Thursday, August 05, 2010

Two views of Tiljander

On Aug. 4 at Collide-a-scape, Keith Kloor presented an e-mail interview of Gavin Schmidt, the NASA climate modeler and RealClimate.org blogger, Gavin's Perspective. It's a generally interesting read. After the post went up, Gavin stayed around to field accodades and brickbats from readers. After some questions about the Tiljander/Mann Affair went unanswered, I penned a Tiljander For Beginners essay at #127, and invited Gavin's thoughts. He duly responded with his view at #188.

The two perspective pieces are side-by-side, below.


AMac's view
Comment #127
Gavin's view
Comment #188
Certain aspects of the Tiljander story are straightforward, and easy for educated laypeople to grasp.

(1) The paleotemperature reconstruction strategy employed by Prof. Mann’s research group in their 2008 PNAS paper (”Mann08″) absolutely required that each data series (”proxy”) be calibrated to the instrumental temperature record that spans 1850 to 1995.

(2) Mann08’s authors considered using the many-thousand-year record of lake-bed sediments that graduate student Mia Tiljander and her advisor collected and analyzed from the bottom of Lake Korttajarvi in Finland. In their paper (”Tiljander03″), Prof. Matti Saarnisto and his student warned that their lakebed data became progressively more unreliable from ~1720 on, as the sediments recorded nearby activites unrelated to climate: farming, peat cutting, roadbuilding, bridge reconstruction, and lake eutrophication.

(3) Mann08’s text acknowledges these issues as potential problems. However, the paper’s authors went ahead and incorporated four data sets from Tiljander’s work into their own reconstruction.

(4) The post-1720 contamination of the Lake Korttajarvi data series turns out to be much, much worse than Mann08’s authors suspected. Anyone who looks can quickly grasp the problem. This JPEG [copied below] is a graph of three of Tiljander’s series: “Darksum,” “Lightsum,” and “X-Ray Density” (in each case, Tiljander03’s “warmer” is plotted “up.”) It shows 2,000 years’ worth of data, using a running 20-year average to smooth out the curves. Look to the 1850-1995 interval at the right of the picture, where the instrumental temperature record is available for proxy calibration. Yikes!

(5) Mann08’s authors found a strong correlation between the rising temperature in southern Finland, 1850-1995 and the rising value of “Darksum” (r=0.3066). They got another one between rising temperature and rising “Lightsum” (r=0.2714). These are “spurious correlations,” unrelated to any pre-1850 climate signals that are contained in the sediments.

(6) In and of itself, this error is not a big deal. Scientists make mistakes all the time. The controversy has arisen over the refusal of Prof. Mann and his co-authors to acknowlege what happened, straighten out the mess, and incorporate any lessons into best scientific practices.

Judy Curry has written extensively about the woeful effects of Tribalism on Climate Science. The Tiljander/Mann affair is an object lesson of its dangers. Each reader will have their own take–and many partisans will bring their own spin to the analysis. In this Comment, I’ve sketched my own view of the matter. Perhaps, with Prof. Curry’s insights in mind, it will be possible to find more constructive ways of handling this (and future) controversies that arise at the interface between professional climatologists and citizen-scientists.
For the sake of completeness, I will simply repeat what I have said before on various [RealClimate] threads, This is drawn entirely from the Mann et al (2008, 2009) papers (so you could just cut out the middle man). I’m not sure why people think that asking the same question a dozen times in different places will get a different answer, but here goes…

1) Varved lake sediments often contain climate related signals, through changes in temperature, local runoff, stratification. Tiljander et al (2003) reported on records taken from a lake in Finland. They also reported possible anthropogenic contamination of their signal in more recent centuries. This makes them potentially useful, but also potentially dubious.

2) Mann et al (2008) used these proxies (4 out of 1209) as input data into two reconstruction methodologies. One (CPS) requires a local correlation to temperature before they can be used, the other (EIV) does not. In CPS, the local correlation requirement fixes the orientation of any proxy – if you have an a priori expectation that it should be a different way, that proxy cannot be used.

