Obviously there’s been a lot of discussion in the last few days about the difference between the CRU 12 and the Schweingruber 34. In making such comparisons, it’s always a good idea to look at the data in detail – something that obviously should have been done by Briffa and the Team before the widespread use of the Yamal proxy in so many reconstructions, rather than this late date, over 9 years since its original use in Briffa 2000.
In a previous thread, I showed a plot of the actual ring widths of the 10 CRU trees ending in 1990. Today I’m going to show a similar plot of the “dimensionless index” for the same 10 trees. It is the “dimensionless index” that is averaged to make the “chronology”.
Recall that in RCS, a “standard” is established for the decline in ring width with age – the decline is assumed to be a negative exponential curve plus a constant (“generalized negative exponential”) and the index is the observed ring width divided by the “age standard” ring width for the age of the given tree in that year.
For comparison, I’m going to do a similar plot for 18 Schweingruber trees (17 sampled in 1990 plus one). The plots are shown on a uniform vertical scale (0,9) and a uniform horizontal scale (1850,2000). I’ve marked 1990 with a vertical red line and a horizontal line at 1 (the overall mean ratio.)
First, here is the plot for the 18 Schweingruber trees. Probably your first reaction is: why did he choose such a squished vertical scale for this graphic – we can’t see this as clearly as we’d like. Your second reaction is probably – well, if there’s a stick in there, it would take something like Mannian principal components to dig it out.
Figure 1. 18 Schweingruber russ035w trees, 17 ending in 1990. Age-adjusted index.
Next here is the corresponding plot for the CRU 10. Without doing any sort of fancy statistical test, one can readily see a difference. None of the YAD** trees on the right are especially old – the graph shows their full history – all start after AD1800. However, instead of the standard negative exponential declining growth, these particular trees started off very slowly, like old trees, and then got a burst of virility when they got to be 100 years old. Benjamin Button trees so to speak. Because of the one size fits all RCS standardization, this post-100 growth pulse is divided by a small standard denominator – YAD06 reaches 8 sigma and is the most influential tree in the world. YAD06 does not always drink beer, but when it does, it drinks Dos Equis. Stay thirsty, my friends.
Figure 2. 10 CRU trees ending in 1990. Age-adjusted index.
UPDATE: Oct 1. Tom P in a comment below asserts:
These plots explain why it would be extremely difficult to extract a centennial signal from the live cores of Schweingruber series. Most are under 100 years old! 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.
I agree 100% that it “would be extremely difficult to extract a centennial signal” when “most [of the cores] are under 100 years!”. However, I disagree that the trees in the CRU archive are “much longer-lived”, other than the trees selected for the modern comparison. The following graphic shows the average of tree by year in the CRU archive, the average age in the “Schweingruber variation” in which russ035w is used instead of the CRU12 to represent living cores. Prior to around 1800, the average age of the tree in a given year was around the 100-year mark that Tom complains about. There is a profound inhomogeneity in the age composition of the living trees in the CRU archive relative to the subfossil archive, which is much reduced in the Schweingruber Variation. Does the age inhomogeneity in the CRU version “matter”? It’s the sort of thing that should have been reported and discussed in a site report, prior to using this chronology in multiproxy studies.
Climate Audit 
314 Comments
So one jolly lumberjack could have changed the Earth’s climate.
a mechanism for a growth spurt like that was described in various material and signs in Muir Woods (a redwood forest): a large old tree falls, creating a hole in the canopy finally providing direct sunlight to older trees which have lurked in the shade, growing slowly for decades.
was there logging in this region starting in 1920?
What do the plots look like for the POR* and YAD* cores that they didn’t select?
My training in forestry allows me to explain Briffa’s treemometers: it often happens that a tree starts growth slowly because it is in the shade of other trees or crowded by neighbors. If it survives this and the overstory trees die, or it emerges from competition with same size trees, it may begin rapid growth even after 100 years (true for many but not all species–some never recover if once they are suppressed). It is a wonder what just looking at the data will reveal!!
One way around the RCS nonsense is to plot basal area increment instead of ring width. For many species, once they become adult trees they grow with a constant basal area increment (plus or minus annual climate effects), so no standardization is needed between trees. I’m surprised no one uses this.
Re: Craig Loehle (#4),
In high latitude “forests” at the treeline, canopy is not an issue. From this, they conclude that there is no inter-tree competition and use one size fits all standardization methods. Same with bristlecones.
With bristlecones, there was a constant competition for water and it seemed to me that there must have been competition for root-space, even if there was no competition for light. I presume that competition arises in some way in these treeline forests as the trees do seem to space themselves.
Re: Craig Loehle (#4),
Now, take a population of trees with a low initial growth pulse. Are these not the worst trees to use as climate warmering detectors? Or rather wont they INHERENTLY over estimate present warmth and underestimate past warmth?
Fantastic.
That was an interesting revelation last night when I was reviewing TomP’s code. Actually, I thought he was right at the time and was looking ever deeper into the code. I single stepped through the function to see what was happening because I couldn’t figure out if the weightings of short series were reducing the Schweingruber impact on the trend. It turned out to be more simple than that but I learned something at the same time.
When the most interesting man in the world reads tree rings he reads them in degrees Kelvin all the while drinking Dos Equis.
People used to believe trees talked…but in this case they confessed!
What have they been fertilizing YAD06 with? IPCC reports? Or is that where the climate science bodies are buried?
Are there crickets in that tree? I can hear them chirping all the way over at RC.
I think YAD061 drinks Carlsberg because its “probably the most impressive tree growth this side of Sheep Mountain”
Dos Equis, indeed. The beer brand was named “Siglo XX” (“20th century”) to commemorate the arrival of the new century, and the bottles were marked with the Roman numerals “XX”, or “Dos Equis” (two Xs). This tree has been celebrating. Can there be a better argument for collecting and reporting meta-data?
I’ve got minor experience with farming, even small shading makes a huge difference in growth rate. The side of the plant shaded makes a difference too.
“Truly, the tree yields bad fruit.” –As You Like It
What does the index look like with and without the Enchanted Larch of Yamal?
Let’s hope those cores were not inverted. That could easily explain the peculiar growth spurts ‘late in life.’
Oh well,
Funny a couple days back I had this thought about this tree, before Steve found it. In a comment made to Tom.
“Tom P does NOT get the statistics behind a reconstruction.
Tom to see your lunacy you only need to extrapolate your reasoning about selecting cores based on correlation with temperature. Briffa selected 12 cores. Cores that correlated well with temperatures. You think this is the right approach. Now consider, of these 12 cores, ONE core will have the BEST correlation. Why use any others? That one core should give the BEST estimate of the temps in the MWP according to your reasoning. Seems a bit odd, doesnt it? “
8 standard deviations? Always remove outliers, that’s what they teach any green, black or other belt. Wouldn’t have gotten that past my MBB, good to see the standards in science are that high. Good thing someone is putting daylight on this, keep it up.
Re: apex (#20),
When I suggested that this should be done I was jumped on as a statistical heretic. But of course if you remove 1 tree for being an outlier, shouldn’t you also remove trees that do not agree with the actual temperature record?
Ah well I suppose the stake has been placed and the fires already lit for a heretics end.
An interesting dissertation:
Click to access KHantemirovRM.pdf
concerning location, tree line, and tree rings from the Yamal over a few millenia.
ends up with a hockey stick in figure 18 though!
Re: thefordprefect (#24),
_what_ actual temperature record are you comparing against? The GISS mess?
Re: Scott Lurndal (#31),
From this document for the briffa yamal stuff:
Click to access 2269.full.pdf
Re: thefordprefect (#34),
Except that comparing the actual summer temperatures shown in the article (Figure 1 a,b) with the Yamal chronology (Fig 3b) it’s clear that the Yamel values after about 1975 are nothing close to the corresponding regional temperatures and shouldn’t have been used as a temperature proxy.
Re: thefordprefect (#33),
How many times does Ford Prefect need to be told that there hasn’t been any significant warming in this area? See the Gisstemp plot for Salehard. Even the Briffa et al paper he quotes confirms this – Yamal is between a blue and a yellow box in fig 1c (summer trends), and the text says the only significant trends are in winter in central siberia and in summer in the East.
None of the graphs look anything like YAD06, which seems to have a minimum in the slightly warmer 1930s and then a growth spurt from 1930-1960 during a slight cooling.
Re: thefordprefect (#24),
Unlike some others here, I appreciate your persistence, and (general) politeness. Civility sometimes gets short shrift in these discussions.
That said, confirmation bias is an ever-present hazard, especially among dendros who seem to have a professional leaning towards cherry-picking.
Not that confirmation bias is limited to dendros, as our host often cautions us.
Thanks for your contributions here.
Cheers — Pete Tillman
Re: thefordprefect (#24),
tfp, if there is population of trees with no MWP temp signal, is it not possible that such a population has uptick “noise” samples which coincide with an MWP?
An Elven ode to YAD06
One Tree to rule them all, let’s go and find him.
One Tree with wide, wide rings, we’ll core and grind him.
In the land of Yamal there’s a hockey stick.
Re: Dave Dardinger (#20), I prefer closer to the original One TreeRing to rule them all, One TreeRing to find them, One TreeRing to bring them all and in the darkness bind them
Re: Anthony Watts (#37), look here at the aerial views close-up of both Yamal and the other areas (thanks to Francis Turner). Yamal has a unique look to me, where microclimate issues could dominate in ways that have not been grasped so far. Eg “live river valley trees” is a different subset to “dead trees that fell in the water”. These Yamal tree communities have very clear boundaries between “forest” and “tundra desert” and individual trees on the edge (that are too far away from the river to fall in) could perhaps show up flukes if they transpose from one microclimate to another.
At WUWT on the AGU presentation, bill (18:57:21) and (19:53:27) has posted links to two papers by Russians (one linked here above). They look interesting and bear out that Yamal dendrochronology is potentially important. They show that a 7310-year Yamal chronology based on 54 living and 452 subfossil larches, averaging 125 rings per tree, max 501, most 60-120 rings.
These papers apparently show “unprecedented warming in the last century” on treering records that has to be explained. I haven’t time so far to wade through 170 pages of one and Russian of the other. Since this does NOT agree with the neighbouring thermometer records I flagged up at the Air Vent, what has caused the recent “unprecedented” spike that disagrees with thermometers, that Hantemirov etc report at Yamal, that doesn’t show up at Polar Urals? Bad calibration? Sheer fluke? Cherrypicking? Local microclimate issues? (this is possible in my opinion, since living trees on the banks are a different subset from subfossil trees in the river.) Bad thermometer records? (one has to consider all possibilities though I think this is the least likely option) Tunguska? The Elvin Tree? What?
Late to the party, but how about…
In the Land of Yamal where the Briffa lie,
One tree to rule them all, one tree to find them,
One tree to bring them all and in the larches bind them,
In the Land of Yamal where the Briffa lie.
Tanfascicular!
========
YAD06 certainly stands out, but the entire set is noteworthy for the consistency of the underlying pattern of warming. Compared to the Schweingruber trees where there is reasonable variance in the trends, there is very little in this set of 10 trees. Clearly there is a need to see the rest of the Yamal trees…
As they say, when it seems to be too good to be true, its probably too good to be true…
Tom P should hone his skillz by doing a sensitivuty to the inclusion/exclusion of the magic tree.
If you spend any time doing morphometric analysis of biological samples, you immediately realize the dangers of having your measurements turn on just a few outliers. If one of my students presented data where the major effect was an outcome of 1 or 2 individuals in the mix, I would be shocked if they did not spot the problem themselves.
http://www.climateaudit.org/?p=7168#comment-357531
If even a non-statistician like myself can spot this tree with a 10 seconds glance at the cores, why can’t the dendrologists?
Re: Clark (#24), Sudden growth spurt? Do wandering bears defecate in the woods? Has soil chemistry for the one “Magic Tree” been normalized for variations in nutrient levels??
It would be interesting to see the Briffa result reproduced without the Enchanted Larch.
dearieme,
You summed up fate perfectly. But I would note that for those determined to find a preferred results, there were plenty of YAD06’s out there to “fit the proper curve”. An army of lumberjacks (why is Monte Python now running through my head) could not stop a determined illusionist.
I see many series capture the mid 30s warming.
These tree rings obviously support the hypothesis that the mid thirties warming was real and the post 1979 data is UHI infected. Magic trees correct the temperature record.
Given the truly extraordinary chutzpah of the Team, it would not surprise me if Briffa re-did the analysis with only YAD06. The results would then show warming that was “much, much worse than we thought”.
mpaul,
Is it me or is the “enchanted larch” just another twist of fate which will bring us all around to the ministry of silly talk????
Python lives! (in the IPCC)
Steve #8,
If we are discussing tundra at high latitudes, of course there is competition. Trees are rare and sparse in this ecological system. What gives a crop of trees the opportunity to thrive while a loan seedling struggles? Find that and you have your answer – and it is not CO2 levels, which would produce massive forests at lower latitudes and/or altitudes found in arctic or subarctic tundra.
The “enchanted larch” is truly the best catch phrase of this entire revelation.
Clark #25,
You ask “how did they not see this”? The question should be “why did they do what they did when they did see this”? Creating science around unique exceptions is sort of an oxymoron, don’t you think?
Shouldn’t the blog title read YAD061 rather than YAD06?
I see in the graph group YAD061 only.
Sorry to appear as a nit picker.
Re: Anthony Watts (#36),
The last digit is the core # – most of the time, only one core is reported. The core is YAD061, the tree is YAD06.
Steve I know you don’t have much metadata, maybe none at all, but what strikes me here is that these CRU12 seem like a small population that had an advantage of some sorts. Could they all be close together?
In my post on Treemometers and Tree Ring Growth we see what Liebigs law defines for limiting factors. In the CRU 12 it seems like the stave for one or more limiting growth factors may have gotten raised.
It would be tremendous if we could place these 12 trees and try to determine if something changed in the environment, especially if they were all in near proximity to count as a small population. This could be something as simple as a reindeer herd dumping a lot of fertilizer nearby.
The Starbucks hypothesis makes me think that there might be a tendency to sample what one can get to easily.
Re: Anthony Watts (#38),
John’s law of dendrochronology states that “the ones most likely to be sampled are always the outliers”
Ed McMahon: Ladies and gentlemen, it is time once again for a visit from that great
seer, that great visionary, Carnac the Magnificent!
(Enter Carnac, tripping over riser)
Ed M.: O great Carnac, I have here in my hand some sealed envelopes that have
been kept in a mayonaise jar on the back porch of Funk and Wagnalls until
this very moment; and I understand that you will divine the answers to the
questions contained in each envelope without ever having seen the question!
Is that so?
Carnac: Yes, O humble announcer, that is so. You do go on, don’t you?
Ed M.: Very well, here is the first envelope, containing a question whose answer
you will divine through your miraculous powers.
Carnac: Enough already. …
Ed M.: Carnac! Carnac!
Carnac: …Wha?… Oh yes: The answer to the first question: “YAD061”
Ed M.: “YAD061”
Carnac: Yes, that’s what I said: “YAD061”
Ed M.: And the question is?
Carnac: And the question is (opening the envelope), “What is Briffa’s license
plate number?”
Oh that’s good
I liked that joke so much I made a plate for it.
Re: Anthony Watts (#43), Anthony, that is beyond good. Fantastic.
Steve, could it help to do a summary post outlining your thoughts and perhaps this might keep a lid on some of the more extreme statements being made based on your finding? I note we are hearing from 1 side and without the counter, it allows others to go a little too far.
ps well done!
Steve,
These plots explain why it would be extremely difficult to extract a centennial signal from the live cores of Schweingruber series. Most are under 100 years old!
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.
Re: Tom P (#44),
Tom you still dont get how RCS needs a large sample. Go read the dang article please.
Re: steven mosher (#47),
I read the article. The total number of trees used in the simulations ranged from 15 to 99 over two thousand years. 15 trees are apparently too few to deduce a low amplitude multicentennial signal but reliability increases with the number of trees, being very good with 99 trees:
“If the model runs containing 99 trees (117 of the 10,000 models runs) are subset, then the mean coherency between the input signals and the RCS chronology for the multi-centennial time scales is
0.90.”
The CRU Yamal archive uses 611 samples from over 2190 years, the Schweingruber series has 34 samples spanning 208 years.
Re: Tom P (#50),
Keep trying tom.
Re: steven mosher (#53)
Steve, instead of sneering, how about dealing with Tom’s very cogent point. The Schweingruber series does not extend far back beyond the period of thermometer measurements (208 years max). How much value does it add to a 2000 year chronology?
Re: Nick Stokes (#56),
If the Schweingruber tree ring series does not make for good thermometers, why is this good for proxy reconstructions from tree rings?
The thing that I get from this discussion is how little is known about tree response to temperature. It just seems like arm waving with comments about divergence, CO2 fertilization, sheep manure etc. This is in contrast to the broad and certain statements that are made by advocates on both sides of this issue.
Re: Nick Stokes (#56),
Nick, you too do some more reading. If you want to screen cores apriori using length as criteria, go ahead.
Do that for all cores and see what you come up with. It would be an interesting sensitivity.
Biffra didnt. RCS gives you better coherence with low frequency variations, which means in a centuries long core it has the ability to pick up a signal back in the MWP. If you think that this impacts its ability to pick up a decadal signal ( the last 20 years, why then go knock yourself and say this. And cite literature please ( oh I get to play at being hank)
Re: steven mosher (#102) and Re: Tony Brown (#73),
I’ve blogged my thoughts regarding Briffa’s response.
One thing you notice when you compare H&S and Briffa’s core counts is that H&S almost always have far fewer. I believe the reason for this is that H&S are only choosing verylong-lived (multi-century) trees. Briffa seems to have included some rather shorter-lived trees too.
I dunna tomRe: Tom P (#44),
Tom,
First, steve plotted 18 of the 34. Second, I believe 8 or the 18 are over 100. yes, 10 of the 18 are less than 100.
Second read steve
“Next here is the corresponding plot for the CRU 10. Without doing any sort of fancy statistical test, one can readily see a difference. None of the YAD** trees on the right are especially old – the graph shows their full history – all start after AD1800.
10 of the 10 are older than 100, one just barely.
So, how do you reckon that this is signal versus noise? Further only one of your 10 actually shows a late term
hockey stick.
Finally if you read the article on RCS you will find that RCS performs more poorly than other methods at detection of low freq signal as the sample size shrinks . the lowest limit they tested was 15.
Now, what idiot decided to use RCS on the yamal 12, 10, 5? WHO? briffa thats who. take it up with him. Before you asked for a cite on numbers and RCS. And said you would throw out the data accordingly. You have your cite. Go on.
Any fertilizer on YADO6 is not likely from a bear, polar or otherwise, but they (bears) do damage trees. Use Google and search using “hemlock tree decapitated by a marking black” – Currently the first link will discuss and show this.
Male canines, however, might regularly add a squirt of liquid fertilizer to a well-placed tree. I’ve had to add strategically placed rounds of firewood near trees and shrubs at ends of rows or corners. Location, location …
There are likely a dozen viable reasons why this tree responded as it has and we are not likely to know anything about it. Only Yado knows.
YAD06 might actually be a sensitive treemometer – Yamal is permafrost region with permafrost layers of just a few meters thickness reported for coastal marshes:
http://www.informaworld.com/smpp/content~db=all~content=a907186912
I assume such shallow layers of permafrost might well react to elevated temperatures within a decade, considerably deepening the active layer above and causing high age growth. Perhaps that’s what YAD061 reflects.
I suppose YAD06 is one of those highly sensitive, thirsty trees discussed in
Tikhonova, I.V.,
and Stolyarova, O.A., Individual sensitivity of Larix sibirica L. in open woodland of the Shira forest-steppe, Contemporary Problems of Ecology, 1(6), 2008.
Tom P:
That would be amazing if we had multi-centennial temperature measurements to calibrate with – but if you think about it, all that would tell us is that the trees found it favourable or unfavourable to grow in any given group of centuries.
Not exactly temperature measurement at all.
It seems it isn’t just one tree that is so influential but just one core of one tree, I wonder if they cored the tree from a different angle as well, I suppose we will never know…
Remember statistic people. You have a range of tree data.
I have been told you cannot cherry pick 10 trees even if they correspond best with temperature. ALL must be used. If tree YAM06 has been measured it must be included or else you are cherry picking!
Re: thefordprefect (#59),
Not quite “must be included” but rather “must be accounted for.”
This is where better metadata becomes crucial for valid scientific study. And multiple cores per tree (which according to proper dendro procedure are a requirement to fulfill the principle of replication.)
It seems that some shortcuts have been taken. Not unusual; it’s what we saw with Graybill’s work as well.
At the very least, the range of results should inform the resulting confidence intervals. This is a topic that’s been done to death here at CA. Unfortunately, it is quite rare to see temperature reconstructions with proper CI bands. A quite difficult challenge to create CI bands from first principles starting with the variance in collected site data. Yet how else can we know if the final analysis has any meaning?
Re: thefordprefect (#58),
A couple of things.
The original Schweingruber survey was a survey of 383 or so sites around the NH taken from sites believed ex ante to be temperature sensitive. This resulted in the Divergence Problem (Briffa, Schweingruber et al Nature 1998b), where the declining RW series is shown in one figure. OTher than this passim mention, the Schweingruber RW results were never published.
In that sense, the Yamal chronology is a total outlier to the Divergence Problem in a very large population – and warranted a detailed examination from the outset before being adopted as a canonical representative of Russian tree ring chronologies,
Secondly, the raw distributions even after standardization (and particularly after standardization) are hugely non-normal.
In “robust statistics” as this term is used in statistics and not as a term of self-praise as in climate science, the mean is not a “robust” statistic (nor is the standard deviation); the median is. Hempel has written a lot on such topics. YAD06’s impact on the median is far more muted than the mean.
In gold mining, you have to use all the data, but there are rules on the handling of outliers (the “nugget effect”). If you get a big outlier, you have to “cut it” to a maximum of double the average grade or something similar; otherwise a “nugget” gives a false bias to the average. You don’t exclude the high value, but you cut it to 2 sigma or something like that.
Obviously the 8 sigma of YAD06 is not the result of a response to temperature according to a linear model. Something else is going on. If you take averages while ignoring confounding factors, you also aren;t doing your statistics right.
If one part of the sample has been selected non-randomly, then you have to account for this as well. There is no “right” way of accounting for such potential bias, if you don’t know what was done. The better approach is surely not to use the site until it has been sampled in a proper way or until the full measurement data from a proper sample is available.
Re: Steve McIntyre (#67), Thanks for the response!
Proxy data is poor at best however grape harvest dates for example seem to be a reasonable proxy (limitation clipping at the extremes) but this is choosing a particular grape (pinot noir) from a particular region – I do not consider this cherry picking just sensible.
Tree rings as you suggest are gathered from all over the world so the first cherry pick is to limit the investigation to a locale. The next is to limit to a particular terrain. the next is to limit it to a particular tree species. the next is to limit to a minimum diameter trunk. I would then assume that “ill”/damaged trees are not chosen.
Thus we have already picked a reasonable sized basket of cherries.
The Russian thesis I referenced also has other growth events noted – late frosts etc where tree rings may be distorted. This is presumably removed by time averaging.
That would be a pretty full basket of cherries by now.
Let us assume that we have a 0.1% accurate temp record for the area of your sampling and for the last 10% of the period you are investigating. You suggest that to pick a couple more cherries by discarding rings that show no correlation with this temperature is going to be invalid. If they do not match the temperature record now why would they add anything but noise? If they match the record then at least you know that the tree, the location, the soil, the water allows the rings to represent temperature over a short period.
To cherry pick a sampling area and then to randomly choose trees (dying, wind blown bog footed, etc) seems odd. This oddness would be further compounded by not discarding those that do not match what we know is correct.
But the you seem to then suggest that real outliers should be discarded or their influence reduced (how far out does it have to outlie before it suffers the wrath of a statistician!?).
This may be the way of a statistician but it is not the way of an engineer. (and I know who I would rather design bridges/aeroplanes! :o) )
Mike
Re: thefordprefect (#75),
Tfp, (Steve or others, please feel free to correct me if I’m wrong) there will be a sampling protocol which qualifies the areas and the trees which are chosen for sampling on the basis of science. Trees chosen under these qualifications are your “population”. That population will have values that follow a frequency distribution. If the nature of the distribution is well defined, strong inferences about the population might be attained – even for sample subsets within the population. Subsets of values with strong insturmental temp correlation should be distinguishable from the the noise of the broader population.
Re: thefordprefect (#75),
Makes sense, doesn’t it! But it’s a bad assumption. They look for trees with “good” response to stress, rather than “complacent” trees that do not vary in growth from year to year. The result: stressed trees are chosen over normal trees.
In the case of BCP’s, they completely ignore whole-bark trees while selecting strip-bark trees.
Don’t feel bad. Common sense has to be tossed out the window to remain sane in this adventure.
Re: thefordprefect (#75), Outliers are by definition statistical oddities. One problem with a small sample is that it is more difficult to determine whether an extreme value/trend is in fact an outlier. YAD06 may not be an outlier if there was a larger sample with the appropriate metadata. We do not know whether someone has estimated the height of mature males in North America by sampling professional basketball players.
Re: Steve McIntyre (#67),
Note that yad061 gets a growth burst in the 1950’s. The 1940’s are the hottest, and growth is minimal then.
Check real temperature in nearby Salehard
http://data.giss.nasa.gov/cgi-bin/gistemp/findstation.py?datatype=gistemp&data_set=1&name=&world_map.x=511&world_map.y=56
Re: Hans Erren (#109),
Thanks Hans. It might be fun to use these cores to test one of Tom Ps pet ideas. Selecting cores that correlate.
Which of the dirty dozen does the best job. Lets say, using the period of 1880 to 1940, see how well the core
predicts the future ( 1941 to the present) and then using the present to 1941, see how well it reconstructs the past.
That would be fun.
Re: steven mosher (#110),
I left a comment here once to ask if anyone had ever taken this approach with any of the multiproxy studies – calibrate on the early part of the instrumental record and verify on the modern years. I don’t think I ever got an answer. Anyone know if it’s been done?
Re: Bishop Hill (#113),
I dont know. Jeff Id could do it I bet or romanM or whatshisface..UC ya he could, Hu too I bet.
There exists a third larch chronology from the area:
Gurskaya, M. 2008. A 900-years larch chronology for north-western Siberia on the bases of archaeological wood of the Ust-Voykar settlement. // Geochronometria 2007, Vol. 28, P. 67-72 (pdf)
Absolutely no hockey stick. Notice that the comparision is to the H&S-version of Yamal.
Re: Jean S (#60),
No temperature reconstruction has been attempted in this paper – they are establishing the dataset:
“The sample replication and high correlation between series make this chronology of high enough quality for consideration of possible climatic reconstruction.”
Hence the lack of a hockeystick is hardly a surprise!
Re: Tom P (#60),
oh really? Do you think Yamal chronology under discussion is a temperature reconstruction?!?
Re: Jean S (#65),
Yes Tom keeps calling it a reconstruction even though I have corrected him several time. He’s stuck on stupid
New forests can be planted by aerial broadcast of seed. Over time, the random seed scattering gives way to a more regular distribution of separated trees (as close and weak ones die), from various mechanisms beyond the obvious that have been touched on already on CA.
There is an Allelopathy Journal that deals inter alia with the release of chemicals from some trees, to discourage close neighbours. http://www.allelopathy-journal.com/allelopathy.aspx?target=aaf1994
There are fungi that develop symbiotic relations with trees. This is important because fungi are among the most important destroyers of woody material; but some, conversely, improve tree growth. See Mycorrhizal Fungi at introductory level at http://cropsoil.psu.edu/sylvia/mycorrhiza.htm
The natural distribution and vigor of such non-fertilizer chemicals and fungi local to a tree can affect the isolation of that tree, its photosynthetic input and its relative size. This touches on the YAD 06 story.
Then there are super trees. In the early days of logging, it was common to fell the biggest and best, a backward step sometimes because these might have had the genes to make future super forests. These days there are active programs to multiply the seeds of identified super trees to giver higher yield per unit time in plantations; and breeding programs to induce hybrid vigour, some of which are emulated in nature by various seed dispersal and fertilization mechanisms to minimise self-pollination.
Not all species exhibit all effects noted above. I have never studied larch, for example and Australia’s abundant eucalypts are a group apart.
There is an interesting reference to larch in Russia (Kamkatcha) at http://www.ecosystema.ru/08nature/world/kam/veg/03e.htm with stark photography of fire effects in 1975-76. Further references can be found on the causes of larch forest density there.
But the point of this post is to note that beyond the obvious influences of temperature, rainfall, sunlight, crowding, nutrition, permafrost etc, there are myriad other confounding variables. Pest infestation, meteorite impact (see further south in Russia in 1908 at http://www.spacedaily.com/reports/The_Tunguska_Meteorite_As_A_Warning_From_Outer_Space_999.html ) fire, are among them.
Therefore it is exceedingly unwise to attribute varying tree ring response to a single variable, especially temperature. Stationarity might well be the exception rather than the rule.
Having said all this, it is not logical to downplay temperature as an influence on ring growth. It is merely hard to isolate its effect. It is also quite hard to find simple graphs plotting one versus the other in the literature. Suggestions welcomed.
IMO one of the biggest problems with Team reconstructions is faux confidence intervals – using incorrect methods or false assumptions, they claim much narrower confidence intervals than their methods and data permit. UC’s view is that the confidence intervals go from the “floor to the ceiling” using Hegerl’s phrase. We observed this in respect to Mann 2008 in our PNAS 2009 Comment.
Sorry if this is offtopic, again.
If one wishes to speculate about the physical explanation of the “divergence problem” at Yamal, then here is some good backround material from the area.
LCLUC Yamal
It seems that the vegetation production at Yamal doesn’t correlate very well with land temperatures.
Steve: If either of you can write up a separate thread on this and email it to me, I’d be happy to start one.
Re: Jarkko (#68),
Don’t know if Steve is interested, but I am.
I’ve posted the following update in the thread as a response to a comment by Tom P about using cores under 100 years old.
Tom P asserts:
I agree 100% that it “would be extremely difficult to extract a centennial signal” when “most [of the cores] are under 100 years!”. However, I disagree that the trees in the CRU archive are “much longer-lived”, other than the trees selected for the modern comparison. The following graphic shows the average of tree by year in the CRU archive, the average age in the “Schweingruber variation” in which russ035w is used instead of the CRU12 to represent living cores. Prior to around 1800, the average age of the tree in a given year was around the 100-year mark that Tom complains about. There is a profound inhomogeneity in the age composition of the living trees in the CRU archive relative to the subfossil archive, which is much reduced in the Schweingruber Variation. Does the age inhomogeneity in the CRU version “matter”? It’s the sort of thing that should have been reported and discussed in a site report, prior to using this chronology in multiproxy studies.
Re: Steve McIntyre (#70),
But the modern comparison was the subject of your original sensitivity analysis that was supposed to have broken the Yamal hockeystick!
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.
Re: Tom P (#77),
You’re still completely clueless.
Re: Tom P (#77),
C’mon! On what evidence do you base this remarkable statement? You read a couple of papers and you are now an expert in the analysis used and in differentiating “noise” and “signal” by mere observation. “Yesterday, I couldn’t spell dendrochronologist and now I are one!”
And which instrumental temperature do you think it doesn’t compare well with? Local, global? have you checked? Presumably you have heard of “teleconnection” (referred to by Tom C. above) which has sometimes been used in proxy selection. Is that what’s happening here?
When you make blanket statements, please provide some good evidence for making such assertions.
Re: romanm (#88),
I take as my authority Steve McIntyre:
Steve further states:
or to put it more directly, he agrees the Schweingruber live-core data comes from much shorter-lived trees than the CRU archive. Look at Steve’s plot above and you will see the average age of the Schweingruber trees (the green line) is mostly much less than 100 years, and much younger than the CRU cores (black line) for the same period.
Hence I am sure he will not dispute that he was diluting any centennial signal that the much longer-lived live cores of the CRU archive might contain by adding in the Schweingruber series.
The temperature data is in the 2008 Briffa paper.
Steve: Tom, again, you don’t know what you’re talking about it. The salient comparison here is between the CRU Archive and Schweingruber Variation. Are you seriously suggesting that CRU archive is more homogeneous in age distribution than the Schweingruber Variation? You’re just being argumentative. To the comparison clearer to such strange misinterpretations, I’ll add a comparandum aging plot of the CRU 12 as well, as it is how the CRU 12 blends with the other 240 relative to the russ035 blend that is at issue.
Re: Tom P (#91),
Steve McIntyre pointed out that the data prior to 1800 came from trees of comparable age to the Yamal trees. If the Schweingruber data is just noise then does not this mean that the Yamal data prior to 1800 is just noise as well and so the Briffa reconstruction is without merit?
I’m a mere mortal layman trying to wrap my brain around this whole thing. I have to ask: Isn’t there an even bigger implication here? That being: doesn’t this cast doubt on the refereed journals themselves and their peer review processes that failed to ensure proper data archival, thus enabling this to go on in the first place? Doesn’t this, therefore, potentially cast doubt on just about everything ever published in said journals? Imagine the black eyes that the reviewers are nursing right about now. I wonder what sorts of CYA postures are in development now?
My previous post may not belong here. Please move as you see fit.
Steve
Can you comment on Briffas response to you as referenced by John at WUWT?
http://go2.wordpress.com/?id=725X1342&site=wattsupwiththat.wordpress.com&url=http%3A%2F%2Fwww.cru.uea.ac.uk%2Fcru%2Fpeople%2Fbriffa%2Fyamal2000%2F
Tonyb
Warming in a tundra region can create unexpected effects. Studies around Hudson bay (small trees, no tree ring measurements) by Payette et al show warming causing dieback, because trees are not covered with protective snow as much and for other reasons. Melting permafrost can ironically drown roots, which kills some trees and slows growth of Larch. Many areas scoured by glaciers have very thin soil (organic mat over bedrock). A very slight warming will dry this out and slow growth.
I’m interested in the impact of the “one size fits all” generalized negative exponential as a correction for the effects of tree age. Seems to me it has to be only an approximation to whatever the real effect of age is, and probably a more or less good one depending on any number of factors – maybe things like species, geographic conditions, and so on. Further, it seems to me that this limits the ability to compare over long time periods – sort of like using a linear approximation to a quadratic curve works pretty well over relatively short segments, not at all over long ones.
Not knowing the details of the RCS method (e.g. the size of the constant relative to typical growth), is increasing error in the estimated “growth signal” due to the use of an invalid correction term a valid concern? Could the error in the approximate correction result in “corrected growth rate” errors on the order of several sigma over a few hundred years?
…CA needs a disclaimer, do not read while drinking milk!
Ow, my nostrils…
I have helped out in teh technical aspects of litigation for patent infrindgement. In one case, a company said that their device was not a computer. It contained an Intel processor, a 20 gig hard drive, a few hundred megs of RAM, a LAN connection for cotrol of remote servers and all of this running with Linux and open high level applicatiosn. However they assured us it was not a computer.
This beings too mind some of rationalizaitons that I am reading here about tree ring proxies and temperature. No matter what, they are going to be valid proxies
RE Steve #70,update
The new graph has red, black, and green lines, but the legend has one red line and 2 black lines. Which of the latter should be green?
And how can the average age of a cored tree be 0? Or is this counting the central ring of an old tree to have age 0?
Wolfgang Flamme
Since noone else has reacted to your seemingly sensible suggestion, I will.
The problem is that the local temperatures have not risen in any significant way. Real temperatures for the area do not match the cherry picked treemometers meteoric rise, and are more or less flat, with maybe a slight rise. Certainly not enough to have a significant noticable effect on the permafrost (never mind being miraculously localised for the particular patch of ground under this amazing tree).
At this point, anybody schooled in science asks themselves “what ? you mean the trees are not correlated to local temperatures prior to being used as proxies ??”
The amazing answer is yes. These climate scientists do NOT correlate their proxies with local temperatures, but with the whole Northern Hemisphere temperature AVERAGE.
The climate scientists have a term for this. They call it “teleconnection”, and do not seek to actually demonstrate that it is a fact, but take it instead on faith that if the proxy has risen in line with the Northern Hemisphere temperature average, then there IS a connection.
Yes, it’s scary that so much fear mongering goes on regarding unprecedented temperature changes when the connections are so weak, but welcome to climate science.
You’ll also find discussing these matters on the blogs of the faithful is futile as your comments will be deleted/edited out of recognition and you will be accused of being in league with the devil, or worse, evil corporations.
I have seen it posted anywhere yet, but Briffa responded here
http://www.cru.uea.ac.uk/cru/people/briffa/yamal2000/
#57
Oh good grief! I had missed this. Does this mean that they are again invoking some sort of teleconnection, or am I missing something?
Re: Tom C (#87), salehard is not in the area!
That is the funniest thing I have read in a long time! Thank you for the comic relief.
TFP, a 2009 paper by Hantemirov, which you pointed me to, has them using Sakhelard for temperature comparisons.
Konrad:
September 30th, 2009 at 7:51 pm
What have they been fertilizing YAD06 with? IPCC reports? Or is that where the climate science bodies are buried?
Possibly a real body (Elven?), slowly leaching nutrients into the groundwater accounts for the magical growth.
Well done Steve, another superb piece of analysis.
Actually, given what has been said about temperatures local to the area it would seem that the Schweingruber series does correlate rather well. And clearly, it is only local temperatures which are appropriate to use for calibration as it is only those temperatures which would actually have had an effect on the growth of the trees.
David, you said: “At this point, anybody schooled in science asks themselves “what ? you mean the trees are not correlated to local temperatures prior to being used as proxies ??”
The amazing answer is yes. These climate scientists do NOT correlate their proxies with local temperatures, but with the whole Northern Hemisphere temperature AVERAGE.
The climate scientists have a term for this. They call it “teleconnection”, and do not seek to actually demonstrate that it is a fact, but take it instead on faith that if the proxy has risen in line with the Northern Hemisphere temperature average, then there IS a connection.”
Is that generally true of the use of Dendrochronology as a temperature proxy by climate scientists? If so, how is it justified? Statistically, how is the calibration accomplished?
Perhaps the preferential selection of long-lived trees for the live-cores used in the CRU archive introduces a bias?
Wouldn’t you expect long-lived trees to also be fast growing?
An Alternate Ending to Dave Dardinger’s
“An Elven ode to YAD06”, #20
One Tree to rule them all, let’s go and find him.
One Tree with wide, wide rings, we’ll core and grind him.
In the land of Yamal where the hockey stick lies.
Schooled in science? Actually, anybody with common sense is amazed at this.
Steve McIntyre,
specifically that portion used in your sensitivity analysis which was for live cores.
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.
Re: Tom P (#99),
That fact that this has nothing to do with Briffa’s selection criteria and the use of HS corridors for his end of century analysis is beyond you I suppose.
Aren’t the chronolgies adjusted for age to begin with?
Nick and Tom,
Don’t get me wrong. I’m all for looking at various ways of looking at the data. One thing that would be interesting is to look at various age adjusting methods for the same datasets. Another would be make various apriori selection criteria, like tree age ( > 200 years) or no subfossile when using RCS etc etc. with and without strip bark, with and without tre rings ( which dr. L and Hu did) These are all things as an analyst that I would do and document. If the data were all archived there are a whole bunch of interesting studies one could do. One could even drop the most influential tree in the world. Now wouldnt that be interesting. One could look at demanding a certain level of correlation with the temperature record ( say over the 1850 to 1950 period) all sorts of things. One could divide all the tree ring samples in half and develop methods with one half and apply them to the second half.
But the question here is
1. Are all the proxy studies really independent? No they are not.
2. Are they statistically robust? I’d withhold judgement on that.
Basically, AGW is true. The MWP is not robustly understood and talk about it is a diversion. A fun diversion
but a diversion.
Steve McIntyre (#70),
Thanks, there is also the issue of using RCS on subfossil specimens that Tom so interestingly avoids. I gave him a paper to read, he cherry picks a quote to suit his taste and runs down another spear. Actually, this is getting fun.
Re: steven mosher (#106),
There is a major problem with Steve McIntyre basing his sensitivity analysis on much shorter cores than the original 12 CRU live cores. I am still awaiting a response from him that addresses this issue.
The core age prior to the period of this sensitivity analysis is a separate issue, and would require a careful analysis of both the longer live cores and shorter fossil cores that form that period of the record.
I suspect Steve may already be working on this – but he will have to admit that his original sensitivity analysis is invalid if he is then going to explore the possibility that the early record has been flattened by the inclusion of noise from two many short fossil cores.
I got just an email from someone claiming their supplement can “Make y0u perf0rm in bedroom like YAD06!”
The GISTEMP data for Sahelard does show warming in the area. The 60s is a bit cooler, and 80s and 90s are warmer. The temp from 1985 is about even, and not higher than the the 1925-1950 period.
Summer anomalies show even less warming in recent decades.
From the Stupid Question Department: One incontrovertible fact is that CO2 concentrations have increased during the industrial age. C02 presumably effects tree growth in a positive way. How do tree ring chronologies separate this effect from the temperature effects they are attempting to proxy?
Re: J. Patterson (#111),
DAGS “Divergence Problem”.
What would it take in time and money to repeat these measurements with something approaching the double-blind rigor of drug testing (and that industry’s reporting of truckloads of laboratory data (not models) to the FDA)? I suspect (if well funded) a lot less than than the 10 years it’s taken Steven to accomplish the most rudimentary of audits.
I’d be happy to contribute to this or a similar effort if a disciplined protocol was published and widely reviewed before-hand.
Re: Ari Tai (#112),
Go to category “Almagre” (popup list at the top left) to read up on an effort that partially answers your question. The Almagre adventure differed:
* USA mountains vs far north Russia
* The goal was not full analysis and publication of proxies
But it was accomplished double-blind in the sense that we pre-committed to revealing all the data no matter what we found, and we cored trees not knowing what their data would look like. I suppose that’s an advantage of being newbies 🙂 … the one experience was enough so that now I could easily bias my data just through coring methodology.
[My own bent would be to again collect a variety of core types, so as to show that variety and the resulting uncertainty levels. The problem is nicely described by CA reader — and enviro plant physiologist — Don Keillor here. The field of dendroclimatology is based on shaky fundamental assumptions. They have a lot of work to do to tighten up the relationship between approved methods of data collection/analysis and outcome validity. I would love to see a serious discussion between biologists and dendroclimatologists! Interested readers can go back to early CA days, e.g. the discussion here.]
#117: Tom, do you not see that the average age of the cores dramatically increases in the last century in the data that briffa used? Therefore, if anything, using Schweingruber’s data has *less* of a bias than what Briffa used (if you assume that tree age affects ring growth, which it probably does). Just what is wrong with the logical part of your mind on this topic?
After wasting an hour plus in reading the silliness on the Briffa Responds thread instigated by Lorax impugning Steve’s motives and ignoring the real issues, I thought it might be more useful to post a segment plot for the Yamal series. I think it is a bit more informative than the average age plot posted in Steve’s comment #70. The plot was done by adapting the function seg.plot from the dplR tree ring library in R.
The ten cores plotted in the original post are highlighted in red. I think it illustrates pretty well just how pervasive shorter cores are throughout the series.
Re: RomanM (#119),
that’s a nice plot. Another interesting R package.
Re: RomanM (#119),
RomanM,
This is a very informative plot. I’d be grateful if you posted the code.
Two additional plots would also be very informative:
1) The same plot with the inclusion of the Schweingruber series
2) A second plot of (1) with all records less than 100 years removed
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…
I strongly suspect Steve McIntyre may already be doing this!
Re: Tom P (#121),
Is this what you had in mind?
Re: conard (#159),
Excellent. The second plot, excluding all the trees under 100 years would also be interesting.
Here is a plot of the number of trees greater than a certain age in the CRU Yumal archive.
http://img194.imageshack.us/i/yamalagedistribution.png/
The general smoothness of the curve is pretty solid evidence that there has been no biased selection of shorter cores. There are a few much shorter cores less than 72 years old. Here they are:
069 L21781 0644 0715 71
150 L21831 1244 1314 70
114 L11831 1022 1086 64
028 L13561 0201 0260 59
056 L08751 0605 0662 57
017 L08651 0117 0167 50
049 _04451 0551 0601 50
022 L15071 0176 0216 40
They seem to be rather spread around, so I doubt their effect would be very noticeable.
As for the sensitivity analysis that I keep banging on about, there would be 198 cores more than 100 years old, 123 cores left more than 150 years old and 65 cores more than 200 years old – all I think pretty solid numbers depending on the overlap.
If you’d like to have a go at this, please go ahead.
Re: RomanM (#119),
This is probably a dumb question: Does each line represent an individual core, and the time period it is shown that correlates with temps? No one tree in both Yamal data sets extends from 0 year to the present?
Re: Sonicfrog (#132),
That’s right. This is a point that people unfamiliar with this data set need to keep in mind – these are not longlived bristlecones, they are relatively short lived. The “chronology” is built up by crossdating trees.
Re: Sonicfrog (#132),
It’s (largely) a wiggle-matched cross-dated assemblage of fossilized wood pieces.
Re: RomanM (#119), Thanks Roman.
I suppose Tom P has something profound to say about RCS short cores and Briffa’s practice here. That’s a pretty damning chart
The R library which contains the various tree ring functions is dplR. The library has a variety of functions available including reading treering files, detrending series (three different modes available for detrending, but RCS is not one of them) and creating chronologies. Be aware that some of the functions will give an error message if the ring widths are not a data frame rather than just a matrix.
The particular function for plotting proxy “lifetimes” is seg.plot. However, since there are 252 proxies, the Y-axis (which lists the proxies by name) looks like a mess. So, I changed the function a little bit to remove the names, add a title and colour the ten modern proxies red.
Here is the code:
I had to cobble this together again for posting. I hope I didn’t mistype anything.
I don’t have time to run any new stuff right now – got some work to do, but will look at it again later.
Re: RomanM (#122),
Thanks.
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!
Re: RomanM (#122),
Great stuff, RomanM. I am amazed how you (and others) find all these useful and applicable functions from R. I know that the more you are familiar with R the easier it becomes to discover available functions. And, of course, it takes some good statistical background to recogonize how to utilize them.
If you get the time I would be interested in the process you used to locate this function. I have trouble (less as time goes by) searching for these specialized functions in R. It might be related to my (wanting) statistical background.
Re: Kenneth Fritsch (#126),
The CRAN website lists a myriad of packages available for isntallation. Google CRAN.
Steve: It was a good find by Roman. I’ve never used it.
Re: bender (#127),
Bender, thanks for the reminder about the project list at CRAN. I did find the library dplR listed and described there. As RomanM has needed to tutor me before, I need to use all the R resources.
Re: Kenneth Fritsch (#136),
Package not project.
Re: Kenneth Fritsch (#136),
In answer to your earlier question, I ran across dplR when I was looking around to see what ARSTAN was like. What I could find was DOS based program that looked a bit lame. The search led me to the Ultimate Tree Ring pages which had a list of programs including dplR.
When in need of something, I also do general searches for key words (e.g. treering) iusing R’s very fine search help feature.
Interestingly enough, I spent the afternoon helping someone cut down a bunch of trees that were closer to my house than I wanted. From personal experience, you don’t get a good set of rings when using a chain saw. 😉
Re: RomanM (#146),
Likewise I spent friday/saturday taking down a ~100yo Norway Maple in my back yard, I tried to read the rings but it was impossible!
Re: Phil. (#284),
Conifers, you could see the rings if your saw is sharp. Hardwoods, you need to take a section, dry, and sand.
Re: Phil. (#283)
My trees consisted of some maples, pines and dead birches. I was surprised that the chain saw completely obliterated the rings. Bender is probably right in saying that the treatmentof the wood is necessary to bring out the information.
In my case, I don’t need the proxies, since I have data on the ice melt dates of the lake I live on. 😉
Re: RomanM (#287),
Yeah, I’d assumed that I’d have to do some polishing to bring it out, I was surprised I could see nothing without doing so. My guess is the tree is about the age of the house (100 yrs) it’s big enough (65′). When digging a foundation for a basketball net about ten years ago I came across a layer of ash and I wondered whether it dated from the clearing of the woods.
I guess fall is tree clearing time, my curiosity is piqued now I shall have to check out the rings now!
Re: Phil. (#292),
Maple (like birch, beech, alder) is a diffuse porous hardwood. The line between years is very indistinct. Not like conifers or ring porous hardwoods (oak, ash).
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 this.
Of course the statistics will deteriorate as the age threshold rises, but looking at figure 7 of this reference
Click to access TRRvol60_2_77-90.pdf
which models the statistics for the reliability of RCS chronology, even a small number of trees (15 to 20 total) with sufficient overlap can discriminate a high enough amplitude signal.
This figure also shows that 12 records at any one time (the CRU 12) is not below a statistical threshold of validity. As good coherence can be obtained for as few as 15 trees of average age 550 years distributed over 2000 years, the threshold can be as few as 4 trees at any one time.
Re: Tom P (#123),
It is refreshing to hear you admit your weaknesses in analytics. It is perfectly reasonable to ask for graphs and analyses and so on. But if you are going to ask, you must be patient. The more clearly and concisely you can phrase your request, the more likely someone will be to provide a reply. Good day.
