The EPA and Upside-Down Mann

In a recent post, Keith Kloor worried about what would happen when a “bad paper spotlighted on a popular climate blog” is not debunked. In fact, even when a “bad paper spotlighted on a popular climate blog” is debunked, that doesn’t prevent subsequent reliance in policy discussions, as we see in EPA citation of the Mann et al 2008 no-dendro reconstruction. This reconstruction, as CA readers know, used the contaminated Tiljander sediments upside down. EPA described this as a “more sophisticated” methodology than averaging of upside-up proxies as done in Loehle and McCulloch.

The discussion arose in EPA’s refusal of various petitions for reconsideration – see here and here.

Various petitions for re-consideration cited Climategate emails as evidence against the Hockey Stick reconstructions used by IPCC:

Petitioners also claim that the Medieval Warming Period may have been warmer than present temperatures, undermining the conclusion that greenhouse gases are a primary cause of current warming. Issues involving the Medieval Warming Period were addressed during the public comment period (see Response 2–62 of the RTC document) Petitioners raise this issue again because of their assertion that the CRU e-mails indicate that the current temperature record may be faulty, which to them gives the Medieval Warming Period new significance.

Response 2-62 in the RTC document made extensive citations from the 2006 NAS Panel and the 2007 IPCC reports. The EPA continued:

In making their case, petitioners cite a temperature reconstruction without tree rings, notably a study that could have been submitted during the public comment period. [24 - Loehle and McCulloch] However, that paper uses an improper methodology, a straight average of 18 proxies, apparently without weighting them to account for geographic distribution or the strength of the data to detect temperature changes. In contrast, another study using a more sophisticated methodology [25 - Mann et al 2088] found that recent Northern Hemispheric warmth was anomalous regardless of whether tree ring data were included.

EPA’s observation that Loehle and McCulloch “could have been submitted during the public comment period” is a procedural point relating to the petitions for reconsideration. As I understand it, a petition for reconsideration requires evidence unavailable in the public comment period. The Climategate dossier was released after the closing of the comment period and was new material for EPA purposes, while Loehle and McCulloch, as EPA observes, could have been cited by commenters in the public comment period.

Even within the standards of the IPCC community, EPA had no basis for stating that Loehle and McCulloch used an “improper methodology”.

EPA criticized Loehle and McCulloch for not “weighting [proxies] to account for geographic distribution or the strength of the data to detect temperature changes”. However, they had no basis for criticizing Loehle and McCulloch on this point for a couple of reasons.

First, prominent reconstructions relied upon by IPCC did not weight proxies to account for geographic distribution, despite claims to the contrary. The most familiar example for CA readers is Mann et al 1998-99. As CA readers know, MBH methodology (particularly using Mannian PCs, but even using centered PCs) did not weight proxies to account for geographic distribution, but, instead, overweighted strip bark bristlecone and foxtail chronologies in southwest US. Mann et al 2008 RegEM methodology also does not weight proxies to account for geographic distribution.

In fact, the even weighting of Loehle and McCulloch prevents the very over-weighting observed in Mann et al. Nor does weighting procedure of Loehle and McCulloch warrant EPA disapproval relative to most reconstructions in the IPCC spaghetti graph, as they also use more or less similar weighting systems: e.g. Jones et al 1998, Hegerl et al 2007, Moberg et al 2005 and Esper et al 2002. (Actually Esper et al 2002 used two nearby strip bark chronologies as separate proxies.) Further some of the IPCC reconstructions using relatively even weighting for the proxies shown had used the biased Mannian PC1 which overweighted strip bark bristlecones e.g. Hegerl et al 2007, Mann and Jones 2003. So there is no valid reason for EPA to disapprove of Loehle and McCulloch on this count. Quite the opposite.

Their endorsement of the Mann et al 2008 no-dendro reconstruction is equally questionable. This has been discussed in numerous CA posts. It was observed early on that they used the Tiljander sediments upside-down from the orientation of the original authors and that the modern portion of these sediments was contaminated, preventing calibration by correlation to instrumental records.