3) Since Mann et al (2008) were very aware of the potentially dubious nature of the modern portion of the Tiljander proxies, they performed their reconstructions without those proxies (and three others with potential problems) in sensitivity tests in the supplemental information (specifically Fig S8). Neither reconstruction (for NH mean (EIV) or NH land (CPS) temperature) is materially affected by the absence of the Tiljander proxies. This is the identical result to what you would have if you had a priori insisted on the opposite orientation of the proxies in CPS.

3) The reconstruction without the Tiljander proxies validates back to 700 AD (NH mean, EIV) or 400 AD (NH land, CPS).

4) If you think the Tiljander proxies are not usable or must be used in a different orientation, then Mann et al (2008) already showed what difference that makes to the overall reconstruction. There is nothing else left to do. All code and all data are available online for people to check this for themselves.

5) Please read the papers. Nothing stated in the RC posts or comments was incorrect. There is a sensitivity to how far back you can go without tree rings if you drop the Tiljander proxies as well. So if you don’t like them, and are convinced that tree rings are useless, these methodologies don’t allow you to say anything before 1500 (compared to 1760 in the original MBH) (though the structure is pretty similar back further (CPS)) (see SI in Mann et al 2009 for the EIV result). Other methodologies may still be useful (cf Osborn and Briffa, 2006; Moberg et al, 2004). If however, you think that tree rings do contain useful climate information (see Salzer et al, 2009 for instance), then you get validated reconstructions back to well before medieval times. All validated reconstructions show late 20th Century warmth as anomalous over the their range of validity.

(Consequences of all this)/(amount of time devoted to discussing it in the blogosphere) = a very small number.

Here's the picture of the three Tiljander data series that's linked within my piece. Darksum, Lightsum, and X-Ray Density are all oriented such that "up" represents "warmer," according to the interpretation specified in Tiljander03. To show them as they are employed in Mann08 (see that paper's Fig. S9), the images of Lightsum and X-Ray Density would need to be flipped so that high data values are near the top (Tiljander03 and Mann08 agree on the orientation of Darksum, as shown). The grey shaded area to the right covers the data on sediments that were deposited after 1720, about the time when Tiljander03 states that non-climate contamination began its growth. The very steep increase in non-climate signals through the 19th and 20th Centuries is strikingly obvious for Darksum and Lightsum.


In the thread following the Collide-a-scape post The Main Hindrance to Dialog (and Detente)”, Lucia memorably noted Gavin’s pronounced tendency to answer questions regarding the uses of the Tiljander proxies… but to not answer the questions that were actually being asked.

In that light, here are the six points I made, and my view of the response contained in Gavin's answer.

(1) Mann08′s methods require direct calibation of all proxies to the 1850-1995 temperature record.
* Ignored.

(2) Tiljander warned about post-1720 contamination in the Tiljander proxies.
* Somewhat addressed -- but "potentially dubious nature”?

(3) Mann08 considered the warnings, then went ahead and used the proxies.
* Tacit agreement.

(4) The 19th and 20th Century contamination was really bad, as a glance at a figure will show.
* Ignored.

(5) The correlations that Mann08 thought they found between 1850-1995 temperature and proxy signals were actually spurious correlations to contaminating non-climate signals.
* Ignored.

(6) The mistake itself isn’t such a big deal, refusing to fix the problem is the issue.
* Ignored.

Which perspective hews closer to the truth, and has greater explanatory power? For many people, I imagine the answer will rather depend on their pre-existing views. Time will tell...

11 comments:

  1. AMac,

    I find the exegetical exercise very interesting! Here's what I am proposing to do:

    Let's look at all the sentences, one by one, in the order they appear. Let's identify what is being talked about, what is being said. Let's interpret what is being done, and the scope of what is being accomplished.

