reconstruction analysis. Say, pick the K=10 conditions from Smith’s Fig. 5.

    5. Proceed with the PC analysis and generate the 1400-1890 reconstruction.

    So far, this is only marginally different from the field’s SOPs.

    6. Now pick a ~ century-long interval where you are fairly sure that temperatures were falling rather than rising; as they were during 1890-1990. The onset of the Little Ice Age, maybe — 1730 to 1830, perhaps.

    7. For the second routine, repeat steps 3, 4, and 5, above, and generate a reconstruction covering 1400-1730.

    If the method is robust, the 1400-1730 reconstruction generated from curve-fitting of the temperature-rising era should be quite superimposable on the 1400-1730 reconstruction generated from calibrating to the temperature-falling era.

    The point is to follow the superior statistical practices that ere are now routinely used in other fields, e.g. clinical trials, where an important goal is to minimize biases in data selection and choice of analytical method — in particular, seeking to avoid post hoc procedures.

    There are obstacles to implementing this idea, of course. The two big ones, I think, would be assembling an unbiased initial cohort of candidate tree-rings, and figuring out what the 1730-1830 temperature anomaly curve for calibration should be.

    I suspect (hey, this is a blog comment, who needs any steenking evidence!) that most of the published paleotemperature reconstructions would fail an exercise such as this.

    --- end Whiteboard comment ---

    ReplyDelete
  13. Jim, do you have any opinion on the climategate e-mail where someone criticizes the various paleoclimate papers for not being based on evolutionary theory? Looking it up, I see evolution was not the primary point.

    "Since I am neither a dendrochronologist nor a tree physiologist, I have a different take on this little brushfire we have going. I find it frustrating that some dendrochronologists stubbornly see tree ring characteristics as being affected by climate. They are not. They are affected by cambial activity. Cambial activity is affected by internalities of tree behavior, mainly hormonal and nutrient fluxes in the crown. Those things are largely influenced by climatic factors. So there is quite a bit of slack between the climatic factor and the ring characteristic. Is this just negligible static? I doubt it. I see this as an oversight by dendrochronologists that weakens their credibility a tad among those knowledgable about tree growth. I also have a quarrel with the dogma of dendrochology that the cambium changes as the tree becomes senescent. I know of no data that trees senesce -- that is, that they undergo changes due solely to aging. This started as forestry dogma, and was accepted by tree-ringers, who then corrected for it. I'm practically the only one who has systematically looked for evidence of senescence (with a Ph.D. student), and we could not find any in young to ancient bristlecones. But tree physiologists do not generally look at such issues because they have become progressively more reductionist. Nor do they try to produce a theory of tree growth based, as it must be, on evolutionary theory."
    Some suggest little is known about wood formation on a cellular level.

    ReplyDelete
  14. Jim Bouldin --

    As far as the relevance of clinical trials to paleo reconstructions, there is an article in yesterday's New York Times about how an innovative concept seemed to work out well for patients. Eventually, it became clear that the underlying science was flawed, and could not be used as the foundation for good policy.

    --- Begin fair-use excerpt ---

    How Bright Promise in Cancer Testing Fell Apart
    by Gina Kolata - July 7, 2001

    ...First, though, [Dr Minna] asked two statisticians... to check the [Duke group's innovative work.]

    [They] found errors almost immediately. Some seemed careless — moving a row or a column over by one in a giant spreadsheet — while others seemed inexplicable. The Duke team shrugged them off as “clerical errors.”

    And the Duke researchers continued to publish papers on their genomic signatures in prestigious journals. Meanwhile, they started three trials using the work to decide which drugs to give patients.

    [The statisticians] tried to sound an alarm. They got the attention of the National Cancer Institute, whose own investigators wanted to use the Duke system in a clinical trial but were dissuaded by the criticisms. Finally, they published their analysis in The Annals of Applied Statistics, a journal that medical scientists rarely read.

    continues ...

    --- End fair-use excerpt ---

    ReplyDelete
  15. Amac, MikeN:

    A quick note here to say that I'm checking in and to clarify some basic points before responding to specific questions and comments.