The “more sophisticated” method endorsed by EPA was to ignore these problems. Regardless of whether the original reconstruction was an oversight, the problem was reported in early 2009 at PNAS and denied by the authors. Much later, the failure of the no-dendro reconstruction to calibrate before 1500 was conceded by Mann et al in the supplementary information to a different article, without a retraction or corrigendum being issued at PNAS or disclosed at the “grey” Penn State website for Mann et al 2008.

Indeed, to my knowledge, there was no public awareness that the SI to Mann et al 2009 withdrew the main claims of Mann et al 2008 until early August 2010, when it was mentioned by Gavin Schmidt inline at realclimate, immediately provoking commentary at CA and elsewhere. This was only a few days before the EPA published its response to the petitions on August 13, 2010.

Had Mann and his coauthors properly retracted the untrue claims in Mann et al 2008 about their no-dendro reconstruction (as they should have), EPA would not have relied on these untrue claims in their Response to Reconsideration to say:

another study using a more sophisticated methodology [25 - Mann et al 2088] found that recent Northern Hemispheric warmth was anomalous regardless of whether tree ring data were included.

36 Comments

  1. diogenes
    Posted Nov 13, 2011 at 11:43 AM | Permalink | Reply

    http://tamino.wordpress.com/2011/11/09/open-thread-6/#comments

    a certain rumnleyfips is making a set of standard “rapid response team” posts on this very subject…a strange cooincidence

  2. JamesD
    Posted Nov 13, 2011 at 11:53 AM | Permalink | Reply

    The points I am about to make are VERY SERIOUS for our future. There are challenges to the EPA going through the courts. In a court case, the truth does not necessarily win. Instead, the party with the better case wins. Of course the truth gives you a better case, but it is not guaranteed.

    This article demonstrates that the plaintiffs have a very serious case. We have absolutely proven failed methodology (anyone can understand upside down Tiljander after about a 5 minute demonstration). And therefore the response to the policy objections by the EPA is seriously flawed. This gives the plaintiffs a very strong case.

    If anyone reading this article has connections with those conducting these legal challenges, please forward them this article and also explain “upside down” Tiljander to them. Or, if you have a reliable means to contact Senator Inhoff (spelling?) of Oklahoma, besides the usual internet means on his site, please notify him of this article.

    • steven mosher
      Posted Nov 13, 2011 at 1:27 PM | Permalink | Reply

      suggest you ask Anthony at WUWT to contact Imhoff

      • ChE
        Posted Nov 13, 2011 at 1:38 PM | Permalink | Reply

        I’m sure CA is daily reading for somebody on his staff, not to mention Horner and several others.

      • steven mosher
        Posted Nov 13, 2011 at 4:01 PM | Permalink | Reply

        sorry I was reading Marc Imhoff last night

        • EdeF
          Posted Nov 14, 2011 at 9:57 AM | Permalink

          Maybe you were channeling Darrel Imhoff, who played for the Lakers in the 60s…

      • Posted Nov 13, 2011 at 7:25 PM | Permalink | Reply

        Not sure why my comment was deleted. The person’s name is Inhofe, not Imhoff, unless you’re not referring to James Inhofe.

        • Hum
          Posted Nov 14, 2011 at 1:04 PM | Permalink

          Nice twit Jeff :)

  3. clivere
    Posted Nov 13, 2011 at 12:31 PM | Permalink | Reply

    Previously posted at AMac’s.

    Pages 30 and 35. How to use a paleo reconstruction to influence government policy.

    http://downloads.globalchange.gov/ocp/ocp2010/ocp2010.pdf

  4. Gary
    Posted Nov 13, 2011 at 1:33 PM | Permalink | Reply

    From now on when I detect anyone obfuscating, prevaricating, equivocating, or just plain lying, I’ll have to remember this is “sophistication.”

