Kaufman’s grudging acknowledgement (see their draft Corrigendum) that they used the Tiljander proxies upside down has not convinced the Team that the identical orientation of the Tiljander proxies in Mann et al 2008 was also upside down.
There has been an active new round of debate in the blogs, with William Connolley endorsing Upside Down Mann. It seems that we are facing not simply an Upside Down Mann, but an Upside Down Team.
Roger Pielke Jr had opined hopefully that this concession would finally settle at least one small point in paleoclimate. Pielke said that “it looks like this dispute will in fact be resolved unequivocally through the peer-reviewed literature, which for all of its faults, is the media of record for scientific claims and counterclaims”. Pielke was obviously aware of the role of blogs (both Climate Audit and in Finland) in this dispute and was here focusing more on the fact that Kaufman was admitting the upside down use in a formal venue, rather than the role of the journals in extracting the admission from Kaufman. This point was misconstrued by Ben Hale here who interpreted Roger’s post as evidence that the Kaufman error had been detected and resolved by journal peer review and due diligence, when that’s not what happened at all. (I posted a comment at Hale’s to this effect.)
The debate over the relationship of blogs and Peer Reviewed Literature is one that Andy Revkin raised recently as well. And it’s one that’s playing out in a very interesting way in the recent Briffa commentary on Yamal, which, regardless of how this issue plays out, is, in my opinion, a more substantive and interesting bit of work of than any of his recent articles in the Peer Reviewed Literature, because it is accompanied on time by DATA, is technical and makes no effort at faux “originality”. (Obviously much more on this in the next week or two.)
Had matters been left with blogs debating the relative contributions of blogs and Peer Reviewed Literature in the settling of this small point, that would seem like the logical denouement of this sorry little episode.
But this under-estimated the propensity of the Team to engage in prolonged trench warfare on the most elementary and seemingly unwinnable points. A Pielke commenter argued that, under Mann’s methods, the “data can’t be upside-down”, adding that neither Pielke nor I were qualified to engage in such a debate anyway:
“Multivariate regression methods are insensitive to the sign of predictors”. Mann et al seem to be saying their methods are invariant to the data’s orientation – perhaps to linear translation? – anyway it means the data can’t be upside-down. Now if the mathematical interpretation placed on the data by their methods conflict with physical information from other sources, it does raise questions. Presumably the next sentences in Mann et al’s reply l refer to this, I can’t see anything in Kaufman et al that illuminates this and it would seem to require detailed knowledge of the field to judge the importance of these issues. And as you said, we (you, me, Steve McIntyre,…) are not professionally qualified to engage in the substance of such a debate.
Despite the worries of Pielke’s reader as to whether Pielke or I were sufficiently qualified to determine whether a series is going up or going down, I can assure the reader that I have enough experience in the stock market (both painful and otherwise) that I know the difference between whether things are going up or down, and I believe that even readers uninitiated into the mysteries of RegEM are capable of understanding the difference.
This comment at Pielke’s was praised by William Connolley, the Team’s representative at Wikipedia, in a recent post decrying Pielke. Connolley then vigorously supported Upside Down Mann in comments at both Pielke Jr and Ben Hale, where the matter has been discussed relatively briskly.
Mann’s use of the Tiljander sediments was originally discussed here on Oct 2, 2008, a post which discussed the multiple problems with Mann’s use of these proxies, summarized as follows:
In Mann et al 2008, there is a truly remarkable example of opportunistic after-the-fact sign selection, which, in addition, beautifully illustrates the concept of spurious regression, a concept that seems to baffle signal mining paleoclimatologists.
The issue with Mann’s use of the Tiljander proxies isn’t just that he used them upside down (which he did). The problem is worse than that. The Tiljander sediments are the combination of two unrelated processes: a presumably climatically driven process in which narrow sediments are interpreted by the authors as “warm” and thick sediments as “cold” and a nonclimatic process in which sediments are produced by ditches, bridges and farming.
Although the following point is not well understood by climate scientists (including, apparently, Connolley and Mann), a “reconstruction” at the end of the day is a linear combination of the proxies. While Peer Reviewed Literature does not require climate scientists to report these weights, in our dissections of reconstruction methodologies, this is the sort of thing that we keep track of.
Leaving CPS aside for a moment, Mann and his defenders say that, in a multiple regression setting, it doesn’t “matter” what the orientation of the series is going in to the meatgrinder. However, this ignores the relevant issue of what the orientation of the series is coming out of the meatgrinder. Connolley and Mann and others seem to assume that the meatgrinder can’t get it wrong. But this is not the case either in principle or in the particular case of the Tiljander sediments.
By examining the code to keep track of weights (as Jean S, UC and myself have done), it is possible to track the orientation of the Tiljander sediments into the final reconstruction and see whether the contribution of the Tiljander sediments to the reconstruction is inverse to the interpretation of the original authors or not. It is definitely and incontrovertibly upside down.
The reason why it is upside down is the spurious correlation between the nonclimatic sediments from bridges and farming and temperature, which confuses the Mannian meatgrinder algorithm. While I confirmed my understanding of the sediment interpretation by email with Tiljander, this is also clearly reported in the original article (See my original note on this).
As I’ve recounted at CA (most recently here), we reported these and other problems in our PNAS Comment. These comments are limited to 250 words, 5 references and no figures. This is far less detail than available in any blog post. Connolley’s most recent argument is that our PNAS Comment was insufficiently clear.
Perhaps Connolley is gradually realizing that the problem is not just the upside down proxy, but a package of issues including modern contamination and spurious regression. Needless to say, Connolley doesn’t blame Mann for making the errors, but blames me for not expressing these points clearly enough that even a climate scientist could understand them.
If he meant what you said, he could and should have said so.
Quite frankly, I’m baffled at what else I could have said. The issues seem very elementary to me and I don’t understand why they seem to difficult for climate scientists. Let me try one more analogy. As noted above, the Tiljander sediments are in a sense a “compound” of a climatic and nonclimatic process.
Consider a series defined as the difference between the incidence of the name Gavin (which increases strongly in the 20th century) and the Central England temperature (scaled to keep on the same page) and feed this into a Mannian meatgrinder. The Mannian algorithm will detect a strong correlation between the compound series and world temperature during the 20th century. In the “reconstruction” period when the Gavin effect wears off and is at a very low level, the compound series is the inverted temperature (upside down). The spurious regression results in the series being upside down in the reconstruction period.
As I said above, it’s not just that the series is used upside down; there’s a combination of problems, ones that, in my opinion, were fully described in posts on the topic (posts that are easily located merely by googling “Upside Down Mann” and following the links.)
For Connolley’s benefit, here is the Oct 2, 2008 post reprinted in its entirety (also see here for a most recent review). The issues are not very complicated.
It’s Saturday Night Live (Oct 2, 2008)
When you’re trading in puts and calls (or derivatives), it’s important to know the sign of the relationship between the value of the derivative and the market. Short positions will go up in value as the market goes down. And, unfortunately, you don’t get to decide afterwards whether you wanted to be short or long. Proxies in climate can, in a sense, either be “short” or “long”, in the sense that the values of some proxies (e.g. coral dO18) are said to go down with higher temperatures, while the values of other proxies (e.g. ice core dO18) are said to go up with higher temperatures.
One feels that it is not asking too much of paleoclimatologists to know the expected sign of a proxy derivative. Traders would like to decide on the sign of a proxy derivative after the fact, by taking a correlation to market performance, but this luxury is denied to them, as it should be denied to climate scientists.
In Mann et al 2008, there is a truly remarkable example of opportunistic after-the-fact sign selection, which, in addition, beautifully illustrates the concept of spurious regression, a concept that seems to baffle signal mining paleoclimatologists. For this example, we turn to the highly HS-shaped Finnish sediment series of Tiljander et al 2003.
Tiljander et al cored varved sediments from Lake Korttajarvi, Finland, going back through most of the Holocene. In Tiljander et al 2003, they distinguished the amount of mineral and organic matter in each varve. The basis for using mineral and organic matter as climate proxies is set out as follows:
The amounts of inorganic and organic matter, form the basis of the climate interpretations. Periods rich in organic matter indicate favourable climate conditions, when less snow accumulates in winter by diminished precipitation and/or increased thawing, causing weaker spring flow and formation of a thin mineral layer. In addition, a long growing season thickens the organic matter. More severe climate conditions occur with higher winter precipitation, a longer cold period and rapid melting at spring, shown as thicker mineral matter within a varve.
The caption to their Figure 5 reports the following link between X-ray density and their climate mechanism:
High X-ray density corresponds to high amount of mineral matter (light grey value tints in X-ray film) and low X-ray density corresponds to dark grey values caused by a higher proportion of organic matter.
Putting the two paragraphs together: warmer climate favors more organic material and thus a low X-ray density. In order to show warm values at the top of a graph, you need to invert the plot (i.e. you have to pay attention to the sign of your climate derivative.)
In the figure below, on the left, I show an excerpt from their Figure 5 which they show vertically (only the X-ray density is shown here – consult the original paper for the other plots.) The left portion of their Figure 5 shows an organic-rich period in the MWP, about which they say:
An organic rich period from AD 980 to 1250 in the Lake Korttajarvi record is chronologically comparable with the well-known ‘Medieval Warm Period’ (e.g. Lamb 1965; Grove & Switsur 1994; Broecker 2001). The sediment structure changes, less mineral material accumulates on the lake bottom than at any other time in the 3000 years sequence analysed and the sediment is quite organic rich (LOI ~20%). Thus, the winter snow cover must have been negligible, if it existed at all, and spring floods must have been of considerably lower magnitude than during the instrumental period (since AD 1881). According to the scenarios presented by Solantie & Drebs (2001), a 2°C increase in winter temperature would decrease the amount of snow in southern Finland significantly. Under such conditions, winter snow accumulation and intense spring floods would be rare events….
The Lake Korttajarvi record also indicates a climatically more severe period in the 17th century. Two periods, AD 1580–1630 and AD 1650–1710, are marked by an increase in both sedimentation (varve thickness) and mineral matter accumulation (relative Xray density). Also, magnetic susceptibility values are high between AD 1650 and 1710, indicating increasing mineral matter input into the lake.
They cite literature, including Hulden 2001, showing mild conditions in Finland in the MWP.
On the right, I’ve plotted the corresponding data so that “warm” grey values are on the top. I’ve also highlighted the (MWP) period identified as having elevated values of organic matter. If you squint, you can satisfy yourself that the left-hand and right-hand panels are showing the same data.
 |
 |
Fig. 1. Left from Tiljander et al 2003 Figure 5; right – plot of X-ray density (inverted).
Plotted according to the climatic interpretation offered by Tjilander et al, the modern warm period shows as colder than the Little Ice Age, something which makes no sense if this data is to be used as a climate proxy. Tiljander et al provide a plausible interpretation of the “divergence” of the proxy from its climatic interpretation as a result of agricultural and construction disturbance to sediment patterns, actually tying several especially thick varves to ditch and bridge construction:
This recent increase in thickness is due to the clay-rich varves caused by intensive cultivation in the late 20th century. …
There are two exceptionally thick clay-silt layers caused by man. The thick layer of AD 1930 resulted from peat ditching and forest clearance (information from a local farmer in 1999) and the thick layer of AD 1967 originated due to the rebuilding of the bridge in the vicinity of the lake’s southern corner (information from the Finnish Road Administration).
Now let’s see what Mann et al did with this data. All of the 20th century values of varve thickness, X-ray density etc go up like crazy with the agricultural and construction activities as shown below for 2 of the 4 series (the other two are similar). Instead of using the climatic interpretation of the data described by Tiljander et al, Mann correlates the increases in varve thickness and changes in density and color, originating from local construction and farming, to world climate.
 |
 |
Figure 2. Two of 4 versions used in Mann et al 2008
By flipping the data opposite to the interpretation of Tiljander et al, Mann shows the Little Ice Age in Finland as being warmer than the MWP, 100% opposite to the interpretation of the authors and the paleoclimate evidence. The flipping is done because the increase in varve thickness due to construction and agricultural activities is interpreted by Mann et al as a “nonlocal statistical relationship” or “teleconnection” to world climate. Mann:
the EIV approach, which makes use of nonlocal statistical relationships, allowing temperature changes over distant regions to be effectively represented through their covariance with climatic changes recorded by the network
A more convincing example of spurious regression in “peer reviewed” literature will be hard to find. After reading through this, I keep expecting someone to say:
Live from New York, it’s Saturday Night.
H/t to Howard Wiseman: http://www.youtube.com/watch?v=M1owcncKCHg
Reference:
TILJANDER, MIA, M. SAARNISTO, AEK OJALA, and T. SAARINEN. 2003. A 3000-year palaeoenvironmental record from annually laminated sediment of Lake Korttajarvi, central Finland. Boreas 32, no. 4: 566-577.
Climate Audit 
377 Comments
Keep on Truckin’, Steve. The mathematical truth in your observations will eventually win out. Metaphorically, the truth is like a tide that has its own enormous quiet power and will eventually wash the beach clear of things that do not naturally belong there. So far, the arguments posed by your detractors at RC seem little more than sand castles.
When you’ve made an error interpreting a message, you shoot the messenger. Logical.
This is absolutely childish. More on your (our) side than on theirs. There is one, tiny, simple issue that is at debate here. We don’t need a dissertation. All we need to do is focus on that one issue and stop calling names and whining and all this other crap.
You have some value. You think it is a proxy for temperature. It may require some kind of transform to convert from its native units to temperature. That transform /may/ include an inversion. This is all basic stuff.
From what I understand, Mann’s “transformations” include linear scaling and /possibly/ inversion. He is treating the values as temperature proxies and using some normalized method to determine its proper transformation. Fine. That all makes perfectly good sense. In his case, then, there is no such thing as “upside down” data.
So the only question, then, is: does the transformation his system infers for this particular proxy make sense from a /physical/ point of view? If so, great, he’s fine. If not, then this is a valid critique of his /methods/ and it is something he should look into.
I don’t know why we can’t just talk about this stuff like adults instead of flinging food across then lunchroom tables at each other.
–t
Re: Tim G (#3),
Ironically, you complain about name-calling, but you just compared us to kindergarten kids. How do you now expect me to respond?
.
You have misunderstood the nature of the problem. It’s not as simple as you seem to think.
Re: Tim G (#3),
The opposition argues that black is white and upside down is rightside up. What to do when facing such irrationality? If you call it what it is, they take offence and get angry and turn to insult. They’re the first to do it, every single time.
Connelly sure has a hard time admitting he is wrong. Especially to SM and RP jr
At CM he appears to admit Mann was wrong to use the upside down proxies:
But then he later seems to excuse Mann’s “bizarre” comment in response to Steve:
So he’s saying yes, the proxies are useless, but then defends Mann and himself by saying it’s okay to obscure the real issue with the “sign doesn’t matter” answer Mann gave.
Please pay special attention to AMac’s explication of the problem at Ben Hale’s blog in response to Connolley’s response to Bender. Link through Ben Hale ‘here’ in the second paragraph of the lead post.
===========================
Its clear why intuitively it is incorrect to use the proxy the way he did. Has somebody run the analysis with the proxy in both directions (as intended by authors and with the Mann inversion) and shown the results?…or is the method shrouded in mystery and so no replication available yet (?) For that matter, I guess one could run the method with all the proxies ‘upside down’…presumably that would change the end result…wouldn’t it? (or do they argue it would not?)
Since analogies seem to be causing difficulties, let’s try this one. You are overweight, and go on a treatment based on increasing doses of a drug with time. The drug doesn’t really work, so for some months there is no change in your weight. Then, your scale starts malfunctioning, indicating that your weight is going down as the dose of the drug goes up. You feed that into the RegM algorithm and get a strong “signal”. Drug approved for market?
It appears to me that climate scientists are using ordinary regression in their analyses. This approach can give spurious results including negative values (or weights). I have seen this happen where the dependent variables are highly correlated. A proper approach in such cases is ridge regression.
Steve can correct me if I am wrong here.
Steve: No. Different issue and problem entirely.
Re: Douglas Hoyt (#8), Doug you might be right if that was the cause of the problem, but the key part of the data that gives the hockey stick is the years in which farm (etc) sediment over-rides the climate signal. You are fitting a disturbance signal and correlating it with climate (spurious regr that won’t be fixed by ridge regression) but in the earlier part of the record even this spurious signal is missing causing you to fix the sign such that the earlier record gets a sign opposite to the physical interpretation.
Re: Craig Loehle (#10),
Craig, my experience with involves the analysis of a Willson radiometer. You can work out the heat flow terms from basic physics. The net result is a major term with several small correction terms. The correction terms in all cases have a negative sign and the negative signs occur because heat flows from warm regions to cool regions.
Now one can also ignore the physics and solve the problem using multiple regression to calculate the correction terms. However, if you do this, some of the correction terms turn out to have positive coefficients, which the physicists looking at the regression “solution” will declare physically impossible. In this case ordinary multiple regression gives spurious and unphysical results. It occurs because there is a high correlation between the correction terms. Hence, the need for caution and ridge regression.
It just seems to me that dendroclimatology could be making this mistake.
Re: Douglas Hoyt (#14), Thank God, someone is finally talking real physics instead of all this statistical mumbo-jumbo. It has always assumed me that we have these statistically obtained “signal” and “noise” without the slightest idea of (or concern about) the physics of the problem.
Re: C Baxter (#71),
Respectfully, the effects Hoyt mentions are a distraction from the Connolley thread, which is purely about algorithms, opportunism, spurious regression, and the failure of peer review. Ask yourself this: why does Connolley leap to the defense of Mann before investigating the details of the case? Don’t bother answering or you’ll be snipped.
Re: Craig Loehle (#10),
Exactly. Read AMac’s note at CM. It is pretty bad luck for Mann that the contamination during the modern era happened to reverse the apparent response to temperature. “Bizarre” is a good word for something so unlikely. OTOH, the problem WAS WELL KNOWN in the published literature. That’s why McIntyre didn’t find it “bizarre” at all.
In the end, Steve, what does it say about the methodology of someone who cannot admit to an honest mistake?
Set the flamingos FREEEE! http://www.youtube.com/watch?v=qJUFTm6cJXM&feature=player_embedded
Suppose these proxies were in fact perfectly correlated with temperature. How can anyone possibly justify applying negative weights to some but not others in a reconstruction? GISS and Hadley may adjust and homogenize individual temperature station data but they certainly don’t invert any of them. Well, Steig did, but that’s a different problem.
A graph of the Gavin name would clear things up.
#3: Well, if you read the post, you’d see that all of this has been clear for quite some time now. Steve and Jean S have verified that the correct physical interpretation is not being used in the output. There is nothing more to say, this is demonstrably true, but the team just avoids the issue or somehow doesn’t get it.
Am I incorrect in assuming that these two statements from WC’s blog boil down the problem into very nice, concise terms?
Mann didn’t flip the signs. His software package did. His software package flips them automatically when needed, so it is therefore insensitive (and robust with respect) to signs.
Re: Jason Lewis (#16),
What is this, an echo chamber? Who do you think you’re talking to? The series’ sense got inverted in an analysis done by Mann. And you are blaming who? The gremlins?
Re: Jason Lewis (#16), Jason, I don’t think even this detail is correct. The program assumes warm is up.
Bender, I should have used a smiley:) This seems to be the defense being used by Mann, Connolley, etc.. here and at Pielke Jr’s blog.
Re: Jason Lewis (#19),
Youi need to do soemthing to indicate sarcasm, because there are people trying to argue exactly as you just spoofed.
Re: Jason Lewis (#19),
They seems to be employing a variant of the well known “Hanlon’s Razor” defense.
“It is pretty bad luck for Mann that the contamination during the modern era happened to reverse the apparent response to temperature.”
You’re far too kind Bender. It’s there plain as day in Tiljander’s paper. Mann was aware of it and mentioned it in his SI. I assume it’s his business to understand a) his algorithm and b) his proxies. Why use it at all? Why use a nine sigma outlier and correlate it to temperature when you know it is due to non-climatic influences? When you (should) know that the climate part of the signal will be inverted incorrectly? I know Steve doesn’t like to speculate about motives…
Re: Greg (#24),
Please tell giano and bent-out-of-shape that I’m “far too kind”. They will tell you I’m far too harsh.
What if all the “proxies for temperature” showed a downward trend in the 20th century and they were all inverted by Mann’s algorithm to give a HS result. Would THAT be OK with the team?
Eric, there’s more than one problem.
1) The use of both orientations in different portions.
2) There’s the modern contamination.
3) There’s the assumption that reorientation makes sense. “Whichever orientation is a better correlation is the correct one.”
1 & 2 are flat-out errors. #1, in particular, is mind-boggling.
On #2: You couldn’t calibrate the proxy we call a “standard bulb thermometer” if we walk up and break the bulb during the experiment – the numbers are just obviously meaningless. Or, if not meaningless, require a much more exquisite rationale with copious backup. Something better than “Well, the height-of-liquid measurements recorded still show a positive correlation with the official temperatures recorded a couple hundred kilometers away… if you invert them.” You haven’t found a credible “Temperature Proxy”, you’ve found a freak event that happens to line up. Which actually isn’t surprising given the inputs.
But the third one is more subtle. I can imagine scenarios where a hypothetical perfect thermometer located dead center in a grid cell and a proxy (Here: measurement of the height of a fluid in a tube) placed at an extreme-but-same-gridcell location might have a far better inverse correlation over a short period. Jet stream fluctuations can overwhelm month-to-month cooling in weather at least. Especially if you’re normalizing all of your inputs. This essentially gets into arguing over teleconnection.
But ignoring #1 and #2 is odd.
Let’s parse the word “used” in “used the proxy upside down”. In the sense of “input to the algorithm” the proxy was not used upside down–it was used as is for input. Thus it is “bizarre” to claim it was (and would imply manipulation of the data). SM means “used” in the sense of AFTER the algorithm computations the sign of the proxy is opposite to the known relationship. If someone feeds the beef into the grinder and only looks at the reconstruction, it may not seem obvious that anything is amiss. Likewise, someone who believes in teleconnections can imagine that a particular lake can always get colder when the world gets warmer. But even for these two (demonstrably false) viewpoints, it is still wrong to 1)calibrate the proxy to the part of the data ruled by a disturbance signal and call that a temperature signal, 2) flip some of the varve proxies but not others and 3) flip signs between different historical parts of the reconstruction for that proxy. Is this hard?
Re: Craig Loehle (#29),
Yes and if you feed it into the packaged statistical algorithm and get negative signs, when they should be positive, then not only are the negative weights in error, but all the positive weights are also wildly off.
The source of the problem could very well be a high correlation between the input series which will lead to unphysical results. If they ignore the cross-correlations, I think all the proxy reconstructions are wrong just from this one error or omission.