    First, to be very clear, I do not consider myself a "specialist" either in dendrochronology or in climatology. I am primarily an ecologist, and as such, have a fairly broad training in a wide range of levels of biological processes, with some background in statistics. That's pretty typical. However, since my dissertation was very nearly a dendro-based topic that I had started, I know more than the average ecologist does about it, enough to make a definite contribution to the field. I've cored my share of trees and read my share of the literature, and in the last year gotten very deeply into the guts of certain analytical procedures. But it's not my "specialty"--that would be the analysis of forest landscape changes since about 1800, particularly in response to land management changes (and more recently, climate changes). The common denominator in what I do, is knowledge of tree biology/ecology.

    Second, we need to keep in mind that dendroclimatological reconstructions involve several distinct steps: some individuals do all of them and others specialize in only a part of them. There is the data collection step (choosing sites, obtaining the cores, preparing the samples, measuring the ring variables). There is then the detrending step (removing the biological growth trend from each tree core). There are then the (several) steps related to calibration of the rings to the environmental variable(s) of interest (climate, for the present discussion).

    My work (right now anyway) is centered mostly on the 2nd step (removing the biological trend), with an additional minor (but important) component dealing with the ring to climate calibration step, at the scale of the individual site. These steps are preliminary to any large scale reconstruction steps. Mike Mann (and numerous others) conversely, work entirely on the 3rd step--large scale spatio-temporal reconstruction (also often involving calibration steps). Then there is the other end of the spectrum--the organismal biologists, who specialize on the physisological/environmental mechanisms of radial tree growth (Fritz Schweingruber being perhaps the world's leading expert). So I am working on a different issue than what is commonly discussed in the blog world. And that's because blog world discussions are always some truncation of the full set of challenges we deal with, and often distorted discussions at that, and sometimes highly distorted.

    So I am somewhere between what Fritz Schweingruber does and what Mike Mann does. I'm trying to reduce the noise that arises from the fact that tree age/size influence ring characteristics and also the noise that arises from considering temperature in isolation from other variables. But it connects in with what Fritz, on one end, and Mike, on the other, are concerned with. You are always going to have specialization, but you need to have people who can bridge between specialties. That is the only way to a robust understanding of any topic in earth/environmental science.

    The comparison with clinical trials is inherently quite problematic--epistemologically far different. I'll try to speak to that when I get some time. I am extremely busy and responses will likely be very spotty.
    Jim

    ReplyDelete
  16. This comment has been removed by a blog administrator.

    ReplyDelete
  17. Test. Something seems to be wrong. I've twice posted a long response and neither time has it showed up.

    ReplyDelete
  18. This comment has been removed by a blog administrator.

    ReplyDelete
  19. This comment has been removed by a blog administrator.

    ReplyDelete
  20. MikeN:

    As soon as I read something like "I find it frustrating that some dendrochronologists stubbornly see tree ring characteristics as being affected by climate. They are not...", I stop reading. Very confused statement, full of errors, self contradictory. An enormous amount of knowledge exists on the formation of wood--as much as almost any plant anatomy topic, because of the economic value of trees. And there are all kinds of models of tree growth and they include a wide variety of considerations.

    Plus the typical red flag words: dogma, oversight, weakened credibility, etc.

    It's a very clueless statement.

    ReplyDelete
  21. Jim Bouldin --

    Sorry that the filtering trapped your comment. These are blogspot's default settings.

    I've put in your original and removed the duplicates (all hail local copies!).

    Thanks for the follow-up.

    .

    "None" --

    I fished your comment out, too.

    Sorry about that.

    ReplyDelete
  22. Amac at 4:00:

    "There are obstacles to implementing this idea, of course. The two big ones, I think, would be assembling an unbiased initial cohort of candidate tree-rings, and figuring out what the 1730-1830 temperature anomaly curve for calibration should be."

    Definitely correct on the latter. It may be possible for some locations, impossible for others, depending on the spatial properties, and quality, of other available proxies.

    On the former, you're missing something important. Dendrochronologists *already* +/- do your step 2--that's implicit in the sampling process, which is based on a judgement of locations likely to be limited by the env factor of interest--that's why they go near upper or boreal treelines to get a T signal, and lower treelines to get a P signal. But it's not a perfect process, because you don't know for sure how any particular site is responding over time, so its efficacy *has* to be evaluated after the fact--by looking for correlations with the env factor of interest at the closest weather stations (or resident grid cells for gridded data). If the field samplers could choose the right stands without recourse to the climate data comparison step, they would. But they can't. It's exactly analogous to having a set of thermometers of known, varying quality. You're going to pick out the ones you know work the best, and leave the rest behind. Exactly analogous. There's absolutely nothing wrong with it.