    • ChE
      Posted Nov 13, 2011 at 1:40 PM | Permalink | Reply

      It’s sophistry, anyway.

    • ChE
      Posted Nov 13, 2011 at 1:42 PM | Permalink | Reply

      Note the common root with “sophomore’.

      • Roy
        Posted Nov 14, 2011 at 6:21 AM | Permalink | Reply

        The word “sophisticated” has several definitions, few of which are positive in modern usage. The more numerous negative ones are clearly appropriate here. From the OED: 1. Mixed with some foreign substance; adulterated; not pure or genuine; 2. a. Altered from, deprived of, primitive simplicity or naturalness. Of a literary text: altered in the course of being copied or printed; 3. a. Falsified in a greater or less degree; not plain, honest, or straightforward; 3 b. Of a printed book, containing alterations in content, binding, etc. which are intended to deceive.

  5. theduke
    Posted Nov 13, 2011 at 1:36 PM | Permalink | Reply

    If it fits the narrative, it’s “sophisticated,” “consensus” science. If it doesn’t, it’s “improper” science.

  6. steven mosher
    Posted Nov 13, 2011 at 1:48 PM | Permalink | Reply

    Steve, you might also note that in Manns most recent effort they made some “adjustments” to handle issues with geeting a good sea level rise prediction using Mann’s proxies.

  7. Robert
    Posted Nov 13, 2011 at 2:39 PM | Permalink | Reply

    Hi all,
    I have a quick question about Mann 2008. First when you go to the following website:

    http://www.ncdc.noaa.gov/paleo/pubs/mann2008/mann2008.html

    You find for example that link to his cps reconstruction (as a .zip file) and when you enter there i’m wondering which series would be used for example for the NH data… I see there is a nhcru_cps_composite csv and a NH_Land csv which has a cps(cru) column in it. The problem is that the two aren’t the same and don’t start at the same time. Similar issues are found in the EIV file. Which is the appropriate one to use?

    Secondly (and more importantly) is a question of whether or not the values recorded as annual anomalies were smoothed prior. I see some evidence that suggests to me the annual values in the reconstruction were smoothed previously and i’m wondering is there anywhere to find the non-smoothed versions, i.e. with the R code here or potentially using the Matlab code is it possible to avoid the final smoothing?

    • Posted Nov 13, 2011 at 3:21 PM | Permalink | Reply

      justg wondering if you have already asked the authors of the paper?

      • Robert
        Posted Nov 14, 2011 at 10:39 AM | Permalink | Reply

        I’m not meaning to speak ill of the authors but I think it might be quicker/easier to get a response here from the people like Steve who have already put in a lot of work understanding these issues.

    • Steve McIntyre
      Posted Nov 14, 2011 at 1:49 PM | Permalink | Reply

      The data as archived is not smoothed. I’ve placed R collations online at http://www.climateaudit.info/data/mann.2008 (which could use a readme). Note that there are peculiarities with the data – Mann’s MXD has “hidden” the decline and replaced it with “infilled” data. I’ve done a re-assembly with realdata – which I could locate. Mann also includes Luterbacher data which incorporates instrumental data and inflates proxy statistics. And the notorious contaminated upside-down sediments. Lots of things to notice.

      Mann has very quirky and ad hoc splicing methods – that have been noted here but not fully parsed in CA posts. I’ve pretty much figured it out , but I don’t think that I’ve written it up yet.

      As to “which is the appropriate one to use” – hard to say.

      I’ll try to revisit this and write an overview.

      • Robert
        Posted Nov 15, 2011 at 2:01 AM | Permalink | Reply

        Okay from what I’m reading from you I’m actually better understanding what I’m seeing in the analysis being conducted.

        I’ll have a look at your functions and try to decipher a bit what i’m seeing.

        RegEM infilling for years 1986-1995… This is based on Schneider’s (Tapio?) algorithm?

