Tuesday, August 10, 2010

Synopsis of some Tiljander-related arguments in a recent Climate Audit thread

This post is a placeholder that will hopefully be superseded by a more careful and thorough description of a discussion of the use of the four Lake Korttajarvi lakebed sediment series (the "Tiljander proxies") in Mann08 (PNAS). This discussion was largely between two pseudonymous commenters at Steve McIntyre's Climate Audit post The No-Dendro Illusion.

I'm putting this together to respond to this recent comment by DeepClimate, at his blog.


Note that I have assembled this version of the Climate Audit thread to highlight the scientific issues concerning Tiljander that were being debated. I've corrected some typos, altered some formatting, and snipped remarks that I didn't think were central to the main thrust of the discussion. Of course, the original remains a click away. If I've mangled any comments or left out any key thoughts, please make a note in the comments.

(In this rough-draft post, I expect there are more typos and formatting errors, and that I'll want to make some content additions and changes. I reserve the right to do these things, without notice. When things settle down, I will put strikes through the relevant sentence.)

- - - - - - - - - -

amac78 -- Aug 2, 2010 at 11:40 PM

The fundamental problem with Mann08′s use of the four Tiljander proxies is this:

1. Mann08′s methods absolutely require that proxies be calibrated to a gridcell derivation of the instrumental temperature record, 1850-1995.

2. The Tiljander proxies cannot be properly calibrated to the instrumental temperature record, because non-climate local factors progressively overwhelm any climate-related signals, in the 19th and 20th centuries.

3. Mann08′s authors mistakenly set aside the Tiljander03 authors’ warnings of contamination, and performed a set of four faux [sic -- substitute "nonsense" -- AMac] calibrations. 1850-1995, the major influences on the varve series were nearby agricultural activities, roadbuilding and bridge reconstruction, and lake eutrophication. These activities caused Darksum, Lightsum, X-Ray Density, and Thickness to increase during the screening and validation periods. The gridcell temperature anomaly calculated by CRUTEM3v also increased during those periods, leading to positive correlations between proxy values and temperature. The correlations of Darksum, Lightsum, and Thickness were judged sufficiently high to be Significant (XRD’s correlation was lower).

All four correlations are meaningless in that they have nothing to do with any causal relationship between any climate-related factor and any proxy-related characteristic. A trivial consequence of this procedure is that Mann08′s authors unwittingly assigned meanings to Lightsum and XRD that were opposite in orientation to the assignments proposed by the only relevant authorities, Mia Tiljander and the co-authors of her Boreas paper in 2003.

[AMac note (10/28/10) -- I was mistaken to take Mann08's authors word for it, and talk about the "four Tiljander proxies." There are only three, as "Thickness" is simply "Darksum" plus "Lightsum". More exactly, Tiljander derived "Darksum" by subtracting "Lightsum" from "Thickness." Either way one looks at it, the degrees-of-freedom argument is the same. See the first graph in the post The Tiljander Data Series: Data and Graphs.]

- - - - - - - - - -

amac78 -- Aug 2, 2010 at 11:52 PM

Links to the Tiljander et al (Boreas, 2003) paper, other primary literature, and data archives are here.

This “Jarvykortta Proxy” post walks through Mann08′s faux [sic -- substitute "nonsense" -- AMac] calibration of the XRD proxy.

This post contains the temperature anomaly that CRUTEM3v calculated for the 5 degree x 5 degree gridcell that includes Lake Korttajarvi.

Arthur Smith’s post Michael Mann’s errors (now closed to comments) contains the most accessible layperson’s discussion of Mann08′s use of the Tiljander proxies. The best arguments of the pro-Mann08 bloggers are on display there, for what they are worth.

Lastly, Gavin Schmidt, Michael Mann’s co-blogger at RealClimate.org, called me out at Collide-a-scape, in the guest post The Main Hindrance to Dialogue (and Detente). His best arguments are there, for what that is worth.

- - - - - - - - - -

AMac -- Aug 2, 2010 at 8:30 AM

I have made an effort to record literature citations and blog posts that discuss the use of the Lake Korttajarvi (Tiljander) lakebed sediment records in Mann08 (PNAS). The compilations may be useful to newcomers to this story.

Mann08 considered four measures that Tiljander and her co-workers derived from their lakebed sediment drill cores. Thus, there are four Tiljander proxies. Two of them are miscalibrated and “upside-down,” one is miscalibrated and rightside-up, and one is miscalibrated and of indeterminate orientation. Thus, miscalibration rather than upside-downedness is the core (heh) issue.

I have walked through the precise issues with Mann08′s employment of one of the two upside-down proxies, X-Ray Density, in the post The Newly Discovered Jarvykortta Proxy — II.

For an exhibit of the best current AGW Consensus arguments in favor of Mann08′s uses of the Tiljander proxies, see the (ongoing) comment thread of Arthur Smith’s post, Michael Mann’s Errors (while a supporter of the AGW Consensus, Arthur has done a commendable job as an ‘honest broker,’ allowing all parties to have their say.)

Within that thread, my comment “It’s certainly helpful to…” describes the three proxies’ assignments in Tiljander03, and contrasts them with the assignments for them in Mann08. (A climate-related discussion of the fourth Tiljander proxy, “Thickness,” was not provided in Tiljander03. Thus, Mann08′s use of it cannot be “upside-down” or “rightside-up” with respect to the earlier authority’s interpretation.)

- - - - - - - - - -

Jarkko -- Aug 4, 2010 at 4:36 AM

Can I suggest something? Take a look at Google Earth. The Street View gives you a good impression of the surrounding enviroment. I live pretty close to Lake Korttajarvi. The bridge doesn’t cross the lake, but the river running into it. The lake is right next to the town of Jyvaskyla, population 130,000.

The coordinates of the lake are 62°20’16.10″E and 25°41’30.14″N.

This recently puplished academic dissertation finds some support for the interpretation of the proxies by Tiljander:

T.P. Luoto. "Spatial and temporal variability in midge (Nematocera) assemblages in shallow Finnish lakes (60−70 °N): community-based modelling of past environmental change." Helsinki University Press, 2010. PDF.


- - - - - - - - - -

scientist -- Aug 4, 2010 at 3:46 PM

1. [snip]

2. I skimmed the midge thesis you referred me to. Can you please identify what parts of the thesis validate “the Tiljander thesis”? Page number? We are talking about two 70 page documents!

3. And does the midge thesis validate the bridge concern, or Tiljander’s overall qualitiative comparison of her core to Roman Warm Period and such?


- - - - - - - - - -

AMac -- Aug 4, 2010 at 10:54 PM

Re: scientist (Aug 4 15:46),

> 2. I skimmed the midge thesis, you referred me to. can you please identify what parts of the thesis validate “the Tiljander thesis”?

Thanks for the opportunity to clarify. I wrote,

It’s highly likely that Tiljander03′s overall interpretations of the Lake Korttajarvi data series as climate proxies pre-1720 are generally correct. Tiljander and her advisor and co-authors are the sole ‘local’ authorities; their views are consistent with other paleolimnologists’. E.g. … Tomi Luoto’s repeated citations of Tiljander03 in his 2010 dissertation [on fossil midges]…

Luoto cites Tiljander et al (2003) seven times. Using the MS pages (not the PDF paging) –

p. 38, “Tiljander et al. (2003) noted from annually laminated lake sediments from southern Finland that during two phases, ca. 1580-1630 and 1650-1710 AD, a colder climate prevailed.”

p. 38, “The study of Luoto et al. (2008) also suggested that the LIA was not uniformly cold… which was consistent with the results of Tiljander et al. (2003).”

p. 39, “the LIA was characterized by wetter climate conditions (Tiljander et al. 2003)…”

p. 42, “The maximal inferred water depth value occurred in Lake Iso Lehmälampi ca. 1700 AD, which is consistent with the results of Tiljander et al. (2003), who suggested wetter climatic conditions at that time.”

p. 42, “The coolest time period of the LIA occurred ca. 1700 AD in southern Finland (Tiljander et al. 2005; Fig. 14)…”

p. 43, “The oxygen levels in Lake Iso Lehmälampi experienced a major increase around 1700 AD, occurring simultaneously with the coldest period of the LIA in southern Finland (Tiljander et al. 2003; paper II).”

p. 45, “The results of Tiljander et al. (2003) from the central part of the country [indicate that] the LIA was a wetter climatic episode in Finland.

These cites do not, of course, “validate Tiljander”–nor was that my claim. They address a narrower but still worthwhile question: how are Saarinen, Tiljander, Saarnisto, and Ojala viewed by relevant paleolimnologists? Are their views outliers; are their results discounted due to sloppiness, or unreliability, or a history of unfounded speculations? Here, we see Prof. Salonen of the University of Helsinki,apparently comfortable with his student Tomi Luoto citing Tiljander03 as an authority for characterizing the Little Ice Age (ca. 1550-1850) as being a variably colder, wetter period in southern Finland.

Are there opposing interpretations from these or other records that suggest something different, e.g. that contest Tiljander03 on climate regimes in southern Finland over the past centures? Judging from this sample, that does not seem to be the case.

> 3. And does the midge thesis validate the bridge concern or Tiljander’s overall qualitiative comparison of her core to Roman Warm Period and such?

No. I wasn’t anticipating such validation. Looking now, I don’t see any.

- - - - - - - - - -

AMac -- Aug 5, 2010 at 12:23 AM

Regarding another question scientist asked (somewhere else on the thread, earlier today) — Here’s something pretty interesting.

I just looked at Fig. 14 of TP Luoto’s dissertation on pg. 36; it includes a water temperature graph of Lake Hamptrask, Finland (just northeast of Helsinki) from about 1350 to the present. It’s inferred from fossil Chironomids. There’s a pretty clear dip where the Little Ice Age should be. From Fig. 5, Lake Hamptrask is about 270 km south of Lake Korttajarvi. At Bitbucket.org, I’ve uploaded a file that shows three Tiljander proxies (Darksum, Lightsum, XRD; 20-year averages, oriented per Tiljander03) and the Luoto proxy (image reversed and rotated from the dissertation).

The file is TiljHamptrask.jpg ... All three Tiljander proxies are contaminated post ~ 1750, but that still leaves ~400 years to compare with Luoto’s record. To me, none of the Tiljander records seem to say much about climate changes (e.g. temperature or precipitation) going into the LIA.

Flipping Lightsum or XRD to make them match with Mann08 instead of Tiljander03 doesn’t help matters.

Of course this is a comparison of trends from places 270 km apart, one near the Gulf of Finland, but still. The LIA was supposed to be a widespread phenomenon.

- - - - - - - - - -

scientist -- Aug 5, 2010 at 1:31 AM

Very pretty graphs. Just shooting from the hip, but:

1. independence

-boy LS and DS seem very anticorrelated. Not perfectly, but pretty close.

-And then LS and XRD are anticorrelated as well (graph of XRD is inverted axes).
-Makes it look like there’s one primary variable in all 3 series.

2. Agreed, you don’t see that LIA-posited dip from Luoto on the XRD.

3. That 1300 excursion in the Tilj series are interesting as well. Was that one super varve driving the whole 20 year block so out of whack? Wonder how she discusses it.

- - - - - - - - - -

AMac -- Aug 5, 2010 at 7:58 AM

[snip]

> 1. Independence

Note that I plotted LS upside-down (heh) to follow Tiljander03′s convention; her “warmer” is towards the top of the graph. XRD as well.

> 2. Luoto

Agreed. We see a ‘clean’ proxy series and go, “aha.” (1) LIA, (2) No 19th/20th Century craziness, where the lakes hafta be hot tubs in the winter or ice-crusted in the summer (depending on which way you orient the graph).

> 3. Yep, one super-varve! [in 1326]

Year A.D.Lightsum (mm)Darksum (mm)
13240.560.96
13250.561.14
13266.1012.86
13270.510.58
13280.470.72

[AMac note 10/28/10 -- Varve values converted to millimeters for ease of interpretation]

A piece of melting, dirty ice that dropped some muck right at the drill site? A hurricane that summer? Chariots of the Gods?

- - - - - - - - - -

PaulM -- Aug 5, 2010 at 8:03 AM

Of course, the series are not independent, and there is really only one variable. That’s immediately clear from the Tiljander paper or the Mann SI. Only a fool would attempt to use these as four independent proxies.

Correction: LS and DS are correlated, not anti-correlated.

This means that the interpretation of Tiljander et al (DS=Temp, LS=-Temp) does not make much sense.

In a more recent Finnish paper (see Aug 5, 2010 at 9:30 PM post) LS, DS and thickness are all interpreted as -Temp.

- - - - - - - - - -

scientist -- Aug 5, 2010 at 1:20 AM

My gut feel is they are not good for much also…or at least Tiljander hasn’t really figured out a way to extract much of a story from them. she had a lot of comments about how her stuff compared to other papers in the RWP [sic: Medieval Warm Period], but I got more the sense, she was reading those papers and then saying “oh…my series corresponds), than that she was really looking at the series, asking what to do they tell me on their own, pause, then compare and contrast to previous work.

I guess you could read and see what she says in Boreas about LIA…

- - - - - - - - - -

PaulM -- Posted Aug 5, 2010 at 9:30 AM

A more recent relevant paper is:

Haltia-Hovi et al (2007), "A 2000-year record of solar forcing on varved lake sediment in eastern Finland." Quaternary Science Reviews Vol. 26, p 678-689. [PDF via "Primary Links" post.]

The data from this paper (as well as Tiljander) was originally used upside-down by Kaufmann et al (2009), see Steve McI's post on the Kaufmann Corrigendum. In the Kaufman SI update correction they say “Record 21 was corrected to reflect the interpretation of Haltia-Hovi et al. (S33) that varve thickness is related inversely to temperature” (it is amazing that Mann continues to deny this after Kaufmann et al have acknowledged it). This paper has some advantages over the earlier Tiljander et al one. There were 8 cores (rather than 2) and there is less of an issue with recent human interference (more remote area). The 8 cores show a consistent picture. They say that thickness and DS are very highly correlated (r=0.99) and it looks like LS is also positively correlated with these. They interpret these as being inversely correlated with temperature. I think the argument is partly that when the lake is frozen over, stuff settles out more easily. They say MWP and LIA are ‘subtly recorded’ by the varves.


- - - - - - - - - -

scientist -- Aug 3, 2010 at 2:49 PM

Some general impressions [from reading through Tiljander's Boreas paper, most of Mann08, and skimming Tiljander's thesis].

- Seems like the parties agree that the no-tree chronology is dependent on the Tiljander proxies.

- Mann seems to have had an overly simple conception of the Tiljander series (bigger number=thicker layer=warmer year). The actual proxies were not just thicknesses, but measures of content.

- Thus, he likely used theses series without enough understanding of their limitations.

- So he probably at a minimum, should have informed the readers than his chronology was dependant on a new geo-isolated set of proxies and that they were themselves debatable. (But I think his failure was a mistake, not duplicity.)

- I would be interested if he thinks the varve series are “good”, “debatable” (i.e. equivalent to the [bristlecone pine series]), or completely unsuitable.

Some more general impressions:

- Not impressed with the Tiljander work in terms of a decisive explanation of the sediments. She does not construct a temperature series. Has a LOT of qualitative references to other studies and comparisons like ‘a lot of gray in period X’ rather than numerical, statistical statements.

- I don’t understand (this may be my limitation) the issue of LS, DS and XRD and wonder if they are all independant or how they would be physically resolved to show exactly from a compositional standpoint (both content and overall density) how these measures are driven. IOW, you have really 2 metrics: physical density and inorg/organic fraction. No?

- I’m concerned about the issue of compaction. How is this addressed? (and if this was addressed in a comment within one of the 20 different blog threads, this is not my “bad” for not reading it, it’s a flaw in having to weed though so much spinach to find the meat.)

[snip]

- - - - - - - - - -

Jarmo -- Aug 5, 2010 at 2:16 PM

I read the Tiljander paper and she is quite clear about the anthropogenic influences. 


On page 9 she noted, “The lake was quite badly polluted by waste waters in the 20th century, but since the loading of polluted waters stopped in the late 1970’s, the quality of the water has improved.”

And on page 24, “Since the early 18th century, the sedimentation has clearly been affected by increased human impact and therefore not useful for paleoclimate research.”

Tiljander may or may not have been aware that Lake Korttajärvi was partially drained by farmers and two smaller lakes were drained completely upstream. This was a common practice in Finland in the 18th and 19th centuries to create new farmland. The geological survey of the area (PDF, in Finnish, with maps) does not state how much the water level dropped.

- - - - - - - - - -

amac78 -- Aug 3, 2010 at 3:22 PM

Re: scientist (Aug 3 14:49),

[These are the central issues:]

1. The procedures in Mann08 for both CPS and EIV absolutely require calibration of all proxies to the instrumental temperature record, 1850-1995.

2. None of the four Tiljander proxies used in Mann08 can be meaningfully calibrated to the instrumental temperature record, the result of progressive contamination of any climate signals during the 19th and 20th centuries by local activities (farming, peat cutting, road building, bridge reconstruction, lake eutrophication).

3. Therefore, Mann08′s uses of the Tiljander proxies rely on faux spurious calibrations.

[Note: My term "faux" was challenged by Spence_UK as poor usage. I amended it later to "spurious" on the strength of a this 1926 statistics paper by Yule (PDF), though "nonsense correlation" seems yet better. MikeP noted Aug 4, 2010 at 2:32 AM

...Yule actually distinguished between “nonsense” correlations – correlations which exist between two quite unrelated variables as in the 1926 paper, and “spurious” correlations which he had identified in 1897 – where two variables are related but only because they both depend on a third variable which is not being considered.
-- AMac]

[snip]

- - - - - - - - - -

scientist -- Aug 3, 2010 at 4:56 PM

My impression (and I’m being honest about my level of knowledge as an outsider trying to decide fairly) is that Tiljander stuff is at a minimum questionable. I don’t think Tiljander can really explain the sediments herself, so it’s not a no-brainer in terms of “only her word counts”, but that her comments need to be taken very seriously. I would probably fall in the camp of “nobody has a good handle on what that series means. I just did not think the Boreas paper was good…way too hand-wavey…and then of course Mike hasn’t even addressed the physical issues at all. Then when you add in that this field seems even more new and evolving than tree rings? If essentially all varveologists (maybe those not Finn group) agree that the samples belong in the “complete junk” category, then I’d be inclined to take the appeal to authority. [Otherwise, I'm forced to read all the references to methodology from Tiljander's paper, get a textbook, etc. to drive myself out of the middle bucket.]

And that to use the series, since they were so crucial to the result, Mike should have disclosed the dependence of results on that proxy particularly, his different usage versus the data collector, etc.

I’m also a little concerned by some other issues.

1. That a substantial part of his paper’s new results depend on one set of geographically isolated samples (even if they were in the “good” bucket)! It’s one place in Finland!

2. That CPS doesn’t work.

3. And that the samples passed correlation screening (does this mean that the tests aren’t tough enough (not enough wiggle matching, joining two linear trends? And I admit to not knowing how to settle this statistically, just highlighting a question). OR does it mean, that really Tilj was WRONG to think that one could not get a real temp chronology out of the samples and too timid and should have taken a swing at it instead of being dissuaded by that 1999 farmer’s comment? Even if so, it still seems that Mike ought to then publish his own paper, just on Lake K.’s sediments, using Tiljander’s data, but giving a different interpretation.

- - - - - - - - - -

Steve McIntyre -- Aug 3, 2010 at 5:34 PM

You’re totally missing the problem of the contamination of the sediments by non-climatic factors such as bridge-building. Tiljander stated clearly that the sediments were contaminated in the modern period. The problem with Mann et al was that they ignored this warning.

This is an entirely separate issue from whether the sediments meant anything in the earlier period either. There are a number of posts at the site about varve-ology, in which individual sites used in Kaufman et al (2009) were examined. But that’s an entirely different issue from the very simple one about Mann’s correlation of sediments from Finnish bridge building with global climate fields – something that may qualify as a classic example of spurious correlation, up with Yule’s original alcoholism and Church of England marriages.

- - - - - - - - - -

Ed Snack -- Aug 3, 2010 at 8:34 PM

[snip]

On Tiljander (the proxies that is) I suggest you’re trying too hard. We have only one set of data and there’s no other source (as far as we know) available. To attempt to justify the inclusion of this data in a way contrary to our only source must surely require at least some form of proof, be it only a logical argument based on some physical mechanisms. I assert you cannot make any scientific claim for the inclusion based on the matching of patterns via a mathematical method without some underlying explanation. Otherwise it is just as valid to claim that the increasing trend in UK marriages (say) between 1900 and 1970 (again, for example) can be calibrated against some other set of data and hence represents a valid temperature proxy.

Do you think it reasonable for the standard of inclusion to be “it’s OK unless you can prove otherwise” on questionable data?

- - - - - - - - - -

scientist -- Aug 3, 2010 at 11:15 PM

There are zero perfect proxies. Would you eliminate them all? I don’t think all should be included or all excluded. This is a tricky issue. You lot should not try to get rid of the stuff you don’t like extra hard, nor should Mann do the reverse. I made my three buckets as a sort of simplification of what is a spectrum.

I think at a minimum, the physical concerns with Tiljander should have been better known by the authors and discussed in paper, especially given the crucial role they have for the early no-dendro EIV recon. Perhaps they don’t belong in at all, but I don’t know enough to say that right away. I mean some here want to get rid of the [bristlecone pines]…but it’s at least possible (maybe even plausible) that they are responding to temps and have done so for a while (the opportunistic trial of sheeps, dry lakebeds, barktype, precip, bad sample versus resample (Ababneh), etc. came across as a willingness to believe in anything wrong with the proxy in a rather hopeful way to get rid of them. Similar behavior with Gaspe.

I think you need to be careful not to use bad stuff or to be prone to grabbing series to support a point of view (Mann, maybe), but also wary of skeptics trying to go after whatever proxies they can with selective criticism. It’s a type I/II issue. You can have a false negative and a false positive. Can screen in bad stuff or screen out good stuff.

- - - - - - - - - -

Steve McIntyre -- Aug 3, 2010 at 11:44 PM

For something to be a “proxy”, it has to be shown to be a proxy. Something doesn’t become a “proxy” merely by being a time series.

If a “proxy” is supposed to be uniquely important, then you should be sure that you understand whether its behavior is properly understood.

[snip]

- - - - - - - - - -

amac78 -- Aug 3, 2010 at 5:34 PM

Re: scientist (Aug 3 16:56),

> I don’t think Tiljander can really explain the sediments herself

I’m not sure what you expect. From what I can see, that paper is about on a par with similar ones. Hank Roberts had a neat link to a report of German lake whose sediments clearly recorded the Younger Dryas, but that was a pretty drastic change of climate regimes (and it’s behind Nature’s paywall). For context from 5 other Finnish lakes, pull Jarmo’s PDF.

> so it’s not a no-brainer in terms of “only [Tiljander's] word counts”, but that her comments need to be taken very seriously.

Agreed.

> [Prof. Mann] hasn’t even addressed the physical issues at all.

Agreed.

> Otherwise, I’m forced to read all the references to methodology from Tilj’s paper, get a textbook, etc. to drive myself out of the middle bucket.

Try looking at the graphs of XRD in the Korttyjarva River post. For Mann08′s authors to flip two of Tiljander03′s correlations around… those are pretty extreme changes in interpretations. Also note that Kaufman09 (Science) corrected his MS to return the relevant proxies’ orientations to those proposed by Tiljander03, when he was made aware of the issue by McIntyre. Mann09 (Science) did not. So we have the spectacle of three recent and near-contemporaneous papers: one revised to be concordant with the relevant authority, and two that are discordant. Can a given value in a lakebed varve–say, an increase in Lightsum over time–mean Cooling (Tiljander03, Kaufman09) and, also, mean Warming (Mann08, Mann09)? How would that work, exactly?

> 1. That a substantial part of his paper’s new results depend on one set of geographically isolated samples…
> 2. That CPS doesn’t work.

I’m uneasy about the methods, but on an intuitive level. FWIW, it seems to me that worthwhile commentary has been written by Steve McI & RomanM (CA), and Jeff Id (the Air Vent); check those sources.

> 3. And that the samples passed correlation screening…

The 1926 paper that Spence_UK links to provides an answer to those questions.

- - - - - - - - - -

AMac78 -- Aug 3, 2010 at 3:59 PM

Re: scientist (Aug 3 14:49) (Further comments to follow #237614),

> Mann seems to have had an overly simple conception of the Tiljander series (bigger number=thicker layer=warmer year).

Not really. The authors accessed the four data series, considered the cautions in Tiljander03 and decided they weren’t crucial, then performed their screening and validation tests as described in the SI. The resulting r values are presumably as described –

0.3066 Darksum
0.2714 Lightsum
0.2987 Thickness
0.1232 X-Ray Density

All are relationships between {proxy quantity} and {temperature} are positive. The issue is that this calibration is to the non-climate-related signals that progressively affected the lakebed sediments in the post-1720 period, according to Tiljander03.

> [Mann & coauthors] likely used theses series without enough understanding of their limitations.

Yes.

> So he… should have informed the readers than his chronology was dependant on a new geo-isolated set of proxies…

If Mann08′s authors were aware of their redefinition of the relationships of {proxy values} to {temperature}, then yes, they should have done so. But since they were likely unaware of the problem, how could they have done so?

> I would be interested if he thinks the varve series are “good”, “debatable”…, or completely unsuitable.

Prof. Mann has declined opportunities to address this issue.

> not impressed with the Tiljander work in terms of a decisive explanation of the sediments. She does not construct a temperature series.

Can you suggest how she was to construct a pre-1720 temperature series, for the portion of the varve record that was least affected by non-climate local signals?

> I don’t understand… the issue of LS, DS and XRD and wonder if they are all independant…

Good question. Lightsum is meant as a proxy of mineral content, Darksum as a proxy of organic content. XRD appears to be a digitization of X-Ray film’s grayscale density, once a core of defined thickness has been X-rayed. It’s unclear to me whether the Lake Korttajarvi record should be considered “one proxy” or “multiple proxies.” (But this is a second-order issue.) For more discussion (but no clear-cut answers) on physical data that can be extracted from lakebed cores, see the PDF linked at Jarmo’s comment, supra.

> I’m concerned about the issue of compaction. How is this addressed?

I don’t believe it has been discussed in the context of Mann08; this seems like a distinctly second-order consideration. It is considered in the sedimentology literature (no citation).

[snip]

- - - - - - - - - -

scientist -- Aug 3, 2010 at 4:27 PM

1. When I said he should have disclosed the concerns, I didn’t mean that he knew it and failed to do so.

2. I don’t understand your comment, first you say that he was well aware of the limitations (in SI) and then you say, not.

[snip]

- - - - - - - - - -

amac78 -- Aug 3, 2010 at 5:10 PM

Re: scientist (Aug 3 16:27),

> 1. When I said he should have disclosed the concerns, I didn’t mean that he knew it and failed to do so.

Agreed.

> 2. first you say that [Mann] was well aware of the limitations (in SI) and then you say, not.

Mann08′s authors were clearly aware of the issues, as they discussed them. They then went on to use the data. Either they knew they were bad (very unlikely, IMO), or they figured that the problems weren’t that severe after all (consistent with what they wrote in the SI). One flag they seem to have missed was that the calculated correlations for lightsum and XRD were inverted with respect to Tiljander03′s interpretations. If they’d noticed these two “upside-down” interpretations, they would have been obligated to describe them. Yes?

[snip]

- - - - - - - - - -

amac78 -- Aug 3, 2010 at 7:20 PM

Re: scientist (Aug 3 18:57),

[snip]

On the science, per MikeN, is “one or the other” wrong? Stoat blogger William Connolley didn’t think so. On this point, he said, “[Kaufman] is right and Mann is right.” (inline remark at Comment #16).


Incidentally, Kaufman09′s authors used a splicing procedure so that they could avoid attempting to calibrate the Tiljander proxies to the post-1850 instrumental record. That is why they could decide to use the Tiljander03-concordant orientation for the proxies on an a priori basis.

- - - - - - - - - -

amac78 -- Aug 4, 2010 at 8:52 AM

On reflection, the exchanges with ‘scientist’ yesterday on this thread serve to illustrate the bind that the authors of Mann08 have placed themselves in.


It’s highly likely that Tiljander03′s overall interpretations of the Lake Korttajarvi data series as climate proxies pre-1720 are generally correct. [Note: since writing this comment, 'scientist's' observations have made me doubt the value of any of the four data series at issue as temperature proxies in the pre-1720 period -- AMac]

Tiljander and her advisor and co-authors are the sole ‘local’ authorities; their views are consistent with other paleolimnologists’. E.g. McIntyre’s cite of Atte Korhola, supra (#237675), Antti Ojala’s citations of Tiljander and Saarinen in his 2001 dissertation (linked by Jarmo #237548), Tomi Luoto’s repeated citations of Tiljander03 in his 2010 dissertation (linked by Jarkko #237696), Ianl8888′s remarks supra.

There are no qualified experts that have expressed contrary views, to my knowledge.

The first part of Mann08′s authors’ self-inflicted bind

If Tiljander and coworkers are correct: Mann08′s authors are wrong.

First, they are wrong because they ignored Tiljander03′s (and Tiljander05′s) warning that the Lake Korttajarvi data series would not be calibratable to a post-1720 climate record. Mann et al. thus calculated four nonsense correlations (Yule, 1926), one for each proxy, and proceeded with their paleotemperature reconstructions.

Second, they are trivially wrong because in calculating their nonsense correlations, they failed to notice that they had turned two of Tiljander’s data series upside-down, Lightsum and XRD.

Third, Mann and co-authors have undertaken a substantial violation of scientific ethics (viz: “Cargo Cult Science”) by failing to acknowledge and correct their mistakes, in their February 2009 PNAS Response to McIntyre and McKitrick, and again (implicitly) when uploading their first re-do of SF8a on their PSU website in late 2009, and again (implicitly) when uploading their second re-do of SF8a on their PSU website in early November 2009.

In addition, it seems fair to suggest that other climate scientists who knew better, or should have known better, have worked to enable Mann08′s authors, helping them to dig in deeper rather than searching for a way out.

The second part of Mann08′s authors’ self-inflicted bind

If Tiljander and coworkers are wrong: Mann08′s authors are wrong.

“Tiljander and coworkers wrong” must mean two things.

First, that their warning about post-1720 non-climate contamination of the varve record did not render their data series unusable in the 19th and 20th centuries, and thus related to the instrumental temperature record only by nonsense correlations.

Second, that Tiljander03′s pre-1720 assignment of the relationship of data series to temperature was correct for Darksum, but flat-out wrong for Lightsum and flat-out wrong for X-Ray Density. “Flat-out wrong” in these two cases means “upside-down”: Tiljander03 asserts that rising proxy values correlate to cooling temperature, while Mann08 claims that rising proxy values correlate to warming temperatures.

To my knowledge, there is no reason to think that this is the pathway that Mann08′s authors took to accomplish their paleotemperature reconstructions. But this is the line of reasoning that [some of] Mann08′s enablers consistently refer to. It is implicit in the “I don’t know, and it doesn’t matter” defense that has been proffered in Gavin Schmidt’s post at Collide-a-scape and by Martin Vermeers and others at Arthur Smith’s “Michael Mann’s errors” post.

So let’s explore it further.

If this is correct (again, I don’t think it plausible), then Mann08′s authors have engaged in a substantial breach of scientific ethics (viz: “Cargo Cult Science”) by citing and quoting Tiljander03 as an authority on the Lake Korttajarvi data series, while failing to note that their paper was departing completely from the cited interpretations. This would be highly misleading to PNAS’ editors, to peer reviewers, and to readers.

Secondly, one would have to ask about this part of Mann08′s bind: “What are the odds?”

While apparently knowing nothing about Finnish varved lakebed sediments, Mann08′s authors defy the sole authority’s warning on contamination.

In constructing their proxies for paleotemperature, these authors then accept the sole authority’s orientation (but not correlation) on Darksum, reject their orientation (and thus correlation) on Lightsum, reject their orientation (and thus correlation) on XRD, and invent a new correlation on Thickness.

And this process yields a skillful and validated reconstruction! Heck: Gavin Schmidt says that these proxies even extend the validation of the essential no-treering EIV reconstruction from 350 years to 850 years!

Wow! Talk About Lucky!

- - - - - - - - - -

That's the end of Part 1 of this exchange. The remainder can be found as Part 2: Synopsis of some Tiljander-related arguments.

1 comment:

  1. Fun chatting with you. Had enough of the Neverending Audit. If you need me, check the strengthosphere or the Nutrisystem forums.

    In health, brotha...

    ReplyDelete