    Now, however, you *are* onto a couple of important points. One, as I mentioned at Bart's, is the issue of whether your deduced calibration relationships are in fact unique, that is, that the relationship meets the definition of a function. Now THAT is important. The second is the issue of how wide the range of env variable states during the calibration period is. The narrower, the more problematic and the greater the reliance on extrapolation. And one upshot of that is this: there has been an enormous amount of ink spilled about how higher-than-current temps of the past might have been missed (under-estimated), not nearly balanced out by the equally great possibility that cooler temps might over-estimated.

    ReplyDelete
  23. last sentence should have been:
    "not nearly balanced out by the equally great possibility that cooler temps might have been over-estimated."

    However, I strike the word "equally". I don't know if they are equal possibilities or not.

    ReplyDelete
  24. This comment has been removed by a blog administrator.

    ReplyDelete
  25. TCO said... (July 9, 2011 11:31:00 PM PDT) --

    Yawn. McIntyre ran a total re-run. Is there anythin new in that post? Even in the comments? Guy has *)&^%$ &^%$**.

    Oh...and which bitch said that the 4 series were not independent? THIS BITCH DID!

    :-)

    - - - - -

    Well that thread where you and I started by slinging insults ended well, it deepened my understanding of the issue. Still, most of the time, getting personal doesn't appeal to me, so much.

    Tilj remains little, but sometimes little things can be the entry point for greater insights. Not saying anyone else has to share my interests though.

    ReplyDelete
  26. Jim Bouldin, could you describe how cooler temperatures can be overestimated? I see that warmer temperatures beyond a certain range and tree growth will not be as high. What is the reverse condition, since you can't get negative growth?

    ReplyDelete
  27. "the sampling process... is based on a judgement of locations likely to be limited by the env factor of interest--that's why they go near upper or boreal treelines to get a T signal, and lower treelines to get a P signal. But it's not a perfect process, because you don't know for sure how any particular site is responding over time, so its efficacy *has* to be evaluated after the fact--by looking for correlations with the env factor of interest at the closest weather stations (or resident grid cells for gridded data). If the field samplers could choose the right stands without recourse to the climate data comparison step, they would. But they can't. It's exactly analogous to having a set of thermometers of known, varying quality. You're going to pick out the ones you know work the best, and leave the rest behind. Exactly analogous. There's absolutely nothing wrong with it."

    But we have a very good theory of thermometers. The situation with trees does not seem to me exactly analogous. Treelines change over time for example. Moreover with the bristlecone pines, for example, there is some evidence that the anomalous growth of the ones used in MBH 1998 could have simply been the response to damage. You need to be able to demonstrate that the background factors are being held reasonably constant over the period of the reconstruction. Otherwise how can you be sure your treemometers are not becoming unreliable over past history. Surely the very existence of the divergence problem demonstrates this problem?

    ReplyDelete
  28. http://pubs.giss.nasa.gov/abs/sc06400f.html

    ReplyDelete
  29. divergence

    http://iopscience.iop.org/1748-9326/6/4/041004

    ReplyDelete
  30. Where is the hockey stick?

    http://mpra.ub.uni-muenchen.de/35565/1/Climate_Change_and_Hockey_Stick.pdf

    ReplyDelete
  31. Thanks for the link, Anonymous. I've scanned the paper -- interesting! -- and quoted the Abstract at an ongoing discussion re: reconstructions at Lucia's Blackboard, here.

    ReplyDelete
  32. maybe you'd like to comment on (see the graph labeled b):

    http://www.realclimate.org/index.php/archives/2012/07/tree-rings-and-climate-some-recent-developments/#more-12427

    ReplyDelete
  33. @ Anonymous (July 11, 2012) --

    The July 8, 2012 RealClimate post Anon links is Tree Rings and Climate: Some Recent Developments. It includes a reproduction of panel b of Figure S6 from Mann et al (PNAS, 2008).

    That panel's legend would be "Comparison of NH mean for combined land+ ocean. Surface temperature reconstructions based on EIV method and using the various proxy networks and target instrumental series as described."