        Steve: yes. however, in the rather different application of proxy reconstructions, the inherent linear algebra can be dramatically simplified. I’ve had a post in the works on this for quite a while but haven’t posted the results. The analysis will also clarify some of the bewildering behavior of this technique.

    • Posted Nov 14, 2011 at 3:41 PM | Permalink | Reply

      Robert (Nov 14, 2011 at 10:39 AM )

      I have a quick question about Mann 2008… [I am interested in] whether or not the values recorded as annual anomalies were smoothed prior.

      If it is helpful, my discussion of the four (actually two) Tiljander data series can be found here. At that post, there is a link to the file “Tiljander03-Calculated_LS+DS.xls”, which contains the annual values from 0 (sic) through 1985 as archived by A.E.K. Ojala. The data as archived for Mann08 are also presented.

      Note the RegEM-based infilling for the years 1986-1995, as Steve stated at 1:49 PM above. Also note that Ojala reported a very thick (~13 mm) varve for the year 1326. The Mann08 archive appears to have put the averages of 1325 and 1327 in the place of the original entries for 1326. This substitution is not described in the text or SI of Mann08.

  8. Craig Loehle
    Posted Nov 13, 2011 at 2:39 PM | Permalink | Reply

    Ha Ha Ha Ha!!! Love it. Criticized for being “not sophisticated enough”. That is rich. When these guys are using data upside down, transposed geog coordinates, giving most of the weight to a few trees in SW USA, ignoring the strip-bark growth response, using six-sigma yamal larch, maybe something simple is the antidote. How can they claim “geographic weighting”? They wish. Just by the fact that most of my proxies were not near each other, their even weighting is more geographically valid than Mann’s work. If I remember the wording, Mann’s comment on my paper in an email of climategate was simply “awful, of course”. You might guess that I was not insulted, but quite the contrary. Better simple than using fancy methods WRONG and making up funny verification stats and withholding adverse results. Doing things you do not properly understand is called putting on airs.

    • Posted Nov 13, 2011 at 3:17 PM | Permalink | Reply

      The problem, Craig, is that I think Mann understood exactly what he was doing.

  9. Kenneth Fritsch
    Posted Nov 13, 2011 at 4:08 PM | Permalink | Reply

    Perhaps the climate scientists/advocates are more in the advocate frame of mind when they publish papers like Mann (08) with very questionable methodologies and the predictable and hand waved conclusions. That a published paper, regardless of merit or sometimes even context, can become a very important document when policy has to be rationalized would appear to me to be a savvy advocacy move. And who can blame the policy makers for not having the technical expertise to understand the weaknesses in the paper cited, although the policy maker who was truly interested in an objective finding would not merely cite references without some detailed discussion/analysis of the reference.

    There was much more wrong with Mann (08) than Tiljander. Without the instrumental record tacked onto the end of the reconstructions one can see that the dendro and non dendro reconstructions meander over the years and both show divergence in the instrumental period. With the Schweingruber MDX series Mann et al simply lop off the proxies at 1960 and without the needed rationalization for that drastic action and replaces it with something not well-explained. Mann used the jiggered Schweingruber series prominently in his screened test and also in an attempt to show that the correlations could not have occurred by chance.

    I think it is very important when criticizing Mann (08) to mention all the problems and not merely refer to it as using an upside down proxy.

    In my view the Lohele and McCulloch reconstruction shows that one can obtain different results depending on which proxies are selected, but not that these proxies have been shown to proxy temperature well. I suspect and conjecture that most, if not all, proxies used in prominent published reconstructions follow some long term persistence model (while reacting qualitatively to extreme events like major volcanic eruptions). Global temperature might have long term persistence but during the recent warming period temperature increases are a result of a deterministic trend in the series and a trend from which the proxies appear to be diverging.