    I see two reasons to do this:

    First, this should give us a better idea of what Gavin is saying. Focusing on what it is supposed to answer, or not, and ignoring everything that does not answer that question, does not do any justice to what Gavin says. Someone interested in Gavin sayings should try to listen to what he says, and not simply run with what interests him.

    Second, this should help the critical apparatus put forward by your exegetical exercise. There are at least three technical issues that spring to mind:

    Numbering issue. - By using the same numbering system for AMac's view, Gavin's view, and AMac's questions leads to confusion. The issues should be named.

    Description issue. - For instance, saying that

    > The mistake itself isn’t such a big deal, refusing to fix the problem is the issue.

    does not identify which issue is supposed to be THE issue. Describing what's THE issue, in this specific case, will help us, among many things, to recognize that this issue is not at the same level of discourse as the other issues that were raised. Identifying this should help the reader follow the shift from technical points to more general methodological conclusions. It also underlines the concept of "explanatory power" in this specific context.

    Naming issue. - The naming should be explicit enough so that it can easily refer to a description. This alleviates memory and prevents degoratory labelling. Numbering is a problem, after a while.

    ***

    This proposal does not need to be done today. It will take time. Doing things right takes time. Besides, this is a hobby. A strange hobby, but still a hobby.

    This proposal does not truly run in parallel. This is what I am proposing to do myself. Other readers will do as they please. As always, I welcome feedback.

    I am formulating a proposal because I am not sure you'd like that to be done here. Its objective is related to the "views" on Tiljander, but the exegetical analysis will quite certainly not have the same flavour as what is done in the post. If you don't think you will like what I am proposing, I will find other venues for it.

    I am formulating this proposal as a token of recognition for your impressive work on the subject, and as a way to thank you for the calmness and the thoroughness by which you undertake to delve into all these issues.

    I shall venture less time each day on what begins to look like climate porn. This seems to me to be the best way out of nonsensical, anticlimatic exchanges. I have no preconcieved idea where this might lead, if it leads anywhere.

    So, what do you think?

    ReplyDelete
  2. Amac:

    This is honest, not said to tie you up, but to try to sharpen descirptions, so that key differences can be understood: I think your point 1, needs to be more precise, just to help the discussion.

    When you say that the proxy has to be calibrated versus instrumental, agreed. Pretty much everything needs that* and it's not really something that either side disagrees with.**

    It's when you say, that contamination prevents the calibration. Well...contamination didn't nescesarily prevent calibration in terms of making it fail some math test or give a poor correlation (or it would have dropped out already). the issue is one of physical rationale. to even decide to look at the proxy at all, there is some sort of physical rationale (we can argue about how good of one it is, how good it should be, but it's at least at the level of an environmental measurement for which some sort of physical explanation of temp influence is tenable.) I think when you say, that it's "impossible" to calibrate, it's not that it's really "impossible" to calibrate...since...er...it calibrates and doesn't just give nothing as an answer...but that it's "wrong" to calibrate. Note this issue of physicality is one that could really be a concern in non-calibration periods as well. what if we had the same physcial debate around Tiljander in the proxy period and the calibraion period was clean? I think the logic of he disagreement would be the same.


    *Yes, boreholes are direct measurments. And one could conceive of some proxies where the recent period was missing, but they were calibrateable based on other comparables which had run the direct calibration (although for tree rings, we really already precombine old and new trees of the same location and species into a series).

    **There are some subtleties like the using part of the period for calibration, part for verification. also for the global versus local of CPS verus EIV. CPS requiring a predetermined orientation, EIV not. Etc. These are fine issues though and I think both sides agree with the basic CONCEPT that the way you use the proxy to tell if a year was warm in the past is based on seeing how much instrumental years move it's value.

    ReplyDelete
  3. Here's some varve stuff I found on the net:

    Dean and Minnesota lake with dust blowing into it:

    http://pubs.usgs.gov/fs/fs-0059-99/
    http://www.sciencemag.org/cgi/content/abstract/226/4679/1191

    I have not read the Science paper yet. Need to get to the library. Don't have paid access at home. Impression I get is that the varves there are more of a precip series. And really based on the dryness blowing dust more. The other thing interesting is a bit more definition of the chemical species in the layers.

    Swiss lake study.

    http://comp.uark.edu/~shausman/paleolimnology/Seebergsee.pdf

    Interesting that they deconfuound climate effects from landuse and seem to see their series as mostly inferring land use issues.

    Here's one to give caution on making sure that the layers are annual:

    http://www3.interscience.wiley.com/journal/119602189/abstract

    ReplyDelete
  4. I'm looking for stuff on the web for limnology 101.

    Wikipedia did not have much and surfing around most of what I saw was old.

    Then there's Colbert: http://noolmusic.com/funny_videos/colbert_report_-_on_notice_journal_of_paleolimnology.php

    A chapter in a book would be ideal.

    ReplyDelete
  5. Have you looked at thickness much? What's the story there?

    ReplyDelete
  6. Thanks for these. I'll have to put up a new category on the most relevant paleolimnology stuff. I just ran into this --

    M. Tiljander, J.A. Karhu, & T. Kauppila. “Holocene records of carbon and hydrogen isotope ratios of organic matter in annually laminated sediments of Lake Korttajarvi, central Finland.” J. Paleolimnology 36: 233-243, 2006.

    Dunno why literature searches missed it!

    As to "thickness", as you see, Tiljander was reticent in Boreas. No assigned climate meaning, thus it can't be "upside down" with respect to Tiljander03. Basically the varves get much thicker as 20th Century contamination/pollution/runoff/whatever gets more intense.

    One thing I have to chase down is the values of the proxies in the final years of the calibration period. Somebody (Jeff Id?) mentioned that Mann08 might have used RegEM to fill in the last bits. What I pulled from NOAA on the other three ended in 1985 not 1995 (IIRC), the paper was 2003 so a record through the late nineties is what I'd have expected. Consolidation might be a big issue here, the first few cm of mud might just swirl up and make a mess, as drilling/coring starts.

    ReplyDelete
  7. Numbering in the post -

    At generalist blogs, I get tempted to go into detail. This isn't usually a good idea; I think most readers want an overview, What are you saying; Why do you think it matters? So the six points were an effort to say a few discrete, declarative things that readers could evaluate (reasonable? correct?), and that Gavin could respond to, if he wanted to.

    Gavin's numbering is as in his C-a-s comment. Keith Kloor doesn't provide a preview, so the double "3)" is an understandable typo. I left it as is, because it wasn't clear to me if Gavin was meaning to do a point-to-point response, or to state his major points, which by happenstance also numbered six.

    Willard,

    Thank you for the kind words. I actually don't know what an exegesis is (I'll look it up). This isn't really a blog, I just wanted a place to post up stuff that I could easily access and point people to, in discussions in the comments at high traffic blogs. My posting record is spotty and will go down as Tiljander (hopefully) moves into the end-game phase. So my suggestion would be to do as I did and sign up at Blogspot or Wordpress for one of their freebies. I'll be sure to link whatever you write. Or email me to discuss more.

    ReplyDelete
  8. Gavin's point number 2 is my real sticking point, and the apparent use of "teleconnections" without using that term.

    I still don't know how it's possible to have a non-local correlation. What is it correlated to if not local?

    ReplyDelete
  9. Amac,

    Fair enough. I'll pursue this project elsewhere.

    Here is a comment you made at **The Air Vent** that relates THE issue:

    http://noconsensus.wordpress.com/2010/08/02/amac-on-tijander-and-hockeysticks/#comment-33714

    Readers might be interested to read it in full. Here is a slap shot:

    > Inerrancy. Cargo cult science.

    Perhaps not a one-timer, but a shot that stays near the ice is always tough to stop.

    ReplyDelete