Re: Douglas Hoyt (#30),
Doug, as I mentioned in an inline comment above, what is happening here has nothing to do with the multivariate issue that you raise (which is an interesting one in other contexts.) Something far simpler is going on here. The algorithm is obtaining a positive regression between nonclimatic sediment from ditches and bridges and world temperature and interpreting it as a “significant” relationship. It’s every it as spurious as the correlation with the name Gavin. Nothing to do with the issue that is bothering you. OK?
Re: Steve McIntyre (#35),
Steve, I am saying even if all the proxies were perfect, then the problem of cross-correlations can give rise to negative coefficients (or weights) in the outputs. All the positive weights will also be incorrect.
Re: Craig Loehle (#29),
McIntyre’s interpretation will seem “bizarre” to anyone who does not understand why the proxy is perverse and what this perversion does when “used” in Mann’s algorithm. McIntyre did his homework. He understood what was going on and why. He reported it. He shoots. He SCORES!
Re: Craig Loehle (#29),
Well, there are some ice-core isotope proxies which might benefit from such a procedure since the sign depends on where (or what direction) the water producing the ice comes from. IF there is a high frequency signal which can be used to determine which direction the low frequency temperature signal should be oriented, then a cut and flip procedure might be of value. But it’s a very big IF. And it would require massive documentation before being used and this hasn’t happened, AFAIK.
As a reader, I have no idea of what Mann et al (2008)’s paleoclimate reconstructon looks like, if the four wrongly-used upside-down Lake Korttajarvi varve proxies are removed from their CPS analysis.
Their Figure S8a is supposed to address that point. It does not.
How will the authors address this apparent shortcoming in their published work?
Re: AMac (#32), Ha ha ha ha etc. You are kidding, right?
Re: AMac (#32),
They will “move on”. Hey, …
Well, even I understand the issues, so I am virtually certain that the climate scientists do, too. Their problem is admitting mistakes. Maybe they can get away with the denial until retirement time 🙂
Re: jae (#33),
The problem is, of course, Copenhagen in December. More than ever there can be no admission that they might have got something wrong – bad timing.
Oh, sign, sign everywhere a sign
Blocking up the scenery, breaking my mind
Do this, don’t do that, can’t you read the sign
Re: Erasmus de Frigid (#39), The year 1972, Five Man Electrical Band, Signs
Got to have a Membership Card to get inside!!!!
I just left a comment at Connolley’s blog about this without realizing you had done a whole post. I don’t recall seeing the post EIV weighting for tiljander, has someone already shown the weightings?
Re: jeff id (#40),
yes, UC put a lot of effort (and time) in order to get the terrible EIV code to run. He finally succeeded, and we demonstrated (graphically) the weights (for Tiljander and other series) in one of the weirdest steps we tested (see here, especially Update). The post didn’t catch almost any attention then, basicly me and UC (and Mark T) having a private chat. Later although I have linked to the graphs a few times, it seems to me that nobody is interested in that. I do not understand why. IMO, the behavior of Mann’s EIV reconstructions in terms of physical interpretation is even weirder than that of Mann’s CPS reconstructions.
Re: Jean S (#65),
Well, the post was very technical and I expect I was like many others and couldn’t really make a contribution, so I didn’t say anything.
Re: Jean S (#65),
Sorry that I didn’t cover this more attentively. Looks like we’ll get another chance 🙂
Re: Jean S (#65),
Thanks Jean,
I was very aware of that particular post actually, although no comments. I’ve probably read it a few dozen times.
There was a lot of unsaid detail which left me wondering about some aspects, the post was short so I read every word trying to get more from them than was there. haha. At the time I eventually got lost single stepping by eye through matlab code for several hours. I’ll leave a comment over on that thread.
And when they were up they were up
And when they were down they were down
And when they were only half way up they were neither up nor down
Simple see, our Gav understands this.
Quick question:
If the data in question is flipped around the right way, what happens?
(Please excuse me if this has been answered earlier)
Re: hunter (#42),
If one had the data and the code one could check this.
Re: bender (#43), Yes, what no one seems to get is that Mann’s paper should be rejected OUTRIGHT because no reviewer can make the most RUDIMENTARY check for accuracy: Does the data used and code used ACTUALLY produce the graphs that are in the article. Well does it?
As a reviewer if I can’t re run the code (AS USED) on the data ( as USED) then I don’t know if the author has
supplied the right graphics. I simply can’t know this most simple fact. I’m left with having to trust
that the author included the right graphic. So as a reviewer I am left with these options:
1. Trust the author has not made some accounting error ( old version of graph, wrong version, etc )
2. Try to EMULATE his method
3. Ask him for Code and data.
4. Reject the paper.
When I look at a post or comment or paper I look at it like a reviewer. Do I have the data as used and code as used?
Nope. What next? ask for the data and code, if the answer is no, then reject it.
You want to watch the team wave their arms so hard they almost fly? Keep asking for data and code.
Re: steven mosher (#59),
I used to keep my code and data to myself, and defended my choice on the same basis: “I developed it, it’s mine to monopolize”. Now I see the value in putting it out there in the bright light of the sun. But there is a disturbing trend in the universities toward stronger IP protection, and this could be a major challenge for the openness movement. University presidents have an important leadership role to play in ensuring professors behave as good corporate citizens: protecting IP where necessary but otherwise ensuring transparency where public policy is concerned. Someone should write the East Anglia president. CRU as we know it needs to be shut down and re-tooled.
Re: bender (#59),
Bender, that is an interesting point, but in this field IMHO, if IP is used as a defense it is a red herring. My alma mater is all to keen to protect IP, but only where commercial exploitation or further research funding are at stake. Paleoclimate risks neither.
Re: bender (#59)
Nothing wrong with proprietary research, and it can be lucrative or impactful or both. Problem is that many seem to deliberately confuse research with Science. The Scientific Method is fundamentally about public verifiability. Journal based “peer review” should be a means to that verifiability. If it’s been corrupted into some sort of private academic exercise, it’s become a perversion of true Science — and is serving some other end entirely.
Re: Kasmir (#63), I never suggested propietary was bad. I suggested it could be used as a cloaking device by the unscrupulous.
Re: bender (#64), “… a cloaking device …”
.
There we have it: Klingon science.
Re: bender (#59), It is possible to get around IP protection even if it did apply by supplying the code to reviewers with a confidentiality agreement. So even IF paleo code were worth something, it should still be possible to review it.
Re: Craig Loehle (#67),
Do you honestly think, if you were to oblige reviewers to audit, that anyone would ever accept to review another paper again? That’s the point Ben Hale isn’t getting. The hot-shot elites publishing in Nature think audit is beneath them. They’re so busy and important. It’s publish or perish. No time for audit. Gotta climb to the top.
Re: Craig Loehle (#67), ya climate science NDAs. SteveMc is essentially doing clean room implementations of the methods using public domain information
Re: bender (#60), yes in my current line of work we do see certain universities making coveting “their” IP more stringently. In one funny instance where a guy came out of Stanford he was very precise in his claim that he came up with his ideas “off campus.” You can well imagine that universities who want to establish centers of excellence ( lets say in climate research) would want to hold the “IP” developed by those centers very tightly, as a guarantee that grants would continue to come their way.
Re: hunter (#42), As I’ve mentioned before, PNAS had a 90-day window for comments and at that time, I , for one, was not in a position to run a RegEM run to see what happened in a non-dendro run without Tiljander. During the Steig commentary, we managed to get a RegEM emulation to work and, at some point, I’ll go back and revisit Mann. His code and methods are so obscure that these sorts of assessments are not completely easy. Plus there are other issues with RegEM such that one doesn’t want to open the file without doing the other issues.
Plus, at some point, climate scientists should be taking responsiblity for their own work. It’s Mann’s job to redo the relevant calculations with the correct Tiljander – he did this with the Rain in Spain falls mainly in the Plain (of Kenya) and could easily do so with Tiljander if the “Community” required him to do so.
Re: hunter (#42),
Mann et al’s Figure S8a shows that removal of seven proxy series that include the four Tiljander varve series leaves the paleoclimate reconstruction virtually unchanged.
If removal of bad data doesn’t improve the reconstruction, I’d guess that flipping the sign wouldn’t help it much either.
But this assumes (1) that it’s clear which proxies are being used in the figure (the 7 proxies are being subtracted from what?), and (2) that the figure shows what the authors think it shows.
In my opinion, neither condition obtains, and your quick question is unanswered.
Re: AMac (#46),
People have already done some simple tests in the context of Kaufman et al (2009), averaging the proxies, but flipping Tiljander upright and then removing the defective tree ring series in a leave-one-out test. IIRC the current warm period drops to roughly equal that of the medieval. No error bars were plotted on those tests. A search of “Kaufman and upside down Mann” should do it.
.
The Mann et al (2008) recon is not so simple.
Re: AMac (#46),
This is not the definitive answer, but it’s a start:
UC’s robustness graph
Re: bender (#49),
you can see why kaufman thought it wise to issue a corrigendum. mann chose a different route.
Re: AMac (#46), To me that would mean the reconstruction is probably pretty worthless in the first place.
Re: Jeff Alberts (#109),
Yes, you mention one obvious possibility. I don’t know enough to say–I don’t even know which proxies are represented by the green line.
I’m new here and haven’t followed this AGW stuff in much detail; just got interested in one aspect a few days ago, and have read out from there.
I am struck by the just-under-the-surface anger and frustration that pervades this debate.
Also by the enthusiasm of many of the supporters of Mann who post actively. There seems to be a drive to oppose any suggestion that Mann and colleagues could have gotten anything wrong.
The territory concerning the Lake Korttajarvi varve series and their use by Mann 2008 really isn’t that hard to master at a basic level. I have seen a number of convincing arguments that Mann got this one wrong, and no persuasive argument by his champions that he might have gotten it right. (It would be better if he made such arguments himself, as he must have the best ones.)
But as far as I can see, the number of minds changed stands at Zero.
The problems with Figure S8a and some of the other later SI figures are more subtle. It’s easier to spot the trouble if one already has a sense of how ‘good’ data tends to behave, and what ‘robust’ challenges tend to look like. If the manuscript had gotten better peer review at PNAS, one improvement would have been that Mann would have had to have been much clearer on what, exactly, he did to build each of the figures. But then, it’s the spectacularly wrong use of the Lake Korttajarvi varve proxies that highlight the wierdness of Fig S8a. If I were a reviewer who was full of respect for the care and quality of Mann’s work, I might have missed this, too.
But for sure, no minds are going to be changed on the defective figures. Not obvious enough.
I’m not saying that the AGW skeptics are a better breed apart from the AGW enthusiasts, I haven’t been around long enough to know. But so far, it does seem that there are more open-minded, let’s-start-with-the-data types posting here and at other open-to-skepticism sites. But I could be wrong.
What are the best (e.g. most open-minded, tolerant to diverse opinions) pro-AGW sites with topical threads?
Re: AMac (#117),
Nice summary, AMac.
I’d love to know as well. My impression is that technically there aren’t any, as given the rules of the game as they currently exist, you can’t be both pro-AGW and at the same time tolerate skeptical questions. Thus while there are many sites like this one that should be classified as accepting of AGW but questioning the details, they have instead been labeled as anti-AGW. But please report your findings!
Re: AMac (#119),
I’m a little late to the party, but the question belies a misunderstanding of the structure of the controversy. Lucia did a fairly accurate characterization of the players as being in three catagories; (iirc) “stone-colders”, “lukewarmers”, and “hell-fire-ers”. This site generally aligns with the lukewarmer camp. There aren’t a whole lot of stone-colder science websites; they tend to be political polemics. The hell-fire-ers tend to be the pro-AGW polemics.
This being the true structure of the issue, the only group where the conclusion isn’t presumed is the lukewarmers. You’re not going to find an open-minded pro-AGW site for the same reason that you won’t find an open-minded stone-colder site; the conclusion is already concluded.
Steve, if you think this is out of bounds, please snip, but I think it’s an important point, and an answer to AMac’s question.
Re: Calvin Ball (#325),
There really isn’t any site where there is the sort of exchange that we would all like to see. It appears that mainstream paleos are simply unwilling to defend their work in online forums where they are engaged by professional statisticians and data analysts.
Part of this is peer pressure. Some paleos that posted here have had to withdraw because of pressure from their peers not to participate here.
I’ve offered unfettered editorial privileges to mainstream scientists in the Community but they are afraid of getting cooties, I guess.
Actually, it even seems that paleos no longer even try to defend their work in venues where they are protected by heavy security and blanket moderation such as realclimate. Then they complain about the “PR Challenge”.
But I don’t want to debate this point on this thread. We’ve been through it lots of times.
“The issues seem very elementary to me and I don’t understand why they seem to difficult for climate scientists.” Not only do I admire the discipline with which you, Steve, refrain from publishing your inferences about The Team’s motives, I also admire the discipline you show in not quite opining on their intellectual standards.
Graph of the name Gavin vs. time:
http://www.wolframalpha.com/input/?i=gavin
Maybe the calculation for Steve’s analogy could be done with WolframAlpha? I’m not familiar enough with the syntax and capabilities. I’m sure someone has R lying around, however…
My understanding of this is that Mann claims to have thrown this proxy into his “meatgrinder” without regard to sign which mistakenly correlated spurious 20th century varves with the temperature record. But since it was an anti-correlation, when the program used it to convert the previous centuries to temperature, it made all the highs lows and the lows highs, which turned Finnish limnology on its ear.
Is this a correct interpretation?
Re: J in L du B (#50), yes
Another example from the real world.
My house is heated with natural gas. Historically higher annual gas consumption is related to cooler outside temperatures (cooler temperatures – more gas consumed), and I have a 20-year history. Five years ago, my wife installed a pottery kiln that uses natural gas. Annual gas consumption has increased considerably. It has also been slightly warmer outside since the kiln was installed.
Should I flip the orientation of the gas consumption data and conclude that gas consumption is now positively correlated with temperature (warmer temperatures – more gas consumed) because of the uptick caused by the kiln factor?
Re: Paul29 (#51), Conversely, I installed a new furnace before the recent cold winters in Chicago, and my heating bills when down while it got colder outside.
Re: Craig Loehle (#66),
Craig, that is to be expected when you have a net gain of 30% efficiency over your old furnace. Did the Gas Co. call to come and inspect?? Usually they do!
Just FYI:
Gavin Schmidt
http://www.google.com/trends?q=Gavin+Schmidt&ctab=0&geo=all&date=all&sort=0
Steve McIntyre
http://www.google.com/trends?q=steve+mcintyre&ctab=180741696&geo=all&date=all
Sadly not enough references to Gavin Schmidt to generate any reasonable webreferenceclimate work.
Re: Stephen Parrish (#54),
every 14 months mcintyre uncovers a major error that is newsworthy? what events are those spikes?
Re: bender (#57),
“every 14 months mcintyre uncovers a major error that is newsworthy? what events are those spikes?”
sounds like an ocean cycle. Does anyone know if Mann is working this into his next reconstruction??
Sorry…
Just Gavin:
http://www.google.com/trends?q=gavin&ctab=0&geo=all&date=all&sort=0
Tim G (#3), you are 70% there. Here it is in simple form. You are correct, the sediment record is inverse, and obviously you need a transform to relate it to temperature. So “High” sediment is “Low” temperature, “Low” sediment is “High” temperature. Now, a curious thing happens. At the sites, after a certain date (we’ll say after 1900 for the sake of argument. I’m grabbing that date out of the air), you get progressively “High” sediment. That means the model should say “Low” temperature. However, the scientists who specialize in this field know that this is incorrect, because after “1900”, man was influencing the sediment record due to construction activities. With me so far?
Ok, so you have two options. You either conclude that this proxy won’t work for temperature, or you try to establish the “cut off” date and only use the proxy before the cut off date. These are the only two option you have. If you choose the second option, you then run into the problem of “calibrating” this proxy with less reliable instrument data, but that is off on a tangent.
Now Mann instead switches the signs, and correlates the temperature rise during the 20th century with the very High (note again this is backwards)post “1900” sediment data. Besides being utterly preposterous and meaningless, here is the part you are missing: “WITH THE HUGE, OVERWHELMING SIGNAL POST 1900, WHAT WILL BE THE EFFECT ON THE MWP TEMPERATURES PREDICTED?”
Doug Hoyt,
You raise some interesting points, and they may find application in CO2, precipitation and Temperature vs. growth in trees, or maybe volcano activity and fertilizer effects, etc… But for this sediment problem, it is pure garbage. They are in a sense trying to get a signal out of man’s farming and construction activities. It is absolute garbage. Unfortunately, the sedimentation makes for a nice “blade”, as it occurs in later history and increased building activities. So the net effect is that the MWP disappears.
The sediment proxy might be a decent proxy, but when you “contaminate” it with this overwhelming signal, it becomes useless.
Re: JamesD (#57),
Yes, I agree that some of proxies have problems. I am pointing out an additional problem that occurs even with perfect proxies and that is, if the proxies are cross-correlated (or non-orthogonal), then it is very dangerous to use any standard multivariate technique. One must correct for cross-correlations. Failure to do so will give you incorrect calculated output coeeficients, even negative ones.
My preference would be avoid these multiple regression techniques altogther because they are so prone to giving spurious outputs.
Connolley, Is that the Opinionator from Wicked Pedia?
Anywayz, from his RC curriculum vitae: “When I joined RC, I was a climate modeller with the British Antarctic Survey. Now I’m a software engineer for CSR. I’m still interested in communicating the science of climate change, but can no longer do so at a professional level“
Re: Hoi Polloi (#68),
Same guy.
Re: steven mosher (#59),
I agree w/steve. As a chemical engineer, all reports that I authored that involved the analysis of a particular process or future projection of a process had to be backed up by accurate data as well as the computation methodology I used, e.g., open and transparent.
I knew that others would take my report and try to replicate the results. If I were to issue reports that used incorrect methodology or data, I would probably have been fired… post haste!
I am amazed at the level of intellectual dishonesty that is prevalent and tolerated in our universities and the scientific community. Its no wonder that our kids don’t want to pursue a career in the sciences anymore.
Kudos for the bloggers that are having to do the verification and validation job that authors and peer reviewers have chosen not to do…. or are incapable of doing.
FWIW I think Connolley is quite a thoughful guy, and I don’t think he would consider himself a member of the “Team.”
Re: Nicolas Nierenberg (#76),
Thoughtful people can have momentary lapses of reason.
In summary:
1. You have to be a climate scientist to make upside down errors.
2. Statisticians must issue apologies for being correct.
I corresponded with William Connolley several years ago. Both then as now, he was very consistent. He never lets the facts alter his beliefs about AGW.
William, if you are reading this, you can no doubt find our email corresponednce in your files.
I too corresponded with Connolley but through Wikipedia ‘discussions’. I remember one conversation where he accused me of making up the concept of tropospheric amplification in the tropics. He had know idea! I’m sure it’s in the talk page on the wiki site.
He talks way over his head all too often.
Connelly at Stoat:
Bender at CM paraphrasing Connelly:
Connelly at CM to Bender:
yikes!!
Re: windansea (#82),
Well it doesn’t end there. I issued a conditional apology and asked him to point out where I was misinterpreting him. That was this morning. Let’s see if he follows up. Or if he moves on.
Re: windansea (#82),
He was probably objecting to my use of the word “perverse”. But is that not an accurate description of a proxy that correlates (positively and spuriously!) with climate during the calibration period and negatively during the reconstruction period? What would be a better word?
Re: windansea (#82),
Connolley at stoat:
But I’m “making things up”, am I?
There’s a new thread at Stoat (Connolly’s blog) called Tiljander. I left the following comment —
I’m a lay person who has looked at Mann et al (2008) (paper, SI, Comment, and Response), and who has followed the last bit of the discussion of the Lake Korttajarvi varve proxy series.
I had thought there was no dispute on the following five claims, but now I am not sure:
(1.) That Tiljander believes that the climate signal in the Lake Korttajarvi varve dataset is that higher local temperatures correlate to thinner, more-organic-rich, lower-XRD varves.
(2.) That Tiljander cautioned that after ~1720, the Lake Korttajarvi varve dataset is likely affected by local-human-activity signals, leading to thicker, more-mineral-rich, higher-XRD varves than climate alone would produce.
(3.) That Tiljander has described two incidents of local activity that led to very thick, mineral-rich, high-XRD varves, in 1930 (peat ditching) and 1967 (bridge reconstruction).
(4.) That the varve proxy was calibrated by Mann et al over a period (1850-1995) in which higher local temperatures were correlated to thicker, more-mineral-rich, higher-XRD varves.
(5.) That Mann et al used the Lake Korttajarvi varve proxy in the reconstruction of the Temperature Anomaly record by applying the 1850-1995 correlations (thicker, more-mineral-rich, higher-XRD varves with higher temperatures) to the varve record spanning 200 AD to 1850 AD.
W. Connolly and AndrewT, would you be willing to say which of these assertions you agree with, and where you disagree with them?
Thanks.
Re: AMac (#86),
I know my answers, but I’m going to withold because I don’t want to be the source of their information.
re 85
here he is at Stoat again: (In oh Dear post)
WC has new thread at Stoat where he says this to Amac’s comment:
Re: windansea (#88),
He wants AMac do his his homework for him. Understandable. But it proves my earlier surmise: he leaped before he looked.
Re: bender (#89),
Homework done, I Googled “tiljander varve proxy mann mcintyre” and the cites came right up. I would have skipped (2.) if I had noticed that Connolley provides it in the body of the post (italicized text).
As I said, I thought those 5 items were matters of fact not opinion at this point. But I guess they aren’t.
I don’t understand “You forgot ‘and those which you don’t know about’.” It’s easy to construe meaning on teh interwebs and then later discover it was meant differently. If important, Connolley will clarify.
Re: AMac (#92),
I think you’re misreading me.
1. I said Connolley wants YOU to do HIS homework for him. (The reality is he wants you to go to the primary source, and not rely on misinformation from that demon of the denialosphere, SM.)
2. Those propositions are factual, not matters of opinion. What I’m witholding is whether I agree.
3. windansea is right; Connolley is suggesting that there are more than just these five claims that matter. (It’s a hand-waving game that people like him and Tom P play. They try to scare you into thinking maybe you don’t know your stuff. They’re upping the ante, thinking you’ll fold. It’s a blind man’s bluff. They usually end up losing badly when dealing with knowledgeable, honest people.)
RE Bender #89
exactly
He defended Mann reflexively without doing a bit of due diligence.
Amac, his reply to you saying:
he’s not accusing you of not knowing, he’s trying to say he didn’t know this
which is weird as this was pretty clear in Steve’s comment to PNAS
Someone may have pointed this out already, but I think there is a mistake in this sentence:
I believe you meant to write the opposite: narrow sediments are interpreted by the authors as “cold” and thick sediments as “warm”.