    • Posted Nov 13, 2011 at 6:43 PM | Permalink | Reply

      it is important if not vital to signal all the errors and misdirections…otherwise you get John McManus puffing awsay on Tamino…chapter and verse at all times

  10. Hu McCulloch
    Posted Nov 13, 2011 at 5:33 PM | Permalink | Reply

    The Loehle and McCulloch (2008) article and my SI are available at http://econ.ohio-state.edu/jhm/AGW/Loehle/ .

    FWIW, my 5/21/09 comment to EPA at
    http://econ.ohio-state.edu/jhm/AGW/EPA-McCulloch-0509.pdf did call their attention to L&M 08, so I’m not sure why they now say that it was not submitted during the comment period:

    In making their case, petitioners cite a temperature reconstruction without tree rings, notably a study that could have been submitted during the public comment period. [24 - Loehle and McCulloch]

    The EPA goes on to complain,

    However, that paper uses an improper methodology, a straight average of 18 proxies, apparently without weighting them to account for geographic distribution or the strength of the data to detect temperature changes.

    In my 2008 SI linked on the above webpage, I did report Weighted Least Squares results that weighted each of Craig’s 18 proxies according to their inverse variance about the common mean. This was qualitatively similar to the unweighted mean (except a little more attenuated), so I was comfortable with Craig’s original equal-weighted mean. In any event, this is what Moberg (05) had done in his “low-frequency” reconstruction with only 9 proxies.

    I agree with Craig that given the good distribution of his 18 proxies, geographic weighting wouldn’t have made much difference. The data is linked on Craig’s and my websites in case someone wants to experiment with this. (I’m busy with other stuff right now, but should give it a try some day.)

    In any future expansion of Craig’s study to include more proxies, they will start to be closer together and geographic weighting will become more of an issue.

    There aren’t enough independent observations to measure an unconstrained covariance matrix, but it should be feasible to estimate an exponential decay model.

  11. KnR
    Posted Nov 13, 2011 at 5:53 PM | Permalink | Reply

    The protection of Mann’s work has nothing to do with its questionable value to science but its value in the politic game the IPCC etc play , the Hockey stick for example is not a pace of research , and a poor one at that , but an ‘Icon ‘ of a viewpoint held so strongly by some that it matches religions extremism . Which means any challenge to it will be meet with the all of the energy and varsity that lead to people being burned at the stack has heretics.

    So its hardly surprising that anything that supports this icon is held up as perfect proof of its validity and never mind its actual scientific value .

  12. Craig Loehle
    Posted Nov 13, 2011 at 6:09 PM | Permalink | Reply

    To reiterate, my purpose in doing the analysis was to see what happens when you do not use tree data (which I believe to be hopelessly compromised for anything more than about 200 years ago). I do not believe the proxies I used are necessarily accurate nor that my reconstruction is “right”–though it does match up with the pre-Mann view of both geologists and anthropologists.

    • JamesG
      Posted Nov 13, 2011 at 7:23 PM | Permalink | Reply

      As far as the mwp is concerned there is not that much difference anyway. The real difference is in the uptick at the end, which is not in L&M.
      http://farm3.static.flickr.com/2666/3914214320_261cba1cf2_o.png

      Hence L&M cannot refute Mann08; only a redux of Mann08 without Tiljander or trees can do that. But even then they’d still splice on that instrumental record to show an acceleration of warming; whether ethical or not.

      Personally I’m of the opinion that all such global recons are nonsense due to lack of data for the Southern hemisphere and the sea. But…we really only need to look at the Arctic and Antarctic (if possible) to determine any unnatural acceleration, don’t we?

      Steve – the issue under discussion here is not whether LM “refutes” Mann et al 2008, but whether the EPA criticisms had any basis.

      • Posted Nov 13, 2011 at 9:51 PM | Permalink | Reply

        the issue under discussion here is not whether LM “refutes” Mann et al 2008, but whether the EPA criticisms had any basis.