    The discussion of this Figure in the RealClimate post is as follows:

    --- begin excerpt ---

    For example, if one eliminates tree-ring data entirely from the Mann et al (2008) “EIV” temperature reconstruction (see below; blue curve corresponds to the case where all tree-ring data have been withheld from the multiproxy network), one finds not only that the resulting reconstruction is broadly similar to that obtained with tree-ring data, but in fact the pre-industrial long-term cooling trend in hemispheric mean temperature is actually lessened when the tree-ring data are eliminated—precisely the opposite of what is predicted by the Esper et al hypothesis.

    --- end excerpt ---

    In other words, as Anon is alluding, Prof Mann is once again discussing a reconstruction -- the blue curve is "Full network without tree rings" -- that depends on the Tiljander data series.

    In another context -- Mann et al (2009, Science) -- Prof Mann and his co-authors note that without Tiljander, a similar EIV reconstruction fails the validation step prior to AD 1500. See the green and blue lines upthread, in the body of this post.

    In other words, the shape of the reconstruction Prof. Mann is discussing in this July 2012 RC post is dependent on the use of uncalibratable upside-down Tiljander.

    If nobody notices, I guess it's okay, especially if the conclusion is on-message.

    State-of-the-art paleoclimatology in 2012. Caveat lector.

    ReplyDelete
  34. Anon (July 11, 2012) --

    Blogspot ate my response to you. Well, here is what I submitted to the moderation queue in the comments of the July 8th RealClimate post you linked, Tree Rings and Climate: Some Recent Developments.

    - - - - -

    AMac says:
    Your comment is awaiting moderation.
    11 Jul 2012 at 4:45 PM

    Prof Mann,

    In the post, you present panel b of Figure S6 of Mann et al (2008, PNAS). This is the EIV reconstruction of “Northern Hemisphere Land + Ocean.” The current post discusses the blue line (“Full network without tree rings”):

    [blockquote]
    if one eliminates tree-ring data entirely from the Mann et al (2008) “EIV” temperature reconstruction (see below; blue curve corresponds to the case where all tree-ring data have been withheld from the multiproxy network), one finds not only that the resulting reconstruction is broadly similar to that obtained with tree-ring data, but in fact the pre-industrial long-term cooling trend in hemispheric mean temperature is actually lessened when the tree-ring data are eliminated…
    [/blockquote]

    This reconstruction appears closely related to the one presented as Figure S8 in Mann et al (2009, Science). The legend to that figure notes that when the Tiljander data series are not used, the no-dendro curve fails validation prior to 1500. In addition, post-1500, the no-dendro/no-Tiljander curve (green) and the no-dendro/yes-Tiljander curve (blue) have entirely different shapes.

    It appears that the portion of the figure that you are discussing with regards to the new Esper paper is entirely dependent on the inclusion of the two (not four) uncalibratable Tiljander data series as paleotemperature proxies in the reconstruction. At this late date, that does not seem to be a reasonable choice.

    - - - - -

    My experience with RealClimate's moderation policy has been that my polite and on-topic but critical remarks fail moderation, are subject to excessive delay, or are aggressively bowdlerized. The last time I tried a submission (2011?), my comment failed pre-moderation! The website swallowed the submission without acknowledging it; must've been some neat trick based on a blacklist of user names or IP addresses. Anyway, this one made it onto the queue, in position #45.

    ReplyDelete
  35. thanks

    best regards, erik

    ReplyDelete
  36. The comment I left at RealClimate (11 Jul 2012 at 4:45 PM) has failed moderation thus far, though submissions 45 through 48 have passed since then.

    Tellingly, RC's owners relegated the following comment by "Benjamin" to "The Bore Hole," despite the cogent and on-topic nature of his remarks.

    Science as it is practiced by the leading lights of paleoclimatology, and celebrated by the rest of the consensus climatology community.

    - - - - -

    I got two questions :

    1/ The Mann08 you show has no tree rings but still has Tijlander lake sediments proxy, right ?

    2/ On the Mann08 graph, the post 1900 part is only represented with instrumental record. Is there a version of this graph without the instrumental record but with the proxies used all the way to 2000, like on Jan Esper’s graph ?

    Thanks.

    Comment by Benjamin — 10 Jul 2012 @ 3:45 PM

    ReplyDelete
  37. The comment I left two days ago at RealClimate (11 Jul 2012 at 4:45 PM) (copied upthread here) has definitively failed moderation.

    State-of-the-art paleoclimatology in 2012. Caveat lector.

    ReplyDelete