Love your blog and think your work is incredibly valuable. Keep it up!
Re: Tom (#93),
There is a certain paper, on my desk, authored by _________. Have you read it? It might change your mind.
Connolley responds:
http://scienceblogs.com/stoat/2009/10/tiljander.php
Given Connolley’s latest, which includes a link to Mann’s supplementary data,
Click to access 0805721105SI.pdf
…it seems that:
1) Connolley is correct on the narrow math point: in this situation, there is no such thing as the signal being upside down.
2) Mann et al were well aware of the problems with the various proxies that Mc has discussed, and dealt with them as best they knew how. Perhaps not up to Mcs exacting standards, but still, they were not caught out entirely by them, as Mc has insinuated.
Bottom line: Mc here as is re-picking a nit Mann himself picked back in 2008. Nothing to see here.
Steve might have originally formulated his criticism differently. Rather than claiming the proxy was used “upside down”, Mann used an unreliable section of the proxy record that the developer of the proxy (Tiljander) reported contained a large non-climatic signal. Linear regression including the unreliable portion of the record produced a coefficient opposite in sign from that expected from Tiljander’s publication.
Due to space limitations, this is difficult to explain in a few words. Space limitations suggest focusing attention on the one or two problems that are most likely to invalidate the paper. Numerous issues make it difficult to choose which ones to focus on. In theory, separate issues could warrant separate submissions.
Proxy record CAN be used in “any” orientation. In some hot arid locations, tree rings probably get smaller during warmer years. If Mann ever used a tree ring proxy “upside down”, Mann would be responsible for providing a convincing rational why this proxy should be used in an unusual orientation in this particular situation. (If such a rational existed, of course, the originator of the proxy would have been the one most qualified to make this judgment.) The tricky problem is forcing Mann into a corner where he has to argue that Tiljander (the expert on the proxy), not McIntyre, is wrong.
Re: Frank (#98),
Didn’t Mann already voluntarily place himself in a corner surrounded by his own wet paint with his “teleconnections” theory which implicitly rejects, without foundation, Tiljander’s well documented man-made sedimentation issues?
Re: Frank (#98),
The rhetorical device that Team is using can be referred to as “throwing up sticks and twigs”. No, Mann didn’t use it upside down, only his model. Whew, what a relief. It wasn’t Mann, it was his model. But then again, take a look at his famous S9 graph. Mann created that graph, so he used the proxy upside down in that case also.
Oh yeah, and who “calibrated” the proxy? So by using a two step process, one calibrating the proxy to the instrument record and then two using the proxy to reconstruct the temperature history, Mann is using the proxy upside down.
But this is all silliness. Mann screwed up.
From Mann’s supporting information:
Click to access 0805721105SI.pdf
What it shows is that, contrary to Mcs insinuations, Mann et all knew about the data quality issues in Tijinder and countered them as best they knew how. So Mc has been doing nothing but re-picking a nit Mann et al picked over a year ago.
Potential data quality problems.
In addition to checking whether or not potential problems specific to tree-ring data have any significant impact on our reconstructions in earlier centuries (see Fig. S7), we also examined whether or not potential problems noted for several records (see Dataset S1 for details) might compromise the reconstructions. These records include the four Tijander et al. (12) series used (see Fig. S9) for which the original authors note that human effects over the past few centuries
unrelated to climate might impact records (the original paper
states ‘‘Natural variability in the sediment record was disrupted
by increased human impact in the catchment area at A.D. 1720.’
and later, ‘‘In the case of Lake Korttajarvi it is a demanding task
to calibrate the physical varve data we have collected against
meteorological data, because human impacts have distorted the
natural signal to varying extents’). These issues are particularly
significant because there are few proxy records, particularly in
the temperature-screened dataset (see Fig. S9), available back
through the 9th century. The Tijander et al. series constitute 4
of the 15 available Northern Hemisphere records before that
point.
It seems Connolley is more interested in the semantics than the substance – and his wikipedia talk page (and ex-admin status) seems to suggest he is unlikely to be persuaded that the use of said proxy is a mistake regardless of the impact of removing it.
Amac nice comment at CM
I agree, what’s behind the green line?
I didn’t use that proxy. Upside down.
From stoat:
1. If Tiljander is “lousy”, what might that suggest about the other lake sediment proxies? Nothing. Nothing at all.
2. “It didn’t matter”? That is not true. A leave-one-out test with Tiljander already out would have caused the crack – err, I mean tree-ring series – to drop out in one of the combos. That combo would have been telling. Alas, a lack of disclosure …
3. I’ve already stated #2 at CM. Rattus doesn’t Readus.
4. I think Steve is probably chuckling, not “fuming”.
Unprofessionalus rattus maximus.
Re: bender (#106),
I wonder what other proxies the team will finally admit to being lousy? This is just a “limited hangout”. Cop a plea to one bad bit of business and move on…
Seriously: calibrating to contaminated data. The failure of Mann to admit and apologize for this basic greenhorn error is shameful. Not a stand-up guy, we should dub him Mann’t. This type of disconnect between the field and the data analysis is all too common in science these days.
Ah there’s that “b” word again, used yet again in a tendentious way. I trust dhogaza’s interpretation of Jeff’s interpetation of Steve’s interpretation of Mann’s summary of his technician’s work. But let’s just verify, shall we?
# 15
First thing we checked. Nope. Spurious inverse correlation in the calibration phase due to signal contamination. Just answering.
Let’s guess as to the lifetime of the “denialist meme” (where do they get these ugly phrases?) on stoat. We’re at #18. Do I hear 20? 30? 50?
We’ll keep a list of the “denialists”.
1. Connolley
2. Bloom
3. BCL
4. Rattus
5. dhogaza
That’s 5, count ’em 5. Do I hear 10? 15? Hank?
Re: bender (#112),
6. Rabett, of course
Re: bender (#112), Did somebody steal my meme of calling AGW folks denialist?
I need to inforce my trademark
Re: steven mosher (#131),
You are the one that gave me this nasty meme infection? Why I oughta …
Re: bender (#133), I think I did it over at lucia where I contrasted two forms of denialism. The forgivable denialism of the ignorant ( folks who deny radiative physics) and the willfull ignorance of those who know better: people who deny that mann flipped the series. And I said that the latter form of denialism had MORE in common with holocaust denialism than the former. For someone who is a scientist to be in denial about the absolute need for others to verify that your code and data actually produce the results you claim is just willful ignorance. And now I suppose someone will quote me on this and attribute this belief to SteveMc merely because he allows me to post here.
But S8 includes bcps! So remove and recalculate. Whoops. No code.
wonder why WC didn’t include this caption with the graph
so…what is “full global proxy network”?
Re: windansea (#115),
There are 1209 of them. You want a full list?
Would there be a large effect on Figure S7 if you exclude Tiljander? Or is that impossible to answer?
Re: JamesD (#116),
Rattus thinks you are asking the wrong question. He thinks you should be looking at S8. He doesn’t understand why you want S7 recalculated – probably because he doesn’t understand that it’s a leave-some-out test that is needed not a leave-one-out test. S7 can’t be recalculated until all the data are available and the EIV code is running. Only Mann can do that at the moment.
.
The crack cocaine in this recon is upside-down Tiljander and the bristlecone pines. Re-instate Tiljander with its proper interpretation (and if it can’t be calibrated due to removal of contaminated data, then leave it out) and remove bcps – that’s the litmus test. All the other flavors of sensitivity test they show are not a strong test of robustness.
Re: bender (#119),
I think pushing for the revised figure S7 is the way to get this litmus test.
Figure S7 is why Mann can never issue a corrigendum correcting for Tiljander. Tiljander is the blade in S7, so if he does issue a corrigendum, he has to redo the dendro sensitivity analysis. This analysis will show no hockey stick without the bcps, and prove Steve right. Mann can never back down on Tiljander, or he is toast. Figure S7 won’t allow him.
Re: JamesD (#118),
That is a surmise. It needs to be checked, not asserted.
Re: bender (#121),
Thanks, good point. I tend towards exaggeration in arguments, and that ends up hurting the cause of truth. Let’s let proper science settle this.
An interesting point is that all natural phenomenon at the macro
level exhibit a quasi-normal or Gaussian behavior, ie peoples height,
tree thicknesses, pond sediment depths, tree rings, etc. Unlike
man-made systems such as stock markets, book sales, millionaires which take on a Mandelbrotian distribution or something close to it.
Now, someone like Steve Mc is using a very small portion of his
analysts tool set on this climate modeling work. You know that Sears
toolbox with the 5 drawers and the tiny one inch drawer at the top
that you have flat wrenches (spanners for some of you’all), well
that is about all that has to be used in this, basically a flat tipped screwdriver, and a crescent wrench. So when you see something
in nature that exhibits 5 sigma behavior, you need to take another
look. That is what happened at Tiljander with the excessive gravel
deposits. Can’t use that portion of the data.
But Rattus was nice enough to state that Tiljander is “lousy”. A nice concession. Which brings us to the needed corrigendum. So to do a proper corrigendum, Mann has to show the impact of the “error” on all the parts of his study that were impacted by it. Since S7 is part of his study, then Mann will be forced to do a sensitivity without the bcps and without Tiljander.
However, I’m not 100% sure this gets rid of the hockey stick, since there are other “upside down” proxies, but I bet it scares the heck out of Team.
It doesn’t seem that Tiljander can be calibrated in a period that ends after 1720 (based on Mann’s quote of Tiljander in SI, pg. 2).
Can a pre-1720 calibration be accomplished? Is there another record from that part of Finland that could link a Tiljander calibration period to a time covered by instrumented records?
How to get the Mann proxies into R is outlined in this comment of Steve’s:
Run both bits of code in the sequence above. The proxies are named in object “mann”.
Wrong. We all read the SI on day 1. Mann was aware of the questionable nature of the lake sediment proxies, yes, but was not aware of the specific problem of sign flipping, and never had any intention of keeping the TWO active ingredients out of the 1209. SNOOOOORT.
dhogaza was counting on a no-show. he’s starting to wonder: “what’s up, do they have something we don’t know about? have we been misled?” Still in denial though. Keep strong the faith.
1. Idiocy? No. Speculation? Yes.
2. More hockeysticks? Yes, at realclimate Hey Ya!(mal) thread, but only because (1) most of those series do not go older than 1600AD, (2) they refuse to fix all those series that rely on Yamal, bcps, Tiljander. Worse, they are censoring every comment suggesting there could be a problem.
Naive question, maybe.
Are the terms “The Team,” “Hockey Team,” and “Hockey Stick Team” considered dismissive or offensive when used to describe an ardent supporter of the AGW consensus position, such as commenters defending the integrity of the Mann et al 2008 reconstruction?
Or are these terms that people are happy to see applied to themselves and their fellow enthusiasts?
Maybe it varies according to context.
(Trying to avoid giving needless or inadvertant offense.)
Re: AMac (#137),
This may give you some background.
http://www.climateaudit.org/?p=568
But it’s sort of like sports. One team or group of fans will come up with something they like or think is clever and their opponents will inevitably find some way to use it derisivally (or vice versa—one story goes that the term “Yankees” for the baseball team was originally meant derisively by other ballplayers who came from the south, and they decided they liked it.)
In blog land? 🙂
But seriously, at this stage of the “game”, you can be pretty sure that folks using the term are not “team fans.”
Re: AMac (#137),
Apologies if someone has replied already by the time I post. This may help you to decide.
OK.. I was wondering why you guys were not praising me for my post at stoat.
Moshpit was very moderate.
William sent me this:
I’ve been trying to keep this thread focussed; I don’t want it to
degnerate into the usual anti-Mann stuff. so I’m deleting your
comment, but you get this copy in case you want to post it elsewhere,
or rework it.
regards,
William
2009/10/29
> A new comment has been posted on your blog Stoat, on entry #136400 (Oh dear oh dear oh dear oh dear).
>
> Aside from misunderstanding the upside argument William you miss the
> most important aspect of this issue. Without Mann’s data ( as used) and
> code ( as used) no reviewer can tell if the graphics supplied for the
> text or the claims in the SI are actually correct. That’s the most
> fundamental quality check that must be done independently.
>
> WRT the upside down series.
>
> 1. The period in which the calibration is done is corrupted by non
> climatic signals according to the peer reviewed literature which
> introduced the series. This would
> suggest:
> A: dropping the series altogether
> B: lopping off the corrupted period
>
> 2. If the orientation “doesnt matter”, then
> A. this is a simple claim to TEST with the code and data in hand
> B. there should be no objection to orienting it in the direction
> indicated
> by the author who introduced the series.
> C: one wonders if varves are skillful at capturing climatic
> signals.
>
Re: steven mosher (#138),
“They rarely venture out of their fantasy world” … and this has nothing to do with the censorship they experience? Look at this. Connolley won’t let mosher describe the core issue here: people like JamesD can’t get the sensitivity test they want to see because Mann’s code is, as usual, unavailable.
Re: bender (#144),
But Gavin assured FredB on the realclimate Yamal thread
Surely he wasn’t misleading him?!?
Re: Jonathan (#150),
Actually, you raise a good point. I haven’t tried, myself, to obtain Mann’s code. I really should do that before jumping to conclusions based on what others have said. Let me retract that statement for the time being.
So do you all see the ILLOGICAL stance william takes.
He Blocks my post because he is afriad it will generate OTHER POSTS which are anti mann.
Posts which he could BLOCK. Now THATS a precautionary principle. I said nothing negative about MANN.
nothing personal, nothing defamatory. William could let my post through and address the issues MY POST
raises. he could block idiots who piled on. But no, I spoke manns name so I am blocked. Funny.
So, you all go support moshpit. Take my comment. rewrite it so it doesnt mention mann. see if William
blocks it? And make sure to give me credit.. moshpit. it’ll be fun if you all do it.
Mosh, 139: The whole thing could be settled if Connolley took time out from his exalted position at his own website to respond to the assertions on this thread. Here. And Now. And while he’s at it, maybe he could bring with him some kind of coherent explanation from the Mann. They are associates, are they not?
I won’t hold my breath.
The fact that they censor and Steve only snips is telling.
Steve: I don’t snip scientific criticisms. 99% of snipping is of “piling on” by supporters or of policy advocacy.
Re: theduke (#141), as a person who has been blocked by RC, tamino, and stoat, as a person who has been snipped and zambonied by steve I can say this. I understand steve’s rules. the other guys? seems either capricious and/or ad hoc.
Don’t forget that WC has “invisible lines”: http://www.climateaudit.org/?p=419#comment-8673
I want to commend AMac for staying on task with a laser-like focus across the several places where these issues are being discussed.
Several of the regulars here, including me, would not have been permitted to participate in the discussions. AMac has superbly filled that potential void, IMO.
Perhaps part of Connelley et al’s confusion is that varve thickness can go either way, depending on how the lake is fed.
If a lake is fed by annual rain and snowfall, a cold year will make a given amount of precipitation be disproportionately in the form of snow. The spring thaw will then be unusually violent, and bring down lots of material to form thick varves. Hence thicker = colder. This is Tiljander’s argument for Korttajarvi.
But if the lake is fed by a permanent glacier, a warm year will cause more meltoff, and hence more sediments and thicker varves. Hence thicker = warmer. This is likely the case for other northern lakes.
Re: Hu McCulloch (#147),
There is no glaciers in central Finland.
I am not 100% sure but i belive there is no glacier at all in Finland.
I sorta see the Teams point on this one.
1) You have a diverse collection of data series that might or might not correllate to temperature, positively or negatively.
2) You have a short series of temperature measurements that you trust.
3) You ignore all physical explanations and scale/flip each series for maximum correlation to the trusted portion of the series.
4) It produces the expected and desired result.
Now people are saying your scale/flip of a series contradicts the physical explanation.
I can understand the position that if the physical explanation fails to match a series of temperature measurements that you trust, then the physical explanation must be flawed. Throw out the data as a set that has been proven to not correlate to instrumentation.
I’m not a warmist, I think there is not a long enough trustworthy record to say what is normal, and clearly some things have happened to cause and then undo several ice ages. But if a series clearly cortradicts the short part of the record you can trust, then be gone with it!
Re: Kevin (#153),
That would be totally against the Team ethos, which appears to be repeat the mistake over and over until its declared “The Paleoclimatic Standard MethodTM”. Either that or attempt to bore your opponents into submission with passive-aggressive contempt.
Nevertheless, in my reconstruction of Michael Mann, I too have found that the sign of the predictor makes no difference to the result:
…and I’m sure you’ll all agree, this is probably the preferred orientation
Re: Kevin (#153),
Sure it’s easy to see the Team’s point. The trouble is it’s strongly tied to the assumption of the proxies actually being there. If instead, proxies have either weak or nonexistent temperature signals, then the correlation to trusted measurements is mostly or entirely random and selecting to match the trusted portion will simply CREATE a hockey stick (or whatever shape the trusted measurements might have had in an alternate universe). IOW, proxies are assumed and then necessarily found.
Now if the original assumption were correct then it would be easy to find an unending parade of such proxies. If so, then why are the same ones trotted out every time? SM doesn’t like stripbarks; fine we’ll present 20 other series with HSs. But it hasn’t happened. Could be laziness, but I think it’s because there are no good proxies to prove the warmer’s point.
Re: Kevin (#153)
“I sorta see the Teams point on this one.”
I don’t. Yes, I can understand how Mann screwed this up in the first place – confirmation bias plus not checking each proxy’s provenance as carefully as he might have could easily result in this one getting used upside down. Scientists make mistakes sometimes. However, on discovering this, his (and the Team’s) reaction to it is just embarrassing. No, it doesn’t matter to the mathematics. But it most definitely matters to the decision on whether to use the proxy at all. According to their own methods, this series should either be discarded because its correlation during the physical temperature record is negative, or else it should be truncated, lopping off the last century or more during the time period where mining and other local human influence rendered it worthless as a proxy.
A reasonable response to this would be “I’ve removed Tiljander, and reworked the reconstruction without it – results archived “. Would admitting (and correcting) a mistake tarnish his record? No. That’s how science is supposed to work. What tarnishes your record is belligerent, arrogant insistence that you’ve made no mistakes, no matter how obvious it is that you have.
Re: MikeU (#166),
Aha. You see, he did do “rework the reconstruction”. But what he didn’t rework is the Fig S7 sensitivity test asked for by James D. You see the pea under the thimble? One graphic in a 27-page packet (is that unprecedented?) of supplmentary information is the graphic that “matters”.
Just posted this at Connolley’s:
William-
Will you be answering AMac’s request for your views on his five questions in #1 above?
Not answering gives a strong impression that you do not want to resolve this question based on the facts of the matter. See #37 above.
If you choose not to respond to his very fair questions, could you explain why?
————-
AMac’s questions were:
I’m a lay person who has looked at Mann et al (2008) (paper, SI, Comment, and Response), and who has followed the last bit of the discussion of the Lake Korttajarvi varve proxy series.
I had thought there was no dispute on the following five claims, but now I am not sure:
(1.) That Tiljander believes that the climate signal in the Lake Korttajarvi varve dataset is that higher local temperatures correlate to thinner, more-organic-rich, lower-XRD varves.
(2.) That Tiljander cautioned that after ~1720, the Lake Korttajarvi varve dataset is likely affected by local-human-activity signals, leading to thicker, more-mineral-rich, higher-XRD varves than climate alone would produce.
(3.) That Tiljander has described two incidents of local activity that led to very thick, mineral-rich, high-XRD varves, in 1930 (peat ditching) and 1967 (bridge reconstruction).
(4.) That the varve proxy was calibrated by Mann et al over a period (1850-1995) in which higher local temperatures were correlated to thicker, more-mineral-rich, higher-XRD varves.
(5.) That Mann et al used the Lake Korttajarvi varve proxy in the reconstruction of the Temperature Anomaly record by applying the 1850-1995 correlations (thicker, more-mineral-rich, higher-XRD varves with higher temperatures) to the varve record spanning 200 AD to 1850 AD.
W. Connolly [sic] and AndrewT, would you be willing to say which of these assertions you agree with, and where you disagree with them?
Thanks.
———————
Connolley does not deserve to be let off the hook on this.
I simply don’t understand the rationale behind folks like Connolley et. al. who argue against such elementary corrections (elementary because it is easy for us layman bystanders to follow).
We are all, be it by profession or sport, in the weather modeling business here. It is patently obvious that current models are diverging drastically from the actual global climate, and our host here does nothing more than the leg work needed to explain why these models are diverging… but these folks still argue that their demonstrably faulty models are made up of perfectly good assumptions and value added data.
When the weather modeling community should be in the “tear down and rethink” phase of climate theory they are instead becoming more insular and defensive of poor assumptions a laughably divergent predictions.
Some of these people could wake up tomorrow to ind a glacier has advanced into their breakfast nook and the first thing they would think is “damn you global warming!”.
Before people create soft pitches, Mann placed code and data for Mann et al 2008 online, a point that I recognized at the outset and for which he deserves commendation. The code is horrendously written and poorly documented and has strange, quirky and undocumented decisions in it – so even with the code in one hand and the article in another hand and considerable expertise in figuring these things out, to borrow Mosher’s phrase, it’s like sticking needles in your eyes trying to figure it out, but it can be done. However, you can only stay underwater so long; we got through the CPS in late 2008. Jean S and UC did a post on EIV/RegEM while I was in Thailand last March and I didn’t follow up on that portion. Combined with work by Jeff Id, Ryan O and myself on Steig RegEM, where we have functioning R emulations, it should be possible to re-visit Mannian RegEM and that is very much on the docket.
But there’s only so much time in the day.
Re: Steve McIntyre (#169),
It was me that threw up the soft pitch. I thought I might have erred, hence the retraction. I was hoping you’d comment. When I say “disclosure” I’m hoping for “turnkey functionality”. What was released was NOT turnkey.
Re: Steve McIntyre (#169),
This was the source of my mistake: I wasn’t certain why you’d moved to emulations in R, whether Mann’s code was unavailable, broken, undecipherable, or just too painful.
.
Jonathon #150, please take note that the code is, technically speaking, “available”.
Re: Steve McIntyre (#160), We also have a much faster algorithm than RegEM that converges to a virtually identical solution – which was based on your initial truncated SVD algorithm. Perhaps when we get the Antarctic paper finished up I’ll generalize both the TSVD and TTLS scripts so they can be more easily used for this purpose.
Re: Ryan O (#194), I also forgot to mention that the script outputs the weights assigned to each series (thanks to Nic L for providing that modification some time ago). It will be rather easy to tell if a series was used opposite of its physical orientation. I should probably turn it into a package and submit to to CRAN so anyone can use it.
Re: Steve McIntyre (#160), Mann 2008 code is up! great. Then William connelly should let my post through and nail my skinny butt to the wall and ask for an apology on his blog. An apology I will promptly and graciously give. in the interim I will upon steve’s word of it’s existence apologize here.
I was wrong. Mann’s code and data has been posted. The ball is in the critics court. Now, mann may
have served up a spitball of code ( yes needles in your eyes) but he served it up.
However my argument still remains. The corrupt section should be lopped off, regardless. Or the series dropped
altogether. after all, it makes no difference.
WOW. The censorship at Stoat is unbelievable. All of the comments that I quoted above are gone. All trace of dhogaza’s obstinant willful ignorance. All of Rattus’s denialism.
Re: bender (#165),
Scratch that. I didn’t realize there were two Tiljander threads at stoat. The second one contains everything that I thought was cut.
Re: willard (#170),
Like I said already: I’m off this case in exactly 30 minutes. Then I go back to reading. Good?
road, posting at stoat:
The more recent explanations are more a propos, because Mann, in fact, is not using a “multivariate regression”. Mann introduced a red herring that caused some oblique commentary. But Connolley won’t clear that up for you. He would rather paint you as quoting somebody ignorant of the facts.
Re: bender (#166),
Bender, did you post this quote of road because you agree with it?
I think if Mann were to remove the Tiljander series and the other 3 series that were flipped by Mann’s code, then, in the subsequent analysis, one or more of the remaining series would be flipped. That is not guaranteed, but is highly likely in my opinion.
Re: Douglas Hoyt (#174),
I wasn’t quoting road, I was addressing him. Because I know he’s lurking here.
road:
Before posting at stoat you should read the exchange between myself and Connolley at CM:
He already knows exactly what the problem is because I told him.
Connolley, get your head out of your ass. The debate isn’t “shifting”. You’re finally starting to pay attention to what Steve’s been saying for more than a year.
Fighting for mathematical truth takes dedication, making scientific observations from 6 am to 10 pm, with only a short pause for supper.
Steve M, you may want to link to the new Tiljander thread at stoat:
http://scienceblogs.com/stoat/2009/10/tiljander.php
The old one is here:
http://scienceblogs.com/stoat/2009/10/oh_dear_oh_dear_oh_dear_oh_dea.php#comment-2032055
I meant to post this earlier so it might be obsolete by now, but it pertains to the argument that the “upside-downness” of the Tiljander proxy doesn’t matter in a regression. It is true that a linear regression is equally able to generate a coefficient with a positive or negative coefficient. But that is the problem, not the solution, when you have a prior theoretical reason to consider one sign valid and the other invalid. I’ve done dozens of system estimations of consumer demand models and producer factor input models. There are theoretical restrictions that need to hold, such as that demand curves slope down not up (i.e. as price goes up, purchases should decline). You can do an unrestricted estimation, but then you need to check that certain coefficient matrices are positive semidefinite or negative semidefinite, as appropriate, in order to be consistent with the theory behind the model. If they are not, you don’t just shrug your shoulder and say, Oh well, it’s a funny old world. No, you impose the sign restrictions using nonlinear inequality constraints and re-estimate. That reduces degrees of freedom and drops the likelihood score, but then the resulting estimations are theoretically consistent. In effect you are using information from theory as well as data to help identify the optimal parameter estimates.
In the paleoclimate case, when a proxy has a definite orientation to temperature, this has to be reflected in the resulting parameters (and the researcher has to check). If the sign is wrong, the model needs to be re-estimated with the appropriate inequality constraints imposed. And if that yields coefficients with confidence intervals overlapping zero, then the reconstructed temperature is underidentified and likely has infinitely large confidence intervals.
AMac at stoat:
1. The renumbering is because of censorship. That’s why I refuse to post there. Making commenters look like idjits is a design feature. RC does it too. But the similarity in policy is a coincidence.
2. He’s in no big hurry to resolve the matter, just as Mann was in no big hurry to resolve the issue a year ago. But the similarity in policy is surely a coincidence.
Thanks to all who responded to my queries in #119 and #139 (gee, the comment #s jump around here, as they do at Stoat…). Particularly Nathan Kurz, John M, Dan Hughes, GP, and Brent Buckner. The link to The Blackboard is appreciated, and the cautionary tale of the deleted RealClimate comment by TCO from November 2005. Note to self, local copy & paste before hitting ‘Submit’.
Re: AMac (#178),
Nothing was snipped. Some OT stuff on motive was transferred to “Unthreaded”.
Re: AMac (#178),
That is why (willard!) I am carrying on my stoat copy-and-paste monologue here. Connolley is censoring liberally. He’s bound to make you look like an idjit, whether he means to or not.
Re: bender (#181),
At the risk of being snipped it should be pointed out that taxonomically a stoat is a weasel.
Re: AMac (#178)
I don’t understand why Steve deletes comments. “Snipping” with an explanation left behind is understandable and completely reasonable; deleting without explanation leaves a detectable trail of disconcerting disruption.
Steve: It takes me longer to snip than to delete. 99.99% of all deletions are commenters piling on or venting about policy and 99.999% of the time it’s by commenters who I’ve done the same thing on multiple occasions. I begrudge the time in dealing with this sort of serial problem.
Re: Kasmir (#188),
Steve has stated before that with WordPress it is much, MUCH easier to delete rather than snip.
Please forgive my under-caffeinated confusion, but I cannot for the life of me this morning discern which words are the “e word” and “c word”. Could anyone give me some clues? Thanks in advance!
Re: Earle Williams (#191),
Charles Darwin.
Re: John M (#192),
Thanks for the hint. Now I feel sheepish.
More coffee! 🙂
Re: Earle Williams (#195). Others of us had the same question, but didn’t ask. You shouldn’t feel baaaad.
Long-time, non-scientist, big-fan-of-this-blog, not-comfortable-taking-a-strong-stand-on-AGW-because-I’m-not-smart-enough-to-understand-all-the-science layman lurker making his first post:
Despite that self-effacing introduction, I do consider myself reasonably good at recognizing when communication styles are working, when they aren’t, when they can’t, and when they might…
Let me also say I wouldnt ask Steve to change a thing about how he facilitates or communicates on this blog: through pure dogged persistence and relentless focus on facts, (which clearly inspires many commenters on this blog) I think he’s increasingly influencing day-to-day social dynamics in a body of science. Might he get more traction if he published more? Perhaps- but I appreciate the point he makes by NOT publishing much, thus placing responsibility on science. In other words, his communication style works.
All that said, here’s my observation:
In following this blog, I’ve always yearned for a new way to summarize and convey the nature of the issues Steve uncovers; something more intuitively accessible to the masses, something not so dependent on either a scientific background or months/years of curiosity-driven persistence and self-study (like I’ve tried). 2 specific examples have struck me as potentially very effective ways to show the issues:
1) the social-network diagrams shown in the Wegman report. a) they make it “visual”, b) they represent “people and process” issues (which is, in my view, Steve’s overarching point) rather than matters of scientific fact. Perhaps there’s a visual way to show all these temperaturate proxy,author,statistical-technique, are-trees-really-thermometers interdependencies that makes the strength (or fragility) of the paleoclimate logic chain visually unmistakeable. I’m thinking Tufte-like visualizations- href=”http://www.edwardtufte.com/tufte/” He’s a master at visually conveying the point of complex datasets. Some of his books include titles like “The Visual Display of Quantitative Information” and “Beautiful Evidence” I’m trying to get my work to pay for me attending one of his courses. I’m certain, in my core, that lurking in all this blog’s evidence is an astoundingly clear visualization of the fragility of some paleoclimate-science community arguments, a visualization that someone like Andy Revkin could really do something with.
2) AMac #86 – For some reason it struck me the moment I read his comment that the technique was powerful compelling, and highly worth imitating. Concise, individually articulated empirical facts that logically proceed from one to the next, leading to inescapable conclusions. No individual point was more than one short sentence. Downstream points remind the reader of points made in previous bullets, carrying the logic chain along. Not an overwhelming number of bullets. I realize that every post on this blog contains examples of logic-laden arguments and bullets. I’m in no way suggesting there’s a lack of logic here- I just know we’re all perpetually looking for ways to make our own work more effective, so I propose that contributors might look to apply the stylistic elements of AMac #86 into ongoing contributions. Let’s just try it and then keep trying to get better at it.
This brings to mind one of my favorite quotes of all time: “Don’t just be right. Be effective.”
If I were smart enough, I’d try this myself- but I’m not. So I invite others here who are clearly smarter than I to consider my suggestion and do with it what you may.
Thanks to all for all your efforts here.
-A thankful citizen
Re: Chris D (#179),
Thanks for kind words.
I think it’s only possible to do this when things are pretty clear cut. There’s a way to phrase the questions about Mann et al’s treatment of the Lake Korttajarvi varve record that goes 1 -> 2 -> 3 -> 4 -> 5. That would be much harder to do in terms of exploring the possible defects of SF7 and SF8a, as the arguments are more complex. But I don’t know how to start that discussion without knowing if there’s agreement on the earlier 5 claims. And–there isn’t.
Steve: snip – sorry, the C-word and discussion is strictly offlimits here.
Re: AMac (#183),
A well-thought out argument, although I, like you, am wary of analogies.
A word to the wise though, the e word and the c word are fodder for the zamboni machine.
If you don’t know what getting zambonied means, think of what the machine does to stuff sitting on the ice that can get in the way of puck.
Though Steve doesn’t “censor” per se, he does like the puck to slide freely.
Re: John M (#184),
> If you don’t know what getting zambonied means
Heh. Apologies for the breach of etiquette.
The title of this post is now ambiguous due to Connolley’s change in position. Connolley now does not endorse upside-down Tiljander, or Tiljander in any form. He doesn’t state whether or not he endorses upside-down Mann’s recon, but he does endorse his SI Fig S8 that removes Tiljander. He still seems to endorse Mann’s PNAS reply.
.
So I think it is fair to say he endorses the use of additive active ingredients in climate science, as long as they’re used rightside-up. That about right, Bill?
.
See you all at Christmas.
The Tijlander proxy was determined by specialists to have one orientation. Mann, however, used a computer algorithm to determine the orientation and this algorithm was insensitive to the particular properties of the varve proxies and of contaminated proxies in general.
There are a couple of obvious issues wit this:
a) why did Mann use a computer algorithm to determine proxy orientation on proxies whose orientation has been determined by specialists
b) Was he aware of the issues that his algorithm had with contaminated proxies in general and if so why did he use such a obviously defective algorithm
JeanS has indicated his opinion that the problem with Mann’s use of Tijlander was the result of a coding issue. I disagree with this. The computer code apparently worked as intended. The problem is not with coding but with the fundamental design of the computer program. I have read in many places that the design of scientific computer programs is somehow different from that of the design of engineering programs. That because they are “trying to find out how things work”, their programs do not have to be designed using proven methodologies.
Now if Mann had used a standard methodology one of his design reviewers would have asked him the reason that he was using the program to determine orientation when this was already known.. One of the design reviewer would have asked him how his temperature responsiveness determination wield handle contaminated proxies.
Perhaps Mann did use a standard methodology and his reviewers were inadequate. Perhaps Mann did not sue a standard methodology and the method that he used was inadequate. The result still is that this is a clear demonstration of the problems that occur when inadequate design methods are used in the creation of computer programs.
in re WMC:
We all know someone who is willing to act like a fool just to gain attention. Quite common.
From W. Connelly on stoat
One of the difficulties with varve proxies is that depending on their source, they may have opposite orientations. That is some may be thick warm and some may be thin warm. If the orientation of the proxy is unknown then how can a cooling trend be differentiated from a warming trend? The same can be said for all proxies some may decrease to indicate warming and some may increase. If an algorithm knows nothing about proxy orientation then how can it differentiate warming from cooling?
JeanS has described how the Mann program detects orientation from the correlation of the proxy with the modern instrumental record. Without this step, the program would not be able to tell warming from cooling. Mann’s regression technique may be insensitive to orientation to determine the magnitude of the temperature bt it is unavoidably sensitive to proxy orientation to determine if that detected change is a warming or a cooling.
The issue is twofold
a) orientation is determined by a program even if that orientation is known to have been determined by specialists
b) the orientation and temperature calibration aspects of the program are sensitive to modern contamination.
There seems to be no acknowledgement in the program of this issue. This points to a poor design and a failure of the design method.
Re: TAG (#189),
Note that my previous post at 189 was posted at Stoat and was deleted/censored.
However Connelly seems to be softening his position that Mann’s program is insensitive to sign or contamination. In reply to nanny_government_sucks Connelly says:
This seems to be an admission that Mann’s program is sensitive to the proxy sign but one that ties to hide the admission.
In is interesting how the fault is shifed to the proxy author and away from the reconstructionist who uses a proxy with known failings
Maybe I missed it amid the noise: but if the upside-down lake sediment record is flipped right-side up AND the questionable tree records are removed –> does the new summation show a warming about 100-150 years ago which is different from the cooling expected via precessional forcing???
Re: bent-out-of-shape (#197),
As I’ve said on several previous occasions, at the time of our PNAS comment, we hadn’t got the RegEM algorithm to work in order to do a sensitivity study. In addition, before doing a such sensitivity study, I want to understand a little more clearly the phenomenon of the “too-perfect” RegEM reconstructions in the calibration period. If Mannian RegEM is effectively splicing the instrumental record in the calibration period – a distinct possibility – then one would take the analysis one way; if it isn’t, then you’d do it another way. Also before I release an analysis, there will be a lot of crosschecking that needs to be done. So it will take a little while before I do this – it will be after Briffa.
Perhaps you should email Mann and ask him. My overriding position is that it is the responsibility of the originating authors to correct their own errors. Mann was able to re-do the erroneous Rain in Spain quickly enough and could do a calculation of the AD800 network with NO Tiljander easily enough. But I will get to it sometime in the next month or two.
Re: Steve McIntyre (#199),
“Always leave the audience wanting more.” You’re killing us. Ha Ha.
Thanks Steve and everyone who contributes for all of this work. Plodding through code and reams of data is no easy task. And it gets confusing with all the errors. Heck youv’e got Urban Heat Island effect, Antarctic “warming”, Briffa, Strip Bark bristle cones, Tiljander upside down proxies, and Mannomatic statistics. I’m probably missing a whole bunch. Oh yeah, there’s Andy Baker SU967. Forgive our impatience, but there is always a thirst for justice, as they say.
So I’m putting a little in the tip jar for you. Thanks again for your important work.
Re: Steve McIntyre (#199),
I thought Kaufman had all of the data available on-line? Do you need to know about RegEM to re-compute the average based on those standardized proxies? I must have missed something…
Re: bent-out-of-shape (#201),
I’ve already done first-cut sensitivity analyses on Kaufman in another post. As I’d mentioned in my very first post on Kaufman, upside down Tiljander in Kaufman is more a matter of craftsmanship; it’s Yamal and proxy picking that is the issue there. But the query was about the impact on Mann et al 2008, which requires analysis of the RegEM methodology.
Re: Steve McIntyre (#211),
“craftsmanship”? I have no clue what it is that means. My query was about Kaufman et al. 2009. If the lake record was flipped the other way around and Yamal is thrown out – such that your issues are addressed – what is the effect on the average? The proxeis are all in the on-line material. I never mentioned Mann or RegEM. You did. Why not just show what the average looks like w/o those proxies? You have spent quite a lot of time, bandwidth, and energy on these issues…
why not audit the climate signal now that you have addressed the proxies?
Steve: See Sep 15 post here here for a preliminary assessment of Kaufman using upside-up Tiljander and using non-Briffa tree rings: Grudd’s Tornetrask, Esper’s Polar Urals and Moberg’s Indirka River. This needs to be updated since only 2 of 3 Finnish series need to be inverted, but the revised version will look a lot like this. I apologize for not being on call 24/7 to do every analysis that readers want within minutes of the request. If you want faster service, you are always welcome to ask the Team.
Re: bent-out-of-shape (#214),
You need to be clearer in your comments. In #197, you say:
I (and apparently Steve) took it that you were referring to Mann 2008. This seems perfectly logical as nearly the entire original posting and almost all the comments have been on Mann. Based on your latest posting, it appears you were actually referring to Kaufman and you use some rather undiplomatic language in response. Please don’t blame others for ambiguity you caused through sloppy comments.
At Stoat | October 30, 2009 4:50 PM
Separately, Rattus Norvegicus is the first to address the five claims I listed in Stoat’s comment #1.
At Stoat | October 30, 2009 5:13 PM
Re: AMac (#202),
For the record, AMac, your five points were factually correct. I forgot to mention that before leaving.
Re: theduke (#209),
I wondered if you would say that and I’m glad you did. 🙂
Re: bender (#210),
Unlike you, bender, most people are not immediately cognizant of mistakes they’ve made. Some people need to be taught, which is why you do what you do, no?
This is all very enlightening (and at times worrying) but I can’t help thinking (as an ignorant lurker): you’ve got (most of) the data, you’ve got the algorithmic ability, now do the science.
Now *that* should impress (and maybe even worry) “The Team”!!! And would be much more impressive! (OK, OK, I know that some of you are working on it, but that is not what 99% of this discussion is about. In my field, you get most kudos for what you achieve yourself, not what you complain about.)
Re: Frazzled (#203),
Please see Steve’s comments on that subject at the following link:
http://dotearth.blogs.nytimes.com/2009/10/05/climate-auditor-challenged-to-do-climate-science/
I would suggest to you that he is doing the science– and doing the science a big favor by showing the powers that be the sorry state of their data and statistical methods.
Only in climate science would you have a discussion that devolves into this kind of twisted dialogue.
When I need a dose of sanity I go over to WUWT and watch Dr. Svalgaard handle questions with
humor, directness, and no BS. I get the same dose of sanity watching Roman work. but watching gavin, connelly
and briffa makes me feel like I’m in some Monty python skit. except they are serious. Is there a cultural thingy
going on here? dunno just asking.
Re: steven mosher (#204),
Here here!
Re: Mad Hominem (#205),
I heartily agree, but must point out that the tribute is spelled “Hear! Hear!.”
In the sense that the “meatgrinder” always gives them the “right” answer, this is true.
Regards, BRK
Never has a little mud been discussed with such intensity.
What I find interesting is that great athletes make mistakes, as do generals, heads of corporations, and presidents. But the Team can not admit a mistake? (except Kaufmann, though his correction is not quite complete yet).
I think I have a tm on the zamboni metaphor as well. I need to check on that one and add it to my growing list of climate science constributions.
Why not just write a paper about this? If the statistical analysis is in fact sound, it could be accepted for publication somewhere, and the argument about credibility would be settled. Or is the continuation of the argument the object of all this?
Re: Simon D (#215), That attitude is part of the problem.
manns paper was published. steve took the “institutional” route to fix it. submit a comment. he was given 250 words
and mann blew him off. Because mann knows that he could. he knows that no journal will accept a piece that merely points out an error. And IF per chance it is allowed, mann will get to review it. and if it gets published a year or more will have passed and they will move on. oh, sorry I’m being cynical and attributing motives. i take it back. Steve should publish. wait, he did already. on this blog.
Re: steven mosher (#217), Perhaps saying “one year or more will have passed” is too optimistic. Here is a story of how an error (violation of energy conservation law) in a biological model originally published in Nature took SIX years to be admitted by the authors via the “institutional route”. And which bizarre forms that took — the model’s authors re-published their slightly corrected model in Science admitting the original model was wrong but not giving the credits to the authors of the critique for pinpointing the error.
hey all,
http://www.examiner.com/x-9111-SF-Environmental-Policy-Examiner~y2009m10d30-Examinercoms-First-Annual-Survey-on-Global-Warming
Tom just put this up. its worth your time and you get a chance to say who you trust on climate issues
Steve:
I am just a old country electrical engineer and programmer (with a degree in geology). I have relatively little experience with high powered statistics, but I have a very good understanding of basic science, data, and logic. I have looked at my fair share of data and graphs over my career !!!
Steve, your original explanation made complete sense to me. If Mann et al could not take your hint, shame on them !!! If their arrogance did not cause them to pause and think (and seek advice), then shame on them. I have been studying the AWG arguments for about a year, and I am astounded at the total arrogance of the RealClimate zealots (particularly with regard to you).
My favorite engineering mentor told us years ago to check the magnitude of our answers, and to check the units. Obviously checking the sign is a pretty good idea also.
So someone suggests using varves as a temperature proxy. Any good scientist would ask, “so what can I measure about varves that might correlate with temperature”. Hopefully one would come up with a quantifiable functional relationship based on physics or chemistry, not just a correlation. But hey we are talking about climate “science” here.
So how did Mann miss the fact that the measurements of the lake sediments suggested that the climate is getting colder in the 20th century. In other words, how did he miss the fact that the 20th century data is crap. It is obvious that Mann did not think too hard. Some folks choose a bigger hammer, he chose statistics. Mann’s arrogance, his excessive reliance on statistical methods rather than on science, and his confirmation bias led him in the wrong direction. Mann’s fundamental belief in the hockey stick graph (ie his confirmation bias) allowed him to go stupid. I am willing to bet that he flipped the graph upside down in his mind (and in his paper) because it confirmed his prejudice.
Had he been a competent scientist, he would have been skeptical (nice concept, being skeptical), and might have contacted Tiljander for a little insight..
It is very scary to think that someone would just stick data into a regression machine without any thought as to the validity of the data, and without any thought as to the sign of the data, or the magnitude of the data.
Yes, I know that regression methods can properly discern the sign of VALID temperature proxy data (as many, many, many Mann defenders have said over and over ad nauseum without pausing to think). I wish that I had a dollar for every time one of Mann’s defenders wrote: “”Multivariate regression methods are insensitive to the sign of predictors”.
They followed him off into a world where down is up and up is down (think lemmings),
Ironically the arrogance of the “peer review” crowd prevented them from seeing simple and obvious truths, even after you hinted and then flat out told them.
This incident is a complete indictment of the peer review process, and a complete vindication of the blogosphere.
In my opinion, your most valuable contribution has been to demonstrate the arrogance of certain academics and certain academic practices, and to totally demonstrate the power of an informed and intelligent community of bloggers.
This is a non sequitur, but my first job after college was analyzing ice cores from Greenland and Antarctica. I might have been famous if I had stuck with it!
Ernie in Utah
Re: Ernie (#219),
I am not a statistical whiz either but even I can grasp what Mann and Connelley cannot (or won’t) admit. I can’t tell why they won’t admit it and can only surmise its caused by self-deception and egotism.
Supposing I produced a dataset showing prices of baked beans during the 19th and 20th Century. A Mannian analysis could pick up that the price of baked beans as a proxy for global temperature by analyzing how closely the rise (or fall) of baked beans correlates with that index
But here’s the kicker. There is no causal relationship between global temperature and the price of baked beans, but the Mann procedure only measures correlation during the calibration period not whether the proxy is meaningful. This is what Mann is telling us by repeating the mantra about “multivariate analysis being insensitive to the sign of the predictand” and what Steve is hammering home is that if that’s so, the Mannian algorithm is virtually guaranteed to produce a spurious correlation between unrelated proxies and temperature.
If the method is insensitive to the sign of a proxy which very definitely DOES have an orientation then the method is spurious and the reconstruction meaningless. This is what Ross McKitrick wrote about in Comment 173:
In other words, meaningless garbage is the only result of a model that is “insensitive to the sign of the predictands”
Why are they unable to understand this? I can understand it. Most people here can understand it. Mann and Connelley can’t admit it lest their credibility (ho ho) is undermined. So they keep digging themselves deeper into the hole and anyone who joins them is guaranteed to get buried. Connelley gets caught out so many times and what he does is bury the evidence and disappear. You watch his blog – suddenly he’ll “tire” of talking about it and its YOUR fault for being so obsessive about trivia.
John A – excellent analogy! Cheap beans – increased methane levels – global warming! QED: Another valuable addition to the proxy palette 🙂
Also, up thread in the pic. you put in 154 – please can anybody point towards a good web ref. on how they measure those non concentric/circular rings to an accuracy down to 0.001 mm? On another thread a while ago someone referenced a standard text book on dendro. but it is was an expensive publication. Any links appreciated.
I do not recall walking through a mirror but I seem to be in a looking glass world. I know it is true that the sign of an individual vector does not effect the produced PCs. That is in the nature of a PC. The PCs represent the key “flavour scales” of the data and the identification of these key scales is not sensitive to sign of particular vectors. But the scalar (EOF) associated with a particular vector is. Tiljander and anti-Tiljander have the same flavour shape but different sign and it is only the shape that is retained in the PC. So the claim that it makes no difference to the PCs is true but this is nothing to boast about as it is a warning that PCs are only an aspect of the picture. If I were to analyse snacks I might find a PC that more or less gave a sweet-savoury scale. Lets say sweet is positive and savoury is anti-sweet. But if I (actually I mean they) just took the shape of the PC I might say I had identified a sweet component in the data, or I might say I had identified a savoury component. Well I (actually they) have done neither. All that has been identified is a key discriminator in the data. That is that the sweet-savoury scale can be used to explain a lot of the variance in the data. I simply do not get the justification for using a PC as evidence of anything without analysis of the corresponding EOFs. For instance if I have an equal mix of x and anti-x, my PC will be the shape of x or anti-x. But if I grind it all up the result is null. I simply do not get some part of the way people seem to be able to use PCs (which are data about data) as data.
Lastly I am really missing something in the use of the term multivariate, I thought this meant regressing data onto a number of selected independent vectors. All I can see is a lot of data and only one vector (temperature). Now hopefully I have got this terribly wrong. But I am in a looking glass world and I just cannot see it.
Alex
Re my own above:
The white rabbit just turned up and put me right. Apparently what I have to be is totally objective. I must not let any of my subjective baggage get in the way. I have made a subjective call in that I assumed that one selected vectors for multivariate regression on the basis that they were plausible causes of the effect. This is a no-no, one must be completely objective. I could not see why one would wish to regress temperature onto data like a varve sequence as there is no way that the varve sequence is going to be a cause temperature changes. I was being criminally subjective. What I should do is forget science (or at least any scientific basis for choice) as it is totally subjective. I simply need to do the mathematics. My idea that one should regress the proxy data onto as many plausible causes of the data, temp, CO2, rainfall, lunar phase, and see what are and are not the most likely causes. What I should do is take a plausible cause (temperature) and see how much of an effect can be found in the data. Then I must objectively state that the concept of cause and effect is subjective, I should then be able to use the results as if made no difference as to which way the regression was performed, or rather that to argue that it did make a difference is purely subjective and hence a no-no. Unfortunately the white rabbit is not here now and I am back to simply not getting it.
Alex
Steve: Regression of cause onto effect is embedded pretty deeply into Team methodology. We saw that recently in an interesting case with Rob Wilson’s calculation of Yamal and Polar Urals Durbin-Watson statistics. Regression of effect onto cause gave a RW statistic way into the problem zone, while the inverse regression of cause onto effect didn’t. Wilson regarded the issue as merely one of preference. It looks possible to me that RegEM, a technique developed for infilling patterned measurements (e.g. temperatures) is a total conflation of cause and effect when extended to “proxies”, made worse by splicing the effect in the calibration period with the reconstruction in the reconstruction period (tho I’m not totally sure of this interpretation.)
Alot of people are hung up on the statement, “”multivariate analysis is insensitive to the sign of the predictor”. I think Mann is saying something simple: If one takes a positive time series that is positively correlated with a dependent variable and then changes the sign of the entire time series to make it a negative time series, regression analysis will now find the “negative” series is negatively correlated with the dependent variable. Because the two negatives cancel out, the regression will come up with same answer regardless of whether one enters the predictor time series with positive or negative values.
This is mathematically true, but is simply not responsive to Steve’s criticism. SM’s point is not that Mann placed a negative sign in front of the entire proxy and that is not what he meant by saying the time-series is upside down.
Steve: It is a typical Team debating strategy – to rebut something that I didn’t say. In the very first paragraph of my very first post on this topic, I clearly and unambiguously distinguished between the two situations in terms then highly topical:
Now the analogies and language may be a little unfamiliar to academics, but they are bang on.
Brian B #190
Great observation. At this point, though, I would say that we have him by the throat.
At Stoat — Posted by: AMac | October 31, 2009 9:08 AM
I seem to be at the bottom of a rabbit hole and still digging myself in deeper. There is obviously something I don’t understand about all this. Something that about it that goes against my grain.
If I was trying to do this using PCA or MVR I would include lots of historic data series that hopefully had nothing to do with temperature. I would then try to identify real measured series like temp, rainfall, humidity, etc., with the PCs produced, or regress all the data on the on those same measured series. I would do this to explain as much as possible of the variance (increase the R^2). Having done this I would be able to build a reversible relationship between the data and the causes. It is not good enough to say that during the period of the temperature record a particular group of trees, varves, etc. correlate well with the temperature record. I need to know how well trees, varves etc., correlate with the temperature record in general. I need all the occasions when they correlate badly as well as the ones where they correlate well to be included in the analysis. I might get a low R^2 but that would simply mean that I have more explaining (of variance to do) before I can come up with a robust result.
Now here I need some assistance, I think that if a regression gives RW(ringwidth) = A +B*T(temp) with R^2 = C, then I would also find T = D + E*RW where E*B = C for a simple bivariate regression. Obviously I want E =~ 1/B (R^2 =~1), if I want to use RW to predict T robustly. I can improve R^2 by adding more measured time series (rainfall, etc) and perform MVR and if I also include a lot of data that correlates well with rainfall but poorly with temp, etc. I can tease a inferences about rainfall, and temp, etc. from the data. Am I just nuts?
Alex
Re: Alexander Harvey (#229),
The figure at this Steve Hsu blog post is on a completely different subject (European Genetic Substructure), but it might be useful to consider.
The diagram is the result of a principal component analysis (PCA) of Single Nucleotide Polymorphisms (SNPs) in the genomes of a couple hundred Europeans of defined and known ethnicities. The investigators made no a priori assumptions about any one of the ~100,000 SNPs that they sifted through. The PCA algorithm selected the subset of informative SNPs, and also assigned the “sign” of each to the principal component. I.e., for an informative SNP, either “A” or “T”, or “C” or “G” might make the component bigger (or smaller) for that given individual’s genome.
The reasoning is that if this particular implementation of the PCA approach produces a useful picture, then therefore the dataset and the methodology are validated.
You can see in the figure that (1) this 2-dimensional PCA does indeed distinguish many European ethnic groups, one from another, and (2) As plotted, the figure approximates a map of Europe, with geographically closer ethnic groups also closer in the PCA plot.
It seems to me that Mann et al used an analogous approach in screening and then in using some paleoclimate proxies.
A question would be the extent to which such an approach is justified in the case of temperature anomaly reconstruction. A “pure” implementation would necessarily involve “ignoring” the physical meaning of the selected proxies. Per the SI pg. 2, Mann et al claim that they didn’t ignore such meanings in the cases where they believed that orientation could be strongly assigned.
Upthread, Steve McIntyre, Ross McKitrick, and others have discussed this general point (in other ways), I think.
[duh] Negative emotions seem to run high in the discussion of Mann’s use of the Lake Korttajarvi varves. I’ve noticed this here, at Stoat, and at Roger Pielke’s blog. More generally, that’s also true at RealClimate.org (I haven’t read a Tiljander-focused thread there). [/duh]
I’m reposting my relevant Stoat comments in this thread, as long as host Steve McIntyre doesn’t disapprove. That’s not to show I share in these negative feelings–my own are much closer to curiosity than anger. It’s done “for the record,” since WM Connolley responded to my concern that the thread at Stoat might be Bowdlerized by comment deletion. He said, “Thats just tough I’m afraid“. To be clear, WMC hasn’t “unjustly” edited or deleted and of my comments at Stoat. But I’m made uneasy by his answer, and by his trimming of this insightful and on-point comment by TAG.
Re: AMac (#230),
Unfortunately AMac, Connelley is an admin at Wikipedia and is a past master at bowdlerizing history through selective deletion. He deletes your comments when he finds (rather quickly) that he can’t answer them without looking like an idiot.
As for the integrity of the historical record, Connelley is shoulder-to-shoulder with INGSOC: “Who controls the past controls the future; who controls the present controls the past”
Re: John A (#232), Was an admin, although that doesn’t stop him from editing it does stop him from banning other people from editing just because he disagrees with them. The nature of his discussions on there makes me disinclined to waste my time with him.
Re: Sean Houlihane (#236),
Its a relief that he’s no longer an admin on Wikipedia. Does this mean that his extensive “Biography of a Nobody” can be edited down to the correct size, and relevant biographical information about his other activities be added? Or has someone stepped up to prevent that?
Re: John A (#246), WC is still editing pages. James Watson and Stephen Schulz canceled my edits to Richard Lindzen’s page.
Copy of my comment at stoat:
dhogaza,
Since as you affirm…
…wouldn’t it be most appropriate to select trees and series based on said physiology, and include the complete data set from said trees and collections? And then to deal with the data that results without removing data sets based on the data itself?
Likewise for varves and other proxies?
I know of no other area of scientific inquiry where the data itself (rather than metadata) is a parameter to selection criteria.
Re:John A #233…having read all of this (Upside down Mann) both on CA and Connelley’s blog I agree with you completely…INGSOC is what comes to mind when you go to Stoat…if he doesn’t like it, it is gone…oh dear oh dear oh dear
Science has taken William Connolley by the brachial plexus, and is relaxing the mailed fist.
===============================
Just had an email from stoat saying my comment had been ” Cut. Needs refs.” So here is what I wrote
dhogaza,
> You say
Researchers understand we have a rich body of knowledge regarding tree
> physiology which enables them to, with some (though not absolute)
> confidence, differentiate trees that respond to temperature vs. other
> limiting growth factors.
>
> I see no evidence of this at all in multi proxy reconstructions. A
> small number of proxies do most of the work e.g the Graybill pines
> which were said by teh original collectors to exhibit sensitivity to
> CO2. It was Mann who first proposed that they were teleconnected with
> global climate. From where I sit it just looks like opportunistic
> correlation picking.
>
> And on what theory did the lake varves get in?
At Stoat— Posted by: AMac | October 31, 2009 2:50 PM
Re: AMac (#240),
I have a contemporary copy of the Mann PNAS SI (downloaded Sep 2, 2008) and it contains the quote in Connolley’s head post.
could anybody comment, if if got the tiljander discussion right ?
1. tiljander proxy is heavily contaminated in recent history and at least this part should not be used in a temperature recontruction.
2. multiplying upside down tiljander by -1 falsely also flips the reasonably correct parts (medial warm period and little ice age) and thus falsely reduces temperatures in the mwp and increaes temperature in the lia.
3. there is no physical law that justifies to multiply a temperature curve with a negative correlation. this is actually opposite to physical laws.
4. in mann’s code proxies may change their flip signs several times during computation.
5. adding a proxie may cause other proxies to change their flip sign.
6. using only proxies with a screened pattern over a recent time window produces hockeysticks with random data a well.
7. using only proxies with a screened pattern over a recent time window therefore requires error bars to be stretched (roughly spoken) to infinity.
8. using only proxies with a screened pattern over a recent time window
and applying sign checking increaes error bars even further.
Re: Manfred (#241), 2 is not quite the position. Mann did not initially multiply Korttajärvi varves (tiljander) by -1 in his input (and indeed doing so would have made no difference).
What happened (as I understand it) is that the human activity contamination effectively gave it a coefficient with the oppositie sign to Tiljander’s physical interpretation in one of Mann’s two methods (CPS) and Mann then used that in a reconstruction, so to the extent it had any substantial impact it was clearly wrong.
On 4, the flipping sign several times is what Jean S identified had happened when looking at Mann’s other method (EIV).
One question, assuming you rerun the sensitivity analysis without Tiljander, is it possible that figure S7 remains unchanged? I’m thinking that the Mannomatic software will just zero in on any remaining proxies with an uptick in the 20th century and increase the weights on those remaining proxies. I believe this is related to how Mann does his standardization, correct?
What is alarming is that Figure S8 has basically no change at all when they exclude 7 proxies with very pronounced “blades”. I would have expected some change, but there is nothing. That is not right. However, there probably was a change, a change to the weighting of the proxies. This whole thing is a mess.
I may be missing the point – some of the arguments are too clever for me to grasp properly. Someone can correct me if necessary …
Surely, the issue is not whether the Tiljander sequence is upside down or right way up. The issue to me is whether or not the sequence is a proxy for temperature. If it is not then it should not be used as a temperature proxy.
As far as I can see, the series could be a proxy for temperature up to relatively recently while it was unaffected by human factors. The original authors seem to think so. They have put forward a case for why the proxy should have the sign they give it: their “Right Way Up”. For the time periods when the sequence has internal evidence that it is uncontaminated by non-temperature factors it could be used as a temperature proxy. For periods where the series can be shown or is suspected to be affected by other factors than temperature it cannot and should not be used as a temperature proxy.
The original authors have been honest with presenting all their data, even when they have evidence that the modern part of the series is contaminated by human actions. If the series is to be used as a temperature proxy it should be cropped at the point where there is evidence that the series is not a temperature proxy. Upside-Down or Right-Way-Up doesn’t come into it until the contaminated portion has been removed.
What have I missed?
Re: Alan Bates (#244), Tiljander is BOTH upside down AND the recent record is disturbed by human development locally.
So, firstly, it should be turned right side up, as Tiljander himself has repeatedly requested Mann, Connolly, Briffa, Kaufman, et al do and which they and the rest of the Team have steadfastly refused to do, refusing to acknowledge even that the data is upside down (until Kaufman’s recent acknowledgement, which the others are not admitting to), AND their presentation of Tiljander should truncate the recent bad data off BECAUSE the recent bad data due to human development sedimentation is what constitutes the entirety of the blade of the hockey stick. Tiljander has stated this, but the Team has refused to comply for obvious reasons.
is it true that if the Tiljander was NOT interpreted inversely as the initial authors intended OR was not ‘contaminated’ with the 20th century artificial ‘signal’, that it would not have been retained as a reliable proxy due to non-agreement in the correlation period? (i.e., its the occurrence of both these, uh, problems, that lead to its inclusion anyway(?)
thanks
Re: Steve Geiger (#247), Yes because the modern contamination was what caused the large changes in the modern period thus producing the hockey stick uptick – which would have been a down tick if climatologists new their ups from his downs
At Stoat — Posted by: AMac | October 31, 2009 6:55 PM
Nice piece on regression abuse at climate skeptic
http://www.climate-skeptic.com/2009/10/regression-abuse.html#comments
which gives a claer explanation of the issues. Though quite why some people are finding it so difficult to grasp I fail to understand.
WC trimmed my comment. My response, which this time is being held for approval rather than allowed through… (readers, note his technique appears to be: cut off any substantial reply, completely removing it from his records! We’ll see if something substantial is actually allowed in.)
On Ben Hale’s blog bender opined that it was creepy the way Halpern always referred to himself in the third person. I agree. Now we read this over at Stoat:
This is creepy. Mann or nothing! Mann uber alles! Mann at all costs! These guys will sell their souls defending the guy.
Re Tom C #252, it is getting to be impossible to parody these guys any more. Are there no depths to which they WON’T sink in their defense of the indefensible ?
Figure 1. A team of bats reflecting on multivariate regression methods.
Re: Steve McIntyre (#254),
I like it! But surely Mannian bats would sleep ‘upside down’, i.e., the right way up?
Re: Jimmy Haigh (#262),
An excellent point. Dunno.
Re: Jimmy Haigh (#262),
As an observer of fruit bats in my youth (when we used to shoot them by the 100s with .22 rifles) I can confirm that bats copulate inverted (a position not described in the Karma Sutra) but that urination and defaction are not pretty sights. They can also die inverted, hanging on with hooks. Indeed, when mortally shot, they seldom fall until days later.
The ground below a bat colony is fruitful for lurking snakes, who tidy up the ones that fall down after becoming over zealous or drunk from fermented fruit. Also, a bat colony smells quite putrid. At least when they dribble while eating fruit, the dribble drops to the ground. This avoids the need to clean stomach fur of dribble but not of urine.
On a quality post like CA, please keep these observations in mind should you be tempted to be reminded of a particular person who fits this unsavoury and ugly description. Polite behaviour reigns, ok?
Re: Geoff Sherrington (#311),
Top quality timing, Geoff – right on Aussie tea-time 🙂
BTW, I was never a good-enough shot with a 22 … preferred an AK47
Re: Geoff Sherrington (#311),
They arn’t dead, they are pining for the fjords.
Re: steven mosher (#327),
Pining for the fjords? This is an ex-parrot!
Re: steven mosher (#327),
Quick and satisfying reference, once I realized you were spelling “fjiord” wrong.
Can you send us more nails please? Geoff.
Tune in tomorrow …. same bat time, same bat channel for another exciting adventure of Bat Mann !
Re: Artifex (#255),
After that, I vote this thread duly closed
By the way I’ll never understand why anyone bothers to comment on Connelley’s blog. Any attempt to get Connelley to admit any failure is just an exercise in masochism on the part of the commenter – and Connelley could care less.
Re: John A (#257),
It was interesting how he was in denial up to the point where I gave him the answer, over at CM. Armed with the facts, denial was no longer a reasonable option, so he had to switch to denigration. But to do that he needed editoiral control, so he took the traffic away from CM and over to Stoat, where he could censor without reprisal.
.
Why comment at Stoat? (1) To force Connolley to censor. As long as people double-post here at CA, let him go nuts. (2) To make zealots like dhogoza say idiotic things like “CA worshippers so rarely venture out of their fantasy world”.
Re: John A (#257), Actually I’d say he couldn’t care less. Looks like you inverted the sign there…
Re: Jeff Alberts (#287), LoL! You know when Americans say “could care less” then actually mean “couldn’t care less”! I’ve never understood why, and I’ve never managed to get anyone to explain it satisfactorarily
– but you’re right it could have some underlying significance pertaining to the understanding of the meaning of a negative vs positive value of a data set! :o)
Re: Phil M (#289), Hehe. I think it’s more because Americans on average aren’t very concerned whether what they’re saying is technically correct. “Couldn’t care less” is a somewhat odd phrase, perhaps a bit archaic. But saying “Could care less” means you care to some degree. This seems lost on most people, probably for the same reason they don’t know why saying “nucyuler” is incorrect.
If Mann’s method is insensitive to the sign of the proxy them how could it determine if Tiljander had it upside down or not?
Re: TAG (#258),
Mann’s method IS “insensitive to the sign of the proxy” … ASSUMING the “proxy” does not switch senses between the calibration phase and the reconstruction phase. Pea and thimble, you see? Steve’s PNAS letter must have seemed “bizarre” to somebody who could not imagine a proxy switching sense at such a pivotal point in the analysis.
NOTICE:
Whilst not having an opinion on Tiljander proxies, Wikipedia needs help. Those who have time to post on blogs may also have time to assist in turning Wikipedia into an unbiased source of information. Whilst it is not for the faint of heart, with more editors we can achieve this together.
Thanks,
Alex Harvey
Re: Alex Harvey (#261),
I used to edit Wikipedia but Connolley and his minions made the experience almost unbearable. Although I may go back if there are reasonable people as yourself.
WC’s latest is that Tiljander after 1800 is useless, therefore any argument of upside-down usage is invalid.
Re: MikeN (#264),
So is Connolley going to encourage Mann to publish a Corrigendum?
At Stoat | November 1, 2009 10:31 AM
A detailed response to WMC’s response to my Five Claims in Comment One will follow shortly. Due to its length, it may get caught in Stoat‘s spam filter. Here is the summary version.
WMC does not explicitly and unconditionally agree with any of the five claims. Neither does he offer any data or arguments that would weaken any of them.
Claim 1 — Requested documentation supplied.
Claim 2 — Requested documentation supplied.
Claim 3 — Requested documentation supplied.
Claim 4 — On reading Tiljander et al (2003) (linked in Comment 2), Claim 4 is revised to read,
(4.) That the varve proxy was calibrated by Mann et al over a period (1850-1995) in which higher regional or worldwide temperatures were correlated to higher X-Ray Densiy (xraydenseave), to higher mineral matter (lightsum), and to greater thickness (thicknessmm). These correlations are upside-down with respect to those assigned by Tiljander. Mann et al assigned a correlation of higher regional temperatures to higher organic matter (darksum), which is consistent with Tiljander’s interpretation. In any case, Tiljander’s Figure 2 strongly suggests that the instrumental temperature record at stations near Lake Korttajarvi show little or no upward or downward trend, 1881-1993.
Claim 5 — WMC’s response is largely opaque to me.
AMac’s Conclusion: The Lake Korttajarvi varve proxy series were incorrectly used in Mann et al (2008). Three of the four were used upside-down with respect to the original authors’ understanding of their meaning. None of the four show a strong and clear-cut correlation to local temperatures 1881-1993, a complication unmentioned in Mann et al’s text or SI.
At Stoat | November 1, 2009 10:42 AM
WMC —
You presented your views on the five claims as bracketed text inserted in Comment 1, at some point this weekend (I think). Thanks; I hadn’t noticed them till just now.
(1.) That Tiljander believes that the climate signal in the Lake Korttajarvi varve dataset is that higher local temperatures correlate to thinner, more-organic-rich, lower-XRD varves.
AMac responds: Tiljander et al (2003) (ref. 12 of Mann et al SI) is linked in Comment 2. On the “sign” of the {organic/mineral :: temperature} correlation, Tiljander writes on page 571, “Periods rich in organic matter indicate favourable climate conditions, when less snow accumulates in winter by diminished precipitation and/or increased thawing, causing weaker spring flow and formation of a thin mineral layer. In addition, a long growing season thickens the organic matter. More severe climate conditions occur with higher winter precipitation, a longer cold period and rapid melting at spring, shown as thicker mineral matter within a varve. … short-term changes (averaged over a few years) could be estimated.”
page 573: “An organic rich period from AD 980 to 1250 in the Lake Korttajarvi record is chronologically comparable with the well-known ‘Medieval Warm Period’ … The relative lack of mineral matter accumulation and high proportion of organic material between AD 950 and 1200 was also noticed in two varved lakes in eastern Finland (ref) as well as in varves of Lake Nautajarvi in central Finland c. AD 1000–1200 (ref). Common to all sites is that the warm period lasted more than 150 years.”
page 574: “According to the Lake Korttajarvi varve record there is a short period, AD 1115–1145, with increased mineral matter accumulation, indicating more severe winters. In the Fennoscandian tree-ring record (Briffa et al. 1990), the largest 50-year cooling trend in the 1400-year long record (-1.78°C) occurred between AD 1090 and 1139 and the coldest 20-year mean occurred between AD 1127 and 1146.”
But note (page 574): “Even though the sedimentation in Lake Korttajarvi most likely reflects relatively long-term changes in local hydrology rather than temperature, …”
“[Tiljander] didn’t do the calibration” is true but irrelevant to Claim 1. Page 572, “In the case of Lake Korttajarvi it is a demanding task to calibrate the physical varve data we have collected against meteorological data, because human impacts have distorted the natural signal to varying extents during the past 280 years and the meteorological data in the Juvaskyla area are only available since 1881.”
(2.) That Tiljander cautioned that after ~1720, the Lake Korttajarvi varve dataset is likely affected by local-human-activity signals, leading to thicker, more-mineral-rich, higher-XRD varves than climate alone would produce.
AMac responds: In the Summary, Tiljander wrote (page 575): “In the Lake Korttajarvi sediment sequence, the effect of human impact has increased since the mid-18th century and has obscured the signal of natural climate variability.”
Supporting quotes from earlier in the paper follow. Page 571, “The thinnest varves accumulated from AD 900–1800 (average 0.68 mm), and during the last 200 years the average varve thickness has been 1.60 mm. This recent increase in thickness is due to the clay-rich varves caused by intensive cultivation in the late 20th century.” Also quote from page 572 under Claim 1.
(3.) That Tiljander has described two incidents of local activity that led to very thick, mineral-rich, high-XRD varves, in 1930 (peat ditching) and 1967 (bridge reconstruction).
AMac responds: Tiljander page 575: “There are two exceptionally thick clay-silt layers caused by man. The thick layer of AD 1930 resulted from peat ditching and forest clearance (information from a local farmer in 1999) and the thick layer of AD 1967 originated due to the rebuilding of the bridge in the vicinity of the lake’s southern corner (information from the Finnish Road Administration).”
(4.) That the varve proxy was calibrated by Mann et al over a period (1850-1995) in which higher local temperatures were correlated to thicker, more-mineral-rich, higher-XRD varves.
AMac responds: Being inspired by Rattus Norvegicus’ careful answers, I’ve (obviously) reviewed Tlijander et al (2003); this causes me to revise Claim 4, as follows:
(4.) That the varve proxy was calibrated by Mann et al over a period (1850-1995) in which higher regional or worldwide temperatures were correlated to higher X-Ray Densiy (xraydenseave), to higher mineral matter (lightsum), and to greater thickness (thicknessmm). These correlations are upside-down with respect to those assigned by Tiljander. Mann et al assigned a correlation of higher regional temperatures to higher organic matter (darksum), which is consistent with Tiljander’s interpretation. In any case, Tiljander’s Figure 2 strongly suggests that the instrumental temperature record at stations near Lake Korttajarvi show little or no upward or downward trend, 1881-1993.
Statistics are not needed to address this point. Most readers possess at least one functional Mark I Eyeball. Application of this tool to the four Tiljander proxy traces in Figure S9 and to Tiljander Figures 2 and 9 (lightsum and darksum) will suffice.
(5.) That Mann et al used the Lake Korttajarvi varve proxy in the reconstruction of the Temperature Anomaly record by applying the 1850-1995 correlations (thicker, more-mineral-rich, higher-XRD varves with higher temperatures) to the varve record spanning 200 AD to 1850 AD.
AMac responds: The meaning of WMC’s remarks on Claim 5 are opaque to me. They don’t seem to call for any further documentation. However, I add (per Claim 4 above) that although Mann et al assigned reversed-sign orientation to the {varve :: temperature} correlation for three varve characteristics, their algorithm picked the rightside-up orientation for darksum (organic matter).
The implications of the Five Claims on the interpretation of Figure S8 I leave for another day, except to say that this figure now appears trivial at best and misleading to the reader at worst. (“Misleading” if Mann et al meant to use it to show the robustness of their proxy selection process, or of the results of that process.)
– – – – – – – – – –
AMac’s Conclusion: The Lake Korttajarvi varve proxy series were incorrectly used in Mann et al (2008). Three of the four were used upside-down with respect to the original authors’ understanding of their meaning. None of the four show a strong and clear-cut correlation to local temperatures 1881-1993, a complication unmentioned in Mann et al’s text or SI.
Re: AMac (#270),
Connolley is being a weasel. He is forcing his commenters to recapitulate McIntyre’s argument, but from primary sources. The other way to understand McIntyre’s argument, of course, is to read his explanation for yourself, as he documents all the necessary primary sources here at CA. Congratulations, though, to AMac for running the weasel gauntlet.
.
I hereby declare that Tiljander was indeed used “upside-down” in Mann’s reconstruction. If someone could now direct the weasel’s attention to Sesame Street, he might there find a suitable demonstration of what precisely “upside-down” means. (Can we drop the quotes now?)
Re: bender (#271),
The only items penned by McIntyre on the subjects of the varve proxies and their treatment in Mann et al (2008) that I have read are (1) his PNAS Comment, (2) a few comments on the two relevant theads at Roger Pielke’s blog, and (3) his brief comments in this thread.
I claim no expertise in this area, because I have none. I have offered no opinion on the broader questions related to AGW and paleoclimate reconsruction: I’m not sure, though I tend to think that (a) worldwide, temperatures are increasing on a scale of decades, and (b) much of this trend may have human causes.
But I could be wrong. And magnitudes as well as signs matter… and are the source of some controversy. Duh.
All of this, of course, is neither here not there as regards Mann et al’s use of the Lake Korttajarvi proxies. To address that, I looked at the arguments and read some of the relevant literature, as a lay person.
To the extent that my reasoning and conclusions in the two comments supra are wrong: Hopefully, those arguments will be made at Stoat or here.
Re: AMac (#273),
AMac, I started into this in a very similar way to you. Incidents like the present one remain baffling to me. The handling is clearly screwed up in Mann et al. Anyone not blinded by loyalty can see that. It’s crazy for Connolley to support Mann on such an unwinnable issue.
Re: Steve McIntyre (#273),
“Anyone not blinded by loyalty can see that.”
– I think that is about the size of it!
– love is blind, I guess!
Re: Steve McIntyre (#273), It’s not crazy for Connelly to do this. Think about this as a tactic. It might me crazy for Mann to do it. But it’s not crazy for William. Its designed to drive YOU crazy.
Re: bender (#270), Yes good job Amac. I ran a similar gauntlet on another site regarding this issue, but it wasn’t a science blog, so I ended up the loser of the argument that the Team engages in obstruction. That’s how I found out about the Kaufman correction.
Ernie and Bert join the Team in an upside-down world:
RE MikeN, #264,
It’s true that Korttajarvi after 1800 (or 1720 even) is useless as a temperature proxy, making it impossible to calibrate to instrumental temperature per Tiljander as quoted by AMac#270.
This implies that Tiljander should not have been used by either Mann 08 or Kaufman 09. The fact that they did use it makes them wrong. The fact that they used it upside down makes them doubly wrong.
Kaufman at least terminated it in 1810 (per CA-initiated objections to the recent period), but Mann 08 actually used it upside down as part of his calibration. This makes Mann 08 triply wrong, and the argument that he used it upside down becomes supremely valid, not invalid per WC.
Kaufman also at least admitted that he used it the wrong way in the SOM to his Corrigendum, and tried omitting it and 3 other series, per my (unacknowledged) suggestion in “Invalid Calibration in Kaufman 09”. He found that it didn’t make much difference to exclude it, since he had already excluded it from his calibration equation.
But Mann 08 is a very different story, since his calibration hinges heavily on the invalid recent portion.
Korttajarvi pre-1720 may still be useful as a qualitative, if not quantitative, temperature proxy, but that’s a different matter.
Re: Hu McCulloch (#274), does Mann’s algorithm calibrate each proxy individually, or does the invalid use of one proxy affect the others?
Re: Hu McCulloch (#274), Hu, I was telling WC that Mann was only doubly wrong. I had missed the extra detail of it being impossible to calibrate the proxy. How did Kaufman handle this?
AMac-
Thanks for this careful and patient documentation of your exploration of this issue as well as your interaction with WC.
If you would like to write up a summary of this exploration — where you started, what you did, what you learned (about the issue and, as well, your debating partner WC), I would be happy to feature it as a post on my blog. You can either email me (pielke at colorado.edu) or post it up here in the comments or on my blog, and I will elevate to the top. It would be good to have this explained rather than spread across several blogs deep in various comment threads.
I agree with Steve’s comment: “It’s crazy for Connolley to support Mann on such an unwinnable issue.”
Re: Roger Pielke, Jr (#275),
“I agree with Steve’s comment: “It’s crazy for Connolley to support Mann on such an unwinnable issue.”
Surely Connolley is behaving in this way because he is a ‘political’ scientist as distinct from a ‘political scientist’ ,such as yourself, and Copenhagen looms!
Re: Roger Pielke, Jr (#275),
Surely you guys have noticed that even arguments they clearly lose (or have lost), they still spin them in such a manner as a win? If enough people say something, they fall onto the side of right by default. It is the same reason there is such importance on getting their flawed studies into the publication arena: once the initial papers are there, they can be cited repeatedly, making future claims “true” simply because they have precedent (that the precedent may or may not have any validity becomes a moot point).
Mark
WC is allowing some of my responses through.
WC says I just don’t understand how proxies are selected.
Phil says I don’t know how real science works, in my claim that scientists outside of this field use metadata rather than study data as a parameter to selecting which data to study.
Phil says there are “many examples” such as in turbulent flow and biology. I responded (below) to his specific bio example; Dan Hughes has asked Phil to be specific in his other claim.
I apologize for being off topic but I think you might be interested to know that a memeber of the Team is quoted in Newsweek saying:
“Gore bothers to come talk to us” “He’s the only politician who’s interested in the nuts and bolts of the science—and the only one who knows what a hydroxyl radical is”
Gavin dixit. This explains many things.
see here for the source
best.
Posted at Cruel Mistress, Curly Wants His Money Back | November 1, 2009 at 1:36 pm
Concerning the narrow issue of the use of the Lake Korttajarvi varve proxies by Mann et al (2008), William Connolley has responded to my five claims on the matter (Comment #1 of Stoat post Tiljander).
My reflections (and tentative conclusions) are way downthread, here in summary form and in the following comment in detail. Stoat slightly mangled my [blockquote] html; the longer argument is more readable as the mirrored version at ClimateAudit.org.
This rather small and restricted issue is quite relevant to the topic of this Cruel Mistress post. Employment of these lakebed sediments in Mann et al’s paleoclimate reconstruction is a restricted and technical question that can be definitively addressed by a layperson who is willing to scan only one paper beyond Mann et al and its related materials.
To date, the manner in which PNAS has handled this modest challenge involving the publication of mishandled data in a peer-reviewed article has been woefully deficient. The authors have apparently chosen to stonewall, and the editors seem okay with that.
In my opinion.
WC defending the Tiljander proxies:
I need a little help here. Is this a photo of Tiljander getting a varve sample, or WMC responding to Mann’s critics? Thanks.
Fundamentally, the AGW-fan-sites don’t want to tackle the data problem directly, because it is very undermining to their case. They want to portray any criticism of AGW as the work of crack-pots & incompitants, as this is the easiest way to brush aside any objections or counter arguements to AGW. So having to admit that SMc has pointed out a very gross error in the data handling & processing by the Hockey Team is potentially very damaging to their cause.
Not only does this error invite ridicule upon the HockeyTeam and suspicion on the HS, it also invites further suspicion of the other data & methods used in the HS creation.
– i.e. if they’ve got this so badly wrong, what else have they got wrong?
– this makes it more difficult to try to dismiss SMc’s other criticisms.
So, I can totally understand why sites like Connolley’s seem to go out of their way to mis-understand the point!
Off topic, but what is the html code to insert a picture into these forums? an excel chart?
thanks
At Stoat | November 1, 2009 9:35 PM
WMC —
Thanks for responding to my comment of November 1, 2009 10:42 AM. I have some further thoughts.
Readability.
While Stoat‘s “Preview” function faithfully renders simple html code, the “Post” step doesn’t handle [blockquote] well. This reduces the readability of my comment of 10:42 AM, earlier today. The inline remarks you’ve added inadvertantly make it even harder to read, unfortunately.
A correctly-formatted copy can be found at ClimateAudit.org. Brackets in the Stoat version highlight WMC’s responses.
Substance of Mann et al’s use of Lake Korttajarvi varve series.
Claim 1 — Apparently no disagreement between AMac & WMC.
Claim 2 — My quotes of Tiljander (pg. 571 & pg. 572) support the claim.
Claim 3 — Apparently no disagreement between AMac & WMC.
Claim 4 — WMC disputes that “These correlations are upside-down with respect to those assigned by Tiljander.” Discussed below.
Claim 5 — None of WMC’s remarks dispute the claim. His original remark “Well I should certainly hope so. That is what they are supposed to do” signals agreement. The newer text, e.g. “I’ll leave you to ponder until you do understand” is not substantial.
Mann et al Figure S8a — Disagreement. Rather than comparing “AMac sez” with “WMC sez,” I recommend looking at the figure and legend for yourself. The Black Line is Mann’s best plot of 1,800 years of Northern Hemisphere Temperature Anomalies. The Green Line shows what happens when the four wrongly-calibrated Lake Korttajarvi proxies are added to the Black Line plot, thus bringing lots of bad data into the calculation. Yet the Green Line and the Black Line coincide! How can this be? Does the figure actually show what the authors believe it shows?
“These correlations are upside-down,” continued
WMC remarked: [These correlations are upside-down with respect to those assigned by Tiljander – well no. Those correlations are in the data, Tiljander can’t assign them any other way. Do you mean, that the sign of correlation of the proxy varies before / after 1800? If you do, I assume you have only indirect evidence for that, instrumental evidence being lacking -W]
I’ll consider darksum, one of the three Upside-Downs. Tiljander states that varves with higher mineral content have higher darksum values (Fig. 9). For the Holocene prior to ~1720: “More severe climate conditions occur with higher winter precipitation, a longer cold period and rapid melting at spring, shown as thicker mineral matter within a varve” (pg. 571).
{Cooler local temperature :: higher darksum}, Tiljander says.
However, local human activities after ~1720 dominate the varve record, mainly by increasing mineral content. Tiljander did not observe a {climate :: darksum} correlation in the varves deposited between 1720 and 1997.
Mann et al “tried to limit [their selection of non-tree-ring proxies] to records that were reasonably well dated and where the original analysts had shown that there was a paleoclimatic signal associated with the proxy” (SI pg. 1). And yet, Figure S9 shows that they were misled by the rise in darksum during the calibration period 1850-1995. While the actual causality was
{More local human activity :: higher darksum}
Mann et al mistakenly found a spurious correlation between darksum and regional temperature:
{Warmer temperature :: higher darksum}, Mann et al say.
Whoops.
But it’s worse. For 1881-1993, temperature in the Lake Korttajarvi area held relatively constant (Tiljander Fig. 2). Thus, establishing a correlation between local temperature and darksum is not possible over this period of time.
Whoops.
Readers who doubt what I’m recounting should follow Rattus Norvegicus’ lead and look at Tiljander’s paper for themselves. The PDF is linked in Comment #2. The picture is pretty clear.
One more at stoat, now time for sleep…
WC, yes, a good proxy will have good correlation. But as your note infers (“the converse…isn’t true in theory”), we can’t pick proxies that way; it’s meaningless. Otherwise all kinds of meaningless data series could be selected as proxies.
We need to have a physical reason — in essence, a set of metadata criteria — to select proxies. And then, all proxies that fit the criteria must be taken. Otherwise, we’re simply fooling ourselves.
Paz –
Got to go deeper on “as long as selection criteria are independent of the effect.”
Metadata isn’t simply independence due to lack of “data effect” bias. Metadata criteria can not be a function of any attribute of the analysis itself (values, the three CI’s, etc.)
Otherwise, what we’re doing is essentially data/uncertainty/model snooping, creating a self-validating hypothesis.
Thought experiment: rather than collecting data from the past, imagine that we’re defining the experiment for future data. There should be no difference in the selection criteria.
Bottom line: to be valid, the selection criteria must not peek, even indirectly, at any aspect of the analysis output. Which means it can’t peek at any of the variables that form part of the analysis.
Blunt Example:
Data set A: 3 2 2 2 2 6 6 6 6 5
Data set B: 3 3 3 3 3 5 5 5 5 5
Data set C: 3 0 8 7 1 2 6 5 3 5
Data set D: 3 5 5 5 5 3 3 3 3 5
Stated goal: avg value of data matches avg of calibration dataset (4.0)
I can choose all kinds of data-based criteria for these data sets that will select the ones I prefer, without touching the “desired effect.” Yet I will greatly impact various characteristics of my analysis by doing so.
Re: MrPete (#294),
Mr Pete. The not so obvious way to show Connelly about metadata is to consider the process
or the practice of holding out part of a sample to do exploratory analysis while retaining the rest of a sample to apply
what you learn in your exploratory analysis. In fact, applying what you learn in one study to future studies is an
example of using metadata. A “model” that is extracted in one analysis and applied to another set of data is an
prime example of applying METADATA.
Re: steven mosher (#301),
Unfortunately, even the obvious version seems to be missed by a heck of a lot of people.
I would love to collect a set of sample climate proxy studies that avoid the problem!
SM also notes
Penetrating observations, both of them. It is a prominent aspect of the culture of the AGW Movement to deny all criticism with equal and great vigor, no matter if the matter is peripheral or central, no matter if the person criticized is senior or junior, no matter if the matter concerned is defensible intellectually or not. These defenses are always characterized by the assertion that whatever is being criticized is of no importance to the AGW theory, which does not seem to suggest to anyone that they must not be prosecuted with all possible vigor. The culture seems to be: never admit any error by anyone, no matter what.
To get a sense of where the phenomenon fits in American culture, read the wry and informative Bill Bryson. Climate hysteria is new, but hysteria in much the same form is a very old and recurring phenomenon in America.
Steve – as someone not 100% au fait with statistical methods (but I’m learning) though with some background in audio / DSP methods it strikes me that the “Team” are doing the equivalent of audio signal correlation (say against a digital-matched-filter). In this, you take the square of the incoming signal sample prior to attempting any correlation match – this removes phase / sign dependencies as they really don’t matter in the case of audio processing (i.e. the output is sign independent as well). It seems to me that this cannot be the case when analysing proxies – surely the sign of the output (or direction) _is_ important and so, therefore, is the sign of the input?
Or am I just being too simplistic and dense (I suspect so)?
Cheers
Mark
Re: Mark Fawcett (#296),
Nope you’ve pretty much got the message as it is. It isn’t hard for most of us to understand that proxies have orientations that must be maintained for the analysis to mean anything.
It’s a shame that this common sense isn’t widely shared by the Team.
I woke up with a fun illustration in mind. I shared it at Stoat; we’ll see if they get the joke…
Here’s a humorous example based on my current favorite illustration of the three kinds of uncertainty (data, model, parameter):
Hypothesis: next-older siblings are twice as old as younger siblings.
Data selection criteria:
* Choose nearest-age sibling pairs
* Prefer larger families (for ease of collecting large samples)
* Select when younger siblings are at 2 years (old enough to have settled in to the “pattern”, young enough to be “growing strong”)
NONE of those criteria “bias” the relative age of siblings. The only problem is they might have a relationship to relative age, but so what? They don’t bias it.
I hope it’s obvious, both how this illustrates challenges in data selection criteria, as well as uncertainty in models and parameters.
🙂
Re: MrPete (#298),
Pete, I especially like your example, because the metric in your hypothesis (I’m twice your age, nyah, nyah, nyah)has a fitting kind of school-yard irrelevance, since every older sibling was (or will be)at some time twice the age of their younger sibling(s).
At Stoat | November 2, 2009 10:58 AM
WMC –-
Your removal of the first half of my last comment (11/1 9:35 PM) makes it difficult for readers to understand me. I hope you will reconsider that editing decision, and reinstate what I wrote.
Inline in the rump comment, you say,
[You seem to be confusing local and regional temperatures. And you haven’t answered my question, which is odd, in one so fond of asking them. If you did answer my question I think you’d find all your newly added text redundant -W]
My use of the term local temperature refers to the instrumental temperature record at Lake Korttajarvi as presented by Tiljander for 1881-1993 in Fig 2. My use of the term regional temperature refers to the temperatures that Mann et al calculated via RegEM for the relevant grid box (Northern Europe?) for their calibration period, 1850-1995 (SI pg. 2).
Re: “And you haven’t answered my question” — There are three questions in your inline remarks in my 10:42 AM comment. Here they are, with brief answers. Indicate which one you meant, and I’d be glad to expand.
WMC: Do you think that Tiljander says anywhere that the proxy is useless for temperature after 1720? After 1800?
AMac: Yes.
WMC: Do you mean, that the correlation (Tilj, Temp) for the period 1850-1995 is of such a sign that it implies that higher temp is associated with thicker varves?
AMac: In my opinion, there is no relevant (Tilj, Temp) correlation for the period 1850-1995.
WMC: Do you mean, that the sign of correlation of the proxy varies before / after 1800?
AMac: I can’t meaningfully answer the question as posed, because a clear climate signal is lacking in the post-1720 varve record.
Re: AMac (#300),
I admire your patience 🙂
You have kept your head against a determined onslaught of obfuscation. Quoting your posts here as well must be unsettling for those determined to censor.
RE Mike N, #278,
Kaufman truncated Korttajarvi in 1810 to avoid most of the modern disturbances, then converted all 23 of his series to z-scores, and averaged over the available series to get a composite z-score for each point in time, with Korttajarvi upside down. Then he regressed this on temperature, and then inverted this regression to convert the composite to temperature.
However, only 19 of the series ran even to 1980, and Korttajarvi wasn’t in the calibration sample at all. My argument in my post was that it is invalid to apply the regression equation to the 23-series index, if the regression wasn’t based on the full 23 series. Kaufman’s Supplementary Online Information paper tries rerunning the calculation using only the 19 that run to 1980, as I had suggested, and finds that the results weren’t dramatically different.
Per your #277, I’m not sure whether Mann 08 does an analogous univariate calibration on an arbitrary average of z-scores, of if he separately calibrates them on a multiproxy basis (sort of like MBH 98 does). But the big difference is that he included the contaminated post-1720 data in his calibration equation, which can have a huge effect on the results either way.
BTW, per your #164,
can you provide a link to where he makes this statement?
Re: Hu McCulloch (#304),
I’m intrigued by Mann et al (2008) Fig. S8a, but don’t know enough to interpret it. It sounds similar in some respects to some of Kaufman’s figures, so I thought I might ask for a reference or URL. A site:climateaudit.org search yielded Mann 2008 Non-Dendro MWP Proxies from Sept. 3, 2008.
(1) What exact point are the authors addressing?
The figure is meant to evaluate their “seven potentially problematic series,” those being the four Tiljander varve proxies (three upside-down and one rightside-up) and three others. See “Potential data quality problems,” SI pg. 2. Presumably, by removing those seven proxies from Northern Hemisphere CPS Temperature Anomaly reconstruction, they believe that they can see (a) whether those proxies are indeed no good, and/or (b) whether the overall CPS reconstruction is so robust that it doesn’t matter if the seven proxies are bad or good.
However, I’m unclear on whether they think they are addressing (a), or (b), or both at once.
Unfortunately the figure’s legend is too skimpy to add much information. For S8a alone, it would read, “Comparison of long-term CPS NH land reconstruction (full global proxy network) both with and without the seven potentially problematic series discussed.”
(2) I don’t know what’s being graphed
The Black Line in the figure shows the N.H. Temperature Anomaly as constructed by CPS from the full global proxy network minus the seven proxies.
The Green Line shows what happens when the four awful Tiljander proxies (and three others) are thrown into the calculation. What happens is, nothing. The two lines are indistinguishable.
What data (proxy series) are the Black Line constructed from? It can hardly be the 8 “non-problematic” screened long-range N.H. proxies featured in Figure S9. I can’t conceive of a calculation that can give near-identical results starting from (8 sets), and from (8 sets plus 3 sets plus 4 wacked-out sets).
But then, what is the “‘full global proxy network’ less seven” from which the Black Line is calculated? The 33 proxies featured in the Flash animation at that 9/3/08 ClimateAudit (less seven)? Every proxy set listed at the start of the SI under “Proxy Dataset Details” (less seven?)?
I’m presuming that the answer to this question is wel-known, even obvious, to those versed in the literature. But as a non-specialist reader (and PNAS is a high-traffic journal targeting a broad readership), I’m stumped.
Re: AMac (#307),
I do not think that this presumption is warranted. For the most part, the methods in Mann et al 2008 are not known in third party statistical literature. They include many strange quirks, so that even with code in hand (which is helpfully available in this case as a result of past campaigns), the content of any graph becomes known only after patient dissection which is highly laborious.
If you asked another multiproxy authors (say Briffa or Crowley or Esper) what this figure meant, I do not believe that any of them would know precisely (or that any two of them would give a consistent explanation if pressed.) You’ve already probably looked at this figure more closely than any reviewer or climate scientist.
Re: Steve McIntyre (#308),
Okay then. Can you steer me towards references or links (at RealClimate, maybe?) that would answer these basics.
For Figure S8a,
* Which proxies are combined through the CPS algorithm to generate the Green Line? From the figure legend, that would be the “Full global proxy network.” What is the full global proxy network?
* How many individual proxy series does it contain?
* To generate the Green Line, is each proxy weighted the same?
* Few proxies reach as far back as, say, 800, and fewer still all the way to 200. If any set other than the set of 15 screened, long-range NH proxies shown in Figure S9 are used, then the Green Line is composited from fewer proxies at earlier times. In that case, why doesn’t doesn’t the removal of the four paleoclimate-irrelevant Lake Korttajarvi data series cause progressively more deviation of the Black Line from the Green Line, as one progresses back in time towards 200 AD?
Re: AMac (#309),
these are good questions. Try looking at last year’s posts on Mann et al 2008 (see Category in Left Frame) ; there’s nothing useful at RC.
I did quite a bit of work on this last year (as did JEan S and UC), but it will take me a little time to refresh on this. ORdinarily I could do this fairly quickly but I’m wading through the new Briffa material right now.
It is odd that it makes so little difference. This is the sort of thing that I like investigating and I will get to it at some point.
Re: AMac (#309),
According to Mann et al text, the “full global proxy network” would include ALL proxies. However, as observed e.g. here, it only includes proxies with the “correct” correlation sign. I guess you are new in parsing through Mann’s prose, so a slight warning: in almost all Mann’s work, it is almost impossible to figure out what exactly was actually done without going through the code (if available).
Depends essentially from the step (which century you are talking about). A rough estimate (upper limit) is given in Table S1 (“Full global proxies/annually plus decadally resolved proxies”). For the better answer, you need to go through rtable1209 (which gives two-pick correlations; row 4), and see which series have wrong correlation (see the category codes for series in 1209proxynames,xls). However, even this might not be the exact answer, there might be additional “tricks” in play: for example, in EIV “final splicing” is not done straighfordwally by taking each splice from corresponding time step, but a splice can come from an earlier step, see the last sentence in “SI Methods” paragraph. I’m not sure if that was used also in CPS, but I suppose so. In summary, I guess the only way to know for sure the exact answer to your question is to run the code yourself (not a trivial task, I warn).
No. Proxies are first averaged within corresponding grid boxies, and then the resulting series are “caliberated” against the instrumental series for that grid box. Thus a weight depends from the number of series in the corresponding grid box, and from the variance of the series during the corresponding step relative to the variance in the caliberation period.
First, notice that the figure S8a is wrong in the (original) SI, see here. Second, notice that three other “problematic” series do not enter in the reconstruction before AD1700 step. So before that only Tiljander series are removed, and those get combined in CPS into a single series in the grid box averaging step. Thus essentially a series is the difference between original and NH-minus 7 before AD1700. Comparing the original figure and the corrected figure (plus explanation for the error), it seems that we have here an example where one of the early steps is propagating towards later steps (notice the sudden change in difference around AD1000 in the corrected figure). As to why the removal of Tiljander series does not have a bigger effect on this (these) very early stage, I do not have a clue. Usually the only way to find out is to actually run the code…
So welcome to Mannian world of wonders!
Re: Jean S (#313),
I forgot that in CPS there is also a 15×15 “regridding” step. In this step, Torneträsk series get combined with the (combined) Tiljander series. So it might be that Torneträsk is here dominant, and therefore removing Tiljander series do not have too much effect. If that’s true, then we need to explain why suddenly around AD1000 there is a clear effect…
Re: Jean S (#313),
The MOST important phenomenon in Mannian CPS is the effect of correlation screening in producing HS patterns. This phenomenon is well-understood at the “analytical” blogs being more or less independently reported on different occasions by myself, David Stockwell, Jeff Id and Lubos Motl. Jeff Id has written some excellent posts on the topic.
There are some other things that are relevant to the weighting – which are in the realm of “stupid pet tricks”. If Mann entered the location of a proxy on an exact grid dividing line (e.g. 15E) – regardless of whether this is a correct, the proxy is assigned to both gridboxes in Mannian CPS and its weight is doubled, as with Socotra in the AD800 network.
Latitude boxes are all off by one degree. Does this “matter”? Probably not, but who knows.
Having said that, we do have control over Mannian CPS and with a little time I can answer your questions – but it takes a few days to pick up the file.
Tiljander has to be viewed both as an error in and of itself and as a sort of vivid example of Mannian method at work.
Re: Steve McIntyre (#316),
yes, the screening is most important, and that’s also the reason why it can be said for 100% certainty that Tiljander series are upside down — no bluff about multivariate regression being insensitive to sign can escape that. I’m well aware of the screening effect, and I also pretty much know how both CPS and EIV reconstructions are done. However, AMac was asking about reconstructions where no “official” screening takes place. As observed, there is still a form of screening (sign of correlation) going on in “no screening CPS”, but that does not have an effect to the inclusion of (upside down) Tiljander series, so the question remains why there seems to be so little effect (before AD1000 in the corrected figure) in Figure S8a if Tiljander series are left out.
Re: Jean S (#317),
Isn’t this the figure where Mann re-inserts the bristlecones? So that the one figure has Tiljander and the other has bristlecones?
As I’ve said before, I’ll do the run in the next week or so, after I finish a couple of Briffa posts.
I also need to re-visit your EIV posts that you posted up when I was in Thailand and didn’t follow up on when I got home.
Re: Steve McIntyre (#318),
yes, that’s the one (Figure S7a in Mann’s revised SI, Figure 8a in the original SI). It is clear that the EIV method (Figure S7b) is just hanging on the bristlecones, but in CPS there has to be additionaly something else going on: the match is just too perfect between original and Tiljander series removed series (even in EIV there is a clear effect). In the corrected figure the almost perfect match breaks after AD1000 although only verification stats have changed (which implies that final splicing is done differently in the original and the corrected figure), but still the match is almost perfect in the AD200-AD1000 part.
Also “not-so-technical” people might want to take a look of that Figure 7 linked above: both CPS (left, Figure 7a) and EIV (right, Figure 7b) are supposingly showing the same thing: NH temperature estimated from the same data! But at least for me they seem like contradicting each other (see, e.g., CPS drop AD800, EIV spikes AD950 & AD 1300, EIV clearly much colder during 1500-1900 etc.). However, without blink of an eye, Mann put the figures next to each other. No problem there to see, I guess.
Re: Jean S (#320),
But they replied to this already,
(PNAS reply)
So, in the case of inconsistency EIV is more reliable (but, at the same time, CPS CIs are ok I guess 😉 )
Re: UC (#328),
oh, I completely forgot that they had published “a formal reply” to that. l suppose “at this point bizarre is not the word any more” as Chris Dudley would put it. 😉 However, my eyes are telling me that there is some “divergence” also after AD1000, which one is “more reliable” then? 🙂
Re: Jean S (#313),
Thanks for taking the time to offer that extended backgrounder on Fig. S8a.
My first thought is still, “If what you say is true, how did this figure get through review?”
It’s one thing for me, a lay reader, to be unable to decipher the data and methods underlying a figure. Not a good thing, but scientific fields grow ever more specialized and complex.
On the plus side, a big reason why online publishing and the use of Supplemental Information was welcomed was for its potential of removing one of the biggest roadblocks to transparency, which is space limitation in the paper-periodical format.
A reviewer might pass a figure (or text, a table, or an entire manuscript) through the peer-review process without comment on the basis of, “Well, data and methods aren’t entirely clear here, but everybody in the field knows exactly what the authors did, and lay readers can find out by checking the reference list.” Not great.
“Well, details of data and methods aren’t really clear to me, and may not be known to other experts, and other readers can’t readily figure out the details by pulling the PDFs in the reference list. And the authors have made no particular effort to lay out a clear summary in English in the SI… but hey, works for me as is!”
That would be appalling. And PNAS is a high-impact journal with access to first-rate reviewers.
The idea that evaluation of figures at a basic level is restricted to expert computer scientists who are well-versed in reading programs written in a particular language — novel, and depressing.
I guess behind that cynical veneer, I’m still pretty naive.
Re: AMac (#321),
Keep in mind that source code is unavailable for 99.99% of all paleo studies. Mann et al 2008 is totally atypical in that respect and this is obviously because of the MBH history. You seem to be coming at this late, but I first requested MBH source code in late 2003 in order to avoid pointless disputes about methodology that was incorrectly, obscurely or not described in the original article. Mann refused. The only reason why source code became available was because Mann was written up on the front page of the WSJ and the House Energy and Commerce Committee took an interest and asked him for the code (even then, he didn’t produce complete code, leaving key MBH issues still unexplained to this day e.g. MBH99 confidence intervals or MBH98 retained PCs).
Add to this the unavailability of important data. For example, if you look at the recent Esper et al 2009 on Siberian trends, a considerable amount of data is unavailable; data that is available is listed in a format that does not link to known ID numbers and to be annoying the listing is done in the photo format that is increasingly popular among paleos as it requires manual transcription to be used; the methods are opaquely described.
Prior to the sort of communal reading of these papers that we do at CA, I don’t think that anyone actually read most of these papers in the way that we do. Their main function seems to have been to provide abstracts that were usable in IPCC reports.
Re: Hu McCulloch (#304), In his Tiljander thread:
Posted by me:
Let’s take a look at the two issues with Tiljander and upside-down usage.
Let’s switch Tiljander proxy to a century scale to explain the issues. Looking at the proxy in Tiljander’s paper,
the century values from 0 to 2000 are approximately
65,60,65,70,70,75,70,65,65,80,50,85,55,75,70,70,75,70,85,95
Warm centuries are the 1000s and 1200s, with the intervening 1100s having the coldest period.
Now, the initial confusion could be that here, lower numbers mean higher temperatures.
Now issue number one is that the last two data points do not represent cooler temperatures, but other factors.
As it is, if you want to use this proxy, you should cut off the last 200-300 years, and use the rest with lower numbers representing warmer temperatures.
Instead Kaufman, prior to his correction, cut off the last 200 years, and used lower numbers to mean cooler.
Mann used lower to mean cooler, and kept the last two hundred years, so the unrelated to temperature lower numbers of the last 200 years correlate with the temperature record.
The last 200 years show cooling in the proxy, but Mann has flipped it upside-down so it shows warming, and the warm parts historically are now interpreted as cold.
[If the proxy is unreltaed to temperature over the last 200 years then it is unusable for Mea’s method -W]
Contra Greg, it is in dispute whether the algorithm will flip a proxy. Some of the proxies are categorized so that the algorithm will not flip them. They are fed in with the ‘correct’ orientation, and Mann has done this upside-down.
[No -W]
Steve Mc. I would like to thank you for permitting posters to take up space in these threads with duplicate posts made at certain other blogs. It is consuming a lot of space here, but I believe it is important to the claims made repeatedly and consistently at this blog that certain competing blogs and personalities intentionally and unfairly manipulate and delete reasonable posts.
At Stoat | November 2, 2009 9:28 PM
WMC —
In my comment of 11/2/09 @ 10:58 AM, I distinguished “local,” instrumentally-measured temperature at Lake Korttajarvi from the “regional” metric derived by Mann et al. Does that seem like a useful distinction to you?
I also addressed three questions you had raised; was one of them the query that you thought I was ducking? Do you agree with my answers?
Needles in your eyes.
Submitted to Stoat‘s moderation queue | November 3, 2009 10:02 AM
WMC —
(1) Inline at my 11/2/09 9:28pm comment, you write,
> You know what my answers are [to the three questions AMac answered at 11/2/09 10:58am], because I’ve said them already.
Scanning this thread, it seems that most of your remarks on the specifics of the Lake Korttajarvi varve series are conditional and hedged with qualifiers (if, may, might). My impression is that you didn’t carefully read Tiljander et al (2003) until yesterday —
> I’ve skimmed [Tiljander’s paper]. What makes it particularly interesting, would you say?
Inline at Phil M’s 11/2/09 8:29am comment.
For myself, mea culpa–I didn’t pull the PDF till reading Rattus Norvegicus, 10/30/09 4:16pm. Checking the source cleared up a number of issues.
(2) By the way, on 12/1/08 Mann et al informally published a correction of Figure S8a at Mann’s website, though it’s not noted at PNAS (hat tip JeanS).
(3) You responded inline to Mike N (11/2/09 9:12pm) —
> So many people seem incapable of seeing the bleedin’ obvious here, its going to need a new post.
Look forward to it; that should clarify positions, and–I suspect–highlight some areas of agreement.
On that subject, MikeN (10/30/09 11:31am) alluded to Kaufman et al (2009)’s corrected use of Lake Korttajarvi data (10/7/09 SI Draft). They now agree with Tiljander that higher varve X-Ray Densities correspond to lower temperatures, 1 AD to ~1800. In their paleoclimate reconstructions, Mann et al use “xraydenseave” in an orientation which is flipped, with respect to how Kaufman now employs it.
Is Kaufman using this proxy in an upside-down manner, or is Mann?
Re: AMac (#319),
Via email, WMC has informed me that my latest comment (mirrored just upthread) has no substance, other than the link I supplied to Mann’s website.
On that basis, it appears that it will not appear in Stoat‘s “Tiljander” thread, having failed WMC’s comment policy.
Remarkable.
Re: AMac (#323),
Let’s see if he’s also going to change the wrong figure in his post (William, why don’t you also show the EIV version?), or just to stick to good old RC style of handling things.
Re: AMac (#319), your patience astonishes me, but is exactly what’s needed. I too look forward to WC’s new thread on this issue, and predict it will be called If I’m so smart, how come I ain’t right?. Note the conditional hedging.
Concerning a image formats an OCR application couldn’t handle them less worse than to copy by hand?
Where exactly it is “demonstrated (in the SI) that the EIV reconstruction is the more reliable where they diverge” ?
Re: UC (#331),
what a bizarre question! Even fossil fuel industry shills see from Table S2 that all skill scores of any merit are remarkably better for EIV than for CPS! 😉
Re: Jean S (#332),
Ah, I see. And from the article:
…
… (my bold: )
So, EIV is better in the case of sparse network. However, in SI it is mentioned that
where ‘seven’ is a robustly estimated median of numbers ‘six’,’seven’, and ‘eight’.
Re: UC (#333),
The first paragraph you cite is an oblique reference to “teleconnections”. Does the concept of “teleconnection” really have any credibility within the general paleo community?
Re: UC (#333),
You note from the Mann et al (2008) SI —
Might this also be relevant to the calibration of the Lake Korttajarvi varve proxies?
As I understand the quoted sentence, it implies that CPS does not make use of nonlocal statistical relationships, and thus does not allow temperature changes over distant regions to be effectively represented through their covariance with climatic changes recorded by the network.
Yet for the CPS analyses (e.g. Fig. S8a), the varve series are apparently calibrated against a set of regional grid box temperature-like numbers that are calculated from instruments that are not-near Lake Korttajarvi, for the calibration period 1850-1995. (Per Jean S #313 supra.)
Would that be a “nonlocal statistical relationship?”
Recall, establishing a “local statistical relationship” between the varve series and local temperature 1881-1993 (Tiljander 2003, Fig. 2) would not be easy, because local instrumental temperature readings near Lake Korttajarvi held relatively constant over that period.
Therefore, for both CPS and EIV, one must use a nonlocal composite temperature record if one wishes to establish a temperature-to-varve correlation over the past ~150 years. This might or might not be acceptable, under a full description of the CPS protocol. It doesn’t seem consistent with Mann et al’s summary description.
#333: It’s amazing, given your paragraph with the bold and known non-linear relationship of many of their proxies to temperature, that they would cite such an assumption and proceed as if the assumption was reasonably met.
Re: Carl G (#334),
The real reason is their unwillingness or inability to calculate valid CIs. We discussed about it some time ago, http://www.climateaudit.org/?p=4833
Re: UC (#337), Ooga chaka.
Mark
Re: Carl G (#334), If you look back at the original assumptions made in some of the first reconstructions, at least the first for our practical concern (MBH98), you’d be surprised how many assumptions are simply swept under the rug and ignored. It’s “Team Science,” so who can really be surprised?
Mark
In making a final defense against(of?) upside-down charges, WC ends up highlighting the main problem with CPS.
Way to go AMac!
http://scienceblogs.com/stoat/2009/11/tiljander_again.php
At Stoat‘s Tiljander, again thread | November 4, 2009 8:37 PM
Tiljander’s Mineral matter couplet thickness varve series
Three of the four Lake Korttajarvi varve series collected by Tiljander are archived here as a text file for years 0 to 1985. Tiljander stated that she believes that accumulation of mineral matter is the result of a harsher climate; i.e. there should be a correlation between lower temperatures and higher “mineral matter couplet thickness”.
In addition, she believes that near-lake human activities have caused increasing deposition of mineral matter since ~1720.
From 0 AD to 1720 AD, the Mineral matter couplet thickness, averaged in 20-year intervals, gives values of 0.10 mm to 0.34 mm.
The sole exception is the 1321 to 1340 interval. That high average of 0.64 mm is due to the 1326 varve, which has a reading of 12.9 mm, seemingly the result of a one-off event.
From 1721 to 1860 (circa the beginning of the Mann et al calibration period of 1850-1995), mineral values rise from 0.16 mm to 0.53 mm, fairly steadily.
From 1861 to 1980 (approx. the Mann et al calibration period) mineral values rise from 0.43 mm to 1.59 mm. The five years 1981-1985 average to 2.14 mm. Human activities mentioned by Tijander are seen for 1930 (4.4 mm) and 1967 (4.6 mm).
A concrete instance of an “A”-type proxy
Thus the Lake Korttajarvi mineral varve proxy seems to fulfill the conditions of Proxy “A” as presented by WMC in the body of the post.
* Prior to ~1720, higher temperatures should correlate with lower values.
* During the 1850-1995 calibration period, human activity correlates with higher values.
* The Mann et al algorithm “flipped it over” such that higher values correlate with higher regional temperatures during the correlation. I believe (but am not certain) that this data was used by Mann et al to build the “tijander-2003-lightsum” trace seen in Fig. S9. (Note that the 1326 event can be identified in the raw data presented by Tijander in Fig. 9, but not in the Mann et al trace. The 1930 and 1967 events can be identified in both.)
(Note also that Tijander reproduces the local instrumental temperature record for the Lake Korttajarvi area, 1881-1993 in her Figure 2.)
Having these attributes of one of the Lake Korttajarvi varve proxy series to hand may help in discussing their use by Mann et al and others attempting paleoclimate reconstructions.
Estimating the effect of the proxy on temperature anomaly reconstruction
What follows is some back-of-the-envelope arithmetic. I claim no familiarity (much less competence) in temperature anomaly reconstruction methods. Still, this might give an approximate idea of the sort of signal the tiljander-2003-lightsum proxy might have contributed to Mann et al’s Fig. S8a, and others.