        And earlier:

        even when a “bad paper spotlighted on a popular climate blog” is debunked, that doesn’t prevent subsequent reliance in policy discussions

        A crushing point, given the fire and fury over Judith Curry allowing public debate of Lüdecke, Link, and Ewert. Thanks as always for keeping our eye on the ball.

  13. Kenneth Fritsch
    Posted Nov 13, 2011 at 8:05 PM | Permalink | Reply

    “As far as the mwp is concerned there is not that much difference anyway. The real difference is in the uptick at the end, which is not in L&M.”

    Upticks at the end are often the instrumental record tacked onto the reconstruction. Reconstructions should stand alone and if they do not extend far enough into the late instrumental period let us wait until someone can produce that data.

    • JamesG
      Posted Nov 15, 2011 at 6:25 PM | Permalink | Reply

      The uptick is produced by overlapping the instrument record with the proxy data. That this overlap period is erroneous is the only point worth making. The EPA was right to reject LM even though they gave a reason that was likely wrong. But frankly, it sounds sensible because we are comparing a study of 18 proxies with a study of 1209. Skeptics need to forget LM; it doesn’t refute anything, it actually agrees with Mann08 as shown above! As such it was a foolish petition.

      Steve: the Mann 1209 is illusory in respect to the medieval warm period. It relies on bristlecones and/or upside down Tiljander. Whatever reasons exist for rejection of LM apply equally to other studies.

  14. genette
    Posted Nov 14, 2011 at 7:49 AM | Permalink | Reply

    Steve, typo alert – you have twice called the paper “Mann et al 2088″ instead of “2008″.

  15. Noblesse Oblige
    Posted Nov 14, 2011 at 1:07 PM | Permalink | Reply

    Re: Whether EPA criticisms of LM have any basis.

    “If you want a stick to beat a dog, any stick will do.”

  16. Posted Nov 15, 2011 at 1:05 PM | Permalink | Reply

    Regarding no-dendro reconstructions, I came across an interesting paper on Scandinavian speleothems, which was unusually frank about their limits as historical thermometers, stating among other things:

    - Every sample has a different response to climate. Even two stalagmites ten centimeters apart in the same cave will have different climate signals.
    - Mon-linear responses should be expected due to the non-linear hydrology of karst (soluble rock).

    Especially interesting since one of the series in the paper, Søylegrotta, Norway, was used to construct figure 6.11 of AR4.

    It has probably been covered here before, but here is the paper:

    http://docs.google.com/viewer?a=v&q=cache:YQwd9hleR20J:www.gsf.fi/esf_holivar/baker.pdf+s%C3%B8ylegrotta+mo+i+rana&hl=no&gl=no&pid=bl&srcid=ADGEESjSo-71tXaLberZydyBdSrCPrcTSaWkRwyVyGBHMxgFbytRntl6P8jABf6XRt9ZoGvxc6ZCUc0RauLOPGzIbLtAIRk7JqhjS6hpS-tjWZxjiNcO2dZkVtbBhbJ-5AImeM029RAa&sig=AHIEtbT0UkR2nzBoy8qu6HDz4Ou4cGR6rw

One Trackback

  1. By Climategate 2.0 « Crazy Jew on Nov 22, 2011 at 12:50 PM

    [...] The EPA and Upside-Down Mann (climateaudit.org) GA_googleAddAttr("AdOpt", "1"); GA_googleAddAttr("Origin", "other"); GA_googleAddAttr("LangId", "1"); GA_googleAddAttr("Autotag", "religion"); GA_googleAddAttr("Autotag", "politics"); GA_googleAddAttr("Autotag", "science"); GA_googleFillSlot("wpcom_sharethrough"); Share this:LinkedInPrintEmailTwitterFacebookDiggRedditStumbleUponLike this:LikeBe the first to like this post. [...]

Post a Comment

Required fields are marked *

*
*

Follow

Get every new post delivered to your Inbox.

Join 3,114 other followers

%d bloggers like this: