Part 1 is here. As with that post, I may clean up formatting and grammar here, without notice.
- - - - - - - - - -
Amac -- Aug 4, 2010 at 1:41 PM
It can be useful to visualize the Tiljander proxies, graphically. This JPEG file at BitBucket portrays 20-year averages for Darksum, Lightsum, and XRD (I don’t have Thickness to hand). [For a clearer view, see the charts posted in The Tiljander Data Series: Data and Graphs.]
[snip]
- - - - - - - - - -
scientist -- Aug 5, 2010 at 1:17 PM
Nice picture. Random thoughts (lot of words and not that much killer insight):
1. Looks like pre-1700s, that there is a lot more relative variability of XRD than in LS or RS. Wonder why?
Actually reloooking at it, I think this is from the axes and driven by the differences in recent excursion. If we cut off the 1700+ part and then standardized each, they might be comparable.
2. I admit that XRD seems the more tractable and physically familiar measurement to me than LS or DS. Kind of am biased to using it.
3. Wonder what is driving the differences in XRD. Is it purely a matter of composition? Higher mineral fraction? Type of mineral? We sure that compaction has no effect? I suspect that some physical insights into what is driving the XRD results would help us with our problem. And just kinda cool to know what is going on physically regardless. I know she does some mag susceptibility and ashing. But maybe quantitative analysis of a few select microtomed sections would help. not the whole core. but some sampling. Just wonder if there is some clue we get from that.
4. I wonder if some sort of approach to exclude supervarves makes sense. Just to clean things up. Not sure if that happens when a pebble hits the core or when a tree falls in the river coming into the lake or a beaver dams the stream or when a fish dies and falls on the lake bottom. And of course, culling outliers is dangerous. But I just wonder if that kinda screening helps get us a more revealing series.
5. Are there no significant biological sources of inorganic material? Just wondered. Some of the other studies talk about diatom shells and the like. I have a simple model in my mind of chemically inorganic metal oxides and silicates being the spring melt DS part of the varve, and then chemically organic goo (fish poop or whatever) being the summer stuff that is LS.
6. Looking at the XRD, looks like there is a big plunge after 1100 and then a different regime from then to about 1700, followed maybe by another from 1800-1950 and then another plunge. Even if you leave aside the post 1700 period, not sure how you explain the phase change from 0-1100 versus 1100-1700 in terms of the typical story of Scandanavian climate. Somethings driving that series. Almost feel like I’m staring at a stock chart (and I hate technical analysis). Really land usage by the nearby population seems more the likely driver? I wonder what is driving the fairly reasonable sized excursions at 20 year intervals as well. With a tree, would think all kinds of biological counfounders could drive it. Would think that the self-averaging in 20 year buckets would eliminate year to to year nuggets from significance. poor resolution of the instrument ought to drive more average of signal. I could see the climate itself having some sort of variability on that timescale, but then the century and millenial scale story doesn’t seem to make sense.
7. Given that Tilj doesn’t really have any kind of clear climate story on it’s own, midge-man’s comments about “agreeing with Tilj” seem a little off (maybe budd-buddyish?). Your comment on disagreement is more interesting…
- - - - - - - - -
AMac -- Aug 5, 2010 at 2:06 PM
Re: scientist (Aug 5 13:17),
3, XRD drivers: I don’t know. Interesting question, I agree.
4, supervarve exclusion: I thought of that too, eg Le Chauvanet’s (sp?) criterion. But I suspect it gets ugly fast, esp. with the time-series/red-noise issues tossed in. To me, intuitively, this actually gets to the overconfident and naive approach that the field takes in general, given that most data series are very noisy, and that some probably don’t have any climate (temperature) signal at all. The subject cries out for interdisciplinary work, where the tree-ring people sit at the feet of actual experts and learn, rather than preach.
5, LS is light-colored minerals (spring runoff) and DS is summertime organics (or perhaps dark-colored clays, I hope Ianl8888 clarifies).
7, Agree. My point in copy-and-pasting the cites stands–nobody is gainsaying Tiljander03. But the more I stare at her series, the less they look like temperature proxies, or even, more charitably, like temp-and-precip combo proxies. This is a pie in the face for Tiljander03′s interpretations. But before Mann08′s enablers break out the champagne, they should think it through. “Hooray, we didn’t use a genuine proxy upside-down, due to carelessness with modern contamination! Instead, we used a series that’s signal-free noise upside-down, due to carelessness with modern contamination! And it’s the unstated properties of this upside-down noise that extend the validation of our EIV reconstruction from 1500-1850 many additional centuries back!” Hmmm, not good.
- - - - - - - - -
scientist -- Aug 5, 2010 at 2:47 PM
[snip]
7. Yeah, pretty much. What was exciting about Mike’s paper were the new data series, and the potential for sensitivity tests sans tree-rings. But the dependence on Tiljander only (and the failure of CPS and needing EIV is a sign of this, I think) and then the physical problems with Tilj (which look worse than tree rings at this point, to me) make the whole thing pretty weak.
That isn’t to say that there isn’t something we couldn’t get out of varves in the future. There’s always hope with new things. But neither Mike’s aphysical math-hopper, nor the Finn’s handwaving and level of analysis are getting us there yet. I almost get the feeling, the poor grad students are faced with a really tricky problem and samples that don’t tell a story, and then just try to weave something out of it, to get done and get the union card. The sad thing is that I’m the type who actually thinks all the data collection, physical analysis and even failed correlations are intersting. And I totally respect publication of failed experiments (within reason, but there’s a way to do it that is additive to science). I know the Finns did some hard work and took a swing at a tough problem, but better almost just to publish data and say one doesn’t have a good physical interpretation.
- - - - - - - - -
AMac -- Aug 5, 2010 at 1:17 PM
At Gavin’s C-a-s thread:
#92 Jay Currie Says (August 5th, 2010 at 12:48 am):
…“what about Tiljander? And is it correct to say that if you remove both Tiljander and the dendro you are left with very little of the hockey stick and, worse, the shaft is rather less than straight?…#94 Gavin Says (August 5th, 2010 at 1:42 am):
#92 There are so many false premises and misunderstanding in your ‘logic’ that I don’t even know where to start. I’ll start off with by pointing out that I was just reading the papers concerned and reported what they said – there was nothing new to my comments at all. If other people had not read those papers, that is not my fault…- - - - - - - - -
AMac -- Aug 5, 2010 at 11:36 PM
Re: AMac (Aug 5 11:51),
Gavin responded to my Collide-a-scape comment #127 at #188. I put the two essays up side-by-side, Two Views of Tiljander. Also see Brian Eglinton’s earlier remarks of Aug. 5 at 10:40 PM.
- - - - - - - - - -
AMac -- Aug 6, 2010 at 12:06 AM
Re: Brian Eglinton (Aug 5 22:40),
I put my #127 and Gavin’s #188 side-by-side at my blog, Two Views of Tiljander.
I find that the six points I made to be simple and in a logical sequence.
1- Mann08′s methods require direct calibration of all proxies to the 1850-1995 temperature record.
2- Tiljander warned about post-1720 contamination in the Tiljander proxies.
3- Mann08 considered the warnings, then went ahead and used the proxies.
4- The 19th and 20th Century contamination was really bad, as a glance at a figure will show.
5- The correlations that Mann08 thought they found between 1850-1995 temperature and proxy signals were actually spurious correlations to contaminating non-climate signals.
6- The mistake itself isn’t such a big deal, it’s the refusal to fix the problem that’s the issue.
Gavin’s points are, um, lawyerly.
1- Varve records often contain climate signal, but they can be contaminated.
2- The Tiljander proxies were 4 out of 1209 proxies used. CPS correlates proxies to local temperature, fixing their orientation. EIV does not.
3- Mann was very aware of the potentially dubious nature of the modern portion of the Tiljander proxies. So he did sensitivity tests without them as Fig. S8. Fig S8 showed that neither CPS nor EIV is materially affected by including the Tiljander proxies.
3.1- Without Tiljander, CPS validates back to 400 and EIV validates back to 700.
4- Regarding Tiljander, Mann08 already showed all there is to show. There is nothing left to do.
5- All RC statements on Tiljander are correct. If you use neither Tiljander nor tree rings, reconstructions are valid back to 1500 for CPS and as far back as is stated in Mann09′s SI for EIV. All validated reconstructions agree that the 20th Century is anomalously warm.
At “The Main Hindrance to Dialog (and Detente)”, Lucia famously noted Gavin’s pronounced tendency to answer questions regarding Tiljander… but to not answer the questions that were actually being asked. Of my six points:
1- Ignored.
2- Somewhat addressed (cf. “contamination” & “potentially dubious nature”).
3- Tacitly agreed.
4- Ignored.
5- Ignored.
6- Ignored.
- - - - - - - - - -
AMac -- Aug 6, 2010 at 9:13 AM
Re: Pasteur (Aug 6 07:17),
My advice on the Tiljander issue is to work up from the basics.
Are the data series any good as temperature proxies? (Maybe.)
Do they have to be directly calibrated, the way Mann08 went about things? (Absolutely, yes.)
Are they hugely contaminated throughout the 19th and 20th Centuries? (Absolutely, yes.)
Does this contamination make it impossible to directly calibrate the Tiljander proxies to the 1850-1995 temperature record in any meaningful way (Absolutely, yes.)
So why all this ducking and weaving about using these proxies? They have to be excluded from all analyses, right? (Absolutely, yes…)
- - - - - - - - - -
AMac -- Aug 4, 2010 at 12:55 PM
scientist,
You see shades of gray when it comes to Mann08′s uses of the Tiljander proxies. Here’s a thought experiment that could lead you to express your ideas more clearly.
Suppose that you and I are working together on a 1,800 year paleotemperature reconstruction. My job is to find candidate proxies, yours is to calibrate them to the instrumental record, 1850-1995.
This morning, I say,
“scientist, I’ve found a reference to four promising 3,000-year-long lakebed sediment records! But there is a big potential problem: the geologists who obtained and characterized the drill cores say that each of the four records has large-scale contamination from 1720 to the present.
“I know the sources of the non-climate signals: farming and peat-cutting in the lake’s drainage area, local road-building, an episode of bridge reconstruction in the 1960s, and eutrophication (there’s a small city beside the lake).
“The geologists’ paper says that non-climate influences on the records have grown over the past 280 years, but they don’t offer a quantitative estimate of how the contaminating signals compare to the climate signals. 50% as large? 200%? 1,000%? Who knows?
“scientist, inclusion of these proxies will make the paper! We really need them!
“Briefly explain to me in simple, non-technical terms how you think you might calibrate these four sediment data series to the instrumental temperature record, 1850-1995.”
Indirect and super-duper complex approaches are out: we know from the SI that Mann08′s authors didn’t use them.
Sadly, then, my response has to be, “It can’t be done.” That ends the thought experiment for me.
Likewise, it’s over if the answer is “I’ll just ignore the recent history of non-climate contamination.” (I was already aware of that method of producing nonsense correlations.)
For us to believe that Mann08′s authors might have a logically tenable defense for their choices with the Tiljander proxies, there must be an answer to this thought experiment.
So, scientist: what’s your suggested approach?
- - - - - - - - - - -
Alan S. Blue -- Aug 4, 2010 at 3:38 PM
Re: amac78 (Aug 4 14:31),
The fundamental disconnect I see is: It seems as if there’s a deep-seated acceptance of the assumption that allows teleconnection. “It’s a valid proxy because our tests of correlation turn out well.” Which is nothing at all like the statement “It’s a valid proxy because we understand the factors involved through causality tests, I can provide a calibration and a calibration bias/error, and it tests well in an out-of-sample period.”
Mann avoids most of that by the reasonable argument “The original authors thought it was a proxy, I’m not an expert on their field, I added it to the pool of potentially useful proxies.”
Once you’ve followed the path “Well-correlated implies good proxy,” you’ve already made the assumption that’s fatal to the case. So the argument “That isn’t a good proxy because of reason X” is always trumped by the inherent assumption. “It was discovered by the proxy-finder, so it’s good.”
This can be seen in the way “correlation” and “calibration” are used near-interchangeably. And also, really, in the way Mann’s method relegates the vast majority of available proxies to insignificant weights. To the point that there hasn’t been a peep about the “lowest ten” proxies according to Mann in years. Given the paucity and skill of surface measurements even at the height of GHCN, the reasoning for discarding those proxies seems ever fainter as time passes.
- - - - - - - - - -
scientist -- Aug 4, 2010 at 4:41 PM
I’m having a hard time following your thought experiment, including the part where you say I would do various things I wouldn’t do (push for different standards for something because it makes the paper), and then want to know what I say a while longer.
I think you could,
1. treat it like all the rest of the proxies, just plug it into the hopper (I mean if your method has a danger of spurious matches, well… that’s what it has… that’s an issue of the methodology. And there are some safeguards… maybe not as high as you want, but again, this is a separate argument.
2. Do as (1), but note the physicality concerns in the paper, highlight physicality caveats.
3. Try to fix it somehow. I would probably avoid that as I don’t know how you fix and it becomes a rathole to go down (in work) if you end up having to do that for every of 1200 proxies. That said, attempts to “fix” like maybe using the pre-bridge years for validation only, might be a method of getting the most signal from noise. Or you could go and do a bunch of work to go to the site and read varveology and get a co-author who understands the issue (or at least consult one) to determine if the confounding is more a 10% issue or a 100% one.
4. Just not use it. I do think, in this case, it’s important to be fair though. Not to bias your final result because of what makes skeptics happy. (For instance, convert it around and say the Tiljander sediments really drove a high MWP result… and was excluded for the unquantified bridge extents. Do you think everyone on the skeptics side would be happy? Buehler?
The extent that you decide which to do, depends on how much you know. If you know the issue the way in your example (and not clear to me that they did), then you need to do something in bucket 2-4 based on some sort of judgment process. I have a hard time answering how to do that. Ideally, you would have everything formalized, and apply it equally to all the series. For instance, saying you intend to use every series of such and such characteristic and then just follow that as a rule. Or if you are going screen things based on physicality (from reading the papers) formalize that.
- - - - - - - - - -
SOI -- Aug 4, 2010 at 4:57 PM
Scientist,
Of course, Mann didn’t do any of the 4 options you state. He tried to do option 1 (or 2), but erred because he used the series upside down. There is nothing in his code that flips a proxy based on which way would pass testing in the calibration period. Nor did he in the paper, or at any time since, ever provide any reasoning to justify why Tiljander’s interpretation was wrong.
This is really quite simple -- Mann screwed up. Your attempts to portray this as shades of gray is quite unpersuasive.
- - - - - - - - - -
scientist -- Aug 4, 2010 at 5:33 PM
Maybe so.
- - - - - - - - - -
steven Mosher -- Aug 6, 2010 at 2:49 AM
[snip]
Absent a cogent argument to wave away the concerns of contamination, a cogent PHYSICAL argument, absent that you are NOT pressure testing Mann’s decision. A cogent pressure test would be this. We have a plausible physical reason to junk the proxy or truncate it. It’s at best questionable and at worst corrupt. If you decide to accept it, you are making a statistical decision with no estimatable probability of being right. Your CI becomes junk overwhelmed by the probability of being wrong about waving away concerns with no basis.
[snip]
- - - - - - - - - -
AMac -- Aug 5, 2010 at 12:56 AM
Re: scientist (Aug 4 16:41),
Hey, nice response to the thought experiment, scientist.
1. Yep, you can just take the data series and throw them into the proxyhopper. Mann08′s authors added a few cavils in the text about possible contamination of Tiljander, but besides that, this is what they did, it seems to me.
2. Highlighting physicality caveats, maybe that helps later if something goes wrong, but what else is it good for? OK, four points awarded for honesty, but minus two for not doing a simple follow-up. As soon as you graph the three series, it jumps out that there’s obviously something very fishy going on with each one, late 1700s on.
One alone or One & Two together are going to lead to a spurious correlation. So that’s what happened.
3. Try to fix it somehow. Well we know Mann08 didn’t. If you’re stuck with using a direct approach (cf. splicing), it can’t be done, as far as I can see.
4. Just not use it. Yep. What you go on to say about “making skeptics happy” is wrong. The need is for predefined selection criteria. The subject is not simple with respect to either practicality or statistics. People in other fields think a great deal about these topics, and develop and revise procedures. Climate scientists, not so much. Their ability to learn useful methodologies from other disciplines, not shown to be high, at this point.
- - - - - - - - - -
scientist -- Aug 7, 2010 at 3:01 AM
I agree the graph looks ugly, with that recent rise. That concerns more than the big spikes. Gotta think something is going on there, but not sure what. My hypotheses are land use or (lack of) consolidation. When you add the appearance of the huge ramp, and then the things in the text about the contamination, it’s a dubious proxy. I think in terms of the recent Zorita paper, that an expert would likely leave it out. I dunno…a lot of these time series look pretty ugly. Maybe they’re used to that, from that.
Still… I would not say absolutely contaminated. Even if you feel the concerns are high enough to warrant keeping the series out (and that would be my bet, right now), I would not overstate the case. I think given that we don’t know what the heck is going on with that series physically-chemically, we should not say things like it is absolutely contaminated. I would reserve “absolute” for where we have a higher understanding of the proxy, physically.
- - - - - - - - - -
scientist -- Aug 6, 2010 at 12:55 PM
It’s not clear to me that Mann or Gavin concede the contamination of the Tiljander sediments. Also, when you say “absolutely”, I would say “probably” contaminated. I mean, I haven’t seen chemical analysis to prove contamination. I don’t think we know what chemical substance is in the sediments to give the dramatic LS/DS/XRD modern changes. Not: I’m not arguing that it’s a good proxy. Even a 50% good proxy. Just that it’s an overstatement to say it is 2 plus 2 = 4 contaminated, “absolutely”. I personally don’t know what the heck is going on with those sediments in either old or modern times. I take the info from Tilj as a very serious concern… but then I’m also not blown away by her thoughtfullness or consistency in the paper. Like I still have a bunch of questions on densification (need to look at thichness values…could all this just be a lack of compression for recent varves?) There’s some kind of long term trend from 1700 to present in that data and it’s probably land usage, but I couldn’t say it’s definitely land usage in the way I could say that Lance Armstrong definitely used EPO in 1999 since he had 6 old samples dug out and tested and they came back positive.
I think that the bigger concern is that the Mannian methods seem to grab anything with a recent ramp in the 20th century and then that basically passes validation (CPS) or gives a good correlation and places into the recon (EIV).* I’m not enough of a statistician to explain what worries me…but it’s something that Zorita would call matching a single degree of freedom (two trends). I don’t really know if the r value handles this, but what I would want is some “wiggle matching”. Like if we detrended each series over the period, would the proxy match the gyrations of temperature? When I look at Cobb’s coral poster and the temperature and the coral, you can pretty clearly see that the coral is “wiggle matching”. Not perfectly, of course, but you can see more than just a match of the long period trend, see some wiggles get matched. I think that Mike does a lot of just matching long trends, so when you get mean on him for the Tiljander, well, he’s all like “that’s how I usually roll”.
The whole “failure” to admit doesn’t bug me as much as the rest of you, since I don’t think Mike concedes the logic/science as simply as the picture of Amac. What bugs ME is that he wrote that PNAS paper with all the Tiljander stuff (quotes from the paper and all) in the SI. That stuff should have been been in the main paper. Essentially what made the new paper was having more series. But a lot of those new series were questionable (as shown by the quotes). That should have been front and center. And then publish in a more serious, specialized, lower level journal. Not the PNAS ego trip.
[snip]
- - - - - - - - - -
AMac -- Aug 6, 2010 at 3:48 PM
Re: scientist (Aug 6 12:55),
> It’s not clear to me that Mann or Gavin concede the contamination of the Tiljander sediments.
No, they clearly don’t. In my opinion, they aren’t playing by the rules of Science, but by the rules of PostModernScience. The First Rule of PMS is, “there’s no such thing as PMS, we are all upstanding traditional Scientists, and I’m offended that you suggest otherwise!” The second rule is, all the methods of scientific inquiry can be used, as long as they lead to the desired result. If they don’t, all the methods of rhetoric and forensics can also be employed, by our side only.
As an example of PMS at work, we have constructions about proxy validity like “I don’t know, and it doesn’t matter.” Even a seemingly simple notion like “I don’t know” takes effort to pin down. “It doesn’t matter” hardly needs its own link at this point. Then there is the logic that goes “maybe X, maybe Y. If X then A; if Y then B, so we must go forward weighing both A and B at the same time” (e.g. this subthread). This PMS strategy abandons the reductionist approach; arguments become so complex and confusing that… uh, scientist, what were you saying, again?
> Also, when you say “absolutely”, I would say “probably” contaminated.
Either the proxies are “sufficiently uncontaminated to enable a calibration to the temperature signal to be performed, 1850-1995,” or they are “too contaminated to be calibratable.” Now look at the graphs again. Let’s pick Lightsum, the series that more or less means “accumulation of mineral.” You think the 1850-1995 signal has temperature information in it? Really? Really? OK, let’s go back a few centuries and ask Excel for the mean and standard deviation; forget red noise and all that–very simple first-order approximation.
Century, mean +- SD
15th C, 2.2 +- 0.6
16th C, 1.9 +- 0.7
17th C, 2.2 +- 0.7
18th C, 2.4 +- 0.9
19th C, 4.8 +- 2.0
20th C, 9.4 +- 8.0 (thru 1985)
[I have expanded this analysis and charted the results in The Tiljander Data Series: Data and Graphs -- AMac]
So you, Gavin, Mia, and I agree that from prior to the 15th Century through ~1720, Lightsum might — might! — contain a temperature signal.
Mia and I look at those 19th and 20th Century numbers and say, “Yikes! Those numbers are huge and variable! And they are huge and variable in a way that they just weren’t, in the 15th, 16th, 17th, even the 18th Centuries! Whatever temperature signal is in there, it’s overwhelmed by a deluge of contamination!”
Gavin looks at those numbers and says, “Hmmm, ‘possible anthropogenic contamination’ in more recent centuries, so ‘potentially useful, but also potentially dubious’. Let’s use ‘em and see what happens!”
And you’re going with Gavin on this, because, despite these ugly numbers, despite the graph that’s level for centuries before going hyperbolic in the 20th Century, despite the obvious impossibility of meaningful calibration 1850-1995, despite the chapter-and-verse of roadbuilding, peat cutting, farming, bridge reconstruction, and eutrophication — I used the word “absolutely.” Without a clear-cut chemical signature for this hypothesized, so-called “contamination.”
> but then I’m also not blown away by her thoughtfulness or consistency in the paper.