From Mann et al Fig. S4a, the Northern Hemisphere Temperature Anomaly was in the -0.4 C to -0.2 C range in 1850-1860. It stood at about +0.2 C in 1985. The varves’ mineral thickness was ~0.6 mm in 1851-1860, and ~1.9 mm for 1976-1985, at the end of the series. That would seem to suggest a that a rise of 0.5 C correlates with an increased varve thickness of 1.3 mm. This is for the proxy in the “flipped over” orientation (with respect to Tiljander’s interpretation) that Mann et al used.
On this basis, if used by itself in this way, here is what this proxy would yield over the 1 AD to 1850 time period.
The largest-magnitude temperature anomalies would be in the range of about -0.5 C for 1000 AD to 1300 AD (avg. mineral thickness 0.14 mm; the smallest 20-year averages are 0.10 mm for 1181-1200 and 1201-1220, and 0.11 mm for 1221-1240.)
Excluding the 1326 varve, this proxy would yield its smallest-magnitude temperature anomalies in the range of about -0.45 C for 581 AD to 800 AD, (avg. mineral thickness 0.27 mm; the largest 20-year average is 0.34 mm for 681-700.)
At Stoat‘s “Tiljander, again” thread | November 4, 2009 10:13 PM
Graphs (small jpeg files) of the Lake Korttajarvi varve “mineral matter couplet” series discussed in the 11/4/09 8:37pm comment are available at Bitbucket.org.
I see on the “Tiljander, again” post that Connelley refers to all this as being a “molehill made into a molehill”.
Seems to me these guys are treating it more like a Pork Chop Hill.
Re: John M (#346),
He wishes. No one knows what size hill this is until supplemental Fig. S7 is recalculated – no Tiljander, NO pine trees. Maybe WMC can do that calculation for us? Should take him but a minute.
Re: bender (#347),
and after one minute gridboxcps.m answers “??? Out of memory”
http://www.climateaudit.org/?p=3844#comment-300596
The “Tiljander, again” psoting is an obvious attempt to maintain the pose of omniscience that climate scientists adopt. The main gist of it is that even if climate sceientis are wrong and their methods are unsound that they are right. And besides that they are in charge because they are smarter than you are and will have the last word any way.
Re: TAG (#348),
They wanted a fresh slate. Wonder why.
WMC: Mann’s algorithm grabbed the stick by the blade and twisted the handle to the orientation that he preferred: upside-down. To spin a hockey stick like that requires quite a bit of torque. A Mann needs his Mannomatic.
At Stoat‘s “Tiljander, again” thread | November 5, 2009 10:13 AM
dhogaza wrote at 11/5/09 8:28am —
I’m unfamiliar with Mann’s sensitivity test. In any case, I doubt I have the skill set to calculate it. I’d be interested in your interpretation of this test, or in a link that you feel describes it lucidly.
The back-of-envelope arithmetic is meant as an order-of-magnitude estimate of what “more minerals, warmer temperature” might mean, assuming that a simple linear relationship is ascribed to the 1850-1995 period and that there is no transformation of the data. I don’t claim that this is an optimal way to handle varve data.
The 20-year averaged data for the three Tiljander varve series and some graphical views of them are now up at Bitbucket.org.
Re: AMac (#352),
dhogaza’s being an ass.
First, he’s pointing out that which we already know:
(1) A graphic has already been produced excluding the defective lake proxies (S7).
(2) Producing these graphical sensitivity tests requires more than a back-of-the envelope calculation.
Second, despite the admission, finally, that an error has been made, there are several dodges:
(1) refusal to acknowledge that McI was right: Tiljander was used “upside-down” (Instead, the argument is “it doens’t matter”, “it’s a molehill”.
(2) This is WMC+dhogaza talking, not Mann. We want to see what Mann has to say, not his minions. (Just as we’d rather hear from Briffa than Gavin and his gurus.)
(3) dhogaza dodges the REAL issue – stated again and again at CA and at Stoat – that S7 needs to be recomputed without pine trees.
(4) It was Mann’s obligation to produce this sensitivity test graphic a year ago inhis response to McI. And he should do it now if he wants to back-pedal and argue that it “doesn’t matter”.
(5) Anyone else arguing “it doesn’t matter” must produce that graphic.
That is the definition of “ass”.
Re: bender (#353),
After all this digital ink has been spilled on the Tiljander proxies, you’d think it’d be interesting to see what the data actually look like. Anyway, for me it helps.
Sensitivity tests are great, but at base you have three or four records of sediments, year on year. In the case of lightsum, you have an annual measurement of how much mineral material (silt) settled on the lakebed, measured in millimeters. What causes silt? Well, Tiljander had a pretty sensible answer. More spring runoff from more snowpack will cause more water to course into streams, carrying more silt with it. Aside from that, more rain in general and more downpours in particular will cause more silt, too.
So she posited a causal link leading from cooler local temperatures and more local precipitation to higher lightsum values for Lake Korttajarvi varves.
So, how does that compare to the instrumental temperature record, for calibration. Oops, we can’t do that properly, as human activities in the watershed have added a lot of silt after 1720. Double oops, the local temperature held steady 1881-1993 or so, so there’s nothing to correlate *to*, locally.
But anyway, the data still shows patterns, 0 AD to 1720, though it sure does bounce around a lot. Overall, there was a lot less silt back then, with the exception of 1326. A hurricane or who knows what.
Since Mann et al *did* calibrate the 1850-1995 varve record to a synthetic measure of regional temperature, I thought it would be interesting to see what a simple look might show, as to the order of magnitude of temperature deviation from the 0 C point of the NH anomaly over the past 2,000 years.
I expect one has to do a log transformation of the data, since obviously a linear “deg C” to “mm” correlation runs out at a varve thickness of zero millimeters–can’t get thinner than that, no matter how much cooler the climate gets. I’m not going to reinvent the wheel though.
The broader point is that the huge increase in silt input to the lake, 1850 to 1985, is going to dominate what this record tells you about past temperatures, if you wrongly calibrate that change to regional temperature increases 1850-1995. As Mann did. And the only thing that it can possibly tell you is that the low-silt distant past was cooler than the increasing-silt recent past.
Since the record is upside-down, to the extent lightsum correlates with temperature, it’ll tell you that times like the Little Ice Age were colder than today (but not that much so), and times like the Medieval Warm Period were colder than today (somewhat more so). Assuming, of course, that such regional warmer/colder regimes affected the Lake Korttajarvi watershed.
1326 was a scorcher, though.
Actually, I could imagine that Mann’s orientation of the lightsum proxy might be reasonable. A northern hemisphere rise in temperature might affect the North Atlantic in such a way that the Gulf Stream alters course (etc.), and a result is that more rain falls in Northern Finland. Irrespective of the temperature in the watershed. Of course, the entire *point* of the screening exercise was to find signals like this in potential long-term proxy series, irrespective of prior conceptions of what something might mean (with the exception, Mann et al noted, of proxies with clear interpretations, like lakebed sediments. Moving on…).
Anyway, if Mann et al’s use of lightsum is reasonable, then folks ought to be pretty busy challenging Kaufman for issuing his recent correction. Why correct what’s right to begin with and flip it upside-down?
Well–I guess it’s possible that *both* orientations make good paleotemperature proxies.
At Stoat’s “Tiljander, again” thread | November 5, 2009 12:24 PM
Hank Roberts 11/5/09 10:48am —
When I look at Figs. 2b and 2c of the Brauer et al 2008 article that you link, I notice the Lake Meerfelder Maar varves getting thicker as the Younger Dryas cooling event takes hold, ~12,700 years ago.
Do you agree? Do you think an analogous {colder temps -> thicker varves} correlation likely holds true for Lake Korttajarvi as well? (Tiljander thought so.)
At Stoat’s “Tiljander, again” thread | November 5, 2009 2:36 PM
Hank Roberts, thanks for the pointers. Still, my impression is that a journal like Nature Geoscience has figures because they can help readers to understand the ideas that the authors wish to present. Thus, it’s reasonable to look at figures with that in mind.
Brauer et al are pointing out that lakebed varves became thicker in Lake Meerfelder Maar at the time that the Younger Dryas cooling began. The red trace in Figs. 2b and 2c don’t refer to the composition of the sediments, although other traces do, and that’s pursued in the text.
Many of the factors that Brauer et al show are important to the Lake Meerfelder Maar varve record — changes in water level, mixing, anoxia at depth — aren’t discussed by Tiljander et al for Lake Korttajarvi. Is that why you commented, “Arrrrrgh”?
#317, 318
Takes some time to remember what was done in MannPnas08; before proceeding it is good to check SI8a without that red paint:
( data source http://www.meteo.psu.edu/~mann/supplements/MultiproxyMeans07/data/reconstructions/cps/ nhcru_cps_composite.mat )
Re: UC (#357),
I looked at the data you linked to, and overlaid it against Mann’s original and corrected Fig. S8a. The results are in the “NH CPS” worksheet in Excel File “Tiljander-cf-Mann-CPS.xls” at BitBucket.
The .csv data you graphed in green, above, matches pretty well with revised Fig. S8a’s “original NH CPS” (green line). But there are some modest differences; I don’t know their significance.
I also rescaled the graph of the raw varve data (20-year-averaged) over 200-1850 (prior to Mann’s calibration period) and matched it with corrected Fig. S8a.
The orientation of the varve data as Tiljander interpreted them is:
* {warm=down} for Mineral
* {warm=down} for XRD
* {warm=up} for Organic
(Mineral and XRD peaks and valleys do trend together, but Organic is hardly a mirror-image of them.)
Summary — Many but not all of the differences between the “NH CPS minus 7” trace and the “original NH CPS” trace in corrected Fig. S8a are consistent with Mann et al using an Upside-Down (with respect to Tiljander) interpretation of the Mineral and X-Ray-Density data sets from Lake Korttajarvi.
Details — Observations from visual inspection.
800-860 is the time of the biggest drop in the temperature anomaly, without or with the 7 records (incl. 4 Tiljander). There’s nothing specially high or low about the three Tiljander series during that time.
1040-1080, adding the 7 records into the CPS process causes the calculated temperature anomaly to fall by ~0.2 C. Mineral and XRD both are at their lowest during that period, and one other (1200-1250). This is consistent with an Upside-Down interpretation (higher Mineral & XRD means warmer).
1120-1160, adding the 7 records into the CPS process causes the calculated temperature anomaly to fall by ~0.3 C. This is at a point where Mineral and XRD values are higher than during the immediately preceding and following decades. The amplitudes of the peaks are big but not huge.
1200-1250, the “CPS without 7” and “CPS with 7” traces aren’t identical, but they occupy the same range. Mineral and XRD varve readings are at their lowest for the 200-1985 span.
1320-1360, adding the 7 records into the CPS process causes the calculated temperature anomaly to rise by ~0.2 C. This is during the period of highest mineral deposition in the varves, including super-high 1326 (they’re high even without 1326). This is consistent with an Upside-Down interpretation (higher Mineral & XRD means warmer).
1610-1650, adding the 7 records into the CPS process causes the calculated temperature anomaly to rise by ~0.2 C. There’s nothing obviously special about the varve records during this time; they are near their averages for the longer 1380-1700 period.
At Stoat’s “Tiljander, again” thread | November 6, 2009 3:52 PM
Hank Roberts (11/6/09 11:47am), thank you for your patience, and for the considered response to my earlier comment. I think I understand your point now.
If I’m correct, you linked the Brauer et al paper to illustrate how difficult it is to understand how sediment series data relate to paleoclimate. In the case of the Younger Dryas, it’s not profitable to wonder whether “thinner varves” or “thicker varves” correlate with cooling. Figures like Brauer’s 2b and 2c are complicated, and thus too easily misinterpreted by non-specialists. Rather than focusing on an illustration from an example, I would do better to learn more about the larger picture of AGW, and then work down from there to analyses of specific detailed data.
Is that about right, do you think?
– – – – – – – – – –
The effect of removing the Lake Korttajarvi proxies from Mann et al’s CPS calculations is compared to the 20-year averaged raw Lake Korttajarvi data in the Excel file “Tiljander-cf-Mann-CPS.xls”, downloadable from BitBucket.
The corrected and expanded Fig. S8a is online at Mann’s website. The expanded version (uploaded on Nov. 4, 2009) seems to address a major criticism leveled by some posters at Climate Audit (as I understand it). Mann et al now show the CPS reconstruction with tree rings entirely removed. The two new traces are “NH CPS w/o tree rings” and “NH CPS minus 7 w/o tree rings.”
Re: AMac (#359),
Ooops, has anyone counted the number of these updates/corrections etc? Of course, it is pure coincidence that these errors are found as the issue was re-raised here. But this time we have something unexpected (my bold):
Interesting, if one checks the reconstruction without UC’s “red paint”, it seems that MWP values are higher than current values, and early values are skyrocketing. So the Tiljander series and bristlecones do indeed “matter”, don’t they?
At Stoat’s “Tiljander, again” thread | November 6, 2009 7:34 PM
Rattus Norvegicus (11/6/09 5:43pm) —
> So AMac, are you willing to accept that Mann’s results are substantially correct?
Our host may find my response off-topic; it may be snipped on that basis, which would be fair enough…
Before last week, I knew nothing about varves’ use in paleoclimate reconstruction. I was broadly familiar with the conflicts between the large majority of scientists working on AGW (and their supporters), and the minority of technically-oriented AGW skeptics (and their supporters). (I exclude those with anti-AGW belief systems as uninteresting.)
I thought (and think) that the broad AGW argument is pretty compelling, but that unresolved issues persist.
Mann et al’s use of varves seemed that it might offer insights into how the “pro-AGW” and “skeptical” camps handle controversy. The cutting edge is usually messy, so I didn’t expect all the evidence would point one way or another.
Science thrives on the free exchange of ideas, and on handling appropriate challenges to the methods and logic used to obtain results.
In my opinion, the Mann group’s handling of the Tiljander proxies is clearly in error, and uncorrected. PNAS’s editors and the paleoclimatology community accept these outcomes.
As you point out, revised Fig S8a shows that Mann’s calculation of the pre-instrumental northern hemisphere temperature anomaly is not greatly affected by the removal of both Tiljander and tree ring series.
Having looked closely at the Mann group’s and the community’s handling of a small methodological challenge, I am less confident of the correctness of Fig. S8a. Irrespective of whether these findings are “right” or “wrong,” the take-away lesson for me is that error-correction mechanisms are not working properly in this field.
Thanks for asking.
Re: AMac (#361),
AMac, I’ve revisited my Mannian CPS emulation and will post up an updated script yielding digital results for the AD800 network (or other networks). Here’s my estimate of the difference from the AD800 Tilj+Dendro Network. It looks like about a 0.5-0.6 deg C swing. This is prelim: I’ll post fresh scripts up in a new post.
Re: AMac (#361),
To answer Rattus: Now that those very fundamental errors – errors that should have been caught in the peer-review process – errors that should have been corrected after being pointed out by McIntyre’s PNAS letter – now that these are “corrected”, we can continue our review of the paper. We’ll let you know when we find other errors in need of correcting.
.
Science is “self-correcting”, right?
Continued at http://www.climateaudit.org/?p=7662.
At Stoat’s “Tiljander, again” thread
– – – – – – – – – –
dhogaza | November 7, 2009 7:34 PM (Comment #33)
– – – – – – – – – –
AMac | November 7, 2009 11:22 AM (Comment #35)
– – – – – – – – – –
AMac | November 7, 2009 1:10 PM (Comment #37)
Re: AMac (#366),
AMac, this is the sort of situation that arises far too often at Connolley’s and similar sites. Your responses were scientific in tone, moderately expressed and deleted, while the hectoring non-scientific comment to which you were responding is allowed to stand. realclimate obviously does the same thing. It seldom leaves the injured participant feeling reassured,
Re: Steve McIntyre (#367),
This behavior is precisely why I abandoned the RealClimate site. You can’t win an argument when their game is to make you look like an idjit, using nefarious methods not at your disposal.
Re: AMac (#366), Re: Steve McIntyre (#367),
Stop going there unless your goal is to become fodder for their disinformation efforts. “Fool me once…”
From Finland: url
Pause, move to Korttajärvi, central Finland.)
VO: Materials for the hockey stick factory have also been collected from Finland.
Reporter Backman, standing on a jetty: “This small Korttajärvi in Jyväskylä has become a focal point in the international climate debate. Based on samples taken from its’ bottom sediments, some foreign researchers claim that, an unprecedented warming occurred at the end of the 20th century. Finnish researchers, on the other hand, have used the lake to show that climate has always changed, even more than recently, and irrespective of human influence.”
VO: Five years ago, one of the Korttajärvi researchers responded to MOT’s question about the IPCC’s claim that recent temperatures are highest in a thousand years.
(Interview footage from MOT archive, 2004)
Ojala: “Based on these studies it seems that this claim is not quite true, at least for the Northern hemisphere, at least for Scandinavia. We’ve clearly had much warmer winters here in the Nautajärvi and Korttajärvi area, than what we are experiencing now.”
Question by Backman: “What’s your estimate, how much warmer was the medieval period in Finland, compared to the present?”
Ojala: “It is difficult to say exactly. But we may speak of half a degree (Celsius), even a whole degree based on several European studies.”
VO: At least two research teams close to the IPCC added the sediment data collected by Finnish researchers as part of their own paleoclimatic model reconstructions. This was done with agreement, but the Finns were surprised to see that in a study published this September, their data and interpretation of its’ meaning had been turned upside down. Here is the millennial temperature reconstruction from Korttajärvi done by the Finns:
VO: And here we have the same data presented by the hockey team:
VO: A nice hockey stick has emerged from the Korttajärvi mud. What in the Finnish study signified cold, had been turned into warmth in the IPCC science and vice versa. This interpretation passed the scientific peer review.
Dr. Atte Korhola, professor of environmental change at the University of Helsinki, is an expert in lake sediment studies.
Atte Korhola: “Some curves and data have been used upside down, and this is not a compliment to climate science. And in this context it is relevant to note that the same people who are behind this are running what may be the world’s most influential climate website, RealClimate. With this they are contributing to the credibility of science – or reducing it. And in my opinion this is alarming because it bears on the credibility of the field, and if these kinds of things emerge often – that data have been used insufficiently or even falsely, or if data series have been truncated or they have not been appropriately published (for replication), it obviously erodes the credibility, and this is a serious problem.”
VO: The author of the September study, Darrell Kaufman, admitted his mistake two weeks ago and sent a correction to the journal Science. But the main author of a previous study, Michael Mann, the father of the original hockey stick, still sticks to the claim that a hockey stick was found at the bottom of lake Korttajärvi.
Re: Steve McIntyre (#370),
That is one of the best quotes I’ve seen. Esperesque in quality, thank god sane people are paying attention.
Dr. Atte Korhola, professor of environmental change at the University of Helsinki, is an expert in lake sediment studies.
Atte Korhola: “Some curves and data have been used upside down, and this is not a compliment to climate science. And in this context it is relevant to note that the same people who are behind this are running what may be the world’s most influential climate website, RealClimate. With this they are contributing to the credibility of science – or reducing it. And in my opinion this is alarming because it bears on the credibility of the field, and if these kinds of things emerge often – that data have been used insufficiently or even falsely, or if data series have been truncated or they have not been appropriately published (for replication), it obviously erodes the credibility, and this is a serious problem.”
The truth will out!!
Fatal error. Assuming the meatgrinder gets it right every time. Not bothering to verify when a mechanic suggests it’s worth a look. Instead, argue the mechanic is defective. Foolish.
I posted the following over at RealClimate, in the “muddying the peer reviewed literature thread”. It is no longer in their moderation queue, and has been censored of course. Too bad, I really wanted to hear more about their side of the story. I guess this is as close as one can get to “no comment” without actually saying it.
At Delayed Oscillator‘s “Talk about Tiljander” post of November 3, 2009 at 4:41 AM
Submitted to moderation as Comment #9 on November 21, 2009 at 1:33 PM EST
Delayed Oscillator — Your enthusiasm for Stoat‘s handling of the Tiljander proxies might have some baiss, after all.
Recently-added Comment #39 to “Tiljander, again” (see link in the first comment here [link]) directs to an alleged series of emails on the Upside-Down X-Ray Density Varve Proxy.
Bolding added.
– – – – – – – – – –
On Thu, Sep 3, 2009 at 4:45 PM, Nick McKay wrote:
On Sep 4, 2009, at 5:24 PM, Nick McKay wrote:
From: Darrell Kaufman
To: Nick McKay, Caspar Ammann, David Schneider, Jonathan Overpeck, Bette L. Otto-Bliesner, Raymond Bradley, Miller Giff, Bo Vinther, Keith Briffa
Subject: Arctic2k update?
Date: Sat, 5 Sep 2009 08:44:19 -0700
Re: AMac (#375),
Credit to McKay and Kaufman for at least having eyes and ears open.
Submitted to RealClimate.org’s Comments on The CRU hack: Context thread, after Comment 88.
– – – – – – – – – –
23 novembre 2009 at 1:57 PM
Your comment is awaiting moderation.
[The comment appears to have failed moderation.]
The previous comment (#377) was accepted by Gavin Schmidt this evening as Comment #128. His commentary is interesting.
– – – – – – – – – –
Submitted to RealClimate.org on The CRU hack: Context thread, after Comment 178.
23 novembre 2009 at 7:47 PM
Awaiting moderation.
Cross-posted to RealClimate.org on The CRU hack: Contextthread
The corrected version of Mann et al. Fig. S8a that is currently at the Penn State website has two Temperature Anomaly traces that are nearly-identical: “original NH CPS” (black) and “NH CPS minus 7” (green).
The corrected version of Mann et al. Fig. S8a that I downloaded from that site circa 11/4/09 had two Temperature Anomaly traces that were not wholly superimposable: “original NH CPS (green) and “NH CPS minus 7” (black).
I have uploaded these files to BitBucket: earlier version and later version.
I am not aware of text at the Penn State website or at PNAS that explains why or when the one file was replaced by the other. I am not aware of an explanation for the discrepancy between the two “NH CPS minus 7” traces (green line in the earlier file, black line in the later one).
Feb. 18, 2010 —
A reasoned approach to the findings of Mann et al. (PNAS, 2008) can only be undertaken once it is acknowledged that the four Tiljander proxies were wrongly used in their reconstructions.
(All four were uncalibratable, thus miscalibrated. Two and probably three of the four were used in an upside-down orientation, such that “warmer” information was taken to mean as “colder”, and vice versa.)
Once this point is understood, Mann et al. (PNAS, 2008) can be seen to be making some important statements about paleotemperature reconstructions, and how they should be interpreted.
My recent post on the subject is The Null Hypothesis.
One Trackback
[…] Curly Wants His Money Back October 29, 2009 Steve McIntyre gives the impression in his recent post that I somehow think peer review is a closed system. He doesn’t say as much, but I gather […]