Obviously, there has been considerable controversy over the past few days over the Yamal data.
First, let’s observe the continued silence of field dendros on the dispute. None have stepped forward so far to support Briffa’s use of 10 cores in 1990 (and 5 in 1995). As others have observed, their silence is rapidly becoming loud. And it’s not as though Climate Audit is so well loved in dendro circles that dendros would not be willing to speak out against it.
The graphic below shows the disparity between core populations at Yamal and the other two sites in Briffa et al 2008. In 1990, both Avam-Taimyr and Tornetrask-Finland had 105 and 113 respectively (51 and 34 in 1995), whereas Yamal had only 10 cores in 1990 and 5 in 1995 (!)
Figure 1. Core Counts in Briffa et al (2008) Chronologies
Despite the lack of endorsement from field dendros, this hasn’t prevented Gavin Schmidt from opining that my observations are incorrect. However, in fairness, in one instance, Gavin included the caveat:
“I’m not a tree ring person, so my opinion on this is perhaps not worth much”.
On this point, Gavin and I can find some common ground. In the absence of any statement from dendros, including Briffa’s close associates at CRU, Gavin seized on an observation from CA and realclimate reader Tom P, who now seems to have become Gavin’s tree ring guru and authority on tree ring homogenization procedures, Gavin endorsing Tom P’s calculations as follows:
“Tom P. above showed that the Yamal curve was robust to homogenising the age structure”
I don’t know how much due diligence Gavin did on Tom P’s knowledge of dendro procedures to determine whether his opinion on these matters was also “perhaps worth not much”. We’ve known Tom P at Climate Audit for a few days and it is my understanding that until a few days ago, he was completely unfamiliar with dendro issues – in other words, well qualified to act as Gavin’s guru and mentor.
Tom’s own contributions to the discussion here have been Monty Python-esque. One could imagine John Cleese playing Tom P in a skit.
First, Tom P seized on the “excellent correlation” between two identical time series – the subfossil portion of the Yamal data set used in sensitivity studies with two different modern samples, one displaced microscopically below the other. Tom gushed that this increased “confidence” in both of them:
The excellent correlation between the two Yamal datasets strengthens confidence in both before 1850.
Next, Tom found once again that identical series have remarkable similarity. This time, Tom compared the post-1990 portion of a “combined” data set in which the only contribution came from 5-10 Briffa cores to the Briffa data set in which the post-1990 contribution came from the same 5-10 Briffa cores. The two were virtually identical!! Tom proclaimed the robustness of the Briffa data set once again.
Gavin had found a soulmate.
Then, Tom proclaimed that the Schweingruber series were no good:
Rejecting the Schweingruber series as a good proxy seems reasonable
While Gavin may agree with his mentor on this matter, I suspect that Briffa’s endorsement will not come quite so quickly. Y’see, Briffa has pretty much built his career on analysis of the Schweingruber network and the Khadyta River, Yamal (russ035) site is used as a proxy in virtually all of Briffa’s publications. The CRU website entitled “Keith Briffa & Tim Osborn: Tree-ring data lists a series of Briffa’s publications relying on the Schweingruber network disdained by Gavin’s new guru: Briffa et al (Nature 1998a); Briffa et al (Nature 1998b); Briffa (JGR 2001); Briffa (Holocene 2002) and Rutherford et al (2005), the last being a joint venture between the Jones-Briffa team and the Bradley-Mann team. The gridded version of the Schweingruber network developed in Rutherford et al (2005) was used in Mann et al 2008.
So I don’t expect Briffa to immediately join Tom P and Gavin in condemning the Schweingruber series as not being a “good proxy”. My guess is that his response, whatever it is, will take a different approach. Just a guess.
The next idea of Gavin’s guru was that joining Schweingruber’s Yamal data with the Yamal data with the Russians was an invalid comparison and was “comparing a signal to noise”:
The Schweingruber series is therefore of very limited utility for a valid comparison with the much longer-lived trees of the CRU archive. Your earlier sensitivity test is comparing a signal to noise…
All you have done is inject noise into the Biffra/H&S series by adding in much shorter lived trees. This also explains why the Schweingruber series did not well correlate with the instrumental temperature…
It’s the duration of the tree cores that’s important to extract a long term signal. The CRU archive during the overlapping period with the Schweingruber series has much older trees in it, as you have already pointed out. Your sensitivity analysis replaces longer- with much shorter-lived tree cores and hence obscures the longer-term signal.
On Sep 29, Tom P observed that he was “patient enough to let Steve to plot his own data when he’s ready”. Awfully generous of him. However, Tom did not live up to this undertaking. A couple of days later on Oct 2, not resting on the laurels of his two previous proofs that identical series had “excellent correlation”, Tom suggested that I carry out a sensitivity analysis only using “trees with ages above a certain value”.
Of course this begs the question of a sensitivity analysis based on recalculation of the Briffa Yamal plot only using trees with ages above a certain value. It would be very useful to see how sensitive the shape is tree age – we’d see how the snake bends as its bones grow older…
A little later, he asked Roman or I do the analysis, noting that it could probably be done easily with the tools that I had provided online (and indeed it could.)
RomanM and Steve McIntyre, Are either of you willing to do the sensitivity test I propose above? I believe this would be unbiased and potentially publishable work. I’d love to get stuck into R and do this, but I face a steep learning curve and little time. I would guess it wouldn’t take many tweaks to the code to filter on the tree-record field to achieve his.
In the early morning of Oct 3 (3:51 am blog plus two hours), less than 24 hours later, Tom started getting impatient that room service had not delivered his requested sensitivity analysis, while still acknowledging the possibility that the kitchen might be otherwise engaged:
The more I think about this core-age sensitivity test, the less patient I am to see the results, but I sympathise with your time constraints!
Less than 7 hours later, Tom P reported to realclimate that he had “lost his patience” with the totally unacceptable delays from room service and “kludged” my code to the sensitivity analysis that interested him:
3 October 2009 at 10:47 AM
I’m afraid I lost my patience and have kludged Steve McIntyre’s code to do my sensitivity analysis (code is posted on Climate Audit).
No wonder Tom was outraged. After all, it was 24 hours since he put in his order. Next time, I guess Tom will order up his data and sensitivity studies over at CRU. I’m sure that he’ll be pleased at the service.
In making this derogatory remark at realclimate, Tom failed to mention that, despite the fact that he had never previously used R, that the tools and turnkey scripts that I had provided enabled him to carry out the desired sensitivity study with an almost immediate turnaround time. Didn’t even leave a tip for room service.
Indeed, Tom was so dissatisfied by the performance of room service that he published his complaint at realclimate before he published a virtually identical complaint at CA a few minutes later (CA blog time plus 2 hours). “Tom’s” code wasn’t actually posted at CA when he notified realclimate. Nor was it available online for a number of hours as it was caught in the filter until I manually retrieved it later in the day. Nor did “Tom’s” code as posted actually generate his sensitivity study. His code as posted was a copy of my code except for one line that didn’t do anything in the code as placed online.
Tom’s supposedly “demanding” test is set out in this comment , linking to a set of figures, of which a recon using Yamal trees over 75 years is compared to the Briffa chronology. In the latter portion of the graphic, the 10 Briffa trees are all well over 75 years in age so that this portion is identical between the two comparisons, a similarity that Tom, once again, finds remarkable.
Figure 1. Tom’s comparison. Original here.
As noted above, Gavin Schmidt, applying normal Team standards of due diligence to results purporting to contradict me, immediately endorsed the findings of his new tree ring guru at realclimate.
Tom P. above showed that the Yamal curve was robust to homogenising the age structure, and frankly I have a lot more confidence in Keith Briffa to do this right than I have in McIntyre. – gavin]
This finding quickly spread around the internet.
However, while Gavin might well have “more confidence in Keith Briffa” to do this calculation right, the calculation was neither done by Keith Briffa nor endorsed by Keith Briffa. It was, in effect, a calculation that, in the terms of Gavin’s head post, appeared “randomly on the internet” with, as far as I can tell, neither Gavin nor any of his colleagues running the calculation by any practising dendros before endorsing them. Briffa might have given them a different answer, as I’ll discuss below.
The underlying complaint of Gavin’s guru about my inclusion of the Schweingruber data was that it had a younger population distribution than the CRU Ten. Drawing on several hours of experience in internet debate on this matter, Gavin’s guru (GG) opined that the inclusion of this younger data added “noise” to the “signal”. Tom proceeded to further complaints about room service, observing that my comparison was “rather uncooked (to be kind)” and that I had “left behind quite a mess”.
Steve McIntyre’s critique was rather uncooked (to be kind) but he appeared to be in quite a rush to find fault and has left behind a bit of a mess.
We saw above that Briffa might not be quite as quick as Tom and Gavin to trash the Schweingruber network. Nor do I expect Briffa’s eventual line of response to follow this approach. Y’see, Briffa’s own protocols for carrying out RCS standardization explicitly preclude the sort of age truncation proposed here by Gavin’s guru, requiring both a “large population” and “a wide range of different tree ages, each distributed widely through time and all drawn from a single species in a relatively small region”. Here is an excerpt from Briffa et al 2001 url :
I also urge readers to read three other Briffa publications on RCS standardization:
Briffa et al 1996 stated:
Unfortunately the RCS approach is far from being a general panacea for the loss of long-timescale information in dendroclimatology. …
The crucial factor, as so often in dendroclimatology…is replication…. In general, much larger sample replication is necessary to achieve an accurate estimate of long-timescale variability in chronologies …”
Briffa and Cook 2008, among other caveats, stated:
We need a different mind set as regards sampling: sample numbers an order of magnitude greater than the “commonly perceived” need for 15-20 trees should be targeted, even at a single site level.
Briffa and Cook recommend the development of regional networks – including Khadyta River would seem to be precisely the sort of thing they have in mind (and did over at Taimyr):
Regional networks of such well-replicated data should be developed – and if possible from different ecological situations (range of elevations, aspect, substrate type, etc.).
Aside from issues of sample size and age classes, Briffa explicitly warns against exclusive reliance on codominant trees in the modern portion, urging a representative age class distribution condemned by Gavin’s guru. Briffa and Cook state 2008 state clearly:
We should not sample only dominant or co-dominant trees and not sample only the oldest trees.
They go on to observe that such sampling will result in “modern-sample bias” (one manifestation of which is described at length in Briffa and Melvin 2009):
The results may, therefore, be affected by a modern-sample bias, bearing in mind the common practice of sampling dominant or codominant trees (Schweingruber and Briffa 1996).
And yet the CRU Ten (originally selected by the Russians) clearly consist of 5 dominant or codominant trees from two different sites – exactly the sort of truncated network criticized here by Briffa on this second count.
There are many other interesting observations in both Briffa and Cook 2008 and Briffa and Melvin 2009. Given Briffa’s failure to archive his Yamal data for nearly ten years after his original publication, his refusal to provide the data to the authors of D’Arrigo et al 2006 and his serial refusal to provide the data to me, the following statement to the PR Challenge conference is remarkable:
The ITRDB is a great resource. It needs to be continually improved to allow easy storage of other than “usual” tree-ring width data. Improved meta data should be sought for all submissions, including tree dimensions and architecture and information on context of measurements (routinely including estimates of missing rings to pith). When standardised indices are archived, precise details of standardisation options should always accompany them. This should include detailed output from the programs used for standardization, auch as the ARSTAN program. Only in this way can others replicate how standardized tree-ring chronologies were developed.
As I’ve observed to readers in the past, when dealing with arguments from the Team (and, I guess, Team groupies and gurus), you have to watch the pea under the thimble. The Gavin guru graphic presented above shows a chronology for trees older than 75 years, but did not show the corresponding graphic for trees younger than 75 years that would support his assertion that the younger trees contributed nothing more than “noise”.
The two graphics below compares chronologies for old (more than 75 years) and young (less than 75 years) for the combined Yamal and Khadayta River data sets. The first shows them for the period from 1800 to the present. As you see, up to the 1970s, the young tree chronology is, shall we say, remarkably similar to the old tree chronology. Thus, if we stipulate along with Gavin and his guru that the young tree chronology is simply “noise”, a similar characterization also applies to the remarkably similar old tree chronology. The most distinctive feature of this graphic is something quite different than the guru reported to us: given the similarity of the two series up to 1970, their divergence thereafter really is quite remarkable. Readers should not presume that it is the young tree chronology that is the exception. Tomorrow I’ll show this with a graphic comparing these two series to the NSIB temperature reconstruction of Briffa et al 2001.
Figure 2. RCS Chronologies calculated separately on old and young trees. See script in comment 1 for further details.
Here’s the same plot for smoothed versions (21-year smooth) of the two series, again showing coherence up to the 1970s, once again showing remarkable coherence between the two series until the 1970s.
As noted above, while Tom P has an undoubted knack for showing that identical series have excellent correlation, his “sensitivity study” proves nothing relevant to the issues surrounding the Yamal data set, except perhaps that Gavin’s opinion on the matter is “not worth much” and their total lack of due diligence in failing to spend even a few minutes examining whether Tom knew what he was talking about, whether his claims held up or even whether his claims were consistent with policies set out in Briffa’s own publications.
Climate Audit 
561 Comments
Here is the code to produce the two new figures. Data uploaded as in other post:
tree=rbind(yamal,russ035)
temp= tree$age<75
chron.old=RCS.chronology(tree[!temp,],method=”nls”)
chron.new=RCS.chronology(tree[temp,],method=”nls”)
#recent plot
ts.plot(ts.union(chron.old$series-1,chron.new$series-1),col=1:2,xlim=c(1800,2000))
legend(“topleft”,fill=1:2,legend=c(“Old (+75 yrs)”,”Young (-75 yrs)”))
title(“Yamal plus Khadyta River”)
#longer plot
ts.plot(ts.union(f(chron.old$series-1),f(chron.new$series-1)),col=1:2,xlim=c(800,2000))
legend(“topleft”,fill=1:2,legend=c(“Old (+75 yrs)”,”Young (-75 yrs)”))
title(“Yamal plus Khadyta River”)
Re: Steve McIntyre (#1),
This was a precursor to the full sensitivity test. The comparison of trees greater than 72 (not 75) years to the full dataset was merely to take eight specific trees out of the distribution which fell off the line of samples here:
http://img194.imageshack.us/i/yamalagedistribution.png/
As I wrote at the time:
There is nothing remarkable about this particular result at all. Except that for some reason it now forms the basis of you selecting a 75-year threshold, and that alone, for your comparisons in your later plots.
And why did my threshold of 72 years become your 75 years? Might I suggest 75 years sounds a little more of a general choice than the value of 72 years I had selected to exclude some specific trees.
The important part of the sensitivity test are by removing larger numbers of trees younger then 100 years, 150 years and 200 years.
Of 252 cores in the CRU archive as I wrote before running this sensitivity analysis:
Removing all the cores more than 250 years old leaves only 32 cores, or an average of 4 at any one time, and insufficient to perform a valid RCS chronology for most of the two thousand years.
Why are you are ignoring these comparisons for trees greater than 100, 150 and 200 years? Here they are again:
Here is where we see the hockey stick hold up despite the removal of the younger portion of the trees.
Does your combined Yamal and Khadayta River chronology show the same robustness?
Re: Tom P (#5),
How have you addressed any of Steve’s points?
1. The briffa problem
2. The two graphics below compares chronologies for old (more than 75 years) and young (less than 75 years) for the combined Yamal and Khadayta River data sets. The first shows them for the period from 1800 to the present. As you see, up to the 1970s, the young tree chronology is, shall we say, remarkably similar to the old tree chronology. Thus, if we stipulate along with Gavin and his guru that the young tree chronology is simply “noise”, a similar characterization also applies to the remarkably similar old tree chronology.
Re: Tom P (#5),
excuse my ignorance, but what exactly your “sensitivity analysis” is supposed to show? And what an earth that has to do with “robustness” as the very few “living trees”, which seem to be the outliers, are old trees?
Re: Tom P (#5),
This statement is confusing to me. Steve’s plot shows that it is the young trees which don’t show the strong late C20th growth. So removing the young trees to make the hockey stick hold up seems to be a complete failure to address Steve’s point.
Of interest to me is that the younger trees seemed to benefit more than the older trees from the strong solar cycles up to 1800 just before both young and old are heavily affected (apparently) by the Dalton minimum.
I wonder which solar indicators might shed light (sorry) on the different and opposite way heightened TSI seems to have affected younger and older trees during the late 1700’s and late 1900’s. Maybe other variables confound the issue too much to draw any worthwhile conclusions.
Re: Tom P (#5),
Tom P.
Interesting graph.
Please document how many total samples in CRU series and how many in each of > 72, > 100, > 150 and > 200 years.
Any suggestions as to the remarkable difference between the hockey stick shape of the CRU series and the non-hockey stick shape of all of Schweingruber’s Yamal data?
Re: Tom P (#5),
Hey Tom, why haven’t you answered the three questions that I asked you? I posted them at RC #458 and again at CA in the Most Influential tree thread at #289. It looks like your are doing the same thing that you are accusing Steve of doing – moving forward and leaving a mess behind. My objection to your method makes clear why your tree removal experiments show nothing at all. You seem to have a habit of ignoring the fact that people are answering your objections and disproving your ideas and you simply move on to the next thing that comes into your head to throw at Steve. Now behave like a responsible adult and go back and answer my questions.
Re: Tom P (#5),
Tom P asks:
Tom, in the MBH discussions, one of Mann’s tactics was to “prove” that the reconstruction was robust to a variety of uninteresting situations, while shouting at the top of his lungs that a sensitivity analysis excluding the Graybill strip bark bristlecones was somehow “mathematically wrong”. In Mann’s case, the reconstruction was “too” robust in the sense that you could replace all the proxies other than bristlecones with white noise and get just as “good” a reconstruction. This seemed an unattractive robustness, but it didn’t seem to “matter” to climate scientists.
In proposing the sensitivity analysis that I did, contrary to allegations at realclimate – that you do not rebut over there – I did not “randomly” pick some data set off the internet. At Avam-Taimyr, Briffa had added a Schweingruber data set of equivalent vintage to the Taimyr data set to increase the population. Schweingruber’s Khadyta River dataset is within the Yamal sampling area, rather than 400 km away, taken by Schweingruber, used by Briffa in at least 5 other studies. The Yamal population is far smaller than the Taimyr population and much more in need of supplementing.
I did this sensitivity because there was an on point precedent, a precedent not merely involving Briffa himself, but in the paper at hand.
In criticizing the under-size of the Yamal population relative to sampling protocols, I am not attempting to suggest sampling requirements to dendro, I am merely applying known dendro procedures (including ones recommended by Briffa) to the Yamal data set.
On the other hand, your sensitivity analysis is one that, to my knowledge, has never been of interest in the dendro literature. Perhaps they should ignore young trees as you advocate, but they don’t.
While this may seem like a good idea to Gavin, who admits to knowing little about the problem, your proposed methodology of only using older trees would have a far-reaching impact on the rest of dendrochronology where young trees have been routinely used for decades. This is the sort of idea that should be presented to field dendros.
Personally I think that dendros will be very reluctant to adopt your idea, meritorious as it may seem to you right now. Your advocated methodology – only use trees that reach a minimum final age of 150 years or 200 years or whatever =- is not one that is used anywhere in dendrochronology. An outcome of your method is that pretty much every chronology in existence would require re-doing. My guess is that dendros would prefer to give up Yamal than re-do every other chronology using only old trees.
IN addition, it seems that you would now be introducing a strange imhomogeneity between subfossil and living cores. Say you have a tree that is 125 years old and your cutoff is 150 years. Who’s to say that the tree won’t eventually live that long – in which case inclusion would be retrospectively required. This doesn’t seem like a great idea.
Your proposed method is such a dramatic departure from present practices that I can’t picture any dendro signing on for your program.
Nor am I aware of ANY article in which your old-tree method has been proposed or used to yield a RCS chronology and a variety of strictures by Briffa against this sort of methodology. CAn you provide me within any references supporting your proposed old-tree method or is this just a personal idea for which you have no dendro reference?
Re: Steve McIntyre (#239),
Steve,
You said in an earlier thread:
This seemed like a reasonable comment, though it seemed a little strange that you should point out a potential problem with the CRU archive without investigating it. I thought it worth seeing if indeed this inhomogeneity “mattered”. The answer – the original Briffa chronology holds up as younger trees are removed from the archive.
Now you say:
How about a bit of consistency here!
You go on to say above:
This is a sensitivity analysis, not an alternative chronology, a point you laboured to make when you first introduced your combined Yamal-Khadyta series:
I’m not advocating ignoring younger trees, but using it as a test to see if the chronology holds up as they are excluded.
While Briffa’s Yamal series passes the test your comments suggest with respect to age inhomogeneity, your combined Yamal-Khadyta series fails such a test. Your combined series produces a different chronology when trees less than 75 years old are excluded from the record.
Finally, do you have any response to my comment (#80) concerning the historical repeated divergence between younger and older trees in your combined Yamal-Khadyta chronology, despite your unsupported claim to the contrary in the head post?
Re: Tom P (#243),
Tom
Why have you made this distinction between trees older and younger than 72 years? Is this a result of advise from dendrochronologists?
Ed
Re: Tom P (#243),
it appears to me that TomP is missing SMc’s point here.
SMc is auditing the quality and utility of these reconstructions. he finds an issue. TomP points out that if younger trees are excluded then the Briffa reconstruction is preserved. SMc points out that in these reconstructions, it is standard practice to use a mixture of ages. Bruffa has stated this in a publication.
So if reconstructions are contaminated with noise by younger trees then this is a common failing of published reconstructions. Hence TomP’s real contribution, if his results are confirmed, is to put into question the validity of common reconstructions. The community that created them to see if they have any utility should then revisit to these reconstructions
Re: Tom P (#243),
The problem is your “tests” are irrelevant. You admit as much here. You suggest there are no consequences of your finding. Who runs tests that have no consequence? Tom PP – that’s who.
Re: Tom P (#243),
There are a number of points you are obviously incorrect on which I’ll point out for other readers – since you’ve failed to see reason to this point.
This image by Steve M on the other thread shows the age of the different reconstructions.
While the Schweingruber trees (green) are of shorter age than the trees they replaced (black – original Yamal series age) the resulting series (red line) is quite a bit more homogenous.
Now if you were a clever guy with a PhD like Tom, you might try to trick people by realizing that the tail in the RCS correction will give an unmodified signal at the end point of each tree. In the middle of the series there are overlapping trees to compensate so no problem there. Being a clever guy you might further realize that the most recent trees are much older than 75 years and setting the cutoff to something like 150 will eliminate most of the older data. You could then claim that the 75 year trees are noise and say the reconstruction is robust.
Of course this is a weak and in my opinion disingenuously put forth argument which like the other two previously discussed arguments attempts to exploit peoples understanding of the detail of the math. Consider that despite the known deficiencies with RCS handling of different age trees Tom’s new position ends up advocating that the Yamal series must have older trees in the recent years than in the historic portion to be accurate.
Does anyone here think that even though Tom didn’t present the less than 75 year old trees he didn’t plot them? Just how did he determine they were noise? Now he’s stomping his feet that nobody addressed that point even though the headpost has already done so. In fact it was done with what in my opinion was a very surprising plot:
I have to say if you would have asked me to guess what the 75 year plot would look like, I would have said >75 years would make a hockey stick and < 75 wouldn’t but in the historic portions of the signal there would be no match whatsoever. Think about how strong Tom’s argument would be if that were the actual case. Instead I was wrong and there is a visually high correlation between different trees in the historic signal but RCS overstates old tree signals in the blade. Sure many of the readers have more experience than me and probably weren’t surprised but I really didn’t expect that much correlation.
The Briffa style Yamal series requires RCS and requires a special set of trees to make the blade of the hockey stick. It’s absolutely clear at this point, fighting it makes no sense to me whatsoever. The data and math are both very clear. Roman’s mean plot Re: romanm (#63), ends the whole thing in my mind.
Re: Jeff Id (#250),
So rather than signal being contaminated by noise as Tom P suggested, it was actually almost the reverse – an upward bias being being “contaminated” by signal when the younger trees are included.
Re: Layman Lurker (#251),
This was, indeed, the conclusion reached earlier. Not on this thread. It was a different one.
Re: Layman Lurker (#251),
That seems pretty close to me.
I’ve got an idea for a reasonable method to do a tree reconstruction now that I’d rather be working on than actual work. Perhaps tonight.
Re: Jeff Id (#250),
Not at all. The end effect issue you mention is known to Briffa and other dendroclimatologists. If you think it is a problem in the Yamal dataset, there is a simple test: try cutting off the data completely at earlier points, say in 1900, and seeing if including or excluding trees younger than 100 years old at that time produces an artificial hockeystick. Until this is shown, it’s only a case that there “might” be a problem.
So your argument against me rests on you eyeballing rather than calculating this difference! Steve McIntyre made the same mistake, but at least he has the excuse that he hadn’t actually looked at the numbers. Here is the calculated difference between these signals once again:
Not really a great match.
There is visually a correlation between the two signals, but this is only at high frequencies. There is no reason for you to be surprised – the trees are growing in the same area and you might expect to see the signals match at shorter time scales.
“Special” in what way? Steve McIntyre has withdrawn his allegation that Briffa cherry picked the Yamal set, so these trees appear only to be “special” in that they show a growth spurt in concordance with the instrument record. Should they be excluded only on that basis?
Re: Tom P (#255),
Spot the falsheood initiated by Gavin Schmidt that has since spread through the blogosphere.
Re: bender (#256),
That’s a slick one. Steve notes an issue in very careful language. Moonbats run around with their hair on fire claiming ding dong the wicked stick is dead. gavin puts the moonbat words in steves mouth. Steve points to what he originally said and
WALA… you have a manufactured recantation.
Re: Tom P (#255),
Tom, can you clarify for me which instrument record you refer to here.
Re: Tom P (#255),
It’s actually quite a good match. There are only a handful of points which exceed 1 SD. 2/3 or more are within .5 SD. I’m not sure what filter or smoothing this is, but there are quite clearly much less than 500 points between 500y lines. Compare this to the figure one in the head post which shows your earlier graph, which probably has a very similar structure (and should have) if it were compared apple – to apple.
Re: Tom P (#255),
I’m having trouble reconciling Tom’s graph in #255 with Jeff ID’s second graph in #250. Am I the only one who doesn’t think they look like they came from the same data? For one thing, in Jeff’s graph, the younger trees appear to stop well before 2000, but Tom’s difference goes all the way to 2000.
Re: Mike B (#266), I had the same issue with it. Since Tom P has a history of buggy use of R I basically don’t even look at his stuff anymore. he failed to admit his errors, doesnt post and comment his code, he’s just a distraction. Class clown.
Re: steven mosher (#268),
Noted.
I’d still like to see an accurate version of the graph Tom attempted in #255, or a reconciliation of the graphs in #250 and #255.
Since I don’t see any volunteers, I may have to slouch towards R.:-)
Re: Mike B (#338),
#250 is a plot of average ring age over time.
Re: Mike B (#338),
I’m not sure why you think my plot is inaccurate – it is simply a subtraction of the two series of the earlier double plot in the head post over the full two thousand years. The reconciliation is straightforward and I’ve already given it:
But to elaborate, the eye concentrates on the spikes, which are aligned in the double plot, but not the difference. By the way, this is why jpeg compression works – it throws away the longer scale information in an image but leaves the shorter scale information which defines sharp edges. The apparent visual agreement of the double plot is really an artifact of the way our brain processes images.
The difference therefore needs to be plotted to be properly appreciated visually:
The two signals are only coherent over short periods, but incoherent over longer timescales – hence the pronounced divergence seen continually over the entire history of the difference signal in direct contradiction to the statement in the head post:
In this case the problem is with the younger-tree series over longer time scales – the segments for the RCS chronology are too short to follow the longer term variations. Why Steve McIntyre prefers the younger-tree series is not obvious, as he recognised when he wrote in the head post:
But my difference plot rather undercut the basis of the analysis Steve proposed to show. The promised graphic failed to appear the following day.
The inconsistencies in Steve’s analysis together with the information from the Russians who actually did the field work have undermined the original case here against Briffa’s Yamal chronology.
I think it very possible that we will see a post from Steve quite soon in effect saying that he only wanted to show that there might have been a problem here. It’s all a far cry from his presentation of
which caused all the fuss in the first place.
Steve: Nothing that you’ve presented “undermines” anything. My surmise that the living cores in the CRU archive had been selected for length in accordance with corridor standardization protocols has been confirmed. The replication at Yamal is far too low for RCS standardization and results were unstable to inclusion of a relevant Schweingruber series. Most importantly, the previous rationales for picking Yamal over Polar Urals have fallen apart.
Re: Steve McIntyre (#328),
Could you then, finally, explain why the divergence between the young and old tree chronologies occurs not just recently, but continually and to an even greater extent in the past over the last two thousand years, in direct contradiction to your claim in the head post?
Re: Tom P (#332),
I’m not sure I follow you. The differences between two series are dropping out whatever common means they might have. Then, there are left (possible) idiosyncratic expected value differences and (possible) divergences between these at short and long frequencies, plus the difference between their idiosyncratic random components, which difference will have summed variance.
You are suggesting that these series have “repeatedly diverged in the past.” On the basis of what statistic? I see some things that look like high frequency wiggles, but I know enough about this sort of thing not to trust my eyes. Do you have a test statistic in mind for delivering on the assertion that these series repeatedly diverge in the past?
When you difference the two series, you remove whatever mean (really, expected value) the two series share at any point in time. Hence, you remove any pattern in their shared expected values across time. So, there is no way to judge from the difference graph whether the (apparent to the eye, but not trusted by anyone used to doing stats) high frequency comovements (you ask us to see as apparent, which I do not think are) are big or small relative to any shared high or low frequency comovements in expected value across time periods.
Try the following. Create two time series of length T = 2000, t = 1,2,…,2000. Make them identical stochastically:
X(t) = 0.003*(t-1000) + ex(t), and
Y(t) = 0.003*(t-1000) + ey(t),
where ex(t) and ey(t) are both i.i.d. standard normal sequences, independent of each other too.
Difference them and graph the difference. Does it look all that different from your graph? Yet the correlation coefficient between the two sequences will be a fairly high 0.75 or so. The human eye easily sees things that large. Yet the graph of their difference will look like noise, because it is noise by construction.
Re: Tom P (#332),
You keep repeating this mantra, but I don’t see it at all. As I pointed out before, the vast majority of the points are inside the +-.5 SD limits and only a handful are outside the +-1 SD level. This is too little to be of significance, AFAIK. I’m not Stats expert though and would like to see what one of the several resident experts has to say.
Re: Dave Dardinger (#336),
it’s actually totally irrelevant how big those wiggles are.
Tautologically, we can write the two series as follows:
O(t) = ES(t) + EI(t) + oe(t), and
Y(t) = ES(t) + ye(t), where:
O(t) = the old tree ring width at date t,
Y(t) = the young tree right width at date t,
ES(t) = The “shared expected value” of O(t) and Y(t) at date t,
I(t) = the “idiosyncatic expected value” of O(t) – Y(t) at date t, a.k.a. the “mean divergence at t,”
and oe(t) and ye(t) are i.i.d. random fluctuations uncorrelated with everything else, perhaps a combination of measurement error and the influence of unobserved (by us) factors (rain, nutrients, wind, etc.)
Notice that when we take the difference O(t) – Y(t) we eliminate ES(t). So whatever pattern was present in ES(t), however big that might have been relative to the variance in EI(t) + oe(t) – ye(t), it has been eliminated by differencing. In other words, make EI(t) + oe(t) – ye(t) whatever you want, and give it whatever pattern you like across varying t: A graph of it tells us nothing about how big it is relative to the variance of ES(t) across periods.
In other words, Tom P’s graph is wholly irrrelevant to the shared variance of the old and young trees. QED.
Re: NW (#337),
Ok, and just to show I read your post carefully, You have I(t) instead of EI(t) at the start of the line defining “ideosyncatic[sic] expected value”. You’re undoubtedly correct that Tom P is trying to throw out the nut with the husk, but I’m still trying to figure out what he thinks he’s seeing in his 75+ – 75- graph?
Re: Dave Dardinger (#338),
Yeah, I(t) should be EI(t) everywhere, and I misspelled idiosyncratic.
I THINK (but I am not sure, so Tom P should confirm) that he thinks he sees that EI(t) has recurring high frequency divergences from EI(t) = 0 in the past, so he thinks that the evidence that “recent EI(t) is nonzero but past EI(t) is basically zero” (I THINK Steve’s assertion, but Steve should confirm) can be rejected. Frankly, I don’t see how that is visually apparent from Tom P’s graph. Hu may have better visual intuition for these things than I do, being a time series guy. I don’t see it. Randomness is more “streaky” than our intuition suggests, a point originally laid out in the early years of Kahneman and Tversky’s “heuristics and biases” research program. So I tend to be very skeptical of visual analysis of long time series.
Re: Dave Dardinger (#338), You and me both. I don’t think he gets what he’s doing.
Re: Tom P (#328),
Tom, thanks for the response. The main reason I questioned not only your graph in #255, but also the second graph in #250 is that your graph goes all the way to 2000, and it does not appear in #250 that both of the series being differenced go all the way to 2000.
I’m not saying your graph is wrong, I just don’t see how the graphs in #250 and #255 can be based on the same data.
Can you please link a table that has the two series and the difference? Or your R code that created the graph?
Re: Mike B (#344),
The R code for the separate plots is in the first posting. For the difference it’s simply:
ts.plot(chron.old$series-chron.new$series))
You really need to quantify the coherence between the two signals to see precisely how the agreement is breaking down. Here is a plot of the coherence against the log of the signal frequency in 1/years (-2 corresponds to 100 years,-1.5 to 30 years and -1 to ten years)
This shows that the agreement between the older and younger tree chronologies becomes progressively poorer at periods much beyond a couple of decades.
Re: Tom P (#346),
Is this your work? Or did you get somebody to help you?
Re: bender (#347),
It’s Tom’s which begs a few other questions 😉
Re: Jeff Id (#349),
No, I’m serious. The guy knows nothing about R on day one and now he’s an expert on spectral coherence?
Re: bender (#351),
No need to express surprise. I’ve worked a fair amount on time-series analysis, but hadn’t used R before – it’s not such a common language in physical science and engineering.
Here’s the code:
oldnew.ts<-ts.union(chron.old$series, chron.new$series)
oldnew.ts<-remove.NA(oldnew.ts)
coheroldnew=spec.pgram(oldnew.ts,spans=c(3,3))
plot(log10(coher$freq),coheroldnew$coh,type=”l”,lwd=2, xlim=c(-2.5,-0.5))
title(“Coherence between the older and youger tree chronologies”)
The fall off in coherence at longer periods shows that the younger trees are not a good choice by themselves for a long-term chronology. It’s not just a recent divergence between the older and younger trees that causes the problem here.
Re: Tom P (#356),
If you have a coherent argument, package it up and present it. You are not gaining an audience, but if you have something worth looking at, you should. I’m trying to help you.
Re: Tom P (#346),
I’m a bit confused on just what is being compared here. Steve’s first reply below the head post uses, as far as I can see, the Old series to be those trees (in Yamel + russ035 older than 75 years while the new series is all of them (i.e. including the young trees.) So only trees younger than 75 years will keep the two series from being identical. Now I’m not sure how coherence (or signal frequency) is defined in this context. Nor exactly what you’re comparing in your plot, but I’d sure like to see your R code so I could figure it out.
Re: Dave Dardinger (#356),
I see my request crossposted with your posting it. But I do have a couple of questions about the code. What is NA in the second line? And what is the equation that spec.pgram calculates?
Re: Dave Dardinger (#360),
That gets rid of missing values.
Re: Dave Dardinger (#360),
R help page for spec.pgram()
Re: bender (#366),
Thanks for the help, Bender!
OK, I’m ready to try seeing what Tom P’s coherence plot is showing.
First of all, I’d gotten confused reading Steve’s lines defining Chron.old and Chron.new. Well, I got confused on Chron.new anyway. It is indeed just the trees younger than 75 years. While Chron.old is all trees older than 75 years. Now these trees of either age range can be found in any part of the tree series consisting of Yamal plus russ035. So the difference between Chron.old and Chron.new is defined at every point, at least in theory. And I gather that the NA procedure Tom P uses fills in missing values by some method which is considered reasonable, though there’s been some discussion here about what’s reasonable.
So we have a nice plot such as in reply 191 or 255 above. But now we come to the graph in reply 346. I’ve been reading a bit about periodograms but I still can’t quite figure out what this graph shows. For one thing, AFAIK the series in question are annual series. And if Tom’s description of the log values are correct, then the values on the right hand side would be for just a few years yet there are many many points. Where do they come from? I realize we’re dealing with somewhat of a statistically imaginary construct here, but it still boggles my mind. If someone can explain just what we’re looking at here, I’d appreciate the education.
Re: Dave Dardinger (#370),
Tom P will have to confirm this surmise, but I think that the coherence is roughly the squared cross correlogram between O(t) and Y(t), that is squared correlations between O(t) and Y(t+k) at varying lag lengths k. k (transformed as log(10)[(1/k)]. It is a little bit more complicated than that, because it isn’t a “raw” correlogram that is providing the input, but rather something filtered through a small window. Is this in principle correct, Tom P?
Re: bender (#364),
if this really tells us something about old versus young series, instead of something about any two series, there is a simple bootstrap to show it:
1. Take the set of all J series, and randomly divide it into two sets of sizes N and M, N+M = J, where N and M are the number of series in the old and young series subsets.
2. Recompute the coherence plot in Tom P’s figure in #346.
3. Repeat 1000 times. Select out the 5% and 95% centiles of the coherence at each k (lag length). Overlay these on the graph in #347. If the periodogram lies between these most of the time, then there is nothing special shown in the #347 figure about old versus young tree series. Rather, it is within the behavior exhibited by any two randomly chosen subsets of tree series.
Re: Tom P (#346),
so, what is the definition of coherence here, and what are we looking at here. Is this supposed to be a cross correlogram with an odd scale, or what? Between what? I prefer math as opposed to R code.
Re: NW (#359),
The variation of coherency against frequency tells you how well two signals are correlated at a particular period of the two signals. It takes a value between 1 (perfect coherence) and 0 (no coherence).
If a signal agrees over a range of periods, say from yearly variations to centennial variations, the coherence should be constant and close to a value of one over that range. Steve McIntyre said
In fact, as I showed earlier by just plotting the difference between the two series, the agreement between the two series is continually poorer over the last two thousand years than recently. The coherency plot explains why: although the coherency is above 0.9 on average for periods of less than a decade, the agreement drops away when longer periods are considered.
We now have another way of testing Steve’s statement above. Here is the plot for the data up to 1960:
It’s not very different from the full dataset – the last thirty years has not noticeably reduced the coherence between the younger and older trees.
Why is this important? As Steve said:
This graphic is still to appear. But we see that the young tree chronology, without the hockeystick, does not agree well with the older tree chronology at just the periods, multidecadal to multicentennial, which are necessary to be able to see identify a hockeystick.
Steve McIntyre is doing a Nelson, putting the telescope to his blind eye so to avoid seeing a clear signal.
Steve: Tom, merely because I don’t respond to every one of your spitballs doesn’t mean that I concur with them. I’m working on other material and don’t have time to deal with your stuff right now. It’s also Thanksgiving weekend in Canada. In due course. But please dial back your rhetoric. I really don’t like responding in kind.
As I noted elsewhere, as far as I can tell, your methods are totally unknown in dendro literature. Can you please provide me with any precedents for this sort of argument in any dendro literature?
Re: Tom P (#372),
Ok. I’d like to emphasize that differencing the series shows nothing at all about the correlation between the series since it by definition removes their shared component at every time t. The variance of the shared component could be very big relative to whatever variance is left over. The squared correlogram (coherence) is a step in the right direction.
But, I still want to know whether the old/young pattern in the coherence plot is special, in the sense that it is significantly different from what we’d see by doing the same exercise on randomly chosen subsets of chronologies, instead of the two particular subsets called “old” and “young.” I outlined a bootstrap for answering this question in #371, and would be interested to see the results. If it is not, then the pattern of coherences across lag lengths you show in #346 is within the universally expected pattern for two randomly chosen subsets of the series. In that event, the evidence in #346 is within sampling variability and no more justification for selecting old versus young trees than any other potential labeling of tree subsets. Or at least so I would argue: Do you agree?
Re: NW (#373),
In this particular case it should be fairly easy to set up in R using the code at hand. Break up the trees into say 25 year age increments and then combine them randomly. If Tom P is right then it should be obvious to the random onlooker what he’s getting at. I suppose I should set up R on my new computer and try it, but I’m sure someone else could do it quicker. There could be a quirk or two like how to say “greater than or equal to” instead of just “greater than” that would have to be learned.
Re: Dave Dardinger (#375),
I’m sure it is perfectly easy to do for the R-proficient. Me, I’m off to Australia to give some papers, and so say goodbye to this fine fun for at least the next 36 hours or so. Happy motoring!
Re: Tom P (#372),
I’d take care of this one for Steve, but nothing else I’ve written has apparently been correct in Tom’s eyes so what’s the point.
Re: Tom P (#372),
and by doing so, Nelson went on to win the Battle of Copenhagen. How apposite! I look forward to Steve’s more accurate gunnery blasting you out of the water.
Re: (#372),
Steve McIntyre, October 9, 2009:
May I refer you to Steve McIntyre, October 1, 2009:
Can I expect a response after the weekend to the persistent lack of agreement between the younger and older trees of your combined chronology despite your claim to the contrary?
Meanwhile, enjoy you Thanksgiving.
Re: Tom P (#390),
Can we expect a definite response after the weekend? You have been asked more than once.
Re: Tom P (#390),
Steve M asked if the profound inhomogoeneity in age structure matters. Tom P confirmed the inhomogeenity existed, and then I think he thinks he’s shown that it does “matter”. I think he’s shown that it can be *made* to matter. But then you have to ask: what does “matter” mean in this case? Which is why Steve put that word in quotes to start with. It matters how you define “matter” – and Tom P has yet to explain what he’s doing and why. Whereas Steve’s posts always clarify why he’s doing what he’s doing. He always closes by explaining the implication of the result.
.
Tom P, please package this up so that it is coherent. If it’s coherent Steve might even let you have your own post.
.
I think Tom P’s objective is to win a game of “gotcha” with Steve. What he doesn’t understand is that that will always blow back in your face if your analysis isn’t sensible. That’s why I call him Tom Petard – for the stink-bomb he left at RC that caught Gavin off-guard. If he’s doing it again, well, watch out Gavin, the great guru’s got game.
Re: Tom P (#372),
Tom P, there is one problem with this bit of your argument. Shall I tell you what it is, or let you run with it?
.
Nah, I’ll be nice. The coherence plot is not going to detect “decoherence” (i.e. divergence) in one piece of the time-series, because it measures average coherence across the full length of the time-series. Therefore your “test” is insensitive to the claim Steve is making re: divergence. The divergence is localized to just one piece of the time-series.
.
But you knew that, right?
Re: bender (#398),
The average coherence over the entire record is not totally insensitive to incoherent stretches. It is always reduced by such. The only question is by how much.
The troubling aspect of Tom P’s presentation of results–besides the illogical log scale of the frequency–is the enormous variability at the reasonably coherent high frequencies. This indicates inadequate degrees of freedom in the estimation procedure. While he claims values there over 0.9, more robust (dare I say that?) estimation would probably show nothing much above 0.8. Thus, at best, only ~0.64 of the variance at those frequencies is inter-related. Also, it’s doubtful that the hypothesis of zero coherence at the lowest frequencies (so vital to climate variations) can be rejected with any high degree of confidence.
Re: bender (#397),
What matters in this debate is that energy is matter. But Einstein told us that a long time ago.
Re: sky (#406),
I think this is pretty much dead on the money. Sky, I THINK from the documentation I find online that “coherence” is already a squared correlation, so maybe coherence = 0.8 is essentially already 0.8 of the variance. See here:
http://www.spectraworks.com/Help/mtmtheory.html
But this is more or less a wrinkle and its basically semantics until Tom says what he computed. But when sky says:
…I grok that completely. This is exactly why the evaluation of a graph like Tom’s in #346 cries out for a bootstrap analysis of the confidence we can place in the coherence estimates for ANY division of the series into two sets, what I suggest in #371. Without that, anyone can put up a graph like Tom #346 for any classification of the series into two sets and tell stories about it. That’s not statistics.
Re: Tom P (#346),
Re: Tom P (#346),
Wouldn’t this coherence curve be a devastating blow to the entire field of dendroclimatology? If tree rings do not cohere then what value are reconstructions made from them?
I’ve said it before but I’ll say it again. ToP seems to believe that S.MC has offered an alternative reconstruction to that of Briffa. He uses his young trees argument to attempt to discount this purported reconstruction. S.Mc has done nothing of the sort. He offered a sensitivity test only and most assuredly did not offer an alternative reconstruction. TomP’s coherence evidence support the thesis that trees do not make good thermometers which does not seem to be the point that he thinks he is making.
Re: TAG (#368),
Bingo! Poor low-frequency coherence sounds the death knell of dendrochronology.
Re: Tom P (#255),
You’ve made an allegation that Steve alleged that Briffa cherry-picked. How about giving us a reference. If it’s true show us.
Re: Steve McIntyre (#239),
My skimpy readings of the literature appear to indicate that larches are somewhat unique in their growth patterns with age. I do not know if the findings on larches age growth patterns came after the original Briffa paper on Yamal or not. Could it be that the RCS was at Briffa’s publishing time a well accepted correction method that was assummed to fit all cases?
Re: Steve McIntyre (#239),
Should I take it you are not going to respond to my comment (#80), the historical repeated divergence between younger and older trees in your chronology, despite your unsupported claim to the contrary in the head post?
As I said before, you’ve left behind quite a mess here.
Re: Jeff Id (#272),
What part of Steve’s words of September 28th
do you not understand?
Re: Tom P (#288),
There are no head posts dated Sep. 28. So which thread is this quote from? In any case Steve clearly had already assumed the cherrypicking would have been done by the Russians as of Sep 26, so a Sep 28 post which elided the Russians doesn’t mean much.
Re: Dave Dardinger (#289),
See link in (#271) above.
Steve McIntyre’s original comment made on September 28th does not even mention the Russians – they only feature in his additional comment on October 4th.
Up to making this original comment, Steve had been rather careful not to accuse Briffa directly of cherry picking. Why he wrote “CRU cherry picking” in this comment to me, only Steve knows for sure. He essentially shifted the blame to the Russians in his comment six days later when he realised what he had written.
Re: Tom P (#290),
There are three obvious reasons in the message itself:
1. Steve started the reply, “It’s too late tonight.” I show the message as 11:26PM [Steve: Add two hours to blog time to get Eastern time. It was 1.26 am].
2. The context was the replacement of Polar Urals with Yamal. This was done by the CRU group so Steve wasn’t perhaps thinking about Briffa vs the russians per se.
3. Since he’d indicated the cherrypicking would probably have been done by the Russians in his initial post on the 26th, he might not have felt it necessary to repeat the whole statement [hence the term “elided” in my post above].
Now on having it brought up later, presumable by you, he did add a note on the 4th indicating his earlier attribution. This should have stopped your bringing up the point again, but no….
Re: Steve McIntyre (#1),
The young-tree chronology is not just simply noise – it contains plenty of information on the decadal signal. Hence there should be good agreement between the older and the younger trees for the higher frequency signals if they are seeing the same growth conditions. The similarity of the positioning of the spikes of your figure 2 and 3 is therefore not unsurprising for trees growing in the same area.
However, as we start to filter out these spikes, the differences between the older and younger trees become more apparent. For instance if the 21-year filter is used (which was your choice until discarded for plotting figure 2 above), we see there is not just disagreement at the end of the 20th century, but also at the beginning of the nineteenth century.
http://img74.imageshack.us/i/yamkhafil21.png/
This disagreement between the younger and older trees is noticeable over many periods in the last 2000 years when a 51-year filter is used to suppress the higher frequencies of variation:
A spurious sudden growth spurt is apparent for the younger cores at around 700 as well as 1800, and there’s frightening drop in growth around 400. Are we to believe these are real and we should discard the older trees in favour of this younger-tree record?
As you look at these two series you state:
Here is a plot of the raw difference between the two series which should show exactly how unprecedented the divergence is after 1970:
Not remarkable at all: there have been differences between the two series over the last two thousand years larger and more sustained than that seen recently.
The young-tree record has been shown to diverge from the older-tree record many times over the last two thousand years for this dataset. As the younger trees of the Khadyta series are discarded, the overall shape of the combined series chronology is insensitive to a threshold for core-age selection. Apart from decadal variations, trees less than 75 years old are indeed the exceptional portion of this combined record.
Re: Tom P (#79), Tom, the major malfunction with your claims here is that, as we see, the divergence only starts happening about 1970, giving us 30 years of data, which should be sufficiently short for a 50-70 year old tree to exhibit the effects of, but you are claiming that 30 years of data is a century scale signal without any evidence to support such a claim. Please start providing references to support your assumptions here.
Wow …now I know why Gavin hates you so much, good job. So did Gavin find his ‘guru’ by reading comments from CA? …I hope not because he describes the CA commentors as the ‘peanut gallery’.
Nice to see Gavin clutching at straws. I have been reading the developments of the past few days and I am enjoying watching the show! Thanks Steve for your work! Your patience and intelligence is impressive 🙂
Steve
You based a lot of this on RC comment 404 (!) dated 4 October 2009 at 5:17 AM
When I read it it stripped out the comments with “my opinion… (is not worth much)” and “TomP showed” and “I doubt very much” and was left with the following paraphrase:
Q: Do McIntyre’s criticisms have any merit?
Gavin’s A: I don’t know
I must be missing something, but how could trees younger than 75 years old provide data back to 1800 (Fig. 2).
Re: Jim G (#6),
The vast majority of trees in the Yamal proxy are dead, as Freespeech mentions above. The trees from the modern calibration period only cover ~1/10th of the total reconstruction period. There is no continuous record of, what some have claimed are, “temperature sensitive trees” and therefore ~9/10ths of the proxy relies on the response of trees whose sensitivity to temperature cannot be determined.
Excellent post. Will Gavin kill Tom P? Will Tom P kill himself? Will Briffa kill both of them? :0
Jim G wrote:I must be missing something, but how could trees younger than 75 years old provide data back to 1800 (Fig. 2)
I’d expect they are talking about dated logs, you know how long they grew by the number of rings and you use dating (either harvesting dates or forensic methods) to figure out when they lived.
I came for the science, but the commentary.. priceless!
You have to give Tom P credit.
Noticing that identical series correlate is a step forward for some.
🙂
Amazing how McIntryre is dead wrong and dendro-newbie Tom P. is on the money.
Really.
Further in the same RC post, Gavin gives a clue to what the Dendro community response will be.
He says “With more source material (which is apparently being processed and added in) the chronology will change somewhat I suppose but I’m happy to wait and see what comes out.”
This suggestst that there will be a new larger reconstruction that will have similar results, though the ground is being prepared for there to be differences. Steve’s points will not be addresased as the community will have “moved on”.
Posted this on RC. I think Jason (George Costanza) Alexander would be the perfect Gavin.
Here’s the last paragraph of Steve’s latest Post….(paragraph follows)
You make a great comedy team. Will there be a series on NBC. Perhaps Jason Alexander is available to play Gavin.
http://www.climateaudit.org/?p=7278
It would be nice to see less snarky quips about Tom P in the comments. He may be wrong or right, he may be jumping the gun rudely as Steve has pointed out, but at least he’s proposing a hypothesis and defending it. That’s science. Shoot at him with facts, not scarcasm…
I think you are wrong here, Mr McIntyre. As I walked by the dendro’s hutch this morning, I clearly heard quite a loud noise. It sounded like ‘La, La, La….’
Re: Dodgy Geezer (#17),
Except that scientist isn’t about “defending” hypotheses, it’s about trying to prove them wrong. If you wanted only to defend your hypothesis, then you would merely stick with the results that prove you correct and conveniently ignore any experiments that might prove you wrong.
Re: Dodgy Geezer (#17),
Agree with this, as I’ve said on another page, Tom P’s analysis is helping me a lot to understand this. I may not agree with his conclusions, but it’s useful stuff. Please keep it civil.
As Dodgy Geezer said – don’t copy RC commenters’ style of discussion – just stick to the facts and hypotheses proposed by Tom P and offer a scientific criticism of what he is claiming. Climate Audit is all about science and showing that in the long run sound science wins over political activism. There is nothing to be learned from unnecessary sarcasm.
This is not RC, this is CA.
Re: Michal Gancarski (#18),
Good point.
So who is Tom P?
In case any one missed Tom’s first comment on Real Climate:
Who needs sarcasm?
AlanB: Yes, I try not to look at RC. I always come away with a feeling of nausea and a desire to throw up.
dodgy geezer – proposing a hypothesis that two identical series are highly correlated, isn’t a hypothesis or even science. A beginner Stats class warns of false correlations as well as points out correlation isn’t causation. But that doesn’t stop “scientists” from stating causation based on observed correlation. I’d like to know how Steve is suppose to use science to point our Tom P’s impatience, light-speed “due diligence”, and summarily dismissing Briffa’s life work in a few hours.
The post is very appropriate and relevant. The ‘commentary’ on RC is inane, reminding post’s you’d seen on a schoolgirl’s Facebook page, not by scientists…
Where are the giants of yesteryear? It wasn’t long ago science giants were always aware that one simple experiment can render their life’s work irrelevant due to unseen error’s in their work or by additional discoveries.
Tom P
Yamal is a reconstruction with a hockey stick that depends on only 10 cores to get it’s 20th Century surge.
The 20th C end segment is neither statistically significant, nor robust, because there are too few cores.
How can removing cores, mostly from elsewhere in the reconstruction, prove it to be robust or significant ?
Falling into sarcasm is very tempting (I do it often enough myself) especially given the unbelievable comments made by Tom P, and some others here and at RC.
I do feel that it risks dragging this blog in the direction of RC (although that is a long way away) and it is not really necessary, as it tends to undermine the strength of Steve’s argument. A strong argument needs no emotion, just a clear head.
I emailed the BBC’s environment correspondent, last week, suggesting that he takes a look at recent threads here, in the light of their continued zeal around AGW. Nothing back yet. Deafening silence (oops, slipping there).
Keep on keeping on, Steve. You are doing a superb job.
Really while the Tm P bashing is amusing, I think you should skip it entitle this Briffa vs Briffa since it seems clear that Briffa himself disagrees with the methodolgy employed in Briffa 2000.
Are there perhaps 2 Briffa pesonalities, Dr Briffa the statistically cautious dendro and Climate Activist Keith the hockeystick miner?
Okay, I think what I’m about to say is relevant. I made a post on R.C. criticizing Gavin’s ‘caustic’ language towards AGW skepticism and to C.A. in particular. You can view it on the Yamal thread over there. But after reading this, I’m a bit disappointed with Steve’s choice of words towards Gavin and his ‘mentor’ Tom P.
This post is downright sarcastic for a site that prides itself on being above the antics of R.C. I can see how this kind of language (let alone the language used at R.C) could dissuade some people from participating here.
R.C’s unscrupulous methods are a joke but Steve’s words here show the subtle application of arrogance and disdain. I’m sure that some of the opposing comments got to him but it is far better not to comment on Gavin and Tom P. In doing so, you are showing that they are important.
While I am here, I believe that *sometimes* Bender’s posts serve a vicarious function for Steve, in that certain comments fly, that normally wouldn’t. This has nothing to do with the quality of Bender’s posts.
I think an outside view is helpful. Steve’s work is far too important to get bogged down in opinion and payback.
regards
S.G
Re: SamG (#30),
I disagree. If Steve leaves Gavin and Tom P crowing about their “success”, then it will rapidly get repeated on all the warmist blogs and in the press, and many readers who are undecided (and unable to follow the detailed arguments) will assume that Steve was wrong, and that this whole Yamal business is just a bit of Exxon-inspired FUD.
And, while you may say that “R.C’s unscrupulous methods are a joke”, the sad fact is that they have been working rather too well.
Re: fFreddy (#34),
That is precisely why I interject when I do, in real time, when some clown starts making false claims on the basis of nothing. The frequency of intervention, to answer kimberley cornish’s question, will be understood by any fisherman: when you get a nibble you want to keep the poor beast engaged.
The tone of the blog would have been vastly improved had the Lorax nastiness been simply filtered away. Steve’s posts are terrific, marked with very dry humour and courtesy. Bender’s vary between wit on the one hand and signs of automatic reflex responses when someone has managed to get to him. Thus Tom P didn’t need any sort of negative commentary at all; it was obvious he came ignorant of R and a good many other things besides, only to proclaim himself as a Briffa champion on the basis of a straight code swipe from Steve. His own self-righteous commentary was far more belittling of himself than anything Bender added. Just allowing their own self revelations to reveal themselves, uncommented, is usually the best method of dealing with such people. But all in all, it has been a most engrossing thread and probably will be historically important as well.
This is a major story by any yardstick. Maybe I’m just dumb (and, by the standards of this milieu, I’m certainly innumerate) but the axes in the charts presented in the text are not labeled.
I have an idea what they are reporting but I can’t be sure. Can anyone confirm? TIA,
Dave
Mockery can be a useful weapon. Tom P’s “finding quickly spread around the internet” so a devastating riposte was required.
From SM’s Blog Rules: “Sometimes I lapse into controversy, mostly after I’ve been slagged in print somewhere, but I try to stay cheerful.”
Re: DaleC (#33),
Perhaps a ‘devasting riposte’ was necessary, but I believe its best when mockery is not factored into the equation. Its can be a turn off to people in the middle who don’t necessarily understand the debate. Steve’s typical no nonsense approach has been appealing to me as I have followed these debates.
Considering though that Gavin shows no respect for Steve I do understand that it would be hard to always turn the other cheek…esp with him pitching softballs with this Tom P business.
Sam G
One of the greatest problems with the english language is that you need nuance, tone and body language to make it work perfect. In other words, to work in a manner that does not offend all. The written language does not carry any of the 3 important parts of the language and we therefore have a tendency to add those 3 when we read. Sarcasm is one of those ‘feelings’ that we add on reading the script. It isn’t necessarily meant to be in the written word. Be careful how you read it can give more information about you than the author.
Re: stephen richards (#35), Re: fFreddy (#34),
fFreddy
people believe what they want to believe, you can’t blame warmist’s for stupidity and ulterior motive.
Re: SamG (#37),
You’re quite right there, unfortunately. Most of RC’s peanut gallery seem to think that the NAS has confirmed the original MBH hockey stick.
Umm, I’m not sure why I shouldn’t blame people for having an ulterior motive.
But the point is, even if you can’t affect the dedicated warmists, it’s the undecideds you’re after, or even just those with an open mind but not enough time to come here that often. That’s why quick reponders like bender are adding value: they help to make it clear that debate is far from over, and that the sceptical view is held by serious people with good arguments.
Re: fFreddy (#56),
I wasn’t clear. The architects of AGW theory do not decide for the general public. Many things have been written about this but it’s completely off topic here.
Re: fFreddy (#59),
The basic conclusion of Mann et al. (1998, 1999) was that the late 20th century warmth in the Northern Hemisphere was unprecedented during at least the last 1,000 years. This conclusion has subsequently been supported by an array of evidence that includes both additional large-scale surface temperature reconstructions and pronounced changes in a variety of local proxy indicators, such as melting on ice caps and the retreat of glaciers around the world, which in many cases appear to be unprecedented during at least the last 2,000 years. Not all individual proxy records indicate that the recent warmth is unprecedented, although a larger fraction of geographically.
Page three NAS. Surface Temperature Reconstructions for the Last 2,000 Years
http://books.nap.edu/openbook.php?record_id=11676&page=3
Sam G
In MOST respects, I agree with your sentiments, I think!! Real Climate makes me sick when I read it. As a scientist I was involved in many heated discussions but no-one ever stooped to the levels of those posting at RC.
The degree of inconsistency that RC exhibits becomes more astounding every day.
How many times have we been told that replication does not require that all the original material be readily available. Indeed, we have been repeatedly lectured that true replication is not obtained if only the original material is used. Yet, Tom P. could not have carried out his rapid-response function if Steve had not made all the material available.
By RC’s usual standards, Tom P.’s results could have been considered true replication only if he had started from a clean piece of paper, gone out in the field and gathered up some tree rings, properly selected the data to be used for analyses, developed the analysis methods, produced the coding for the solution of the methods, and then carried out the analysis. ( I’ll let Tom P. skip the Verification and Validation procedures as Climate Science does not do these. )
More and more, it seems to me that RC’s views are shaped by what is convenient in contrast to what they say is the correct approach.
By the usual RC standards, RC cannot accept Tom P.’s analyses and conclusions.
/\ Sorry for the confusion. re. fFreddy
IF SteveMc is annoyed with TomP I can well understand it. TomP has made some very basic errors in his maths and then used those errors to severely criticise steve elsewhere. I consider those action extremely rude and in the old days, waranting a smack in the mouth behind closed doors. You do not run to ‘mum’ to criticise someone, you do it to their face and you do it constructively and with solid evidence. TomP has failed on all counts and then run over to Papa at RC to announce his success and SteveM’s failure. That sort of behaviour IMO, deserves evrything it gets, BUT you are right, in the current environment it is better to not stoop to the same level as RC but it will take longer to reach Joe Public if you do not. A good slanging match is what the media loves. Blood and guts is good for news.
Re: stephen richards (#40),
Someone left a snipped comment in another thread about Tom.P and his purpose being to ‘obfuscate’ matters. I felt that at this point he was receiving far too much attention from Steve and others.
Re: SamG (#41),
You can say Tom P received “far too much attention”. On the other hand you got to learn quite a bit about Gavin Schmidt’s judgement and expertise. The backslapping at RC was laughable. I think this whole episode was quite fruitful. We should do it again next month.
.
Great post, Steve.
Re: stephen richards (#40),
Yes Stephen there is a lot posted over a RC where the only appropriate response would be an immediate punch in the nose if said in person. Steve Mc’s restraint is admirable, even if it appears to be through gritted teeth as above.
What of the other (non-Yamal-based) Hockey Sticks Gavin cites?
Re: Rene (#44), those are even more hilarious, of course!!
Re: Rene (#44),
Look out Jeff Id’s shredding of all but the last hockeystick:
and the last one is from Lucia
Re: Rene (#44), the other hockysticks at RC are impressive in showing that they “match” the results called into question by Steve here. I guess the point RC is trying to make is that even if one set of measurements is discredited they have many others that have not been. Cue the Jacksons: “One bad apple don’t spoil the whole bunch, girl!”
As a lay person who is “math challenged” I can’t independently verify any of the data. But it seems to me that Steve has done a valid critique as evidenced by the admission of “one bad apple.” The original researcher is going to take a look at it again, so at least he is responding as you would expect someone trained in the sciences would.
Steve’s critique came only after quite a struggle to get the original data. It took the Royal Academy’s insistence for the data to be published, and when finally done, Steve ran the numbers and found the problems in short order. That process has to be completed for each of the “hockeysticks” provided as “good apples” at CA because, as this controversy shows, climate scientists are not evaluating the data closely enough.
Re: Frank H (#95),
Whether the one bad apple spoils the barrel depends on data and code that is not available because of the same foot-dragging that prevented Yamal from appearing for ten years. Wher eis Lonnie Thompson;s ice core data, for example? Is it ok if he never publishes it? What if it’s a divergent result? Shouldn’t we know?
The several layers of irony involved in this issue are astounding. TomP arrives, makes demands, snarks because he didn’t get his demand fulfilled within 24 hours, is able to make a nice graph because McIntyre published his own code, and concludes McIntyre is “dead wrong”, and “full of hubris”, making a show of what a complete [auto-snipped] someone can be. Well, we’re all only human, after all, and this is to be expected time to time. One could say, it’s a troll, it’s a wannabe. The hilarity only arrives when we see Gavin supporting this gentleman!! Ahh! Long live the internet! Where would we be without this to entertain us?
Could Tom P have been a “Seminar Caller” from the Team? If you don’t know what a “Seminar Caller” is, then you do not listen to Rush Limbaugh.
An impeccable response.
It amazes me that the Team & Groupies think nothing of the consequences of their tactics. Discredit everything and anything if there’s somehow a chance to discredit Steve. This all reminds me of Mohammed Ali rope-a-doping George Foreman. We all know who was left standing. Steve has vividly demonstrated that there are some out there who are just throwing wild punches in the dark.
I’m neither a dendro, nor a statsitician, but I’m wondering why Tom P. assumes that the rings of 70 year old (young!?!) trees grow in response to temperature in a “noisy” fashion? Presumably these trees are affected by the equivalent of teen-age anxiety, and only settle into a regular growth pattern after age 72.
But, then, even a non-statistician can see that removing variability from the data set will probably lead to amazing correlations.
Re: Tamara (#51), he doesn’t need a scientific reason, it’s an unsupported excuse to cherry pick. RC guys like Tom P and Gavin hate it when someone like Steve chops down the cherry tree.
Personally I think Steve’s response is well-measured, witty and deserving. It’s certainly leagues above what we’re all acustomed to seeing from his opponents.
Re: Gavin’s ‘other’ Hockey Sticks
Luis (or anyone) : could you elaborate please?
Eli’s new guru is none other than Tom P:
Quote from Ben Hale’s hockey stick thread
#22: I think Jean S’s point is the only salient one in the comments. Just what exactly is Tom P doing anyway, besides throwing crap at the wall and hoping some of it sticks? If you have a network of trees that on average are not long-lived, and you add in old trees at the end of the record (cherry-picked old trees, no less), clearly this is a bias. A “sensitivity” test would be to REMOVE these older trees and look at the same type of record as has been using for the entire length of the reconstruction. This is what Steve did, although not originally with that purpose in mind.
If you want to look at these new trees and call it a “sensitivity test”, you’re going to have to sample 100s of old trees that cover the length of the entire record. Come on, this should be obvious… arbitrary cutoff points (72, 75, who cares), which are themselves cherry-picked, don’t show anything.
Here’s a writeup about Steve, Briffa, etc. in the American Thinker
http://www.americanthinker.com/2009/10/un_climate_reports_they_lie.html
It may be useful to reiterate on this thread just why TomP’s supposed test of “sensitivity” is worthless. Of the ten living trees in Yamal, 9 have a hockey stick shape. Therefore, unless all of them are removed, the overall reconstruction will have a hockey stick shape (at least at the modern end). Tom’s graphs ended when five of these trees were still in the mix. One would expect therefore that there would still be a hockey stick shape to the reconstruction. This is case as far as the end of the graph is concerned, but due to the thinness of the remaining trees, there were other points which exceeded the end of the graph and TomP used this as an excuse to terminiate the experiment. What’s he shown? Only that as long as half of the only living trees are still in the mix, they rule the roost. In Steve’s terminology, he’s shown that identical time series correlate well. Of course 5 out of 10 aren’t quite “identical” but they’re pretty close in this case. And you don’t need R to show it, just your own eyes looking at the graphs of the 10 trees. Then look at the tree charts by Steve and Conard and see which of the ten trees are still around at each graph TomP did. Answer there are always at least 5 of them and those 5 all have a hockey stick shape. End of story. All the rest is obfuscation.
Re: Dave Dardinger (#53),
Others should thank you for the recap. The main thrust of Tom P’s “analysis” has nothing to do with tree age (Sorry Carl G, but that’s just a sideshow). But of course he may try to steer the story in that direction because he now knows his primary argument – “Briffa’s chronology is robust” – to be ill-conceived. He proved that Briffa’s chronology is internally consistent, which we already knew based on Steve’s plot of the raw data, and which further stregthens Steve’s claim that it is not representative of the patterns in Schweingruber’s chronology. Leading to the question: how did Briffa come to ignore Schweingruber’s data?
.
Is YAD061 “the most influential tree in the world”. Until Tom P finds a more influential one, I do believe Steve may have identified it.
TomP,
In your post you admitted to cooking the data! You selected 72 years to remove some trees causing your results some problems. And this is supposed to salvage your busted credibility??? I seriously hope your day job doesn’t involve science or engineering. Cooking the data is not conducive to rewarding career.
Although we have been discussing the validity of the end proxies, we have not really looked at the effect that the RCS adjustments have on them. To examine that aspect, I calculated the series consisting of the means of all of the proxies available for each year. The comparison between the mean series and the Briffa RCS chronology is enlightening:
The two series have been standardized to have a mean equal to 1 for the period 1000 to 1500, but they have not been otherwise rescaled in any way. The middle portions of the series are very similar (although the differences are larger than they look). However, what is very obvious is that the two ends vary quite drastically as a result of the RCS processing. A plot of the difference looks like this:
What is the ostensible reason for this? My guess would be based on an issue raied in comment 289 by Tilo Reber in the YAD 06 thread.
The RCS adjustment is age dependent. As observed by Tilo, all of the trees at the beginning will be young, since only fresh “new” cores are being used (i.e. there is no overlap with any older cores). The effect of the adjustment seems to be to lower the effective ring size. Similarly, at the recent end, older cores are predominant and the effect has been to raise the values upward rather substantially. In the middle, there appears to be an overlap of cores of mixed ages and the resulting differences are smaller.
In my book, this raises some theoretical issues with the RCS methodology. Does anybody have a link to where these issues might have been dealt with?
Re: romanm (#60),
They’re fitting a negative exponential detrending over time to a HS shape. Look at the coefficients of A, B, C from the function RCS.chronology. What is that going to do to the residuals?
Re: bender (#61),
i.e. Has anyone thought to overlay the fitted curves on the original series?
Re: romanm (#60),
Nice job Roman, I also looked at some of this. The difference in species of tree and whether certain resources are limited in certain regions has got to also have some affect.
Steve, I haven’t read the RC thread but I’m surprised that Gavin supported Tom’s work. He’s been obvously flailing about trying to strike some portion of this work at random, criticizing everything he can find along the way. Gavin has not checked any of this and is using his ever weaker credentials to make any argument which comes his way.
Re: romanm (#60),
While I am new to RCS methodology, the Briffa and Melvin Chapter that Steve listed as recommended reading (Briffa and Melvin, 2009 in Dendroclimatology: Progress and prospects edited by M. K. Hughes, H. F. Diaz, and T. W. Swetnam), deals with some of these issues. In section 5.4.1 the authors make the following observation about the series that they created to illustrate the “Trend-in-Signal” bias, which seems to match your observation:
“When the index series are realigned by calendar year, each series systematically underestimates the
magnitude of the ideal forcing in its early section and overestimates the signal later, a potential
medium-frequency bias. In the average chronology (Fig. 5.2f), the original overall signal trend is
captured by the differences in the means of the index series. In our simplified example, the bias in
the trends of individual index series cancel to some extent by virtue of the compensating biases in
overlaps between early sections of some index series and late sections of others. In situations where there is a good overlap in many series, this potential bias could be averaged out. However, this cannot happen at the start and end of the chronology. In the case of a long-term declining signal, the chronology will, respectively, under- and overestimate the ideal chronology at the beginning and end.”
Re: wbabcock (#73),
Thanks for the information. It looks like a possible systematic flaw in the methodology which needs rethinking.
Re: romanm (#60),
Excellent Roman.
Let me make sure that I understand:
Briffa uses a methodology that assumes (not completely unreasonably) that older trees will grow less (in the absence of temperature change).
Therefore, growth spurts by older trees count much more heavily than younger tree growth.
At the end of Briffa’s proxy, the included trees are all at their oldest, so RCS treats growth occurring very recently as much more significant than growth occurring earlier in the study.
WITHOUT THIS AGE ADJUSTMENT, THE BLADE IN BRIFFA’S HOCKEY STICK ALMOST ENTIRELY DISAPPEARS.
Is that correct?
Is that why TomP doesn’t want to use younger trees? Because RCS can not be used to exaggerate their growth as much?
Re: Jason (#92),
There seem to be a number of different adjustments possible for taking into account the fact that the growth rate of a tree will change over the tree’s lifetime. Some of them use a negative exponential model with three parameters which can loosely be interpreted as an itial level, a rate of change per unit time and an asymptotic level reached as the tree ages.
These adjustments could be applied to trees with separate parameters individually or the same single adjustment made to all trees or something in between. TThis choice might be made differently for different types of trees. However, there are drawbacks due to loss of information on the effects due to time from each method.
The reference given by wbabcock – comment (#73), implies that for RCS bias due to age of tree can be a problem resulting in the type of effect visible in the graphs I put up.
My view is that Tom P. is pretty much out to lunch on his home-grown “sensitivity testing” which really shows nothing more than the blade doesn’t disappear until the end proxies which formed it are removed.
Re: romanm (#63),
I don’t think people should discount Roman’s mean plot here. This is the same result I got in my latest post so it is verified and any method which dramatically changes the shape of the mean needs strong justification and verification. The hockey stick is actually flipped backwards in time because there were a number of trees with higher growth rates 2000 years ago in Yamal than today.
RCS mutes these trees and amplifies the most recent for some reason. I doubt that every tree ring study with RCS would suffer the same problems, it seems a bit lucky this time. It’s the lack of due diligence and low tree count on what is primarily an RCS based result which bothers me. If it were my data (and I believed this was temp), I wouldn’t have published in this form because I would be concerned about the methods which have obviously distorted the signal. If it’s the right method, they certainly made no convincing argument for it.
Re: Jeff Id (#221),
A point I’ve made before and will make again – the Yamal RCS reconstruction, like Mann’s PC1, was never presented in an on-point peer reviewed journal article (Where things like core counts would have been required by a reviewer). It was introduced in a very broad review article. The site article (H and S) presented a different reconstruction.
Re: Steve McIntyre (#223), I see.
If you want another good example of how Gavin operates look at his response to co2isnotevil (number 66) in the Hey ya! (mal) thread:
[Response: Clueless. Absolutely clueless. – gavin]
Then look at post number 70.
[Response: Because we are not even on the same page here. I have no interest whatsoever in photosynthesis kinetics. I, my work, this site, the whole debate is about the role of CO2 as a greenhouse gas. Something Arrhenius knew very well and which still appears to be a mystery to some. – gavin]
Essentially, Gavin had no understanding of the point being made in #66 but had no problem attacking the poster by projecting his own cluelessness. That became obvious in #70 when he admitted he didn’t understand co2isnotevil’s point. In my few trips to realclimate I have seen these kind of narcissistic projections over and over. By comparison Steve is very honest about what he knows and what he doesn’t.
Personally, I find sarcasm to be necessary part in any response to this kind of behavior.
Sam, your lucky to get your comments posted at RC. Mine don’t even get past the “gate keeper” at RC. Seems they didn’t like (understood) modern signal conditions methods for information extraction.
So how do Briffa and Cook’s remarks square with what was done and concluded for the Yamal series that Briffa reported? It appears that age is (could be) a critical factor in larch tree response to climate Since this relationship does not hold for some other species one it appears should not generalize..
Does not the lack of age discrimation have an important bearing on Yamal? Does it not also give evidence in direct opposition to what Tom P concluded about the series being robust?
Critic’s of Steve’s style here do not appreciate good rhetoric. The invective aimed at him in the RC post is “sound and fury, signifying nothing”. Steve, OTOH, wove Gavin’s and Tom P’s posts together skillfully to demonstrate 1)his mastery of the proxy literature, 2) that they didn’t know what they were talking about, and 3) that Briffa’s practice contradicted his published principles. Discerning readers grasp the difference and that is all that matters.
What’s harder to forgive is his frequent use of the nominative personal pronoun where the accusative should be used, e.g. “,… he asked Roman and I…”
Re: Tom C (#66),
And plurals don’t require the apostrophe, which is indicative of possesion. But, yeah, its art, ain’t it? (Everyones a critics.)
Re: Tom C (#69), absolutely. Steve I haven’t laughed so much for a long time. And it felt clean.
Re: Dave Dardinger (#57), well said Dave.
Re: Jonathan Bagley (#79), What matters is the truth and relevance of what’s posted, not the person posting it. What also matters is the learning that’s happening here, and TomP, despite his peacock nature and his flight to his chelas at RC, has actually contributed to that learning process. Here cats may look at kings and I think this is the healthiest possible response to the closeted situation in Climate Science and the plaudits-not-audits of RC.
Bender I reckon you’re a cat.
Re: Lucy Skywalker (#132),
I reckon he’s more catamount than cat. :>)
As for this post of Steve’s, best smackdown I’ve read in years.
Re: Tom C (#69),
The science may not be settled? I questioned a Ph.D about the “… nominative personal pronoun …” issue and they replied:
Wouldn’t it be more productive for Steve to continue his analysis of Briffa, rather than to teach a statistics class to Tom and Gavin? Just saying…
Re: Robinson (#68),
I’m sure it took Steve no more than a couple of hours to review the case and write this up. And the value is there. I can spot a clown from a mile away (some would say it takes one to know one), but no one can write like Steve. Me protesting Tom P’s antics does nothing. But it does bring it to the attention of Steve, thus minimizes his time investment determining if the thing is worth pursuing. The whole exchange was rather revealing, wouldn’t you say?
Re: bender (#74), It was, yes. I have no objection to the content. I’m learning about this stuff along with Tom ;).
#67
Ouch!
Robinson,
You miss what I believe is Steve’s key point in critiquiing Gavin’s apparent endorsement of TomP’s analysis. He supported an argument before he had carefully checked it out and, yet again, failed to acknowledge TomP’s indebtedness to the way Steve provided everybody with access to the code Steve used to do his analysis. The lack of transparency (not to mention common professional courtesy) among many Team members is the theme that recurs throughout CA.
Like you I am looking forward to more information on Briffa’s analysis – but I think much is now dependent upon what additional information Briffa decides to share.
Re: bernie (#75),
And let us not forget that the topic here *was* Kaufman, before Steve discovered the undisclosed archiving of Yamal. So while waiting for more on Briffa, why not discuss Kaufman and the robustness of varvology? Maybe Tom P can rescue that piece of work too? One can hope … 🙂
[mod: Bender, quit the name games.]
Re: bender (#79), Excellent point. As SM said we have to watch the Pea.
Now we know what SM was doing all that time while Tom P was vogue-ing down the internet runway: he was giving him enough rope to hang himself.
Re: theduke (#77),
He also indicated that his grandkids were visiting. Give the guy a break!
I can’t help but take more seriously the participants in these disputes who don’t choose to remain anonymous. Any member of a traditional academic community, rather than this mad-house, would ask who these people, for example Tom P, are; and if an answer was not forthcoming, move on to something else. Might the decisions of world leaders depend on the calculations of anonymous blog commenters? That’s pretty funny to me. And who the heck is Tamino? I suggest everyone writes under their own names. It tends to concentrate the mind – not making a public fool of your self.
Re: Jonathan Bagley (#78),
And the role of court jester in the time of kings was … ?
So now that Gavin’s willingness to accept uncritically fairly nuanced arguments that come his way, what might this say about his approach to GHG modeling in his beloved GCMs? Perhpas he can reply here? I’m sure Steve would give him the stage. I have the rope.
WOW !! Great work Steve
What I find puzzling how everyone (on both sides it seems) is not undertaking a thorough quantitative analysis of what the concomitant instrumented SATs where doing in the region of interest. What where they doing, and were they consistent with either one (any) of the dendro analyses? I have asked this before, and it is my recollection that it went unanswered. Why the silence from SteveM et al. on the fact that the MWP is no warmer in their application of the data warmer for the duration of the MWP? It is hard to tell from the figures (poor quality), the chronologies seem to be very similar during the contentious MWP. Could you actually quantify the differences in number terms instead of traces. How do these alleged transgressions by Briffa affect the global chronology/reconstruction. Instead of getting bogged down in details, regardless of how important those details are, it is also critical to consider the bigger picture and discuss how this all relates to what happened globally in the late 19th and 20th centuries. Otherwise, you are operating in a vacuum of sorts and losing sight of some critical implications of this “audit” for the global SAT record.
Steve: the Yamal data became available to me a very short time ago. Before discussing the implications of Yamal quality defects, it’s first necessary to verify the details. Also this discussion is hardly in a “vacuum” for readers of this blog (and I realize that you’re relatively new here.) If you look under the category “Briffa”, you will see many posts on Yamal over the past few years, some of which show its impact on canonical chronologies. The question for me in prior posts was the validity of the switch from Polar Urals to Yamal as a canonical site in this region. Take a look at the prior posts and it will help you place things in context.
It is impossible for me to answer every comment and I make no undertaking to do so. I respond to as many things as I can, but, as you see, I’ve been busy.
Re: Lorax (#87),
Read the blog. Do you think that Kenneth Fritsch and MikeN might have comments on that exact topic. Nah …
Re: bender (#88), You are showing admirable restraint.
>Except that scientist isn’t about “defending” hypotheses, it’s about trying to prove them wrong
I don’t have a problem with TomP auditing Steve’s work. He just needs to think things through.
Tom P has had quite a history at WUWT prior to doing battles here.
My first experience with his argument style came here, when he went off topic and insisted that boiling milk modeled positive atmospheric feedback.
Follow the comments. Very enlightening.
Re: Anthony Watts (#93), took a looksie over at WUWT. I can see why ‘Tom P’ would want to remain anonymous. If he wants to be respected, he might consider changing name from ‘Tom P’ to something else entirely in future comments.
I wonder if Gavin has read the “boiling milk” comments?
An interesting test:
If we ever get access to Briffa’s exact methodology/code, I would love to run the following test. Replace all of the actual raw data with one single realistic “robot tree” profile, cloned for all trees in the reconstruction. The ages of the trees would vary randomly, as would the time period when they lived. So a tree that started life in 1750 would have exactly matching data as a tree that started life in 1920, just different starting points and lifetimes. In theory this test should result in a perfectly flat lined temperature reconstruction; I’d love to know whether Briffa’s method provides this flat line or not… If you get a hockey stick out of this I’d be floored!
Re: hmmm (#100),
If you got a hockey stick out of that, it would lend validation to Roman’s post #64.
Tom C wrote:
Thus, in Team-Speak, Steve’s argument is robust! 🙂
Bernie #90, you are kidding right? Bender is pretty hopeless at showing restraint– read bender’s long series of juvenile quips here and on unthreaded. If those quips are alleged examples of restraint, I’d hate to see what bender writes when bender actually loses it.
Talking of restraint, it is completely lost on you folks, but the one person showing truly remarkable restraint here is Briffa. You guys, not so much, this blog reads more like a mud slinging free for all, except when it comes to like-minded thoughts, and in that regard you all seem to be cut from the same cloth. Very little meaningful debate, very little critique, just high-fives all around on how wonderful your illustrious leader SteveM is and how daft and amateurish the other lot are. Hardly a “professional audit”. Tis is what happens when an “audit” is done in a public circus…..
[RomanM: Your views have been expressed and noted earlier. When you continue repeating yourself and baiting others through derogatory OT comments, it becomes trolling and frankly, rather than watch you instigate a food fight, I won’t wait for Steve to clip your comments – I’ll do it myself..
If you have something to say relevant to this thread you are welcome to do so]
Re: Lorax (#105), [RomanM: clip – don’t bite on the previous baiting]
Re: Lorax (#106),
This not a particularly useful thread to make my point, but I’ll try it anyway. To analyze threads on this blog, you first need to see what the head post discusses and then see which, if any, posts advance the points made. Here’s what Steve is talking about in the head post on this thread:
1) “the disparity between core populations at Yamal and the other two sites in Briffa et al 2008”
2) “Despite the lack of endorsement from field dendros, this hasn’t prevented Gavin Schmidt from opining that my [Steve’s] observations are incorrect.”
3) “Tom P seized on the “excellent correlation” between two identical time series – the subfossil portion of the Yamal data set used in sensitivity studies with two different modern samples”
4) “Briffa has pretty much built his career on analysis of the Schweingruber network and the Khadyta River, Yamal (russ035) site is used as a proxy in virtually all of Briffa’s publications.”
5) “Less than 7 hours later [+~24], Tom P reported to realclimate that he had “lost his patience” with the totally unacceptable delays from room service and “kludged” my code to the sensitivity analysis that interested him”
6) “Tom failed to mention that, despite the fact that he had never previously used R, that the tools and turnkey scripts that I had provided enabled him to carry out the desired sensitivity study with an almost immediate turnaround time.”
7) “I [Steve] also urge readers to read three other Briffa publications on RCS standardization.”
8) “Briffa explicitly warns against exclusive reliance on codominant trees in the modern portion, urging a representative age class distribution condemned by Gavin’s guru.”
9) “The two graphics below compares chronologies for old (more than 75 years) and young (less than 75 years) for the combined Yamal and Khadayta River data sets.”
Now actually I was surprised by how many different major topics Steve brought up here (and some of them could be broken down further.) So just how many of the follow-up comments actually addressed one of these points in a constructive way, as opposed to either just agreeing or disagreeing? I think I’ll try to see just which comments (up to your post) do so, but it may take a while so I’ll post the list and then later repeat it with each point followed by a list of applicable comments (if any)
Re: Dave Dardinger (#123),
Here’s what I get concerning matching comments to the main points in Steve’s head post. I did it pretty quickly and some posts could be applied to more than one point. I have 22 of 105 posts which have IMO advanced one or more of Steve’s points. As I said, this wasn’t a particularly useful thread as it’s somewhat fluffy in some respects and a lot of posts, Yea, Boo, OT or just fluff. But 20% useful comments on this thread isn’t “very little meaningful debate”, IMO.
1) “the disparity between core populations at Yamal and the other two sites in Briffa et al 2008” – 6, 8, 11, 27, 57
2) “Despite the lack of endorsement from field dendros, this hasn’t prevented Gavin Schmidt from opining that my [Steve’s] observations are incorrect.” 4, 66, 76
3) “Tom P seized on the “excellent correlation” between two identical time series – the subfossil portion of the Yamal data set used in sensitivity studies with two different modern samples” – 5, 22, 23
4) “Briffa has pretty much built his career on analysis of the Schweingruber network and the Khadyta River, Yamal (russ035) site is used as a proxy in virtually all of Briffa’s publications.” – 58
5) “Less than 7 hours later [+~24], Tom P reported to realclimate that he had “lost his patience” with the totally unacceptable delays from room service and “kludged” my code to the sensitivity analysis that interested him” – 26
6) “Tom failed to mention that, despite the fact that he had never previously used R, that the tools and turnkey scripts that I had provided enabled him to carry out the desired sensitivity study with an almost immediate turnaround time.” -38, 45
7) “I [Steve] also urge readers to read three other Briffa publications on RCS standardization.” -63, 64, 68, 74, 102
8) “Briffa explicitly warns against exclusive reliance on codominant trees in the modern portion, urging a representative age class distribution condemned by Gavin’s guru.”
9) “The two graphics below compares chronologies for old (more than 75 years) and young (less than 75 years) for the combined Yamal and Khadayta River data sets.” – 1, 80
Tom P’s argument:
1) old trees trend differently than young trees, and
2) it’s excusable (and even preferable) to change the weightings of these two groups over the course of a 1000 year chronology, and even exclude one from the last (and most important) century of measure, when the only value of said chronology lies in the assumption of homogeneity
is laughably amateurish. To defend this method by arguing the superiority of one category over the other – old trees versus younger – is to invalidate the study he’s trying to defend. When he (or Briffa) reworks the chronology to exclude the first 75, 100, 150 years of every core used – from every time period – then one may be able to do something with the result.
Such as ascertain its “robustness”.
To forgo the mechanical integrity of a study in favor of adjustments and post-rationalization is to do one of two things; to bend the results to where they’re desired, or to give the impression of breadth where it doesn’t exist. Or both.
The strategy of sending in the weakest members of the team into battle is WELL ESTABLISHED. Come on guys we’ve seen this over and over before. Why send in Briffa when a Tom P, who learned everything he knows about RCS from one article I pointed him at, will do? And so your time and energy is spent correcting Tom P’s obvious mistakes, while gavin works away in the wings. Tom P is expendable. His role..delay, obsfucation, etc. The gavin comes in. Same tactic: borehole hockeysticks, C02 hockey sticks . This topic which really was about DATA SELECTION has been reframed as a topic about hockey sticks.. in particular the BLADE. So, send in the clowns and hope that while briffa heals the conversation can be derailed. There is another point to sending in non experts. the can make mistakes with impunity and the team is not damaged.
Re: steven mosher (#111),
bingo
.
Tom PP is a sentinel, an expendable freak sideshow. Bonus: Maybe he can trip up the casual reader, bait them OT, and make a mockery of the “frauditory” process here. (Nice, Hank. Nice.)
mosh: what’s his next take-home assignment?
.
The issue here is Briffa’s SELECTIVE USE of an undisclosed data set. Not disclosure. Not Tom PP’s incompetency, or Gavin’s lack of good judgement.
As an interested outsider, may I ask a dumb question? I see that all the argument is about one area of the world, Yamal. Does a study of tree rings anywhere else show the hockey stick? Alaska/Yukon? Equitorial areas?
[RomanM: Please ignore him in this thread]
Jim Brock>> The 2007 IPCC report states that app. 30 (thirty) locations on earth have been used as proxy sites for the period of years ~1000-1400. Where 3 (three) are on Greenland for instance, where 0 (none) is close to the coastline of Greenland.
None in the oceans (of course).
From these 30 locations science has been established, stating some certainty about global temperatures during 1000-1400.
Would you trust a global temperature estimation as of today if only 30 thermometers in different locations of the earth (and none in the oceans) where used? 🙂
I believe there is a basic flaw in this proxy method and it spells: i-n-t-e-r-p-o-l-a-t-i-o-n.
For every location inbetween these 30 places on earth we know absolutely nothing.
I see two things here that are not being discussed as part of the analysis and why the dendro guys really need to chime in.
1) The graph is pretty flat across both the MWP and the MIA. I would want to know if a population of larches near the extreme of its range tends to “clip” grow on warm or cool cycles. You could get to a point where “too cold” is a constant and then you get just minimums of equal size, but it does not measure the depth. The same could be true for “too warm”. This falls into why some trees just make poor thermometers.
2) Can you use younger trees for a sustained growth based on enviroment? Like animals, plants also have grow cycles. Growth and reproduction can be inverses. Since an immature tree is not focused on reproduction, it could show strong growth. Hopefully, the differentials on older, mature trees can show similar phase shifts. But it can lead to considerable bias using the inner rings exclusively as a proxy. And considering the differences between the mid 1800’s and the mid 1900’s because of the MIA and exit from it, and the fact that these trees are at the edge of their range, can you make a valid comparison.
Also note that trees are just as individual as humans. There are individual trees that can vary a USDA Zone or more than their siblings in tolerance. Add unknown environmental variables such as shading or lack thereof, microclimates, etc., small samples can make such outliers shine.
But based on all the comments I have seen to date, Briffa’s response has been the most reasonable in allowing for the possibilities that there could be an issue with the dataset. But not having the data and methods for this long… That was an automatic F when I took stat and econmetrics.
Donde esta los dendros?
Re: George Ellis (#118),
My post is a good argument for a browser with spell/grammar check… “econometrics”, “tends to “clip” growth during warm or cool cycles” and I am sure something else.
Mosher “This topic which really was about DATA SELECTION has been reframed as a topic about hockey sticks.. ”
[RomanM: OT. Try “stick”ing to the topic at hand].
Re: Lorax (#119),
[RomanM: OT continuation. Not here.]
Anthony, that boiling milk and the saucepan analogy is interesting since it is in response to climate-skeptic, who frequently uses the same analogy to say that a step change in solar output caused higher temperatures in a non linear fashion.
Wake up and smell the coffee gavin…
Next hockey stick please.
#104 SonicFrog
[Slap to forehead] How did I miss that? Looks like Steve’s rhetoric is “even better than we thought”.
steven mosher (#111): “There is another point to sending in non experts. the can make mistakes with impunity and the team is not damaged.”
Unless of course someone like Gavin endorses their silliness, thereby making it his own. 🙂
<a href=”javascript:edInsertContent(edCanvas,’Re:%20bender(#69)
should be “(Everyone’s a critic.)”
Heh. I was doing the same thing. i appear to have been more lenient on determining what, exactly, the poster might have been addressing.
.
1) 12, 58, 80, 88, 96, 118,
.
2)
.
3) 5, 25, 26, 58
.
4) 63,
.
5) 5, 22,
.
6) 38
.
7) 63, 68, 74,
.
8) 27, 55, 63, 68, 73, 80, 92, 102, 110, 127,
.
9) 5, 12, 23, 27, 51, 55, 57, 60, 68, 80, 92, 105, 110, 118, 127,
.
10) Code/procedure, 1, 6, 8, 11, 32, 38, 53, 60, 64, 65, 86, 92, 96, 100, 102
.
11) Congratulatory, 2, 3, 19, 31, 32, 50, 52, 56, 69, 87, 94,
.
12) Cross-blog chatter, 4, 15, 21, 24, 28, 54, 59, 62, 66, 67, 76, 85, 93, 97, 107, 111, 115, 119, 121, 125,
.
13) Snark, 7, 9, 10, 13, 16, 17, 33, 42, 45, 46, 48, 78, 81, 83, 84, 89, 99, 103, 104,
.
14) Heck if I know, 14, 20, 39, 71, 72, 75, 77, 82, 91, 124,
.
15) Civility, 16, 17, 28, 29, 30, 31, 34, 35, 36, 37, 40, 41, 43, 47, 49, 61, 69, 79, 88, 114,
.
16) Meaning of science, 25, 26, 79, 90, 101, 103, 109,
.
17) Wider Briffa/Hockeystick issues, 44, 46, 62, 94, 95, 109, 111, 112, 116, 117, 120,
.
18) Substantive Comment: 63,102 (romanm), 74 (wbabcock), 5,80 (TomP), 109 SteveM
.
19) Grammar Nazi: 70, 98 (bender), 122 (self)George Ellis, 126, BKC,
.
20) Vitriol: 88,106 (Lorax), 108, 113,
Re: Alan S. Blue (#129),
The “grammar nazi” incident coincides with a couple of your “what the heck is going on” and that was caused by some comment-renumbering when Lorax was snipped. Lorax getting snipped cause several now incomprehensible exchanges.
Jason:
“Briffa uses a methodology that assumes (not completely unreasonably) that older trees will grow less (in the absence of temperature change).”
That’s not quite correct Jason. Craig Lohele’s studies indicate that Larches (the topic of this debate) become more temperature sensitive as they get older. So older trees will be more likely to reflect a warming period. At the end of Briffa’s chronology he has a set of live trees, and the trees at that point have their maximum age. The earlier portion of their cores is not represented in the later part of the 20th century. But when you go earlier in time, all the trees are randomly overlayed. This means that at any point in the earlier history you will have that point represented by both old and young tree rings. Whereas at the end, the points are only represented by old tree rings.
Re: Tilo Reber (#135),
What Jason said was not incorrect. Perhaps you meant “incomplete”?
COMPARING TREE PROXIES and LONG TERM TEMPERATURE DATA
With the debate about tree ring data and “global warming, I though I’d compare tree ring data to long term temperature data. The tree ring data I found from http://www.climatedata.info
With tree ring Nor. Hem. proxy data shown below, using the 20 year MOV Norway, Sweden & Russian data, since they were more compatible to Ave14 defined below:
http://www.climatedata.info/Proxy/Proxy/Proxy/treerings_northern.html
Next I took the 14 longest temperature records from http://www.rimfrost.no/
plus the east English data starting in 1659. I averaged the whole bunch up to form a composite average Ave14. This is shown below:
I then added 40 year filtering consisting of a MOV, Fourier filter, and a 2 pole reverse Chebushev filter. The later is found in MATLAB as “filtfilt”. Basically the later filter is run forward and then backward to compensate for phase delay. Unfortunately the end points generally will have a significant error, but is a good cross check for date in the middle of the sample. The Fourier gives much better end point results, comparable to the EMD method.
The figure below compares the 20 year MOV averaged tree ring data with the 20 year MOV Ave14 data. The tree ring width is plotted against temperature.
For what it’s worth, it’s in the region where it “kinda looks” correlated, but would need more sophisticated analysis to show anything definite. From this short analysis, if I were a betting man, I would not bet the chicken coop on tree ring data, much less the farm .
Mike Roddy at Revkin’s blog:
Who can argue?
The only question is: how do you quote the silence one gets in response to some questions?
Hi all,
Ok..I have to ask…and please don’t tell me not to be a lazy so & so and look it up myself (thats you Bender – Heh!)as I’m pretty sure I’d fail to understand it if I did anyway…so..in laymans terms, these 30 proxy areas that Patrik refers to in 117…
Q…are these all tree ring proxies
Q…is one of them the famous Graybrill/Bristlecone thingy
Q…are the others ‘robust’ to use the in-word of the moment
Q…has anyone audited the others.
You see, I’m uncomefortable with someone telling me that the MWP/LIA did not really exist, and when presented with the hockey stick drawings I flinch and look to someone who can deconstruct the stats (thank you Mr McIntyre et al)and show that the stats are indeed dubious, and therefore the MWP/LIA did actually exist as my own history learnings indicated.
However, the other side of the arguement seems to suggest “yeah, well, maybe this one is wrong, but all the others are robust”. I admit I find it hard to understand how anything definitive can be drawn from tree growth – any gardener would be skeptical I think, its far too chaotic surely? – but the ‘other side’ does have a point if they can point to another 28 studies showing hockey sticks that are not contested.
Ok..hope its not going OT this, but its been on my mind since reading point 117.
Anyone care to enlighten me (please don’t make me do my own research….ahem)
Re: mike f (#138),
Generally speaking, a fair amount of the confirming evidence is confirming the blade portion of the hockeystick. That is – it doesn’t directly address either the LIA or MWP.
.
Jeff Id wrote a partial answer to that question on Oct 1, and hosted another yesterday.
Thks Alan.S.Blue 140
I did read that earlier..so I read it again. But whilst it states the position/arguement/logical inference it was not really an answer to what I was looking for. Maybe I’m too dim..my eyes were glazing over!
What I was after was in line with what I think Steve M’s message on this whole topic is ie “if you are going to use trees as a proxy record it needs to be massively robust in terms of qty of data used and selection process defined”…which even a layman like me can see the Briffa data fails.
So hence my concern regarding these other 30 or so proxy data sets. If they are all tree’s, do they too wipe out the MWP/LIA or is it ONLY the Briffa & Mann papers that did this. The fact that proxys show a warming in recent years is hardly rocket science, history tells us that much compared to our Victorian ancestors. But its that vital MWP/LIA CONTEXT that is important yes? If the other 30 studies do not cover the MWP/LIA then they are just pointless I think?
If the others DO show 1000 yr records as hockeysticks, then Gavin gets to say “it does not matter” unless these other proxys are found to be as ‘unrobust’ as Briffas appears to be at the moment?
“First, let’s observe the continued silence of field dendros on the dispute. None have stepped forward so far to support Briffa’s use of 10 cores in 1990 (and 5 in 1995). As others have observed, their silence is rapidly becoming loud.”
Well, that might be because they’re busy with other things besides reading and/or responding to your blog Steve.
But as for me, my reply to a number of your stated issues have been, or will be soon, posted at RealClimate and you are welcome to read them there.
Re: Jim Bouldin (#141),
Exactly why couldn’t you have posted the responses here?
Re: Dave Andrews (#143),
Maybe because Gavin needs his help more than Steve?
Re: bender (#143),
No, this way he doesn’t have to justify his post and they can control the responses to his post 488 on RC.
Re: Dave Andrews (#142),
Exactly why is that a problem for you?
And bender, if you think you have some dendro- or climate-based counter arguments that will fly, let em fly my man.
Re: Jim Bouldin (#149),
From your post:
Which temperature is that?
Re: Jim Bouldin (#149),
On the contrary, Jim, you make some excellent points. I see no reason at all to engage in conflict. As a man of knowledge, perhaps you would like to weigh in on the idea of selecting chronologies that yield a desired 20th century signal? Is that something you do regularly? Is it something you see quite often in your readings? I’m intrigued. I love to learn.
.
And what are your views on data dislosure? Eli Rabett suggests academics have a right to secrecy, even if their data is being used to inform public policy. Do you agree? Obviously, there is no right or wrong answer, so it can’t hurt to share your opinion.
Re: bender (#151),
“On the contrary, Jim, you make some excellent points. I see no reason at all to engage in conflict.”
Good deal!
“As a man of knowledge, perhaps you would like to weigh in on the idea of selecting chronologies that yield a desired 20th century signal? Is that something you do regularly? Is it something you see quite often in your readings? I’m intrigued. I love to learn.”
Oh hell yes, Jesus we do that kind of thing all the time. That is, when we don’t just fabricate the data outright you know. In fact we get paid to do exactly that kind of thing.
“And what are your views on data dislosure?”
Why, you have some to share?
“Eli Rabett suggests academics have a right to secrecy, even if their data is being used to inform public policy. Do you agree? Obviously, there is no right or wrong answer, so it can’t hurt to share your opinion.”
You know bender, with all due respect to the Rabbit, I’m not at liberty to answer those kinds of questions, because that would be violating the secrecy code that we are sworn to, upon penalty of death if violated. Sorry.
Gotta run bender my man. Take ‘er easy.
Re: Jim Bouldin (#155),
So informative! Thanks ever so much. I’ll be here for your next drive-by. Until then, best wishes.
Re: Jim Bouldin (#155),
Thanks for the answer to my very serious question about the temperature. Hey, this is climate science at its best.
Re: Jim Bouldin (#156),
Seriously? You are asked about the very important issue of data disclosure and get flippant? Really? I guess science just is not very important to you. As long as data disclosure is not practiced in climate science, there will always be skeptics.
Re: Ron Cram (#399),
Pretty bad, hunh?
Re: Jim Bouldin (#141),
As others suggested you can post them here. Posting them @ RC effectively cuts off discussion. As I can personally attest my posts there have been blocked. Even when I say I believe in AGW as I do. Even when I tried to THANK HANSEN PERSONALLY for releasing the Gisstemp code. Like we lukewarmers say: free the data, free the code, free the debate.
here is Jim Bouldin at RC:
“Questions to you, Gavin. Do these specific criticisms of the science in Briffa (2000) have any merit? If they are without merit, what are the specific rebuttal points which defend Briffa (2000) against them? If they do have some merit, how much do they weaken the scientific conclusions of the Briffa (2000) paper?”
A few points here in response to this, and to other various posts that I don’t have the time to cite:
1. Briffa did not collect the original data. The Russians, Hantemirov and Shiyatov did, in the 1980s and 1990s. Briffa’s modern sample was thus limited to a maximum of 17 tree series. McIntyre says straight up that he does not accuse Briffa of cherry picking the 12 of these 17 series, which is good, because it is not even clear that Briffa had all 17 to begin with.
2. The Russians state that they use 224 series in their chronology, chosen based on being the longest and most sensitive of the larger set of series, with sensitivity defined by inter-annual response to instrumental temperatures. It is thus highly likely that they chose their 17 modern cores based on these same criteria. This is perfectly valid dendroclimatic practice, and indicates that the 12 cores Briffa used in the modern chronology are sensitive to temperature.
3. Briffa’s response to Mcintyre states that these 17 trees were taken from at least 3 locations in the Yamal sampling area. Conversely Schweingruber’s 34 cores according to Briffa, came from just one location in the area. The latter would thus lack the spatial coverage of the H & S sample, even though the sample size was higher than the 12 Briffa used. Microclimatic effects could thus play a role in any of the 4 or more sampling locations.
4. The proper weighting of the modern trees, from a spatial perspective, would thus be at least 3:1 (Briffa:Schweingruber), although perhaps with some adjustment for any differences in variance between the two sets. This is why Briffa mentions in his online response that it did not appear that McIntyre weighted the modern series properly, although McIntyre’s response that he could levy the same argument against Briffa is also valid (but not beside the point, as he states; it is entirely germane to the composite graph McIntyre created).
5. McIntyre’s focus on the decline in the sample number from 1990 on is almost completely misplaced. I say almost, because yes, it would be great to have a good sample right up to the present to see how the trees are tracking the instrumental record, and you can bet the Russians are collecting same. But that’s not McIntyre’s objection. Rather, he seems to think that low sample sizes in these years will strongly affect the RCS standardization procedure, when in fact it will have almost no impact at all, because there are already over 225 cores used in the RCS sample. This is because all cores are used in RCS standardization, not just modern cores. It is not clear that McIntyre understands this, or he wouldn’t be making such a big deal about it.
6. McIntyre, in spite of making 9 posts now on this topic, has still not explained exactly why he thinks non-homogeneity of tree ages through time is a critical issue, other than vague references to Briffa’s discussions of this issue. The vast majority of the series used were from short lived trees, so there’s not much Briffa could have done about that, and long lived modern trees allow an easier crossdating of the dead material, especially given the large percentage (5-10%) of missing rings in these trees.
7. Analogies of subsequent users of the Yamal chronology to crack cocaine addicts is completely out of line, and additional oblique references to the sloppiness of the science (”…But maybe this is a coincidence. One never knows – it’s climate science”), reference to “the Team” etc., further undermines the supposed neutrality of McIntryre’s position. The condescension towards Gavin and Tom P in his latest post, which is in fact a diatribe, essentially seals this.
8. The more appropriate place for this debate is in the peer reviewed literature. This should be obvious.
Does that help?
Comment by Jim Bouldin — 5 October 2009 @ 4:04 PM
Re: windansea (#145) and others including Tom P,
Tom Bouldin on RC states:
Since IIRC the metadata of the Briffa (2000) data was not archived, we don’t really know which of the three locations each of the 12 that Briffa used came from. Accordingly, following Dr. Bouldin’s reasoning, the best we can do is divide the number of cores used (12) by the number of locations (3) to yield an average sample size per Yamal location of 4 (we could call the 3 Yamal locations Yamal(a), Yamal(b) and Yamal(c)).
Statistically, a sample size of 4 is virtually identical to a sample size of 0, so, following Dr. Bouldin’s reasoning, all 3 Yamal locations should be discarded because the sample size might just as have been 0. That would leave only the Schweingruber 34 or some other data from the general vicinity.
With all due regard to Tom P’s diligence, I would submit that any further discussion of the Yamal data would be of limited usefulness regarding Briffa (2000).
sorry, this part at top is not Jim Bouldin
“Questions to you, Gavin. Do these specific criticisms of the science in Briffa (2000) have any merit? If they are without merit, what are the specific rebuttal points which defend Briffa (2000) against them? If they do have some merit, how much do they weaken the scientific conclusions of the Briffa (2000) paper?”
“First, let’s observe the continued silence of field dendros on the dispute. None have stepped forward so far to support Briffa’s use of 10 cores in 1990 (and 5 in 1995). As others have observed, their silence is rapidly becoming loud.”
Well, that might be because they’re busy with other things besides reading and/or responding to your blog Steve.
But as for me, my reply to a number of your stated issues have been posted at RealClimate and you are welcome to read them there if you like.
Ok so I’ll ask a question starting with my understanding of the process. None of the trees cover the entire period of interest because they don’t live that long. Trees in the current period are selected for sensitivity to temperature based on local temperature records with the theory being that if they are sensitive to temperature during the time when we have measured temperatures they will also be sensitive to temperature in prior periods. Since the trees don’t go back far enough the tree rings are matched up with trees that lived during prior periods so that the record can be extended back in time.
Here’s the question. If a subset of the trees in the current period have to be selected due to sensitivity, then how do you select the sensitive trees in the prior periods? The fact that a selection has to be made means that we can’t assume that all trees in prior periods are sensitive to temperature. It seems to me that this would mean that the record might be interesting for the prior time during which the selected trees were alive, but the extension back in time wouldn’t be meaningful.
The point is so obvious that there must be an answer, I just haven’t been able to find it.
Re: Nicolas Nierenberg (#153),
Dr. Bouldin may have you believe otherwise, but I think that you are correct. He states in the quote of his RC post:
Since most of these 224 series are well outside of ANY instrumental period, I would ask him to explain how it is possible to make that judgement.
If in fact there are differences in “temperature sensitivity” and the claim above is false, then picking sensitive proxies only for the current time period would in fact raise the possibility of serious bias because the current sensitive proxies are being compared with a mix from population with a wider range of lower sensitivity thereby depressing the response of the sample during the earlier period.
A simple example of this effect would be the following. Suppose you are doing a sequence of of exractions from samples of ore to remove the content of a particular mineral. The samples all come form the same ore body with some given average content level. Now at some stage, you decide to implement a method which will extract a higher percentage of the mineral (read: is more “sensitive to temperature”). Even though the average amount has not changed in the samples, you are now getting a higher amount of ore from the samples even though the underlying situation is unchanged. You will decide that the ore samples actually contain a higher amount unless you can adjust for the difference in the use of the new method. This same bias could be created by merely selecting the new proxies based on their sensitivity to temperature and no other reason.
So your question is an excellent one and unless there is a valid method for doing what Dr. Bouldin claims (which is why I would like a genuine answer), the question would remain with just what the chronology was measuring.
Re: Nicolas Nierenberg (#153),
Nicolas,
Exactly how the sensitive trees in former times were determined would be a question for the Russians who did that task. But basically it would be follow this general idea:
The ring width data from all the series, past and present, are lined up by cambial age chronology (cambial age = the relative age of each ring, from the pith or tree center). The average of the ring widths for each ring, over all the trees is then computed. This creates a “standard curve” that reflects primarily the size-dependent part of the growth response of the trees. This mean value series is then subtracted, ring by ring, from the actual ring widths of each tree, thus removing the diameter-related ring width component from each tree, since the goal is to isolate the environmental signal. The residuals from this detrending are then examined to see how “complacent” they are, meaning how much they vary from year to year. Those that vary the most strongly are the most sensitive to the environment, and whether they were responding to the same environmental factor is assessed by looking at the spatial similarity of the variation pattern across trees, across the area of interest.
Re: Jim Bouldin (#161),
So why is this temperature instead of some other factor?
My question with regard to “sensitivity to temperature” is very genuine. I will state it in more detail so there is no misinterpreting what I am asking.
Is Yamal supposed to be sensitive to local, Arctic, NH, global or some other temperature? Annual? Summer or what?
Re: romanm (#164),
So why is this temperature instead of some other factor?
My question with regard to “sensitivity to temperature” is very genuine. I will state it in more detail so there is no misinterpreting what I am asking.
Is Yamal supposed to be sensitive to local, Arctic, NH, global or some other temperature? Annual? Summer or what?
OK roman, I’ll take you at your word, you seem honest here, and then I HAVE to go.
The Yamal chronology is designed, like most of those in northern Eurasia that Briffa, the Russians, the Swedes etc have been working on, to detect temperature variations. Any given chronology, say the Yamal or the Tornestrsk, are designed only to evaluate the temperature history in that specific area. None of them are meant to represent the global pattern–they can’t. That must come by the slow accumulation of many such chronologies all over the world. As for the temporal question, that depends heavily on whatever signals are evaluated to be the strongest in the comparison of current trees and the instr record. In the case of the Yamal it’s primarily June and especially july temperatures that the ring widths are recording.
Jim Bouldin is this Jim Bouldin . He also has done guest posts at RC.
I am new to climatology.
I do not know the usual procedures. Being a biologist, I prefer to use large, huge datasets to extract biologically relevant data. I use a lot of statistical methods. Why has so far as I know nobody tested the Yamal dataset using LOO or Random Forest (No joke). And hiding data and algorithms is a sin, regardless of what you are doing.
.snip –
Steve: This is about Briffa and Yamal. Please do not digress into gernal discussion about dendro.
1st – i agree with Steve in the header, why no dendro replies? why the silence? are they all in a room somewhere thinking Briffa has to deal with this on his lonesome when he is well again (or is something afoot that we are not aware off)? – dunno (as bender might put it)
2nd – back to the science/data/method
it seems clear to me after all this kerfuffle that we should be careful of basing our understanding of historical climate temps on treering data. (many or few, unless no other data is available).
3rd – however, someone mentioned treeline data somewhere in these threads & implied NH treeline was much further north in yrs 600-700 ago (AD1300-1400) roughly, can’t find it now, but think it was worth further comment.
that seems to me to be a more reliable signal for climate that a few cores, a whole treeline moving would convice me that we have a problem.
Sorry to linger and spoil your triumphant party by answering your exact question there but ever heard of the uniformitarian principle roman?
Really gotta run now fellas. Miller time ya know.
Re: Jim Bouldin (#166),
That’s a really great answer. Yes, I actually have. You choose one set of proxies by comparing to temperatures and another by comparing a bunch of long dead trees to each other (and not the previous set of proxies)… and the uniformitarian principle guarantees that you are looking at the same thing. I like that. If only the rest of the world was that unifrom.
Did you answer the question that I asked separately? Not a hope. Why, do you already know the answer maybe or does it require some thinking?
Real class, Dr. Bouldin. Before I formally retired from the academic community, I would (fortunately rarely) encounter this type of arrogant behaviour. Good reminder that it still can be found.
[Addition: This was cross posted with the previous post]
How does the uniformitarian principle account for YAD06?
I got a bit cranky with the snark and your post only appeared as I posted mine. I appreciate the response and withdraw the previous personal comments.
This is the answer I wanted. I have been looking at the monthly temperature CRU record for the two 5×5 degree grid sectors which cover the Yamal region. So far, there was nothing like the hockey stick shape of the chronology, but I will look at it some more tomorrow.
Should have known better. Can’t play answer man any longer. You’ll have to do some work on your own if you want to understand the topic I’m afraid roman. But I doubt that you really do
Scott Brim if I get some time in the next couple of days I’ll try to answer your good questions.
RE: Jim Bouldin, who says:
“7. Analogies of subsequent users of the Yamal chronology to crack cocaine addicts is completely out of line, and additional oblique references to the sloppiness of the science (”…But maybe this is a coincidence. One never knows – it’s climate science”), reference to “the Team” etc., further undermines the supposed neutrality of McIntryre’s position. The condescension towards Gavin and Tom P in his latest post, which is in fact a diatribe, essentially seals this.”
This is completely disingenuous of course. Gavin’s site is replete with condescension and diatribes, and frequently bars comments from critics, yet Jim chose to post his response there as if it is not. I think many people would have more confidence in the science if the scientists were less prone to such simplistic and obviously hypocritical characterizations.
oops, ok we’re going back and forth missing each other’s honest attempts roman. i’ll take my part of the responsibility there.
later
Re: Jim Bouldin (#175), 🙂
Dr. Bouldin,
A couple of questions for you when you get a chance.
1. As you mention, Briffa says there were at least 3 sites where the 17 living trees were from. Steve was able to identify 10-12 as being from two groups based on tree numbers. Do you have any insight as to what trees might be the other group(s)? And why they wouldn’t show up when Steve went searching for living trees?
2. You say,
and later imply that the sensitivity for older trees is just a generalized response, not related to instrumental temperatures since there is none available. Wouldn’t it make more sense to use this latter sensitivity definition all the way instead of picking only modern trees which relate to a particular signal? At least wouldn’t it be valuable to see what difference it would make? If most all living trees which are sensitive in the second sense are also sensitive with respect to the instrumental temperatures, this would be an important validation of the concept of sorting by instrumental temperature sensitivity. OTOH if, say, only one in 10 sensitive trees are also sensitive to instrumental records, then it might be questioned if older temperatures can be obtained using this method.
Re: Dave Dardinger (#176),
They are looking for trees that are sensitive to local conditions.
Trees in the middle of a forest full of similar trees tend not to be sensitive to their environment.
If the temperature goes up a little or down a little, they will still be happy.
If there is more water or less water, they will still be happy.
Dendros are supposed to look for trees right at a temperature driven tree line, an imaginary line on one side of which it is too cold for that species to be happy, while on the other side that species is happy. The line is more likely to be a combination of factors (height, water, wind, etc.) but for the purpose of analysis will be treated as a temperature driven line.
Thus very small temperature variations in the selected trees should have a very large impact on their happiness.
Of course, this approach depends on being able to accurately identify which trees are sensitive to temperature. If a scientist never even collected the original data (as is true in Briffa’s case), he really has no basis for establishing this relationship. The Briffa 10 display growth patterns that could certainly result from the death of surrounding vegitation (thus exposing them to more sun after an extended period of relative shade), and on Watts’ page there is an account from somebody who has seen similar patterns caused in exactly this manner.
So when Bouldin refers us to the Russians, he is reminding us that the relationship between tree growth and temperature depends on a tenuous set of assumptions, many of which are often false, and that only the scientists that collected the original data are in a position to speak to how they ensured that the cores they collected came from “treemometers”.
In the case of the Briffa 10, we now have an additional problem. Briffa (unlike the Russians), adjusted the cores based on the age of the trees. Older trees were assumed to grow more slowly in the absence of temperature changes. Younger trees were assumed to grow more rapidly. This is a widely accepted assumption by dendros.
The degree of adjustment was determined by analyzing the cores, and guessing an adjustment that best fits the actual observed data. Obviously there are numerous scenarios in which this adjustment may be improperly selected due to intervening (non-age, non-temperature related factors).
Roman has now shown that virtually the entire signal from the Briffa 10 comes from this adjustment; an adjustment not used by the Russians who collected the data.
In sum there is a very tenuous chain of assumptions linking the data from the trees, to the temperature changes calculated in Briffa’s paper. By and large, Briffa has no personal knowledge on which to ground these assumptions (having not collected the data in the first place). Additionally, the data itself seems to indicate that Briffa’s assumptions of roughly Gaussian noise and roughly linear temperature response were unreasonable.
Other than that, Briffa’s reconstruction appears to be on solid ground.
snip: sorry, Nicolas. Editorially, I’m with bender on this. Ask bender on Unthreaded.
Jim Bouldin, please don’t bother answering basic questions that are easily answered by a read of a textbook. Especially if you are going to be condescending about it. Why not tell us something we don’t already know? Like details on the Hantemirov & Shiyatov plot network, as you seem to know something about it. Any unpublished studies that you know of that could shed light on the missing metadata from the Briffa sample?
If Steve’s not asking the questions or Rob Wilson’s already answered it, then you’re wasting the opportunity.
Romanm, what data are you looking at when you look at CRU gridcells?
Re: MikeN (#179),
Crutem 3 5×5 degree gridded data from here or from KNMI.
So the question remains, why do Schweingrubers results change the Briffa results so dramatically? Doesn’t Schweingruber pick trees using the same standards that Bouldin described above? Don’t the trees come from the same area? In recommending a 3:1 weighting, I don’t understand why more locations are important, but more trees are not. How does the uniformity principle stand in light of the modern divergence problem? And how do we know that there was no divergence problem in the past? What about the stacking problem? You have trees that have climate sensitivities that vary with age. At the 20th century end you are using trees that are at their oldest and therefore at their most sensitive because that is a feature of larches. In other worlds all of the tree rings at the end of the period are old tree rings. The earlier periods have tree rings that are a mixture of old and new. How does that not break the uniformity principle?
Re: Tilo Reber (#180),
He has already stated that the H&S network covers a much wider area than Schweingruber’s sample (so that’s where Briffa’s comment about appropriate weighting comes from). The question is how does he know that, and how come we don’t? He may know what’s different about Schweingruber’s site choice. But if he knows that it’s again a question of how does he know? What metadata does he have that we don’t? Inside knowledge from H&S, or is there some metadata out there we haven’t come across yet? Some here have been making a big deal about the size of Schweinguber’s sample compared to Briffa’s subsample of H&S. If all Schweingruber’s tress are clustered close together then it also may not be as representative a sample as some are supposing. So many are leaping to conclusions when they don’t know what Briffa knows. Steve is walking a very fine line. Many here are falling off that line. Some don’t care if they go a bridge too far. Others should. Remember that Briffa knows quite a bit that he is not revealing.
“Indeed, we have been repeatedly lectured that true replication is not obtained if only the original material is used.”
I agree, that seems crazy. If you use different trees, even from the same area, you are going to get different results. Think of Ababneh getting completely different results from Graybill even though she went to the same place and took samples from the same kinds of trees. I guess using their standard of replication, we would have to say that Graybill was falsified.
Dr. Bouldin’s comments on this thread were disgraceful. Troll-like, even.
Jim Bouldin,
some of us less knowledgeable types here are still waiting for an answer.
How do trees on the Yamal Peninsula show a temperature response to a Globally Averaged Hockey Stick Value rather than to the much flatter regional temperature of the Yamal Peninsula??
I’ve got some interesting results messing around with RCS
It in fact has an interesting meaning with respect to Dr. Bouldin’s contentions about RCS. Just lucky this time.
Re: Jim Bouldin (#159),
Readers, please do not raise generic issues about trees or dendro on this thread. This is about Briffa and Yamal.
Re: Steve McIntyre (#188)
[snip – way out of line with blog policy – calls for harm- Anthony]
Re: Steve McIntyre (#188),
I see you’ve put up a new posting before addressing my points concerning either:
(#5) the age sensitivity of the combined Yamal and Khadayta River chronology, or
(#80) the historical repeated divergence between younger and older trees in that chronology, despite your unsupported claim to the contrary in the head post:
75 years) and younger (
Have you decided to “move on”?
Re: Tom P (#191),
I’m busy on personal things today. I will reply to this point, which is on topic. Tom, in the mean time, I would urge you to read some of the back threads on Briffa at this blog (use the Category feature in the Left Frame – there are many of them – and the original Briffa articles, as it will give you a little better context and save me having to review a lot of past issues.
Steve McIntyre,
You made me laugh !
Re:windansea (#145),
This seems to lead to the same problem Jeff Id has been exploring, except that the “calibration” to temperature is being done manually. The “most sensitive” would be those trees that showed correlations with temperature. If the temperature is going up during this period, the procedure generates the blade of the hockey stick (as Jeff has shown). But if the temperature record is flat, no hockey stick. Could it be that H&S were using local temperatures (flat?) while Briffa used global temperatures?
Re: davidc (#192),
If they cherry-picked 224, how large was the full sample?
Re: bender (#193),
I have seen a number, but don’t remember. And I do recall someone posting local temperatures that were flat. In that case the “cherry picking” is harmless and would only amount to discarding data.
Re: davidc (#195),
No, you can cherry pick to match white noise. it won’t give you the same strength of HS as a red noise series trending up. But it will inflate the significance of your calibration statistics.
Re: bender (#197),
But that (HS) won’t happen if the calibration period is flat. All you would be doing is discarding those that went either up or down during the calibration period. But if Briffa came along with a calibrating temperature that went up, he could get a HS from the same data.
H&S 2002 says: a total of 2171 sawn wood samples has been collected.
Re: Alan S. Blue (#194),
Hmmm. But then how many samples did they process? Then how many of these did they measure? Then how many did they successfully crossdate? Each stage of processing could have some whittling down.
Re: Alan S. Blue (#194),
So about 1 in 10 showed some reasonable agreement with the temperature, with reasonable being defined by whoever was doing the manual matching. Seems reasonable.
Re: davidc (#200),
1 in 10? You don’t know that all 2171 sawn samples got polished, measured, and crossdated. Do you?
Steve’s first post on the recently de-sequestered data has snippets of that information Bender.
Re: Alan S. Blue (#198),
Not really. And I’m not too interested, actually, now that I read those notes more closely. The 224 samples were fossilized wood pieces. My guess would be the best of the best. The other 1900 could easily be useless scraps that never get processed. That they were forced to supplement with 17 old living trees supports this possibility. Steve surmises that the 17 may have been chosen from a larger sample. While it is true that Rob Wilson could core 17 in a half a day, it is possible that this was an after-thought on the last day of sampling. It is possible they were specifically looking for a tremendous motherlode of fossilized trees, and having found little more than scraps, took a few living trees to make sure they had something publishable.
.
This is all speculation of course. But it serves to highlight why it is important not to make any wild accusations about cherry-picking samples within a chronology.
.
Chronology SUBSTITUTIONS – now that’s a different issue. There’s fertile ground there, for sure.
Davidc, they looked flat first, but it’s a bit trickier than that. Temperatures rose, but to the same levels in 1925-1950 as 1980-2000. It also looks like Briffa didn’t use the two closes weather stations to compare temperatures at Yamal.
Re: MikeN (#201),
It would be surprising if he searched elsewhere or more broadly. I am skpetical of that possibility.
1. Tom P: Maybe it would help you (and me) understand what is going on here if I put it another way.
IPCC AR4 claims huge recent warming in “North West Russia”. On inspection, this is Briffa’s Yamal series. But there is much data from NW Russia that was omitted from this series. When those data are added back in, the huge recent warming disappears.
Digging further, we find the huge recent warming depends on the 5-12 trees left in the recent period of Briffa’s chronology. This is too few trees by normal standards. Briffa (still) needs to offer a convincing explanation of why these trees and no others in the recent period. If there were really no others suitable, he would need a large error bar around the recent data, or to delete it altogether (especially the 5-tree bit).
Your sensitivity tests essentially just remove trees from earlier periods. As there are still plenty of trees left in those periods even when many are excluded, the chronologies only change gradually compared with the huge recent uptick. But that does not prove that the huge recent uptick is real. It is that end part of the graph which needs to be justified.
2. Gavin’s first graph reminds me that he has already conceded McIntyre and McKitrick were right about the Gaspé cedar and the non-centered PCA. The arguments are just as good against the bristlecone pines, and may well be just as good against Yamal. A simple idea for a scientific article would be to carry out a sensitivity test on recent reconstructions deleting these series and eschewing the centering error. (Steve is not getting enough good ideas for extra work, and he is far too dumb to think of this one, so I know he’ll appreciate my guidance…)
3. We are all trying to be objective scientists here at CA. Still, if we get a HS in the end, we may be Shiyatov.
http://delayedoscillator.wordpress.com/2009/10/05/yamal-emulation-i/
http://delayedoscillator.wordpress.com/2009/10/05/yamal-emulation-ii-divergence/
Re: Scott A. Mandia (#208),
Thanks Scott. That’s a nice start. You should try detrending the data. The correlation with temperature means nothing if you havent detrended the data. But it’s a good start. Oh, one little warning, if you use RCS to detrend the Yamal series as briffa did you will be in uncharted waters WRT the sample size. But have a go.
Cross post from RealClimate, reference to Jim Bouldin
Re [Jim’s post on RC number] 492
Jim, thanks very much for this. A measured, helpful response to my post. I agree with most of your points and find them useful. Cross posting to CA as well.
I can see where you’re coming from. The objective is to find the “climate signal” in a low signal to noise environment. Your approach (and H&S’s approach) was to pre-select the older trees which may be best responders to the signal. That should pull more signal out of the noise (the non-responding, younger trees). I’m coming from a signal processing background where to get more signal, you increase your sample size and do more and more averaging. Different approaches.
Your approach kind of makes sense to me if we have a luxury of a large sample size. However it makes me very itchy when we have a very small sample set. The sampling error becomes tremendous.
>> McIntyre’s focus on the decline in the sample number from 1990 on is almost completely misplaced. […] Rather, he seems to think that low sample sizes in these years will strongly affect the RCS standardization procedure, when in fact it will have almost no impact at all, because there are already over 225 cores used in the RCS sample. This is because all cores are used in RCS standardization, not just modern cores. […].
I see where you’re coming from, but I disagree. Only the modern cores are used in the calibration. As I see it, RCS is an algorithm whose inputs are 225 tree cores and some calibration factors. Its output is a temperature record. The problem I see is that the only information from which to derive calibration factors is contained in trees which were alive in the instrumental period. This information set is very limited, and no algorithm (RCS or otherwise) can increase the amount of calibration information.
Let me try and explain my point in more detail. Sorry for the mickey-mouse language, it’s the only way I can think.
The Briffa (2000) methodology goes as follows
(1) take a set of tree samples
(2) define some a-priori selection criteria to select “good climate responders”:- eg species, location, age
(3) extract the set of tree samples (“responders”) which meet the criteria:- 225 trees in Briffa (2000)
(4) confirm that the set of tree samples does show a statistically significant climate response
(5) confirm that any climate response observed in (4) is statistically strong enough to be extrapolated across the entire “good responder” set.
(6) if (5) is valid, use the entire “responder” set as a thermometer.
The only information we can use to carry out steps (4) and (5) is embodied within trees alive during the instrumental record:- the “calibration trees”. We must infer the properties of the rest of the set (“long dead trees”) from the calibration trees. In step (4) Briffa (2000) must show a statistically valid correlation between the calibration trees and instrumental temperature and in step (5) it must show that it’s statistically valid to pass properties observed in the calibration trees to the long-dead trees.
So the results stand or fall on the information embodied within the calibration trees. RCS can’t add more calibration information than there is. The fact that there are 225 trees in the “responder” set isn’t relevant to this part:- RCS is inferring the climate response of around 200 long dead trees from only the information contained in the handful of calibration trees.
And to paraphrase McIntyre “how can you infer anything useful from so few trees?”. This small a set seems sensitive to sampling errors. Is it enough to disprove spurious correlation in step (4)? Is it enough to be confident we can infer the properties of the long dead trees from the calibration trees. Tom P’s first graph shows this sensitivity. Removing just one of the calibration trees (albeit the most important one, YAD06) changes 20th century results from 2.75 to 2.35 “units”. That seems to me to indicate a large sampling error. When we draw error bars on the Briffa (2000) results they are going to be pretty big – which weakens its conclusions.
Do you think I have a point?
Mark
Re: Mark P (#220),
The way you see it is incorrect.
[See how easy it is to be snarky and dismissive, Jim Bouldin? Aren’t I authoritative. I sure feel big now.]
If you are talking about the dead parrot sketch, I can just imagine Tom P playing John Cleese. The parrot was obviously dead, even though he was told repeatedly it wasn’t. The Yamal issue is a non-issue.
To get this post back on track I think Tom P’s post in 191 needs to be answered.
I appreciate deeply that “the devil is in the details” and have endeavoured to follow the details here. But the converse problem is “not to see the wood for the trees” when impressive displays of narrow expertise are used to silence the debate, while in fact something different, and far simpler, is perfectly valid evidence. Gavin has been shouting, you can’t throw away the HS because there are others – and, by implication, these are all underpinned by correlation to temperature records that have shown a large twentieth-century uptick. But YAD061 etc are being related to the CRU Pan-Arctic temperatures. For a start, much more local temperatures should have been used to correlate Yamal. CRU will not release their stuff, and many suspect that their overall temperatures may be based on a conglomerate of stations improperly adjusted for UHI and other factors, which could render the temperature record null and void as calibrator.
Hantemirov and Shiyatov say, in the same 2002 paper, that the treeline in medieval times was a lot further north. I did the page on temperatures circling Yamal using individual station records that had both longevity (spanning 1860-2000, which covers the crucial time when uptick supposedly started, say 1900-1950) and a likelihood of trustworthiness. It clearly shows a total non-correlation with the Yamal excesses. Obviously all this is subject to further audit. But it all constitutes serious evidence. This evidence is not disproof, but since it challenges the “dirty dozen” trees, it does mean there should have been a far more extensive recent dendro sampling, with data and metadata publicly available, before jumping to conclusions. And since all this evidence was publicly available all along, the implicit requirement of a far higher standard of evidence has been there all along as well as well.
I’m not convinced any of this debate is going anywhere. As I see it, McIntyre’s original criticisms of Briffa (2000) were that the calibration tree set
(1) is very small
(2) “looks different” to the rest of the sample set (the “dead tree” set)
Note in (2) I use the word “looks”, not “is”. Indeed, it does look different. There may be valid reasons for this, or it might be spurious. It might matter, it might not (indeed McIntyre made this specific point). The fundamental problem seems to be: there is insufficient information in the calibration tree set to have confidence one way or the other.
What Tom P and others are doing is repeatedly removing information from the dead tree set until it too “looks different” in the same way as the calibration tree set and then discussing whether the result shows any robustness or not. This is in no way addressing the lack of information in the calibration tree set, it’s simply removing information from the dead tree set.
My simple question is this: is the information content of Briffa (2000)’s calibration tree set sufficient to show that
– the calibration tree set’s correlation with temperature is not spurious
– the calibration tree set is statistically representative of the rest of the set (the “dead trees”)
?
I don’t see how messing with the information in the “dead tree” set can answer this. We need to address the sufficiency of information in the calibration tree set. How can we do this? Are there standard statistical tests we can apply? In my old job, I’d have seen how much calibration information I need to delete for the “effect” to disappear. Is this scientific?
Tom P.
If removal of a small subset of trees from an age compensated dataset (34 Schweingruber trees + 12 Yamal), completely flips the late 20th century portion of the graph upside down the study would appear too sensitive. Look at the graphs at Romanm #63 to see the problem with Briffa’s RCS method. Read Tilo Reber #134. Too sensitive.
Briffa himself states that a wide sampling of tree ages is important, why are you insistent on taking that away? And where do your filter lengths come from, 21 years, 51 years, 72 years, what’s next 34.5674 years? – STOP CHERRY PICKING and stop ignoring Briffa’s own dendro guidelines. The Schweingruber study contradicts the Yamal study, which is more robust?
Commenters – there are general questions about tree rings and there are particular questions about the Yamal sample and Briffa’s handling of it using RCS standardization.
I realize that many of you wish to question the whole idea of tree rings as a climate indicator. However, editorially, that results in every thread being identical. If you have something to say on Yamal, RCS standardization, please say it. If you want to talk about dendro in general, for the 100th time, please do not do it on a thread where a narrower topic is at issue. I’ve moved some comments accordingly, but the server is very slow right now and it’s a waste of my time.
Steve – I agree. What is your response to Tom P #191?
It was surprising to find that applying RCS methods to just the 12 trees rather than the whole set seriously cut down the blade of the hockey stick. RCS, while reasonable, is simply not a good method to apply across large sets of trees. There isn’t enough flexibility in the method to account for different growing conditions and it does have a substantial effect on the result.
It will be interesting to see what Briffa comes up with next. After this incident, he should be ready to move on.
Re: Jeff Id (#219),
Jeff, before RCS, there was “ARSTAN” standardization, which is probably more widely used than RCS. It standardizes trees on a tree-by-tree basis. The problem with this method is that it was impossible to recover centennial-scale variability.
The difficulty with one-size-fits-all (RCS) standardization is that it is extremely sensitive to inhomogeneity in population – a point that Briffa makes over and over. In this particular instance, Yamal does not meet Briffa’s protocols for RCS standardization.
This is a point that a field dendro could agree with – even if we disagreed on the merit of dendro as a climate proxy (and I’m not as discouraged on this count as many readers). Because there is a narrow but important issue here that could be agreed on, I’m taking a hard editorial line on people injecting generalized worries about dendro into the thread.
Can you imagine the personal discomfort you would feel, going to print with a dataset where the blade is created by the methods and known to be from too few trees. I thought it was the data before but actually it is from the difference in tree response on the RCS curve. I would have been nervous as heck and would certainly have done the mean as a sanity check, it’s hard to imagine that Dr. Briffa wouldn’t have done that. The more I’m understanding the more broken Briffa’s Yamal is.
Re: Steve McIntyre (#222),
Thanks for the ARSTAN reference, I’ve seen you comment on it before. It’s clear that individual standardization could create problems as well. It could be kind of a fun math problem actually.
Re: Jeff Id (#225), jeff did you read the paper that I pointed Tom P at? It has some things to say about RCS.
Re: steven mosher (#240),
I’m sorry, I missed the link. I’ve had several papers recommended recently that need reading and some extensive work on Antarctica to look over. Would you mind telling me where you left the recommendation, it doesn’t seem to be on this thread.
“I see you’ve put up a new posting before addressing my points concerning either:”
I see that you’ve put up new postings before answering the three questions that I gave you two days ago. Stop crying and expecting other people to follow standards that you are not willing to follow yourself.
Jim Bouldin:
I can understand how the evaluation method will help you to sort trees with unique growth characteristics from those that are responding to temperature. Namely, temperature covers all of the trees, so those with unique characteristics can be filtered out. But what about variation in rain. It seems like that element would cover the same trees as temperature in a particular chronology. What about CO2 feeding? It would seem to have the same problem when trying to distinguish it from temperature. And the uniformitarian principle would seem to break down in cases where you are calibrating with tree rings who’s average age is much older than the tree rings in the rest of the data set. This is, of course, the case for the Briffa larches. Even pulling the younger trees out of the later dates doesn’t help because within the calibration period the oldest tree rings are stacked into that period.
Is Jim Bouldin Gavin’s new guru? But what’s wrong with Tom PP?
Lucy Skywalker (#213): “But YAD061 etc are being related to the CRU Pan-Arctic temperatures. For a start, much more local temperatures should have been used to correlate Yamal. CRU will not release their stuff, and many suspect that their overall temperatures may be based on a conglomerate of stations improperly adjusted for UHI and other factors, which could render the temperature record null and void as calibrator.”
So the disclosure required by the journal’s data archiving policy is still incomplete? And will remain incomplete until CRU releases their own data? Daaaaaang.
I hope that we all realize that the RCS was not intended to handle the larch problems of changing responses to climate with age as reported by Craig Loehle elsewhere. Briffa and Osborn make note of two different types of growth: linear and exponential that apparently can be handled by separate RCS corrections, but I cannot see how the larch growth that contains growth, both linear and exponential within the same tree that changes with age can be adjusted by RCS.
Do not we need to proceed with a proper larch age sensitivity test in Yamal series?
A new guru, perhaps someone should invite him over here?
He claims the divergence comes from endpoint smoothing.
Re: MikeN (#232),
How can divergence come from “endpoint smoothing” when it’s observed in unfiltered tree ring data from around the world?
Re: MikeN (#232),
He’s PERFECT for Gavin.
Steve, this is one of the finest entries ever on CA. Two thumbs up. Others may find the tone off-putting, but I find it extremely effective at making the point particularly when dealing with asinine hardheads. I applaud your patience with Tom. I gave up reading anything he posted shortly after I realized all he wanted to do was somehow justify the exclusion of perfectly valid data because it didn’t correlate with temperature. That by itself tells you all you need to know.
Get this stuff published. It’s important….
How to Recognize Trees from Quite a Long Way Away
Oh my, I misidentified the error. No way Gavin touches that. Sorry I even posted it. That should be like a test case. Find the error or no R code for you.
Francis posted a useful and relevant map of the areas on Bishop Hill’s blog.
.
Yamal Area
.
Bishop Hill on The Yamal Explosion
Funniest and most illuminating discussion for ages. Thanks Steve, Jeff, Roman, Bender etc.
As usual when these discussions arise, because I’m no Dendro, I turn to Fritts “Tree Rings and Climate” for a refresher. [Hint to some here].
However, I’m puzzled by something that hasn’t been discussed here unless I missed it.
Gavin says H&S is a winter temp recon, whereas Keith Briffa’s is a summer temp recon according to Jim Bouldin. If we’re looking for the AGW fingerprint, shouldn’t we therefore be looking at H&S temps rather than Keith Briffa’s? Maybe I’m just thick, but I can’t seem to find anything to substantiate Gavin’s claim.
Re: Pompous Git (#242)
I can’t seem to find where Gavin claimed that. Reference?
Tom P (#243), give it up.
Your ad hoc “sensitivity test” has no theoretical basis and you don’t provide any reasonable statistics to evaluate what you think you are achieving by doing it.
You don’t seem to realize that the a key problem here is the use of RCS standardization . Steve is quite right in his assertion that the hockey stick is created in this case by the use of the 12 selected long-lived proxies.:
Yes, it does! If you look at my graph in comment #63, and in Jeff’s thread on tAV, you will see that the reason is that the exponential adjustment which is calculated from all the proxies (including a very large number of shorter ones) exaggerates the effects in the tails of the longer lived one. Since the recent proxies are predominantly long, the blade is large. Hu’s observation earlier that this type of tree exhibits a stronger response to temperature late in life indicates that what we are seeing should not be a surprise.
More analysis of the effects of RCS should produce some useful insight.
Re: RomanM (#246), Thank you Roman. Earlier I reference a paper that did a simulation of tree growth in response to temperature. They were basically evaluating RCS versus other methods. The lowest sample size they tested was at 15 cores. The other observation I wondered about was the existence of trees with lower than expected early growth. This in my mind would really throw a negative exponential fit out of wack, especially if you are looking for rapid recent growth.
Nick Stokes @ 245
“I can’t seem to find where Gavin claimed that. Reference?”
My bad. It was *Mike* @ RC.
“…the migration of the treeline depends on factors such as permafrost distribution which is greatly influenced by winter temperatures. Tree-ring growth in these regions, however, is generally reflective of summer temperatures. So even if a quantitative reconstruction from treelines were available, it wouldn’t even be comparable in terms of the seasonality reflected by the record. In short, there is nothing there that challenges the quantitative climate reconstructions provided by tree-rings. Take your talking points elsewhere. -mike]”
Re: Pompous Git (#257),
To the extent that winter and summer temperatures are correlated over the time scale of interest, it doesn’t really matter. The thing that affects the extent and depth of soil freezing is the number of days – regardless when they occur – below zero. For permafrost, presumbaly the shoulder seasons are the swing vote, not summer, not winter. But temperatures in the middle of the growing season will affect the rate of cell division during its peak. I’m not sure it pays to pick away at this point.
Re: Pompous Git (#257),
Quoting Michael Mann
How many non seqtiturs are in this statement? To me, this is pure nonsense. I recall Joliffe’s comment about, in paraphrase, people shouting louder and longer to make points in climate science. I would tend to agree with him in this case. If a treeline reconstruction as available which showed that winter temperatures were warmer and this was in distinction to a summer reconstruction from tree rings then these reconstructions would not challenge each other? I find this difficult to understand.
Re: TAG (#262),
Thank you, TAG, for correcting me. I did not see the non-sequitir there; and the context (always missing in a quote) is important too.
.
To argue that treeline reconstructions are irrelevant is disingenuous. We are interested in all-season GMT, not just whether mike could wear a swimsuit in a given summer. His suggestion that someone was working from a set of “talking points” is speculative and pretty offensive.
.
Whether treeline reconstructions reflect winter temperature and growth-based reconstructions reflect summer temperature is a separate matter. If he’s right, it would make sense to consider both lines of evidence jointly, as they are, by his own logic, orthogonal (independent components of the annual mean). And if they correlate, great. So much stronger the evidence. To say that the two are not comparable is “bizarre”.
You might want to look up the topic ‘YAD06 – The Most Influential Tree in the World’, but I doubt you’d understand it because…
Talking to you is like talking to a brick wall, you have similar reading comprehension skills. Until you can understand where you are wrong in the above, basic statement, there’s no reason anyone should pay any attention to anything you have to say.
Re: Eric JD (#259),
On the contrary I think Tom should take his argument over to RC and show it off. They might think very differently about it.
Tom P, I’m still trying to understand what you did.
Did you for each test, remove the trees that lived to less than a certain age, and than plot the remainder?
bender (#256),
Steve MacIntyre at
http://www.climateaudit.org/?p=7168
Bender, I’m tempted to say you are the court jester, but you lack the requisite intelligence normally associated with the role. However, in this particular court you’re a fair match to your king.
Re: Tom P (#271),
Tom, he never made the accusation in the first place – nothing to withdraw. Are we supposed to believe you can’t figure that out either?
By the way, the original authors claim 17 trees were available and transferred to Briffa and we see that only 12 were used by Briffa. I’ve asked a question of Steve on the other thread about it but I’m curious what happened there.
Re: Jeff Id (#272),
Arrg Jeff I can’t find your other comment .. here is the paper
Click to access TRRvol60_2_77-90.pdf
Maybe they will release their code that simulates trees..
Re: Tom P (#271),
Did you not read his comment: “as noted in other posts and comments …”? Not surprised.
.
I have more for you later. But I will let you wallow in this error of yours until you get up to speed as to why semantics matter. There’s “picking”, and there’s “picking through”. VERY different things, we are told.
.
Have you read the “Yamal IPCC AR4 reviewer comments” thread? Please try to keep up.
Re: Tom P (#271),
Wow, do you love to cherry pick your quotes. Steve clearly stated in the original post
That’s on Sept. 27th. Yet you decide to pull a quote from something he posted at 11:30 PM, a week later, which he quickly corrected. His original intention is clear, he blamed the Russians.
snip – polite language please.
Tom P, can you please answer the question I posed in 258? I’m really just trying to understand. Thanks.
Re: Michael Smith (#274),
Maybe I can try to answer that for you:
Maybe not – not a lot of hockey sticks here.
From Jim Bouldin (#166):
June and July look pretty flat to me. Maybe if I tilt the monitor a little. let’s see …
Re: RomanM (#276),
Maybe that’s why the actual correlation – the one incorrectly reported as 0.49 in 4AR – is only 0.12. Oops, heh heh.
Re: RomanM (#276),
Right. A global annual temperature reconstruction that derives its shape from one stand of treemometers working 5 weeks out of 52.
Thanks very much, RomanM.
I wonder what the correlation would be between the temperature record provided by RomanM and the Schweingruber trees that Steve used in the sensitivity test he applied to Briffa? Or has that already been calculated and I simply missed it?
bender:
“Right. A global annual temperature reconstruction that derives its shape from one stand of treemometers working 5 weeks out of 52.”
Were there any years in which the trees wdid not grow at all because of low temperatures all year? If there were, how could you tell how many of these years there were and when they happened? Maybe your average tree has lots of ‘missing rings’. Just a thought…
Re: Jimmy Haigh (#282),
Steve has, thankfully, asked us not to discuss very basic dendro issues that anybody could look up with a bit of effort. [Missing rings do happen and are dealt with by crossdating and inserting placeholders where years are strongly suspected missing.]
Re: bender (#283),
P.S. Note this implies a fairly fundamental nonlinearity in the treemometer concept, not at the warm end, but at the cold end. You can’t have negative increment. So from the get-go, the linear models that are always used are an approximation.
Re: bender (#284),
That is one of the several reasons that I am of the opinion that a log transform might be applied to tree ring width model before doing the subsequent analysis and then transforming the chronology back at the end of the process
If Steve wrote “CRU cherrypicking” but not “Briffa cherrypicking”, what is the point of quoting “Briffa cherrypicking” (especially in square brackets] in your complaining post? These days, the belief that CRU has a preference for alarmist data over more pedestrian data is in any case not an opinion to be dismissed as extremist or unreasonable. This change in public opinion has been brought about by the unit’s own activities and its reluctance to support the call for public data archiving and code access.
Tom P
Briffa is guilty of the use of cherry picked data and cherry picking which chronologies were used to beef it up (or not, as the facts stand).
Steve’s comment can certainly be read in that context, and is absolutely not a claim that Briffa cherry picked the specific tree cores. So, no, he has not withdrawn the claim the Briffa cherry picked the tree cores because he never made it.
A big name climate scientist has hidden for almost 10 years the fact that his widely used temperature reconstruction had a 20th C surge which depended on a sub selection with a mere 10 tree cores.
Truly shocking, and yet the focus at realclimate is to persuade it’s gullible followers that Steve M is merely bad mouthing prominent scientists. Amazing.
Tom P at last appears to be grudgingly acknowledging the falsehood of his statement in #255, after 40 comments. Of course it was pointed out immediately by bender in #256.
Tom P, #288, what part of the difference between “CRU cherry picking” and “Briffa cherry picking” do you not understand?
Re: PaulM (#293),
Steve McIntyre doesn’t seem to have the same problem as you in identifying Briffa with the CRU archive. Just look at his code:
Re: David (#292)
Steve first “badmouthed” Briffa for cherry picking data, and then the Russian dendroclimatologists who made the core measurements in the first place. But his basis for doing so are unravelling fast.
Re: Tom P (#294),
(1) Steve M carefully avoided accusing anyone of anything. Gavin S was the one who started “bad-mouthing” by putting words in Steve’s mouth.
.
(2) Playing this never-ending gotcha game of “he said she said” is not going to help anyone. It looks ugly to any decent human observer not familiar with the details of the case.
.
(3) It is the nature of the data that has brought suspicion upon itself. Who can deny: the sample *is* suspicious.
(a) It looks like Mann’s hockey stick. We all know the result of that audit.
(b) The sample is small.
(c) One of the trees has an 8-sigma uptick unlikely to be the product of temperature rise.
(d) The tree-temperature correlations reported in AR4 are incorrect, inflated by a factor of four.
(e) The Polar Urals-temperature correlation matches the one reported in AR4.
(f) The CRU dozen may have been substituted in place of Polar Urals.
No one’s accusing anyone of anything. But we clearly have a problem here. Someone did something to arrive at this suspicious result.
Re: bender (#299),
(g) The CRU Yamal larch diverge sharply from the Schweingruber sample from a nearby location.
(h) It is exceedingly unlikely Briffa was not aware of the Schweingruber sample.
(i) The discrepancy was never reported until two weeks ago.
(j) The small sample size on the CRU Yaml larch was never reported until 2 weeks ago.
Re: bender (#301),
(k) The net delay in reporting the sample metadata – NOT the data itself – just the metadata – was 10 years.
.
The length of time Steve M has been waiting for the actual data – Rabett’s new yacking point – is not nearly as relevant as the length of delay in reporting the sample size. 10 years? Given the divergence between CRU/Briffa vs. Schweingruber, this lengthy delay is indeed suspicious.
Re: bender (#299), Bender, my point is that Gavin didn’t start the “bad-mouthing”, nor Tom. Long before, many of Steve’s supporters, including a CA moderator, had taken him to imply cherry-picking, or worse, by Briffa. And that didn’t seem to be seen then as “bad-mouthing”. From the Register, for example, it was welcomed.
Re: Nick Stokes (#303),
And my point is: you are wrong. The chronology published by Briffa has brought suspicion upon itself. Accusing Steve of “bad-mouthing” is shooting the messenger. Or are you in denial that there is a problem here?
Re: bender (#304), Bender, you’re getting muddled. No-one has accused Steve of “bad-mouthing”. You accused Gavin of “bad-mouthing” Steve, and I was saying that many SM supporters had earlier drawn the same inference from his posts, which was not then thought of as “bad-mouthing”.
Re: Nick Stokes (#306),
No, it is you who is “muddled”, but I can help you. Gavin put bad words in Steve’s mouth. Does that help clarify?
Re: bender (#304), bender you should know from nicks past performance here and his performance over at lucia in defense of Tamino that he is the type of person who is incapable of uttering the words ” I was wrong.” Except perhaps in mathematical areas. But when it comes to semantics and interpretations he puts to shame any scholastic I have ever had the occasion to “reason” with. Angels on the heads of pins kinda guy he is.
Re: steven mosher (#341),
I will be disappearing in short order. Once things have settled down. Then I won’t have to worry about answering to spinmeisters. They can misquote me all they want in my absence. The facts speak for themselves … for anyone willing to listen.
Re: Nick Stokes (#303),
I have difficulty with the statement ‘bad mouthing’ when it’s actually a known practice to sort chronologies in dendro. We all think of the practice as disingenuous but if that’s the norm, how can it be bad mouthing. Even H % S sort the data for high variance.
Re: Jeff Id (#310), Jeff, I don’t entirely disagree, but Bender was accusing Gavin of “starting the bad-mouthing” by saying Steve said this. And yes, there is an earlier reference to “bad-mouthing”. But the issue here is not whether that would have been a bad thing for Steve to imply, but on Steve’s rather odd insistence that his post didn’t carry that implication. And the even odder accusation o gainst Gavin and Tom of “foulness” or whatever, in believing, along with so many others, that it did.
Re: Nick Stokes (#311),
The CRU chronology is suspicious. Steve uncovered the suspicious sample. Gavin claims no expertise. I suggest that Steve has the expertise that Gavin claims he lacks. These are the material facts of the skirmish. The rest is irrelevant gamesmanship.
Re: bender (#312),
Indeed, bender. That’s the most succinct resume of the recent discussions.
I would so much more prefer reading about the possibilities, problems and caveats when using tree ring data for reconstructions than the endless and tiresome talk about why SteveMc did this or that or if he should be blamed for other people’s reading of his blog or not.
Re: bender (#299),
#(e) Is the match exact? Do we have real “finger print” evidence for just suspicion?
Re: charlesH (#327),
Ask Steve how close the match is. Don’t quote me, but I thought it was at least two decimal places. I would think even a single decimal place match would be considered disquieting, as there is some non-trivial slop in how these series are assembled and compared. I wouldn’t look for “fingerprints”. I would look at the facts: an overstated calibration, by a factor of four, is somewhat suspicious. Perhaps it was a typo. Maybe the correlation was mistyped as 0.49 instead of 0.19. Who knows, except the text authors?
PaulM: unless you can quote Steve saying that Briffa cherrypicked data, what you are doing here is simply foul. Would you please provide chapter and verse for your accusation (omitting the very dirty square bracket business implying that he said something he didn’t). That is to say, if you are a gentleman, you ought either provide the supporting quote for the accusation or else acknowledge its foulness and apologise.
Re: kimberley_cornish (#295),
This is getting really over the top. Tom has done that. But many supportive blogs and press reports have seen Steve’s posts as implying cherrypicking by Briffa. Anthony Watts (who has been moderating this very thread) put up a WUWT post calling on Briffa to explain or resign.
Orlowski, in the Register, was quite explicit. He said
“The implication is clear – the dozen were cherry-picked”
And since it was Briffa who chose a dozen, there’s no doubt who is being pointed to. So did CA demand a retraction of this “foulness”? No, it put up a gleeful post “The Register picks up the Yamal story”, quoting the Register’s title “Treemometers: a new scientific scandal”.
Is it only foul when Tom says it?
Re: Nick Stokes (#298),
Read the blog. Then you will see the logical reply to this. If you are upset at how someone has interpreted, or extrapolated from, Steve’s words, take your case there. You should know better than to quote from third parties. Come to the source. That’s what real journalists do. Propagandists … they work differently.
The “Paul M” of my previous, should, of course, have been “Tom P”. My apologies to Paul M for the oversight. So again, to Tom P, may we have chapter and verse, please, for where Steve is alleged to have accused Briffa of cherrypicking his data?
Tom P
“But his basis for doing so are unravelling fast.”
Don’t quite follow. Nope, Steve never accused Briffa of cherry picking the cores, and has explicitly several time said that is not his view. The CRU data though HAS (everyone agrees) been cherry picked, so reference to the CRU cherry picked data is entirely accurate, and that statement does not by any stretch of the imagination imply that Briffa cherry picked the cores.
Again though, you are very focussed on trying to prove a very minor point (whether someone thinks or not that Briffa cherry picked the cores), when the main point which again is not disputed by anyone is that the study using these cherry picked cores is dependent for its 20th C surge on 10 cores. It is not statistically robust AT ALL – a fact which has been hidden for almost a decade, despite it’s use as a corner stone for climate alarmism.
Again, it’s terrifying that something like that has taken so long to come out, and yet you and other real climate devotees seem intent on turning this into an issue of whether or not Steve accused someone in particularly of making the cherry pick. It’s like a sketch from Monty Python.
If Briffa had revealed that his chronology was built on 2000 samples from across the region then we would have a minor disclosure problem. That it was based on just a dozen changes everything.
TomP is being quite astute here. He can’t deny all the problems with the data so has switched his attack to SteveM. Stokesie is just a troll being sucked in by the opportunity that TPP is offering and the whole thing is degenerating into an unending slanging match taking the thread away from the analyses and the REAL problem.
Bender if you can, and in the past you have been able, ignore the trolls and stick to what you do really well, the analyses.
Phil.: It’s the second sentence, fourth paragraph after the second blockquote of text from a Tom P. comment — if that helps.
Jeff, there are 5 more trees in the 1900s that might be the other 5 cores.
Re: MikeN (#342),
Thanks Mike but here’s the quote.
If the trees are living they should reach the end years of the graph right? I think there may be other explanations but I’m really curious.
Re: Jeff Id (#348),
I have not been closely tracking the details. But “living” could mean “uprightly rooted and died only recently”. Remember, they are trying to distinguish these samples from the subfossilized wood *pieces*. In that case it might make sense to refer to all long sequences as “living”. I wouldn’t make any hard and fast assumptions about these mere labels that must pass for proper metadata.
Re: Jeff Id (#326),
Here is my surmise as to the other 5 “living cores”. Cores M021, M202 and M331 all end in 1963, indicating that the cores were taken in 1963 and cores X02S and X13 end in 1978 and 1983 respectively. Other than the CRU 12, these are the only cores without the subfossil prefix. So it looks for sure that these are the other 5 living cores.
NCDC series russ006.crn and russ007.crn (chronologies without measurement data) are larch chronologies from Khadyta River, Yamal taken by Shiyatov and ending in 1963 and 1964 respectively. One site is marked “near spring” and the other is marker “intermediate moisture”.
I think that it is a reasonable presumption that the M-series cores come from the Shiyatov 1963/1964 campaign. It looks like at least 33 cores were taken (M33), tho not necessarily all of them were crossdated.
The location of the Shiyatov samples on Khadyta River were 66 55 N 69 10E, while the Schweingruber Khadyta River samples were 67 12N 69 50E – further demonstrating that the Schweingruber site was well within the Yamal perimeter.
I’m going to give Tom a break, as he doesn’t understand the implications of cherry-picking proxies. Steve’s quote even in context can be read as talking about Briffa. CRU vs Briffa is irrelevant, unless the Russians were part of CRU. It is not the cherry-picking of Yamal vs Polar Urals, which may be Steve had in mind, being discussed. He is talking about a combined chronology, and saying I would leave out the CRU-cherry-picked-data. In that context, leaving out a cherry-picked sample vs a full sample, then the cherry-picked-data should be understood as Yamal, and specifically the modern cores.
bender: “To the extent that winter and summer temperatures are correlated over the time scale of interest, it doesn’t really matter. The thing that affects the extent and depth of soil freezing is the number of days – regardless when they occur – below zero. For permafrost, presumbaly the shoulder seasons are the swing vote, not summer, not winter. But temperatures in the middle of the growing season will affect the rate of cell division during its peak. I’m not sure it pays to pick away at this point.
TAG: “How many non seqtiturs are in this statement? To me, this is pure nonsense.”
Many thanks 🙂
[Please note that any references to Russians, Keith Briffa or Steve McIntyre in my statement above are entirely in your imagination. Please see a doctor if your delusions persist.]
TomP, what would your chart have had to look like for you to conclude the two signals were coherent?
re: TomP 328
Isn’t it a result of the two plots having similar shapes?
Sorry, the second link in my previous message doesn’t seem to work. Either (a) try the link
http://img13.imageshack.us/img13/3089/milkfeedb
or (b) click on the first link, go to the identified comment, and click on the link in that comment.
Again I’m sorry. I’ll try one more time. Try
TomP,
The “disquieting” image:
shows the difference between Yamal and Schweingruber.
The end of Yamal does not match local measured temperatures, and is dependent on only 10 cores for its exponential response. You simply cannot claim therefore that whatever selection process which was taken (and whoever did it) giving the subsample actually improved the correlation with temperature. It does not.
Yet it is Yamal which is (ab)used in the proxy studies, frequently in such a way that it gets a heavy weighting.
Steve was right, it’s very disquieting.
I’ve found Hantemirov’s PhD Thesis in Russian. Mostly it is as in Hantemirov and Shiyatov 2002, but now they have 7000 year chronology and much more trees (compare Fig. 2 in Thesis with Fig.6 in HS02). Growth of tree rings of larch reflects temperatures between June 16 to July 30 (Page 18). In the conclusions, Hantemirov states among other, that the temperatures unprecedentedly increased in the last 150 years, which led to overall improvement in the larch growth and forest density. The new version of Yamal is at the page 19 (with 50, 100 and 200 year low-pass filter)and may be compared to Fig 11 in HS02. It is definitely more hockeystickish.
Re: EW (#342),
the link is given in Steve’s latest post. IMO, the smoothing in the thesis may suffer from the end-point-problems, and I’d like to see the unsmoothed version before making any statement if it is anything like Briffa’s chronology. Can you figure out what was the smoothing used in the thesis?
Re: Jean S (#343),
Is it?
I would swear that when I started to compose the posting, the Yamal divergence article wasn’t up yet. I just searched Russian portal Jandex for words Yamal and Hantemirov, found the .pdf and tried to find out what’s new when compared with HS02, that’s all.
Anyway – except for the use of low pass filter as mentioned above, there are no further statistical details and also no mention of archive, where the data might be found. They analyzed 1103 trees consisting of 120 samples from living trees and 983 from semi-fossils.
The reconstruction on Fig. 5 is defined as that of mean summer temps from June 16 to July 30, where correlation coefficient between index of width of tree rings and mean temp of the said interval is 0.71. Fig. 6 shows between-year changes of temperature expressed as running values of standard deviation in 100-year intervals (?? maybe my translation abilities aren’t exactly up to this…). The author concludes that the larches liked more, when there was stable climate and much less when there were great summer temp differences between years.
The problem is that this .pdf isn’t the complete Thesis, but some sort of short summary.
Re: EW (#342),
The divergence is still there, in Fig 2.
Re: bender (#348),
Figure 2 – Distribution of the number of samples of wood used
for climate reconstructions based on the width of annual rings
(thin line shows the proportion of samples from living trees)
Looks like that, but I think it’s an overview of specimens over the centuries. Y-axis is No. of specimens.
Re: EW (#354),
Thanks. That’s why I ask. It didn’t make sense to me to be plotting important data as early as Fig 2.
Re: EW (#342),
Briffa will likely have had advanced knowledge of this result. If it withstands scrutiny, his non-disclosure of the earlier small sample size may become academic. The broken Yamal stick may be reparable. *IF* this result withstands scrutiny.
Re: EW (#342),
Figs 6-7 are pointing at the number of days with temperatures above a threshold? If the process limiting Yamal larch growth is thresholded (e.g. soil temperature limiting root initiation) then the calibration must be nonlinear. Watch for this when the translation comes out. Use of a linear calibration model for a thresholded/nonlinear process could lead to some interesting extrapolation effects.
Figure 6 – Interannual variability of summer temperatures, expressed in
moving standard deviations for the 100-year intervals
Figure 7 – The number of extremely cold summers (with temperature below the average for more than 1.5 ° C plotted as sums per centuries (?). Dark part – seasons with temperatures more than 2 ° C below average.
Hantemirov notes that the forest density changed throughout the centuries as well.
I agree with Tom P on one thing – R is not generally used by anyone except statisticians.
But on the more substantive point, his graph simply does not show what he claims it does.
EW – thanks for the helpful translations of Hantemirov’s captions. Are you Russian?
Re: PaulM (#358),
Yeah, but it’s free.
Re: PaulM (#358),
No, I’m not Russian, I’m Czech who learned Russian years ago. Therefore I can read or search in Cyrillic and catch, where Google translator does not get the proper context.
I might add that most time series will have an OWN correlagram that decays with lag length too. So, if your assertion is that this is a cross correlogram, and the correlations get poorer at lower frequencies, that’s just not interesting. Again it would be true for the cross correlogram of most time series with themselves. Oh and by the way: Nothing is evident to my eye. I want a test statistic. Do you know how to make them?
Re: NW (#361),
What the plot shows to me is that the short sequences (young trees?) may crossdate well with the long sequences (oder trees?) because of the stronger (but how strong?) high frequency coherence on the right side of the plot, but do not contain the same stregth low-frequency trends and vacillations (supposedly climate signal) on the left side of the plot.
.
But I think Tom P is getting ahead of himself here. He needs to construct a full argument. I don’t see that it’s relevant. But I’m willing to be convinced.
.
Tom P: In putting your argument together be sure to label your axes.
Gavin’s other guru is Jim Bouldin. I thought his name was familiar.
Check out this other drive-by of gentleman Jim at lucia’s blog:
Read your head post! Was that reluctantly written?
I’d like you rather to respond to the substantive issue I brought up four days ago (#80), to which you have so far turned a blind eye.
Tom P – please can you replot your coherence graphics with the horizontal scale expanded by a factor of (say) 3? Thanks
Tom P.
You are not dealing wioth raw data and I don’t see where you took into account that there is an adjustment for the growth curve. This adjustment changes pretty radically radically if the mix of ages is changed. Perhaps you could point out how you took this into account.
TomP: Not sure what you’re looking for exactly. Steve has already responded to your “substantive issue I brought up four days ago (#80), to which you have so far turned a blind eye.”
Steve wrote: “As I noted elsewhere, as far as I can tell, your methods are totally unknown in dendro literature. Can you please provide me with any precedents for this sort of argument in any dendro literature?”
Soooo… you created a new method for analyzing tree ring data that’s never been seen or tested in the dendro literature before…But, gosh you got the signal you were looking for! Seems that your comments and time would be better spent discussing the issues with the professional dendros. Who knows? You may have actually come up with THE method for identifying temperature found in tree rings (when analyzed after the fact)! Of course, if that’s the case, then that means that all the previous dendro-reconstruction literature is wrong.
Bruce
Identifying signal from nose in these threads is definitely frustrating work.
As far as I can tell, the question “If one eliminates data from trees that do not, for unknown reasons, correspond to current SAT – how then can that be done with confidence for the previous +1800 years ?”
seems to have gone begging. Yet this is the critical question.
Damn – #381, shold be noise, of course, not nose
Unless I am missing the full significance of the concept of “corridor” … does this concept mean that samples are ring-correlated over time, ie. ring patterns are correlated from dated fossil wood to dead trees to living trees to construct a time series based on these dated, correlated fragments ?
Given the iconoclastic nature of the AGW posts here, I expect several answers of the “do your homework” genre. Ho hum
Tom P,
Come on. You can’t expect to write the arrant nonsense you have done so on this blog (identical series are well correlated with each other!!), followed by caustic over-reaching comments on other web sites and not expect some response from Steve. If you have been subject to mockery, it is self inflicted.
You continue to state that Steve has “left behind quite a mess here”. Maybe you are right, but you certainly don’t appear to be in a position to judge.
From the head post quoting a Gavin Schmidt comment on RealClimate
Does anyboby know what this means? It seems backward. Isn’t the finding that TomP is touting is that the Yamal result is not robust to changes in age structure. The Yamal result disappears as the age structure is made less homogeneous.
Dr. Schmidt’s use of the term ‘robust’ seems to be backward. If the result is only found for homogeneous age samples then the result is not robust with respect to age structure.
Re: TAG (#387),
I urge caution in discussing this work that Tom P has done. He has not presented a coherent argument worthy of scrutiny. Which is not to say I don’t encourage him to do so. It would be nice to know what he thinks he’s accomplishing.
.
I say this because Gavin’s remark was in reference to some work that Tom P did a week ago. The work that Tom P is presenting now is different. So be careful not to distort what Gavin said. He’s not using the term “robust” incorrectly. It’s Tom P that’s not being clear about what his various “sensitivity tests” are showing, what *he* means by something being “robust”. It’s just not worth saying much more until Tom P gets his act together.
Bender,
It’s obvious what Tom P is trying to accomplish. He introduces something that looks like it could be relevant, refuses to explain or defend it, and then announces victory far and wide because SM won’t drop everything to refute what even Tom P can’t explain. It’s a common political technique; keep shouting indignantly for answers but never answer any yourself.
Steve made a mistake in engaging him with sarcasm in the head post. Not that I blame Steve, I would have wanted to slap him too. But the end result was that people here paid more attention to Tom P than they would have otherwise, simply because he annoyed Steve. To top it off, all that attention gave him ammunition for the other sites.
Tom P accomplished exactly what he wanted.
TomP – is that a “no” to the question posed?:
Re: curious (#391),
Not at all:
“Climate signal age effects—Evidence from young and old trees in the Swiss Engadin” Espera, Niederera, Bebic and Frank, Forest Ecology and Management, Volume 255, Issue 11, 15 June 2008
From the abstract:
The same cannot be said of the combined Yamal and Khadayta River data sets that Steve advocates.
Re: Tom P (#393),
If that is so, should they (both Yamal and Khadyta) not be discarded and Briffa’s work thus be declared worthless, if not misleading? The very obvious “change in climatic signal” with using various subsets both Steve and you have demonstrated by very different methods ought to prove that there is very little “climatic signal” in ANY of these subsets. In my own field of audiovisual signal processing, one way to recover a signal swamped in noise is to average several noisy sources (datasets) containing the same signal: The noise will even itself out while the true signal is reinforced. Now all I see in both Steve’s and your comparative graphs is that the “interesting parts”, namely the 20th century uptick and the elevated MWP, appear and disappear at random depending on what selection criteria are used – ergo, they must be nothing but artifacts of noise. These groups of trees simply do not seem to contain much “climatic signal”, or all cores young and old (except maybe very few defective ones) should have obvious common features – like those looked-at by Espera et al. apparently had.
I am wondering more and more who IS this “Tom P”? For all I can see he might as well be Gavin (or Briffa) himself, or someone in his pay, posturing as a layman in an attempt to disrupt and ridicule the science done here. Steve, have you done some research on the e-mail address and ISP he uses to verify he is a real person? If not, IMHO it would be high time to do so and post the results here.
Likewise, I very much doubt that the “Manola Brunet” posting offensive BS on the “Core Counts” thread is really the eponymous researcher (unless she happened to be very drunk at the time…). Could it be that the two are related?
TomP – thanks for the reference. The complete abstract is:
I haven’t followed your argument blow by blow (as an aside: I think you should take bender’s advice and pull it together into a single post – I’m pretty sure Steve offered the facility of a guest post to those with contradictory views – it would help clarify and consolidate your position) so I’d be grateful if you could comment on the bolded suggestion of separate calibrations for different age categories against the instrumental record. Are you in support of this? Apologies if you have covered this elsewhere and I’ve missed it.
(Also I’d still appreciate your coherence plots with expanded horizontal scales).
TomP:
Since TomP seems to think he’s expert all of a sudden, we should demand that TomP point out to us what Steve has done wrong or demonstrate that what Steve is doing is “totally unknown” within the dendro field.
TomP could then explain to us what is wrong in and of itself with a method being novel. It’s of course irrelevant whether it’s common practice or no, what’s relevant is whether it’s been correctly done.
Re: Carrick (#400),
Carrick,
Interesting mistake. Your point is valid if you just swap Tom P and Steve McIntyre.
Steve McIntyre doesn’t actually think what I’m doing is wrong – he used my analysis to try prove a point in his latest post! Steve also doesn’t seem to keep up with, or maybe recall, the dendro literature – it didn’t take me very long to find the Jan Esper paper that had previously used a similar approach to mine.
Steve really should have remembered Esper doing work on age sensitivity. Just six months ago he posted on Esper’s work on Morocco cores,
http://www.climateaudit.org/?p=5694
In it Steve writes:
Six months ago Steve was arguing that it didn’t make sense to combine signals from different sites in Morocco because of a divergence in the ring widths. But apparently what wasn’t plausible in Morocco is now the right approach in Siberia. First he combines the divergent Yamal and Khadayta River data sets. He follows this by what is no longer apparently a “peculiar” stratification by age to obtain the young-tree chronology without a hockeystick of his figure 2 above.
Steve is giving the unfortunate impression that his choice of assumptions rather depends on what he wishes to prove.
Re: Tom P (#404),
TomP please comment on this:
Esper as quoted in the post that you reference:
Doesn’t this observation by Esper argue against the point you are ttying to make about ring widths in young trees
Carrick, the above quote is McIntyre on Tom P’s methods.
Tom, you never answer my questions as to what you are doing.
Is your original ‘robustness test’ a calculation of the RCS chronology with trees that lived to more than age x only?
In your graph of the difference that you say shows incoherence, what would the graph have to look like for us to conclude it is coherent?
Steve has explained why the chronology is not robust. Your test does nothing to change that. Perhaps your first test can show nonrobustness but not robustness, especially since Steve’s objections are not related to your test.
TomP:
I apologize for misattributing it to you.
But I agree with you, the point is still valid regardless of who uttered and who it was directed to.
There’s nothing wrong with novel methods per se, and if somebody is going to criticize you for a particular technique not being canonical, seems to me they at least need demonstrate that what you are doing is neither canonical nor perhaps even statistically valid.
Re: Carrick (#408),
Isn’t that proving a negative or proving that thre are no white ravens?
TAG, you can’t start out saying a technique isn’t canonical, then when asked to substantiate it, say that this is impossible to do.
If you can’t “prove the negative”, then I guess you didn’t know whether the technique was canonical to start with, right?
If Steve thinks that TomP isn’t using standard methods, perhaps he could at least point us to which analysis methods those are, and further explain to us why he thinks that TomP’s nonstandard methods are necessarily worse than standard dendro practice.
The only novel method I see here is Steve’s. Breaking a coherent chronology and throwing noise into it, while removing the data used in the calibration results in garbage. delayed.oscillator runs the numbers in a slightly more appropriate way. He includes the Khadtya river cores and recomputes the chronology w/o removing the recent data from the Yamal series and gets only a small difference. If the question Steve is asking is why didn’t Briffa use the Schweingruber series, the answer is that a) it seemed to have been collected for different purposes (looking at a host of climatic variables) and b) including it didn’t make much (if any) difference if you didn’t use the “McIntyre mendacity”.
Re: Rattus Norvegicus (#411),
Don’t be silly. The point of Steve’s method is to show that the Yamal chronology isn’t robust to the removal of the Yamal 10 (or 12). Delayed.oscillator is, like Tom P, trying to keep the small band of live trees because no matter what else is done you’ll still have a hockey stick. 10-12 trees, especially when their cores have been run through RCS, can’t give a stastically valid result.
Re: Dave Dardinger (#412),
Please read http://www.treeringsociety.org/TRBTRR/TRRvol60_2_77-90.pdf and then explain why the authors are wrong and you are right.
Re: Tom P (#413),
Is this the only paper you’ve seen on this topic? From the referenced paper:
So you are saying that a paper based purely on modelling foxtail pines with artificially generated “tree rings” with trees whose ages range from 300 to 1700 years each is the definitive statement on actual yamal larches whose lifetimes are mainly less than 300 years?
And how do you get a sample of trees from a Normal(550,250) to have a range from -1 SD to 4.6 SDs?
Re: RomanM (#414),
There’s little reason that the absolute ages of the trees would make any difference to the analysis – it just affects the time axis. Hence if you’d like to apply the results to trees half the age, just divide all relevant times by one half.
The important parameters are both the absolute number of trees, used to get an exponential fit to the growth curve, and the number of trees at any one time. As the Yamal dataset has 400 trees in total, this first parameter is not at issue. This paper shows that when there are less than 20 trees in total, or an average of five at a time, the ability of the RCS chronology to pick up centennial periods is impaired. But even then the paper says that above its minimum number of 15 trees “at high enough amplitudes signals are generally detectable regardless of period or the number of trees.”
As to your last question, I presume they simply truncated the gaussian.
Re: Tom P (#416),
This is a tatuology: “at high enough amplitudes signals are generally detectable regardless of period or the number of trees”.
.
Now you have to prove you have a signal of “high enough amplitude”. You can’t just assume it.
Re: Tom P (#416),
Nonsense. There are very real differences in the growth rate curves of the Yamal series over proxies of different age lengths. These age differences would not be the same in the artificial data. I agree with bender. Just what is the point of the paper and what is the relevance to the issue of the appropriateness of the Yamal 12?
Anybody who knows even some elementary statistics would be aware that truncation would not be responsible for proxies that are 4.6 SDs away from the mean. Such values occur about 1 in 500000 tries.
Re: RomanM (#420),
The Yamal chronology is insensitive to the exclusion of younger trees, so this does not appear to be an issue for the series under discussion.
Because as I have previously explained (twice) the paper shows that 5 cores at any one time are sufficient to be able to identify a signal using an RCS chronology. The size and periodicity of the signal to be identified is shown in figure 7.
Nevertheless, if you have a good reason why 12 trees are statistically insufficient, please explain. The size of the number, though, is not in itself a good enough reason.
Re: Tom P (#423),
As I pointed-out, the smallest number of trees they looked at statistically was 15-20 (subsets A and G…it’s spelled-out on the figure 7 you refer to), and their only conclusion for this size is that RCS and NECS mean values of squared coherency were basically the same. Does that say you can get a statistically valid signal from 5 cores?
The only mention of 5 of anything that I see anywhere is when the authors state, “…we arbitrarily truncated our chronologies to a sample depth of five series…” And if you look at Figure 2, you’ll see the sample depth for a given calendar year in their example year of 60 cores is 20+ the vast majority of the time. It drops below 10 cores once in that simulated history.
Their conclusion concerning sample size: “Although our results indicate that there is no statistical difference between RCS and NECS for a chronology with 15 trees, we are reluctant to give a recommendation on the number of series needed for an accurate regional curve…”
Why don’t you find a paper that says things you claim it does?
Re: Michael Jankowski (#425),
You do actually have to do the relevant but simple calculation. If for instance you have 60 cores of mean age 550 years, then in a 2000 year period there will be on average 60 x 550/2000 or an average of about 16 trees at any one time. For twenty cores about 5 at a time, etc.
There is no one figure for a minimum number of cores. As I quoted above from the paper, it depends in the strength and periodicity of the climate signal you are looking for. Look again at figure 7 for a full analysis here.
Re: Tom P (#428), You claimed, “Because as I have previously explained (twice) the paper shows that 5 cores at any one time are sufficient to be able to identify a signal using an RCS chronology.” Where does it say that in the paper? I am also failing to find where you’ve “previouslyexplained (twice) the paper shows” this. You brought the paper up in #413 and never explained any such thing on this thread.
So there is “no one figure for a minimum number of cores,” yet you claim in the next sentence that I caan “look again at figure 7 for a full analysis here.” Well, does figure 7 have “a full analysis,” or is there still “no one figure for a minimum number of cores?” Make up your mind.
You also claim, “As I quoted above from the paper, it depends in the strength and periodicity of the climate signal you are looking for,” but I don’t see any quotes on this thread from you involving the terms “strength” or “periodicity.”
Re: Tom P (#428),
(Emphasis added)
Well, yes, if “the climate signal you are looking for” is expected to be so strong as to be uniformly detectable in every tree, then you can use a small sample size. Is that what you are claiming Tom P?
Because even if that is what you are claiming, one look at the “CRU 10”, which includes YAD061 as well as YADs 041 and 121, and you know that the population you are sampling has some amount of variation that cannot be ignored in deciding how many samples to take.
So then the question becomes, what sample size is required to properly estimate the signal strength and periodicity in this particular population of trees? Are you saying that the “CRU 10” would convince you that a smaller sample size, say a sample of 5, would be sufficient to represent the population?
If so, then I don’t understand that reasoning at all. Dendroclimatology must be fundamentally different compared to the industrial production sampling I’ve worked with.
Re: Michael Smith (#432),
This is why Tom P is going to have to present a coherent argument. All we have is a scattered collection of drive-by assertions that are often contradictory. Contrast with Steve’s posts which have a point, are internally consistent, use the right analysis to answer the question, provide documentation, etc.
Re: Michael Smith (#432),
The Bunn paper made no such assumption of a uniform response to climate of the trees to validate their sample size. Look at Table 1: their initial tree-ring width varied over a multiple of 2.5, their decay coefficient over a factor of 3, and mean width at asymptote over a factor of 9.
Re: steven mosher (#450),
Truncating the last 5 years in a 250 year chronology is not going to make much difference to the correlation function.
Just throwing up a multitude of links without indicating the point you want to make is not very helpful to the discussion. Where exactly is the minimum number of trees suggested in the literature you cite?
Re: Tom P (#451),
The minimum number of trees for EPS depends on the type of tree it is. I haven’t found briffa and jones 1990 paper online… The chronology stripping paper (pg 55) has the clearest indication citing briffa 1990. a range of 4-25 trees
( or 4-20.. cant recall) however the lower bound of 4 is only recommended for semi arid conifers of western US.. Anyways, the calculation for EPS is shown in the other links and a suggested threshold of .85 is recommended, Looking at some of the curves I’d hazard that you’ll run into some problems there.. there is also a dependency on the form of standardization. Simply put, I think we can safely say that with such a low core count ( tree count?) that before the series was used it would have to be defended separately in the literature. This is a point that Steve has made. Trying to defend it post hoc, as you are doing, will always look suspect. With respec t to the actual correlation, I’m waiting. also dont forget the auto correlation hurdle.. did you check that by eye as well?
Re: steven mosher (#452),
No, this is entirely different from Steve McIntyre’s view that such a small number is indefensible:
And Steve is definitely wrong here. See dendrochronologist Delayed Oscillator’s posting at:
Re: MrPete (#453),
You are being deliberately (I hope) obtuse. There’s no failure of correlation with temperature in the data that was lost through truncation.
Re: Tom P (#456),
Back from a FRIGID Thanksgiving.
Tom you quote my comment:
and say:
I don’t know how you purport to know my state of mind and to presume that I wouldn’t be flabbergasted. Please note that the sentence read “10 trees in 1990”. This isn’t the same thing as “10 cores in 1080”. It’s ridiculously easy to get more than 10 cores in 1990 – that’s why I do not expect any dendrochronologist to support the use of “10 cores in 1990”. Thus far, I haven’t been put to the test since no dendrochronologist has come forward to support Briffa’s use of 10 trees in 1990. Nor do I expect any to do so – particularly, ones that are familiar with RCS methodology.
Re: Steve McIntyre (#457),
Did you read Divergent Oscillator’s words on the subject? I take it you are flabbergasted, then.
Re: Tom P (#459),
Once again your reading comprehension skills are severely lacking.
I’ll try to explain it to you, although I imagine you still wont get it.
Steve would be flabbergasted if there was a dendro out there who would advocate using only 10 trees from 1990 to draw a conclusion. Yet you link to a guy commenting on how Steve did one of his sensitivity tests as proof that someone is advocating exactly that, when he is doing no such thing.
Your posts are nothing but a collection of you misunderstanding what you’ve read and then beating down the resulting straw man argument that you’ve erroneously built up as a result.
Re: Tom P (#459),
Tom P., regarding Steve’s quote, could you point out where exactly “Divergent Oscillator’s words on the subject” (supporting Briffa’s use of 10 trees) are posted?
Re: Tom P (#459),
Leaving aside the issue of whether this guy is qualified on the topic, he did not endorse Briffa’s use of 10 cores in 1990. Your comment is totally offpoint. That is the dendro part of the matter. You haven’t seen Rosanne D’Arrigo or Jan Esper or Rob Wilson or any of the leading RCS dendros defend 10 cores in 1990. If they do, as I said before, I will be “flabbergasted”.
On statistical issues, I don;t place any particular weight on dendro opinions, though I pay careful attention to what they appear to be doing.
As a statistical issue, if one has reason to think that there may be an issue with the how one part of a study was sampled e.g. potential (inadvertent sampling) bias, the usual course of action would be to try to crosscheck it against other data. To do the sensitivity study, your primary analysis is of the contrast with data not including the data in question. That’s elementary data analysis.
If you concluded that there were problems in the sampling, one perfectly logical course of action is not to use the data. There is no legislation requiring the use of the Yamal data and, if reconstructions are as robust as advertised, then surely no one would mind if (say) the Briffa 2000 reconstruction was done without Yamal, and with Polar Urals. 🙂
As others have observed, I’m not trying to argue that the “Schweingruber Variation” is a true history. My point was that the Yamal chronology is based on a pathetic number of 1990 cores, that the abysmally low size of the modern portion of the Yamal sample was unknown to specialists in the field, and that the Yamal chronology is not “robust” to the use of cores in the same area that Briffa himself has used in many studies.
Re: Steve McIntyre (#465),
Well, a quick look showed Esper used less than ten live trees in some of his Asian series in 1990 (10.1007/s00468-006-0104-0). Briffa’s Yumal series is certainly not a one-off in that regard.
As you’re back, though, how about a comment on the marked repeated disagreement over all the record between the young-tree and old-tree chronologies in your head post, despite your claim to the contrary?
You’ve managed to produce not just a recent but chronically divergent series with your choice of trees less than 75 years in your combined Yamal-Khadyta series. That the Briffa Yamal series fails such a sensitivity should be to its credit!
Re: Tom P (#468),
Gosh I google that Tom and I got this
Abstract We present an analysis of 28 juniper tree-ring sites sampled over the last decades by several research teams in the Tien Shan and Karakorum mountains of western central Asia. Ring-width chronologies were developed on a site-by-site basis, using a detrending technique designed to retain low-frequency climate variations. Site chronologies were grouped according to their distance from the upper timberline in the Tien Shan (∼3,400 m a.s.l.) and Karakorum (∼4,000 m), and low- and high-elevation composite chronologies combining data from both mountain systems developed. Comparison of these elevational subsets revealed significant coherence (r = 0.72) over the 1438–1995 common period, which is inconsistent with the concept of differing environmental signals captured in tree-ring data along elevational gradients. It is hypothesized that the uniform growth behavior in central Asian juniper trees has been forced by solar radiation variations controlled via cloud cover changes, but verification of this assumption requires further fieldwork. The high-elevation composite chronology was further compared with existing temperature reconstructions from the Karakorum and Tien Shan, and long-term trend differences discussed. We concluded that the extent of warmth during medieval times cannot be precisely estimated based on ring-width data currently available.
is that the right paper with 28 trees?
Re: Tom P (#456),
There’s a failure to meet the stated requirements. Either it meets the requirements or it doesn’t.
And as I suggested in my followon comment, I believe their after-the-fact filter based on the data should not have been included anyway. But hey, I’m no math expert so am quite interested in others’ take on this.
Re: Michael Jankowski (#425), hehe. I was quite pleased to find that paper and point Tom at it, knowing he would just scan it for sound bites. First I tricked him into using Steve’s R code and he borked up his first program. This is fun. Now he’s inventing new analysis approaches for dendro..If Bender is right. Tom will come up with something where he hoists himself on his own petard.
Re: Tom P (#423),
No, the Yamal chronology is sensitive to the inclusion of the Briffa 12 or a major subset thereof. The living trees were chosen for high-age trees. When five of them are removed, as in your test, the other 5 are still there and dominate the reconstruction. If all ten are removed, then the HS disappears. But whether there are longer-lived trees in the schweingruber set, I don’t know. Has that been looked at in this thread? I.e. has there been a graph of schweingruber trees older than 250 years but with all Briffa trees removed?
Re: RomanM (#420),
On your last minor point:
They ran their model 10,000 times with an average of 57 trees created randomly in each series. 570,000 trees were therefore synthesised and that one was 4.6 standard deviations form the mean is what might be expected. They would have had to do something to stop synthesising trees with negative ages, though, and that might well have led to a simple truncation at 300 years.
Re: Tom P (#416),
This is where the difference between simulation and real data kicks in. On top of the Larch growth anomalies already mentioned, there’s another key issue never raised in that paper. I haven’t the resources to find out if this has been mentioned anywhere…it is simply an observation from the field:
Living trees, by definition, provide tree ring data from early in their life until their current age. And, it is obvious where the current growing area of the tree is located (eg stripbark trees).
Dead trees do not necessarily do so.
a) The outer rings can be eroded over time
b) The final growing area (non-stripbark) may not be clear
For both of these reasons (and more I’m sure), it is quite possible that data from dead trees may miss the growth in the final years of the tree’s life.
As we have seen among stripbark BCP’s, it is quite possible for the final years of old-growth trees to have a significant uptick in growth, independent of climate.
The result: it is possible for otherwise-neutral sampling scenarios to produce a current-time uptick, independent of climate.
None of this is emulated or described in the paper referred to by Tom P.
Perhaps that’s because I’m an ignorant fool. Or perhaps it is a subtle and inconvenient observation that hasn’t been studied.
Or both. 😀
Re: Tom P (#413), the authors seem to consider a minimum of 15 trees in their simulations. The only statistics they seem to mention with regard to the number of trees have to do with the difference beetween RCS and NECS at various numbers of tree samples, and they explicitly state that they do not make a recommendation at all as to the number of minimum trees/cores.
Re: Tom P (#413), How about you summarize what you think is relevant in that paper first? So that the discussion is focused? This is the second time you’ve drawn out attention to that paper without clarifying why.
Re: Tom P (#413),
The following is from page 2 of the paper quoted by Tom P.
“”Foxtail Pine is a high elevation temperature sensitive species form the Sierra Nevada, USA that has been critical in many paleoclimatic studies and has shown substantive low-frequency variation(Graumlich 1991;Graumlich 1993; Scuderi 1993; Lloyd and Graumlich 1997). All growth and chronology parameters were built using existing foxtail pine cores from Graumlich’s unpublished data.””
The first thing to note here is foxtail pine. There are many dendros who have said that foxtail pine and bristlecone pine are not a good proxy for temperature. Google Steve’s threads for “foxtail pine” from the last few years and there is enough there to cast three grains on salt on the pines.
In addition they say “Foxtail pine is a high elevation temperature sensitive species, etc”. WRONG!
Re: Tom P (#413), Tom, when you cite that paper please give me credit for pointing you at it. And listen to Roman.
Re: steven mosher (#430),
I asked Tom P if this was really his work or if he was working with someone else. He hasn’t replied.
Re: steven mosher (#430),
I thanked you when I first cited the paper for bringing it to my attention. Is that not sufficient? I’m a little taken aback by your apparent continual need for my gratitude.
Re: bender (#434),
All mine, as I explained earlier (#356).
Re: Tom P (#435),
thanks. i missed that. my apologies for suggestng you hadn’t replied. but please, can you write this up?
Re: Tom P (#435), Well when you do a proper write up as bender has suggested I would expect full acknowledgement of all the people who helped you. I’ve posted a link to another paper on divergence which lists a whole set of criteria for core selection. They make no mention of the type of test you suggest. They do however set an expectation that a sample must have 10 cores all the way to the present day. In particular because the divergence problem is most pronounced in the last two decades. They also set criteria for correlation with the temperature record.
Re: steven mosher (#438),
I can’t find the link on this thread. Please could you provide the reference?
Re: Tom P (#439), Sure.. Sorry I have a hard time keeping threads straight.
This comment is a snippet from the paper
http://www.climateaudit.org/?p=7320#comment-361062
The full paper ( rob wilson and esper, vose, etc ) is linked in the comment following the one above.
I would think it’s an easy matter to take the yamal series and apply their test. On core count it would appear to have an issue since the count drops to 5 at the end. Nevertheless you could apply the correlation test ( pick the right temps ) and then I suppose one could ask the question. which series do you add to yamal to increase the replication? I need to reread the paper, just scanned it, but it might have some guidelines on combining series from different sites.
To other readers, yes this raises the issue of selecting cores that are correlated with temp and yes it raises the UHI issue and yes it raises the uniformtarian principle all topics for unthreaded I think, but bender will correct me if I am wrong.
Re: steven mosher (#441),
So here’s the reference:
Click to access Wilsonetal2007b.pdf
These paper tries to avoid the divergence problem by setting some pretty strict criteria for inclusion of any series in a temperature proxy.
As you said:
Here these relevant criteria from the paper applied to Briffa’s Yamal and Schweingruber’s Khadyta series:
That would exclude Khadyta straight off – it hasn’t got 10 series until after 1790 and stops in 1990. Yamal would indeed have to be truncated in 1990 when it falls below ten series, and hence excluded. But let’s consider the truncated series as valid over the relevant time periods and continue:
Yamal (see fig. 6 of Briffa et al., Phil. Trans. R. Soc. B 12 July 2008 vol. 363 no. 1501 2269-2282) will pass this criterion. Khadyta, with its divergence after 1960, will fail.
So on the basis of the strict temperature correlation criterion in the article you brought up, Steve can’t consider the Schweingruber’s Khadyta series for inclusion in the first place.
Re: Tom P (#445),
Has anyone checked Steve’s threads for an analysis of any of those series?
Re: Tom P (#445),
I think we would need to see you calculate that correlation with the truncated series. Further you have the whole issue of which form of age adjustment to perform. As Briffa notes the selection of chronology methods comes after an EPS calculation.. if I read him right. Plus last I looked at the temperature series it was unclear ( upon visual inspection) that the truncated series would pass the test.. so, I’d suggest a some math to back up the claims. One issue ( which you havent addressed because I just gave you the links ) is the mimimum number of trees ( not cores) required to get you a signal.
Reductio ad absursum. If a tree were a thermometer there would be maximum amplitude in signal and therefore just one thermometer is enough. That’s why I have only one thermometer at home. I don’t need 15.
tom P #416,
‘…. to apply the results to trees half the age, just divide all the relevant times by half.’
I am not sure I follow you here.
Re 412 432
Exactly, FWIW.
Could someone R-proficient extract the error bars based simply on the SD of the samples and do a plot?
I’ve done a bit using Mk I eyeball and Excel on the plots at http://www.climateaudit.org/?p=7241
But I’m pretty bad at this kind of thing
Cheers
here Tom.
http://books.google.com/books?id=yJBOd4WNxqUC&pg=PA81&lpg=PA81&dq=briffa+and+jones+1990+expressed+population&source=bl&ots=LIISQ3JzI6&sig=SntzicpcqlQJxstfj_oxBcL3JaU&hl=en&ei=6ozTSryyPJKgsgPsxJjwCw&sa=X&oi=book_result&ct=result&resnum=1&ved=0CAwQ6AEwAA#v=onepage&q=briffa%20and%20jones%201990%20expressed%20population&f=false
And for an application of EPS
Click to access TRRvol59_2_53-62.pdf
Briffa, K. and P. D. Jones. 1990. Basic chronology statistics and assessment. In Cook, E. R. and L. A. Kairiukstis, editors. eds. Methods of Dendrochronology Boston Kluwer Academic Publishers. 197–152.
Click to access Briffa2001.pdf
some more references on EPS.
http://books.google.com/books?id=zr8Ucld6FYcC&pg=PA146&dq=expressed+population+signal&ei=mZPTSvfrO5WGlATPsfGUDg#v=onepage&q=expressed%20population%20signal&f=false
page 153 begins the data analysis approach.
http://books.google.com/books?id=zr8Ucld6FYcC&lpg=PA146&dq=expressed%20population%20signal&pg=PA153#v=onepage&q=expressed%20population%20signal&f=false
and like Briffa says… in conclusion. never use any computer program as a black box. Having said this in print one wonders if he releases his source code so that others
a. don’t have to use code as a black box
b. dont have to try to emulate it
http://books.google.com/books?id=zr8Ucld6FYcC&lpg=PA146&dq=expressed%20population%20signal&pg=PA161#v=onepage&q=expressed%20population%20signal&f=false
Tom P,
That is the Texas sharpshooter fallacy in full dress.
You exclude a posteriori trees that in the current temperature series don’t match, so you increase fidelity of the calibration dataset. There is no way to do this filtering for “matching” trees in the fossil set, so the fossil set will always have a lower fidelity. Due to larger noise in the signal.
So a posteriori selection creates a hockeystick. You must include all available modern samples because you also include all available fossil samples.
It’s that simple.
Re: Hans Erren (#447),
Not true at all. Failure to meet the criteria, the two above plus three others, causes rejection of the entire series for consideration by the authors, not individual samples. Please read the paper.
Re: Tom P (#448),
If I agree to “read the paper”, will you respond to my questions in 432? Or just continue to ignore them?
Re: Tom P (#448),
But by truncating the series, you’re performing another type of Texas Sharpshooter fallacy.
At the extreme, you’re essentially saying “if a (significant) portion of a series fits the criteria, I’ll use that portion.”
I don’t buy it. A digital thermometer that randomly goes out of spec is hardly trustworthy.
Re: Tom P (#448),
Not true at all. Failure to meet the criteria, the two above plus three others, causes rejection of the entire series for consideration by the authors, not individual samples. Please read the paper.
How can you tell the fossil trees don’t meet these same criteria? The fossil trees are not selected, then why would you do it with living ones? That’s applying a selection on modern trees which you impossible can perform on fossil trees. You can’t explain why one set of living larches would be a better thermometer than another set of fossil larches in the same river valley.
The underlying assumption is that the fossil larches respond identically as the living trees on the same site, yet we are looking at a floodplain where the river has meandered frequently affecting the sites, so the sites themselves are far from static, which would be a requirement for a climate proxy site.
The mere fact that recent divergence exists on such a short distance shows that there is a big problem with the fossil record in the same area.
Steve, do you have a “grass plot” available? That would help tremendously showing my point.
Re: Hans Erren (#473), ya hans when I looked at the satallite view of the site I wondered about how the river meandering would efect tress/moisture etc. seems unstable wrt treeline.
I skimmed the paper and have a couple of “dumb” questions:
1) Series vs Chronology and selection. Did they pick the series that matched best, or accept/reject sites and/or chronologies as a whole?
At the very least, I saw one site (from Wilson, who occasionally comments here) where they used the MXD data but rejected the RW data.
2) Is this simply a groundwork paper for what they hope will eventually become a priori species/site-selection criteria?
The fundamental issue I have with this paper is that it uses a posteriori selection criteria. As has been frequently stated by the statisticians here (and elsewhere, and again by Hans above), you can’t pick and choose among your data after the fact.
We’ve seen the exasperating claim that sharpshooter data selection is a “unique advantage” of TR research (according to Esper IIRC). Sooo frustrating. I wish this were made more public…other scientists might have a bit of an opinion about it.
CAUTION: I’m hoping the stats experts can weigh in on the following:
In hopes of improving my own understanding… it seems to me there’s a huge difference between selection criteria based on metadata vs criteria based on data values. It’s fine to select data sets based on metadata info, as long as the metadata doesn’t reveal the data.
In other words:
Potentially reasonable criteria, even after the fact:
– What species, what location, how well defined
– How many cores, what proportion of cores sampled well, is it stripbark or wholebark
– How many years are covered by series, chronologies, site
– Proportion of collected cores dated [if we are cautious about “the system” then it behooves to find out why other data was excluded — there are good and bad reasons]
Potentially UNreasonable criteria because they are data-based:
– Correlation to temp, autocorrelation, etc.
Sneaky criteria, actually unreasonable because they can infer bias,
– Who collected
– Acceptance based on this chronology or related chronologies being published or referenced
I’m suggesting any data inclusion/exclusion criteria is suspect if in any way the criteria are data-dependent.
So now (sharp)shoot this down 🙂
Just looked at Delayed Oscillator – this from comments
jmsully October 10, 2009 at 9:28 pm
delayedoscillator
October 12, 2009 at 10:49 pm
Re: AlanB (#462),
This says absolutely nothing. It conflates the issue of an alternative reconstruction vs. comparing Yamal and Schweingruber which ends in 1990.
Tom P. I’m still waiting on that calculation.. taps foot. I’ve been kind enough to research some criteria for you. Can you please calculate the EPS on the cores in question? and can you calculate the correlation. Thanks.
Tom. I’m growing impatient here. Can you please support your claim of correlation with a program and some charts. Thanks
Speaking of the age of treemometers, it’s interesting to note the differences in warming between old thermometers
( those that have been around a long time) and young thermometers.
Tom Since you want to talk about signal strength, try this to make your case
Click to access Den_2001.pdf
Re Tom P #468, Steve Mosher #470,
The paper in question appears to be Esper, Frank, Wilson, Buentgen, and Treydte (2007) at http://www.st-andrews.ac.uk/~rjsw/all%20pdfs/Esperetal2007.pdf.
They are talking about 28 sites, not 28 trees. Many of the sites had too few trees to work with, so they merged them into two groups by altitude. There were 778 trees in all, most or all of which appear to have been alive in 1990 (or time of sampling, which sometimes was earlier), to judge from the paper’s figure 2.
Re: Hu McCulloch (#472), sorry I was wrong. Tom P can you provide the right link?
or were you wrong?
Re: Hu McCulloch (#472), Hu
They also dropped one site because it had but 8 trees. I need to read the article again because they seem to rather cavalierly aggregate the 28 sites into two sub-sets which they then use for their analysis. For example they drop the site with 8 trees but not the site with 12 which may be justifiable but I did not see a rationale.
Re: bernie (#477),
If you look at the coordinates of Esper’s sites, they are grouped entirely differently than Briffa’s “Yamal”. Hantemirov’s location map shows that samples are not from a single site. The location map shows samples taken over an area that is perhaps 100 km by 50 km. The five YAD living cores were probably taken from the Yadayakhodyaka R and the five POR living cores from the Porzayakha R, a long way from the three 1963 Shiyatov cores (the Schweingruber Khadyta R site is closer).
Some of the Esper sites have exactly the same coordinates. If the Esper sites were grouped the same way as Hantemirov, the Kyrgyzstan sites for example would be grouped into one Yamal-scale site. The final chronology appears to use over 100 cores in 1990.
Esper the Non-Archiver is probably the very worst dendro in the entire world about archiving at ITRDB. Maybe he archives in the super-secret password protected Euro tree ring site, but I don’t even know about that. So it’s hard to say for certain how many cores are used in his 1990 overall chronology, but it would be in the hundreds, not Tom’s 10.
Re: Steve McIntyre (#480), I am still confused as to why one site with a small number of trees was dropped but not others with less than 20. They dropped Hun because “since the site was too poorly replicated (n = 8 series) for RCS.” This suggests that there is some metric used to include or exclude a site. Shouldn’t such a metric be available for all sites?
Re: bernie (#520), RCS is an neg exponential fit ( err right bender?) to a population of trees. Just thinking of the math of that makes me wonder how sensitive the fit is to early growth data scatter as opposed to late growth data scatter. I kinda struggling thinking outloud about this. I’m wondering how the data is aligned before the fit is generated and just how sensitive the fit is to varying numbers of cores at various times in the series. Maybe Hu or Roman can put math on my words. It would seem to me that if the cores are aligned at their year zero and then a neg exponential is fit to remove the biological growth/age effect, that you have a confounding problem.
For now, I’m noodling around in this which looked interesting.
http://www.bioone.org/doi/abs/10.3959/2008-6.1
Re: bernie (#477),
The Hun site Esper dropped had eight trees in total, with only four of them extending to the present. This is the site that Esper describes as “too poorly replicated (n = 8 series) for RCS.”
The other 27 sites, in comparison, are considered valid for individual RCS chronologies. These included sites that ended up with just 5 or 8 cores, although the total number was certainly larger than 10. It is this total number, rather than the number extending to present, that Esper is using for his validation of a site.
Re: Tom P (#482),
The whole discussion is about the fact that the Yamal series used in a lot of reconstructions, only comprise 10 cores (from 10 or less trees). Even if this may be borderline valid for a one site RCS chronology, this is far too small to be valid to represent the temperature trend of an area.
The minimum that Esper used for one site RCS were 12 trees. All were living trees, but as Esper said himself, to have a valid reconstruction, one need a mix of dead and living trees. That was solved by him by combining several high elevation sites and compare the result with several low elevation sites. The endresult is from 27 sites and hundreds of trees combined, not from one site with only 10 cores…
There is a direct link between number of sites (or number of trees/cores if you wish) and variability of the trends, leading to increased divergence between the two combined groups of sites:
Thus when the total number of sites/trees/cores dwindles, reproducability goes down.
Conclusion: one can’t base (the last years of) a local/regional/global trend on only 10 cores…
Re: Ferdinand Engelbeen (#483),
Just to clarify, the Yamal series has 253 cores. It’s the cores used at any one time that’s one of the issues here.
It would be surprising if this wasn’t the case! The question is the threshold below which you have enough cores to produce a valid result.
I’d certainly agree here with global trends, but I don’t think anyone is saying that. As for regional, it rather depends on the definition.
But most importantly nobody is basing the last years of any trend on tree data – we’ve got thermometers for that. It’s the pre-instrument record that is of prime importance here. And this is where there seems little doubt about validity of the statistics of the Yamal series.
If Steve really wants to invalidate the Yamal chronology, he would have to find another set of cores that also gave good correlation with the instrument record, but indicated a previous climate comparable or warmer than that seen today.
Steve’s introduction of Schweingruber’s modern, divergent, Khadyta series tells us nothing about the pre-instrument chronology of Briffa’s Yamal series.
Re: Tom P (#485),
It would have been courteous to spend a few minutes bringing yourself up to date with the past discussions on Yamal at this site before making such definite opinions here and elsewhere.
One of the ongoing questions here about the Yamal chronology was the “Yamal substitution” in which Briffa switched from Polar Urals to Yamal as a type site for this region. The Polar Urals update has elevated MWP and virtually identical correlation to instrumental records. In my AR4 comments excerpted recently, I raised this issue on a number of occasions. I would appreciate it if you familiarized yourself with previous comparisons of Polar Urals-Yamal before discussing this further. We discussed the Yamal Substitution with Rob Wilson here briefly in Feb 2006, but he gets into trouble from the Community if he talks here very much and had to sign off. The Phil Trans B information showed that the modern replication of Yamal was much less than Polar Urals, a point unknown in 2006 to any third party specialists, including Wilson and the other authors of D’Arrigo et al., as Briffa had refused to let them see his data and they used the Yamal data more or less blind – a point made to Wilson by a CA reader in Feb 2006.
Re: Steve McIntyre (#486),
Ah, so could it be said that there are two divergence issues?
1) Tree ring width as a temperature proxy diverges from thermometer temperature from late 20th century onwards.
2) Tree ring proxys that DO match the thermometer temperatures don’t match each other in the MWP.
Or is number 2 more of a regional vs global issue?
Re: Patrick M. (#489),
Divergence is not the same thing as difference. The modern divergence is a very convincing pattern where the difference grows incrementally every single year. The difference in reconstructed temperatures 1000 years ago does not arise as a result of a consistent and ever-growing year-to-year difference.
Re: bender (#490),
You’re assuming that temps and tree rings won’t converge again, right?
Re: Patrick M. (#491),
Where do I make that assumption? I am simply describing what in historical fact has led the use of this specific term.
Re: bender (#492),
My mistake. I thought by saying “consistent and ever-growing year-to-year difference” you were implying that it would continue on into the future.
Re: Patrick M. (#496),
Nope. The differences have been ever-growing through the divergence phase. Will they converge? Maybe. Would such a convergence imply the problem is past? No. It depends on what caused the divergence.
Re: Tom P (#485),
That would constitute “invalidation”, would it? Warning: you’re about to walk into a post.
Steve: 🙂
Re: Tom P (#485),
*Little* doubt? Can you quantify the uncertainty for me? Can you put some confidence intervals on your chronology, and then also on the regional temperature reconstruction derived from it? I would like you to teach yourself something about doubt. I can help. TIA.
Re: bender (#488), I’ve asked Tom P for his correlation many times, even provided him with cutoff criteria from Wilson. He seems unaable or unwilling to produce it. Perhaps somebody can supply him with the R code.
Re: Tom P (#485),
Tom P.,
The Briffa Yamal series used 12 cores from living trees in the calibration period. Hantemirov’s thesis shows that there is some difference in growth pattern of living vs. fossil trees, which makes the calibration for the whole length of the series quite suspect.
In comparison: Esper combined over 300+ (low altitude) and 450+ (high altitude) cores and even then found some discrepancies between the two combined series when the numbers got lower at the end of the past century.
Thus what makes you confident that there is “little doubt about validity of the statistics of the Yamal series”?
Re: Ferdinand Engelbeen (#512),
Where? I’m wrestling with translation (finished up to Chapter 2 of the .pdf – Mat$Meth.)and it says approx. this:
Next, to assess patterns of tree growth in height, data on the growth of 13 living and 13 sub-fossil model trees were used. There was a very high correlation between growth of tree diameter (at a height of 0.2 m) and height (correlation coefficient = 0.97). It was found that there are no differences in the patterns of the growth of modern trees and those that grew in the past.
Re: EW (#524),
EW, possibly an error in the Google translation (I don’t speak any Russian!)? Google translates the same sentences you mention as:
Maybe the “no” is missing in the translation?
Re: Ferdinand Engelbeen (#526),
При этом не было обнаружено различий в закономерностях хода роста современных деревьев и тех, что
произрастали в прошлом.
Word for word:
In that not was observed differences in patterns (of) course (of) growth (of) recent trees and those that grew in past.
Can’t change that – the NO is there and the sentence is clear even w/o the Google translator helping.
I noticed that Google translator has issues with different order of words in Russian and English sentences so that it indeed changes the context.
Hopefully Hantemirov will publish the extended chronology soon so that my Russian-English attempts will no more be necessary.
Re: EW (#539),
Thanks EW, strange that Google misses the “not” when whole sentences are used, but does translate it when individual words are fed into the translator…
But it is clear now that living and sub-fossil trees have the same pattern. Thus I have to retract my interpretation that there are troubles with the appliance of the calibration from living to sub-fossil trees.
So far, in two dendro drive-bys – by Jim Bouldin and Luminous Beauty – we have 0 endorsements of a sample size of ten. Only Tom P has suggested otherwise.
Re: bender (#476), bender, 10 may be OK under certain circumstances, but Briffa’s 10 trees have such a large variance that a much larger sample is warranted.
Re: bender (#476), i think Tom misread “Tien Shan” for ten shan and hurried off a post without double checking. has he hoisted himself again? snip
Steve McIntyre said:
Until recently, many people assumed that Briffa’s Yamal series contained more than 12 trees, too.
Until some journal follows it’s rules and forces Esper to provide his data, we’ll never know exactly how many trees/cores Esper used.
Tom P, I would suggest that rather than continue in your quest to educate yourself further in dendrochronology and statistics that you first take a reading comprehension course because you suck at it.
RE Tom P #485,
Not true — the instrumental period proxy values are in fact essential for the reconstruction, since they must be used to calibrate the proxy. If the proxy has excessive measurement error during the calibration period due to a small number of trees, the calibration uncertainty becomes larger, and this must be taken into account for a correct calibration.
Since individual tree measurements are highly non-Gaussian, being both skewed and leptokurtic (heavy-tailed) even after taking logs, it is more robust to focus on the median of the individual measurements rather than the mean. In comment #155 over on the Yamal and the Divergence Problem thread, I show that using a sample of size 10 as in Yamal 1990, a conservative 95% CI for the population median extends from the 2nd to the 9th largest value.
Rather than try to take this great uncertainty into account with a complicated Errors in Variables estimator like Total Least Squares, most dendros (according to Steve) simply go for a much larger sample size than Yamal’s before drawing conclusions. With a sample size of n = 100, for example, the 95% CI for the median extends only from the 40th to the 60th percentiles of the data, as I show in the indicated comment.
Re: Hu McCulloch (#493),
Thank you for saving me the trouble of having to continually point this out.
Re: Hu McCulloch (#493),
You are taking my comment that was in response to Ferdinand Engelbeen (#483) out of context. There is nothing in your first paragraph (apart from the “Not true”!) that I would disagree with.
As to your comments about the problems with the variability of tree-ring data, Bunn’s paper did try to take this into account with the broad range (#451) of parameters they fed into their simulation which involved modelling full RCS chronologies. The aim of RCS is to reduce the effect of the variability seen in the raw data on the chronology. Hence I’m not sure of the relevance of your comments based just on the tree-ring variability.
Re: Steve McIntyre (#486),
It looks like Wilson’s problem with his own Ural series compared to the Briffa’s Yamal series was the increased variance in the Ural series before 1100. Is this large variance just at the MWP still an issue?
Re: Tom P (#500),
How can you “agree” with Hu when you said that the pre-instrumental period is of prime importance and Hu says it’s the calibration period? You are dodging the fact that you don’t understand how hockey sticks are generated – by inflated calibration statistics during the instrumental period. You still don’t understand this, do you? Admit it!
Re: bender (#501),
Pre-instrument period prime, calibration during instrumental period essential. What exactly is your problem?
Re: Tom P (#505),
My problem is that Hu contradicts you and then you turn around say you agree without clarifying what exactly you agree with. Thank you for clarifying that you were just splitting hairs over what you think is important. If one is “primarily” interested in the modern vs MWP comparison, then the calibration period is of “prime” importance. This was Hu’s point and we’re all glad you agree.
Re: Tom P (#506),
A day without a Tom P report is a day without warm sunshine. 😦
Tom P, where’s your latest? You get that thing rolled up into a proper argument yet? 🙂
Re: bender (#538),
Maybe he’s actually fixing his R code?
That could take awhile.
Re: MikeN (#541),
He is one of the few that actually do the homework and I applaud him for that.
I also note that it takes guts and a thick skin to learn a new language in such a public environment and he deserves some measure of respect for his efforts.
Re: bender (#538),
While the organ grinder refuses to admit his instrument is badly out of tune (#80), there seems little point in staying around just to listen to the squealing of his monkeys.
Re: Tom P (#545),
There was a guy here named Dano used to toss around links like that. What are you saying? Can you write this up as, say, a grade 6 level science report: questions, materials, methods, observations, conclusions?
Re: Tom P (#545), So much for my hope that you figured out your R code was flawed. You really should try to learn R to the point where you can redo the RCS chronology.
In fact, if I get around to learning R, that is one of the things I might do to test my abilities.
Re: Hu McCulloch (#493),
Your essential points are worth noting. As a matter of terminology, however, it is platykurtic distributions that are fat-tailed.
Re: sky (#508),
Not.
Re: sky (#508), dude. platy and lepto kurtic are two of my favorite words. I’m pretty confident that Hu got it right. I prefer skinny tails to fat tails. ahem. cough cough
I’m beginning to think that Tom P is a sort of double agent who has assembled this massive corpus of comments as an embarrassment to the pro-Briffa cause.
Re: Tom C (#497),
Like I said, Tom P is my little finger.
Re: Tom C (#498), hehehehehe
Re: steven mosher (#513), now that’s a really nice way to learn stats, thanks 🙂 Homoskedatic too!
Some questions for the new guru, Jim Bouldin:
.
Are you saying that the selection for “sensitive” (wider tree-ring variance) tree series would be unnecessary at a “regular” site like Schweingruber Khadyta River? Or to put it differently, should the utility of such a site for paleoclimatology be evaluated on an “all or nothing” basis?
.
As far as I know this site has not been used in a published study even by Schweingruber himself (at least I can’t find one). Is this likely because of the apparent “divergence problem” at this site, or some other reason? (I guess that calls for speculation, and I probably should ask Schweingruber that question … still I’d like to know your thoughts).
.
Assuming the Schweingruber site were considered useful, is there a “proper” way to include Schweingruber into the H&S Yamal network (e.g. weighting), or are the incompatibilities between H&S approach and Schweingruber too great?
Bouldin’s dodge #1: bender (#367),
Bouldin’s dodge #2 Jim Bouldin (#156),
I shall continue the tally.
Deep Climate:
I note that Bouldin has yet to endorse the use of ten modern samples to support a regional-scale temperature calibration and reconstruction. That silence continues. Contrary to what Deep Climate suggests. These drive-bys have been notably void of content.
Tom, I’m still waiting for the charts on the correlation. You’ve had plenty of time. What’s the delay? taps foot.
386. Michael Smith
Sharpsburg, GA
October 14th, 2009 3:26 pm
Here’s one possibility: luminous beauty fears that he will be unfairly snipped:
Are his fears baseless? I have challenged luminous beauty to debate me here, but he has thus far refused.
.
I pointed out that getting snipped at CA would be pointless – and this is one of Revkin’s points – because anybody is free to double-post at another blog, thus revealing the censorship.
.
Crickets.
Re: bender (#510),
Bender, I agree — I see no rational basis to fear that one would be “unfairly snipped” at Climate Audit, not given the repeated postings of Tom P., Phil.:, thefordprefect, Bouldin, and any number of other critics who come here and have their say.
Re: Michael Smith (#513),
I have been watching here for years. Look at what Steve M allowed Martin Juckes to say ad nauseum without snipping a thing. He had every opportunity to debunk Steve’s work and dodged every single issue Steve raised. To me, this is exemplifies Steve’s editorial tolerance. These fears that people will be snipped are baseless. Avoid policy and religion and impugning motive and there is no risk of being snipped.
Re: bender (#514),
If you get snipped at CA, go trumpet it at RC.
Re: bender (#510),
And thus most likely revealing that said posts were in fact ad hom or one of the things not allowed here. In the unlikely case there was actually a serious point being supressed, being snipped would be a plus, not a minus. Of course in such a case, the valid point would have to be addressed here, so there’d be no point in Steve doing it. But first there has to be a valid point.
What, just because I said I was here to make sure his funding got pulled?
I have finally gotten around to doing a sensitivity test for the Yamal tree ring ages in the form that I preferred. I did this by removing tree ring ages, not individual trees by age as Tom P did, in increments of tree ring ages over 75, 100, 125 and 150 years. In order to do make this work with the Steve M’s RCS chronology algorithm, I had to reset the ages in each group by off setting the tree ring ages by 74, 99, 124 and 149 years respectively. The algorithm evidently does not work without an age 1 starting point.
.
I have posted links to the graphs of the modified RCS series below with each including the archived Briffa chronology for reference. Additionally I have posted links to the tree ring counts per year over the series for all ages, over 74, over 99, over 124, and over 149.
.
What one sees as obvious is that as one goes to older tree ring ages the HS appearance of the series with all ages begins to erode and finally vanishes all together. At the same time one sees that coverage for some years becomes rather thin as the younger tree rings are excluded from the series. However, with the trending, I would judge that this sensitivity test is indicating what Craig Loehle’s excerpt at: Post #202 above
http://www.climateaudit.org/?p=7241#comments in that larches are better indicators of climate as they become older. I do not consider the responses in the series with the older tree rings ages noise as Tom P has indicated.
.
Since I have not studied the Steve M RSC chronology method, I am not certain whether the modification I used is legitimate. I think it is, since the higher frequency responses appear to be similar to the Briffa RCS chronology. I borrowed much code from Steve M and list below the code I used for the graphs. I also looked at the tree ring widths time series before and after putting it through the Steve M RCS chronology and found (not shown in this post) that the RCS methods greatly change the shape of the tree ring responses. That might be fodder for later discussion of that method and perhaps alternative methods.
.
What ever I accomplished by the analysis reported here and whether it is legitimate, I found by going through the process I learned a great deal more about the RCS process and R coding than by simply reading the post introductions and responses.
.
Code for tree ring age graphs:
.
Code for tree ring counts by year graphs:
.
RE Sky #509,
This is perhaps getting a little OT for a thread on “Gavin’s Guru”, but just for the record here is Student’s [W.S. Gosset] mnemonic for platy- versus leptokurtosis, published in 1927 in Biometrica (pp. 151-164):
Student’s “β2” is what is known today as the (gross) index of kurtosis (i.e. not the “net kurtosis”, from which the “mesokurtic” value of 3 has been subtracted).
RE Mosh #513, It is, admittedly, more politically correct today to refer to “fat tails” as “heavy tails.” See, e.g. A Practical Guide to Heavy Tails, edited by Adler, Feldman and Taqqu, to which I contributed 2 chapters on the fat-tailed stable distributions.
Re: Hu McCulloch (#525),
Mnemonics are not the issue. “Heavy” or “fat” can be taken to mean appearance, rather than the area under the pdf at high sigmas. The perils of casual language! But funny things are happening with postings here; where is your original (now-invisible) #493 post?
Re: sky (#531),
“Fat” tails mean long tails. If they’re stubby, then by definition there’s no tip of the tail to be “fat”. It may sound counter-intuitive, but it’s the accepted jargon. And it goes to show you why jargon is a loathsome thing.
Re: bender (#533),
I agree that jargon is a loathsome thing. The jargon that many of us might agree upon as being least ambiguous for leptokurtic distributions is “long-tailed.” “Fat” associates too readily with “stubby.”
RE Tom P, #501,
Wilson’s Polar Urals graph appears in Steve’s Feb. 22, 2006 post, “Wilson on Yamal Substitution”:
The series is admittedly a lot noiser, as well as higher, back before the LIA. Perhaps the warmer conditions made these trees more prone to irregular growth spurts.
The series could be made a lot more homoskedatic (equal in variance) if logs were taken. However, this is RCS output, rather than raw TR widths. Does Steve or anyone know whether logs would already been taken here? Or if not, why one wouldn’t just take this obvious measure before applying RCS?
Re: Hu McCulloch (#527),
I’ve been looking at the data over the past week. There are very good theoretical reasons for taking logs before doing all of the statistical calculations.
As it is right now, they are implicitly treating the model as multiplicative, not additive (because they are dividing by the fitted values rather than subtracting them). However, the non-linear least squares fit is done additively, thereby emphasizing extreme values a lot more. As well, the calculation of the chronology is also done additively when averaging the adjusted ring widths.
In my opinion, all these things need to be done after the log transformation.
I have been working on an implementation of this within Steve’s RCS program structure, but I can’t get it to converge when the ring widths have not been divided by 1000 to convert them to millimetres. I’m still working on it.
As it is, there also appears to be an undesirable effect on the data due to the one-size-fits-all growth adjustment. Trees under 100 years lifetime and trees longer than 200 years seem to be poorly fitted within the results that Steve derived in the chron.yamal fit. This suggests that there are differences in the growth patterns of proxies with different lifetimes within the briffa dataset.
Re: RomanM (#528),
RomanM, my analysis presented in the last post at the thread linked below shows evidence – if my analysis is legitimate – that tree ring age has a large affect on the shape of the TR width response over the time series of Yamal. I have also noted in analyses not posted that putting the raw TR widths through the RCS chronology has very profound effects on the shape of the time series and that the effects vary with the TR ages selected.
I hope that you continue with your analysis and report the results here.
http://www.climateaudit.org/?p=7241#comment-361525
Re: RomanM (#528),
I’m glad you say this. It’s what I was going to say but you happily pre-empted me. For biological growth processes you often expect a variance proportional to the mean, and if log or sqrt stabilizes that variance-mean relationship, it should be used.
Re: bender (#534),
The simple question to ask is whether variation is a plus or minus of a fixed amount or whether it is percentage of a given value. In the latter case, it implies that the variability is proportional to the mean. A log transformation usually produces a homoscedastic linear situation which is easier to analyze in a reasonable fashion.
The initial uses of the RCS methoology were in what I term the BC era (Before Computers became as useful as they currently are for doing statistics). I suspect that it governed the choices which were made in analyses to some extent.
OT, bender, how many of you are there? You are prolific! 🙂
Late – I’m outta here.
Re: RomanM (#536),
Dendroclimatology was a poor man’s science until AGW came along. The old methods and software were developed by guys working out of a hole in the wall. The science is experiencing growing pains.
RomanM – sorry if this is too simple but this caught my eye:
Are the original data in m or m^-6? Are you dividing or multiplying in the right direction? I say this because I quickly read one of the dendro papers recently and reference was given to ring measurements being taken to an accuracy of 0.001mm ie m^-6. I found this pretty astounding and suspected they meant 0.001m but did not bottom it out. Elsewhere in the paper there is a figure which shows mean ring widths of less than 1mm – again this seems way off but maybe I don’t understand the definition of ring width? I’m sure you will have this covered but thought it worth posting – apologies offered in advance!
Re: curious (#529),
The Briffa data is in multiples of 10 from 0 to the multiple 1000s. It is my impression that to turn it into mm’s you need to divide by 1000 *or in some case by 100) since the data is typically given in .01 and .001 mm. formats. The problem is that with 40800+ observations in the nls procedure, the SS’s are of the order of 10^10 and I believe that this may calculationally have an impact on the convergence. Reducing it to mm’s does allow convergence, but I am still checking the results before I post anything.
Re: Kenneth Fritsch (#530),
Yes, you are very right that the series is dependent in a big way on how the adjustments are made on the various trees.
I have been looking at the distribution of the treerings after adjustment for growth and also adjustment for “climate” as measured by the chronology. I was hoping to possibly write up a short post on this if our host does not object.
… suggesting Tom P’s protest about unsuitably high variance during MWP is moot. (And: did they really reject this because of the variance inhomogeneity issue? Or is that a convenient cover for rejecting something that just won’t generate hockey sticks?)
Speaking of the divergance problem (earlier), Jim Bouldin weighed in with his professional opinion on the subject from RC…
This kind of logic is exasperating.
Gavin’s 3rd and 4th gurus are “Deep Climate” and “Delayed Oscillator”:
Check this post from realclimate:
So:
1. DO’s proclamation that one reconstruction or another is more accurate or precise is based on unspecified criteria.
2. There is no instrumental record of the past, so no such validation is possible, yet that seems to have escaped DC.
3. DO is a hero for his auditing job, but Steve M is a villain.
I think DC and DO need to go back to dendro school and try again. Maybe LB can lecture from his high horse.
Hey, DC, when you need help with the “corridor method” maybe mosher and me can help you out.
Can’t wait for #5.
Bender, I tried to help Deep along in his understanding, but I don’t think he ever totally figured out that there was nothing in this post. He did figure out that he put up a wrong comparison chart. He put up Steve’s Red and black, when DO was graphing the green in-between work.
With my amateur eyes, I’ve now looked carefully at essential details of what TomP is saying here at #80, and Tom’s stuff don’t hang together. But let’s see it all juxtaposed to make this clear. Now Tom claims to show the chronologies of old trees (>75 yrs old) and young trees, (<75 yrs old)
then claims to show the difference between them,
and then claims that this shows that recent divergence is totally non-exceptional. But hey, make both graphs visible together, and it’s clear in the first graph that the recent divergence is about twice each of the three earlier largest divergences in around 700, 1400, and 1700 CE. That’s an order of magnitude in my books.
So it seems that Tom’s second graph, the comparison, is NOT the comparison he claims. Is he just trying to get away with sleight of hand?
Or have I idiotically missed something? In which case, apologies in advance, Tom. Please explain in layman’s terms.
Re: Lucy Skywalker (#550),
The top graph is smoothed and the bottom one is not.
Re: romanm (#554), thanks, I did wonder… but surely this is still a visual sleight-of-hand because the bottom graph should then also be smoothed and would then show the same patterns discernible in the first. Which would surely be… that the recent divergence IS far higher than earlier.
Re: Lucy Skywalker (#556),
You’d think that they would both be done the same way, but hey, who did them? 😉
Re: Lucy Skywalker (#554),
I am going to ask a silly question. If trees are diverging today, why is it surprising that they diverged in the past? And why do some people seem to claim that this supports the effort to derive temperature reconstructions?
It would seem to do just the opposite. Trees diverge for some unknown reason or reasons. The relationship of trees to external factors such as temperature, precipitation etc is not understood. This is especially true if these factors interact
So if trees diverged in the past and are diverging today, for reasons that are not understood, then how is this accounted for in temperature reconstructions?
Re: TAG (#560),
Here is the issue with divergence bias over time. Suppose divergence today implies divergence 1000 years ago. If you accept only the positive divergers today (what Briffa did and what Wilson half advocates), and these did not diverge positively 1000 years ago, you will bias toward generating a hockey stick reconstruction. Thus you need to know the casue of divergence. You need to know if it is a tree characteristic, a site characterisitc, an interaction, or something flakier and more random.
oh, add 400 and 450 CE.
I’m still waiting for Tom to do a simple corelation with the local temp for Yamal. Gosh, I wonder why he hasn’t.
1. he dont like taking orders from me. Last time he did he wrote some bad R
2. he’s ignoring me. sniff sniff
3. he did the correlation and it doesnt measure up.
4. it’s being peer reviewed.
anyways, 10 second tom seems to have forgotten how annoyed he got when steve Mc would not jump to and post a graph for him the minute her requested it
Re: steven mosher (#556),
Steven M, please give it up. I think way too much bandwidth is wasted on the likes of a Tom P, a Lorax and even a Nick Stokes. If those diversions inhibit or detract from some real analysis from the likes of a RomanM, I say we need to rethink the usefulness of some of these exchanges.
Re: Kenneth Fritsch (#558),
You aren’t interested in seeing Tom P package up his wares into something intelligible?
Re: bender (#559),
I think I know what he has done and I suspect others here do also. It is his apparently disconnected conclusions that we do not understand. What he will do is merely repeat his calcultions and conclusions. Who needs that. The attention he is paid here gives him recognition way beyond what he deserves – based on any contributions he has made to the knowledge base here.
Re: Kenneth Fritsch (#569),
Agreed.
.
By the way, you should look at Esper et al (2002) Fig 3 “Low frequency signals in long tree-ring chronologies for reconstructing past temperature variability”. They plot confidence intervals on their RC chronology. And, man, are they wide.
Re: bender (#570),
In that paper:
Schweingruber was a co-author on that paper. Briffa was not.
Re: bender (#571),
I leave the reading of p. 2252, centre column, for homework. Well worth it.
Where is Tom P at in this post-climategate world of ours? I can’t believe I actually miss him. The good ol’ days.
TAG, are you asking me or is it a general question? Have a look at my post that’s up at Jeff Id’s on the Dirty Dozen. What became clear is that we have twelve highly anarchistic trees, no real longterm patterns deducible at all – but they do spike in concert for some individual years.
Seems that their short-term correlations makes them work for dendroCHRONOLOGY, but that does not extrapolate to dendroCLIMATOLOGY.
Re: Lucy Skywalker (#561),
I read TomP postings and am at a loss to understand them. Some of them seem to arguments against his own position but he announces his results with great confidence.
Re: TAG (#562), 🙂
To Jim Bouldin
Not sure if you are reading this site. Gavin at RC has started snipping my questions to you. I was trying to understand what kind of spatial selection criterion H&S might have had & why the paper didn’t mention it.
I’m not going to play games with Gavin, so this is me signing off.
I’m sure you will be very relieved, you can get back to doing dendro/drinking coffee now!
I do thank you for your patience. I was genuinely trying to learn and was getting ratty purely because you were talking about the importance of a spatial selection criterion, and I was reading and re-reading H&S (2002) without finding any mention of it.
Thanks very much for your time, sorry to waste 1000’s of your words.
Mark P
Re: Mark P (#565),
Are you getting your answers? It seems they are (or were) more interested in dismissing you than answering you.
Mark P
email him. It’s not like he’s hard to find with a web search
Some gratuitous slagging of McIntyre by dhogaza:
Interesting how he’s never been able to point to any errors in Steve’s work, but slags him anyways.
And I notice he doesn’t apologize to Mark P for his ugly demonizing.
I think dhogaza is just upset that he’s not gavin’s guru.
Re: bender (#568), dhog is a pretty vile person. I’m on file as vile, I should know.
Re: bender (#568),
Does anyone know how H&S (2002) selected their dead trees? Jim Bouldin tells me that there has to be a well defined spatial criterion in addition to the (contentious) age & variability ones.
Without a spatial separation criterion, the whole thing falls over.
CAN ANYONE FIND IT IN THE FREAKIN’ PAPER? I’ve been over and over it. I don’t even see how you would define and test a such a criterion:- in death the trees can be a long way from where they grew. I’m REALLY interested to read it.
That’s why I got so hung up on selection criteria at RC. Now Gavin’s started snipping me so I’m not playing any more.
Re: Mark P (#574),
You are asking the right questions, Mark P. Jim “guru” Bouldin talked, for example, about someone sampling from dry vs. moist microsites and stratifying samples accordingly. First, there is no evidence this was done in this particular case. Second, if such stratification was (and is ever) done, rarely are there objective indicators of microclimate recorded at the same time; they are usually subjective indices of microsite type. Third, beyond such stratification there is little or no guidance at all to how samples are taken in 2-D space. One suspects a “haphazard” scheme that would be replicable only to the extent that an investigator could replicate a random-but-learning-walk = hunt for long-sequence samples.
.
In short, the recipe you are searching for is probably absent from any dendro paper you choose. The GAP is the “haphazard hunt”. Show me one exception and I’ll buy you lunch. Your RC gurus dance around your point because they fear you may be on to something. They cover up even when they don’t need to. It looks bad. It feels bad.
Re: bender (#575),
Yep, I’d come to the same conclusion. Just one sentence in H&S describes the entire selection methodology on which the whole tower of cards depends.
When I read a paper like that (I’m in engineering now) I always look for very short sentences, cos that’s the weakest bit that they want to rush over. Then you look for arbitrary numbers like 224 samples – that’s clearly a POMA (Pulled Out of My A**). Fun!
You’re a smart guy, Bender. On another thread you said that maybe the reason dendros don’t use error bars is because they have come from the background of building a chronology where statistics and confidence limits aren’t important. I think that’s a great insight (at least, I’d come to the same insight after going round, and round, and round the difference in methodology between chronology and climate reconstruction, see my 1000’s of words at RC).
Cheers, Mark
Re: bender (#575),
It might (or might not) amuse you, one of the posts Gavin snipped was the damning evidence Nature presented after it had audited Benveniste’s work on “homeopathy”:
– Benveniste’s experiments were “statistically ill-controlled”, and the lab displayed unfamiliarity with the concept of sampling error. The method of taking control values was not reliable, and “no substantial effort has been made to exclude systematic error, including observer bias
– “interpretation has been clouded by the exclusion of measurements in conflict with the claim “.
(my emphasis)
Sound familiar? I was trying to make a very serious point to Jim B that even the best scientists can get their statistical knickers in a twist. But for some reason Gavin didn’t like it….
Keep up the good work.
Re: Mark P (#577),
Forest mensurationists wil have areal photos, with clusters of tees defined by size, species, topogrpahy, and will have an a priori plan how many stems to sample per strata, how much uncertainty is tolerable, and so on. In contrast dendros will walk around looking for old trees that yield long sequences. They spend as much time searching as sampling. It is definitely more art than science. There are no calibrated growth responses to help guide the search for those “sweet spot” microsites. (Naturally. It’s hard to step up controlled experiments with 1000-year old trees.)
.
I couldn’t tell you what Hantemirov did because I’d be guessing. And maybe Jim Bouldin has talked to the guy and has insider knowledge. But odds are that he did a “haphazard long-sequence hunt”, i.e. ill-defined, and Bouldin doesn’t really know how, or care, to communicate this to a lay audience. Does this mean Hantemirov’s methods are unreplicable? No. Because of the presumably high degree of correlatedness among samples just about any sampling scheme would probably give you Hantemirov’s curves. The landscape I saw portrayed in pictures didn’t suggest there was a ton of microsite variation that could act as a systematic and predictable source of bias.
.
To really answer your question would require interviewing Hantemirov at an AGU poster session. Why the gurus are hesitant to tell you these facts is anyone’s guess. Probably just too busy to care.
3 Trackbacks
[…] er som är intresserade av att följa bråket om Yamal har Steve McIntyre ett underhållande inlägg på sin blogg. Dessvärre drar sig inte Steve för att förlöjliga sina motståndare, vilket antagligen roar […]
[…] (often begrudging) from many professionals, you need look no further than his latest post on the Yamal controversy. Some people won’t like his tone and others won’t like how his work is used and spun in […]
[…] P. does not meet RealClimate’s Bona Fides Requirements I made this comment at Climate Audit. I’ll repeat it here: The degree of inconsistency that RC exhibits becomes […]