OK, you convinced me of that. Now I wonder whether Lightsum and the rest are any sort of temperature proxy at all, because maybe Tiljander et al were scraping the bottom of the interpretation barrel. Um, that’s not an endorsement of Gavin’s position! Having a reconstruction whose validity is improved by adding upside-down (with respect to Tiljander03) noise isn’t better than having that validity bulked up by reliance on upside-down (with respect to Tiljander03) signal…
Are you sure you don’t want to change your vote?
> I think that the bigger concern is that the Mannian methods seem to grab anything with a recent ramp in the 20th century and then that basically passes validation (CPS) or gives a good correlation and places into the recon (EIV).
Well that is something of a problem, and I can think of four or so examples offhand.
> what I would want is some “wiggle matching”.
Agree, especially now that we’ve seen some.
> The whole “failure” to admit doesn’t bug me as much as the rest of you, since I don’t think Mike concedes the logic/science as simply as the picture of Amac.
I understand that Prof. Mann doesn’t concede anything to my view. That explains his conduct, but it hardly justifies anybody else’s. If he decides that the moon is made of green cheese and Apollo 11 be damned, will PNAS’ editors send that paper out to peer review, and publish it? Will Gavin then declaim, “I don’t know, and it doesn’t matter”?
- - - - - - - - - -
AMac -- Aug 7, 2010 at 9:37 AM
Re: scientist (Aug 7 03:07),
> I agree the graph looks ugly… My hypotheses are landuse or (lack of) consolidation… it’s a dubious proxy… I would not say absolutely contaminated… I would not overstate the case… we don’t know what the heck is going on with that series physically-chemically, we should not say things like it is absolutely contaminated. I would reserve “absolute” for where we have a higher understanding of the proxy, physically.
I agree narrowly but not broadly with your general thrust here. You’re being nice, and it’s nice to be nice, but we also have to remember that guy Feynman, who was eccentrically charming but didn’t approach technical issues in terms of the importance of being agreeable with our scientific betters.
You’d have the basis for making a reasoned argument if Mann08′s authors and champions would offer some justification for the idea that Lightsum (to stick with that example) is a plausible temperature proxy. Then we could go back and forth, a “preponderance of the evidence” discussion. But Prof. Mann ducks the question, preferring to describe the fossil fuel lobby’s anti-Mann conspiracy; listen to Chris Mooney’s painfully fawning podcast interview if you’re feeling cruel (link at my blog). Co-authors, “Maintaining radio silence, sir!” When Gavin pronounces, how much prep work does he have to put into those legalistic turns of phrase and lawyerly grammatical ambiguities, do you think? Mann08's other defenders are flyweights on this issue (name one who isn’t?).
So please take “absolutely” as my shorthand for “overwhelming evidence at this point that post-1720 Lighsum is invalid as a proxy for temperature, with no plausible rationale on offer that post-1720 Lightsum might be an acceptable temperature proxy (although new theories and evidence might change this picture).”
Saying you don’t like “absolutely” doesn’t cut it. You have to (1) Dispute the multiple lines of reasoning that each show Lightsumm to be an unacceptable temperature proxy (er, you’ve added to the pileup!), or (2) Propose a hypothesis by which Lightsum is (or might be) an acceptable temp proxy. Remember, the big piece of gristle here is that for Mann08, “acceptable” means “directly calibratable to the 1850-1995 CRUTEM3v record.”
PMS adepts loathe Occams Razor, but I kinda like it. WWWilliamS? “Dubious”? Hardly–that’s a pleasant euphemism. “Must discard?” Yep.
> the thing got through peer review without anyone worrying about Tiljander… I think the review was fine.
I’d be interested in seeing the peer reviews and editors’ comments on Mann08. I have a pretty clear suspicion as to what they looked like, and it doesn’t map to yours. This was a shallow, perfunctory review process, “Yeah, looks good to me,
- - - - - - - - - -
That will wrap up this synopsis of part of the very long comment thread at "The No-Dendro Illusion."
No comments:
Post a Comment