Re: bender (#125),
No, just in R. As you are apparently far superior here, would you be able to run the sensitivity test I suggested?
Re: steven mosher (#135),
I’ve already said something here!
Again, if you’re proficient in R I would really appreciate you running this sensitivity test. It would be an easy way to demonstrate that the Briffa chronology was itself corrupted by the inclusion of too many short cores.
How come that the most tree-ring samples is taken far from places with long temperature record?
It is then impossible to correlate any result from tree-proxy.
Would it not be wise to first take samples located very close, say some 100 meters or so from stations with long temperature records to show if there is any correlation with temperature. Then widen the circles to more remote places. Right now does this science try to investigate historical climate where there is no measured temperature data. Very difficult to prove any correlation then.
Steve: It’s not anybody’s fault. They are looking for trees at northern treelines on the basis that they are more temperature sensitive. These are lightly settled places.
Re: Cold Lynx (#128),
Great question. Easy answer. Because treeline sites (alpine and tundra) where trees are most sensitive to temperature are typically in locations where nobody cares about the weather. Remember that the world’s met stations were never located in such a way as to estimate global climate change. They are located in places where people live (non-alpine, non-tundra) and care about the weather.
Tree rings sample from living trees close to Uppsala, with a measured temperature record from 1722 would be very intresting, and not especially hard to find. Located 59,51 north. Plenty of forrest and parks. It would probably be a good proxy for UHI as well.
How do we now that the treeline sites is most sensetive to temperature? The HS? Is that not the other way around? It may show that it just survive and is not sensitive to small temperature changes as the stick show. The Yamal HS is not showing the MWP, and we may have to ask ourselves why.
Re: Cold Lynx (#130),
1. Uppsala is not treeline.
2. Treeline sites are cold. That’s why growth at those sites is expected – and has been shown – many, many times – to be correlated with temperature. Just as tree growth in warmish valleys has been shown to be fairly insensitive to changes in temperature.
3. No, not in any of the multiproxy temperature reconstructions. In the canonical literature, which I cited earlier (Salzer, Kipfmueller, Wilson, many others, etc).
Re: bender (#131),
Well here is some that share my argument to use trees with longer growing period to have a proxy for average temperature
From a comment at WUWT
“If you want your trees to validate global warming, then I agree. If, however, you want your trees to measure average temperature throughout the year, then having them be dormant over 90% of the time is not going to tell you anything.”
Re: Cold Lynx (#147),
The point has been made a kjillion times. Temperature reconstructions are for the growing period only. The hope is that over a 2000-year period where temperature variations are very high, summer temperature is strongly correlated with winter temperature. I strongly suspect this is the case wherever global-scale orbital, solar, and volcanic forcings are the driver.
Hmm i do not agree this time with You bender.
Tree rings from trees in tempered areas is very easy to see. And measure. And they do have different with depending on growth conditions.
Tree rings from treeline areas have very narrow tree rings that MAY be more difficult to measure and get a trustworty signal out of noise.
There is tree rings studiesthat show LIA and MWP
http://www.worldclimatereport.com/index.php/2008/02/04/1500-years-of-cooling-in-the-arctic/
2 and especially 3 is not at treeline areas.
Why is Yamal series not showing this?
Excellent post at WUWT “Response from Briffa” by Caleb (18:59:33) which bears repeating in full here IMHO:
I’ve worked outside since I was a small boy in the 1950’s, and have cut down hundreds of trees. I always check out the rings, for every tree has its own story.
I’ve seen some rather neat tricks pulled off by trees, especially concerning how far they can reach with their roots to find fertilizer or moisture. For example, sugar maple roots will reach, in some cases, well over a hundred feet, and grow a swift net of roots in the peat moss surrounding a lady’s azalea’s root ball, so that the azalea withers, for the maple steals all its water.
I’ve also seen tired old maples perk right up, when a pile of manure is heaped out in a pasture a hundred feet away, and later have seen the tree’s rings, when it was cut down, show its growth surged while that manure was available.
After fifty years you learn a thing or two, even if you don’t take any science classes or major in climatology, and I’ve had a hunch many of the tree-ring theories were bunkum, right from the start.
The bristlecone records seemed a lousy proxy, because at the altitude where they grow it is below freezing nearly every night, and daytime temperatures are only above freezing for something like 10% of the year. They live on the borderline of existence, for trees, because trees go dormant when water freezes. (As soon as it drops below freezing the sap stops dripping into the sugar maple buckets.) Therefore the bristlecone pines were dormant 90% of all days and 99% of all nights, in a sense failing to collect temperature data all that time, yet they were supposedly a very important proxy for the entire planet. To that I just muttered “bunkum.”
But there were other trees in other places. I was skeptical about the data, but until I saw so much was based on a single tree, YAD061, I couldn’t be sure I could just say “bunkum.”
YAD061 looks very much like a tree that grew up in the shade of its elders, and therefore grew slowly, until age or ice-storms or insects removed the elders and the shade. Then, with sunshine and the rotting remains of its elders to feed it, the tree could take off.
I have seen growth patterns much like YAD061 in the rings of many stumps in New Hampshire, and not once have I thought it showed a sign of global warming, or of increased levels of CO2 in the air. Rather the cause is far more simple: A childhood in the under-story, followed by a tree’s “day in the sun.”
Re: Lucy Skywalker (#139),
This is wild speculation at best. Highly unlikely tundra larch are growing in any shade at all.
Re: bender (#140),
Maybe not a forest grove, but why not two or three trees which grow up together. Just how are larch seeded anyway? And what about root shoots anyway? Root systems hate to have to die just because their visible light gathering system was killed by lightening. Why not the occasional shoot which gets started but held back by the old tree(s)? Anyone, how about a quick link to some photos of these northern larches?
Re: bender (#140), look at the aerial photos and video hyperlinked from my Yamal page. Mostly no trees; in some places trees are sparse like you suggest; but in other places (esp by rivers, which are in the map ref. for the Yamal site) they are more crowded and all one sees from above is green. And sunshine is fairly horizontal at these latitudes too. Sure it’s only speculation, and many other factors are possible, but from a lifelong timber man I wouldn’t call it “wild”. Just opening up possibilities.
Re: Lucy Skywalker (#144), Lucy and Bender, both good points
The Hantemirov and Shiyatov tree coring expedition (paper below)
Click to access Holocene_v12a.pdf
was down the river by rubber boat, it seems reasonable to conclude that they didn’t venture far from the river. Many of the cores may have even been done while in the boat, as we see in this photo from the expedition:
But this photo from the Hantemirov and Shiyatov paper lends some credence to what you say:
Re: Anthony Watts (#146),
I should add that there is absolutely NO telling under what conditions fossilized larch bits were growing 1000 years ago. My point was regarding living trees with upticks in the 20th c. – the sort of trees that matter most to this discussion. Treeline trees are not light limited.
Re: bender (#148),
Bender, thanks for the search tip on the other thread. You are quite right there is lots of reading. Besides threshold temperatures, perhaps ecotone areas are subject to more growth pulses as well. Something like an edge effect with crops. I wonder if any possible ecotone issues are dealt with in sampling protocols.
Re: bender (#148),
Of that there is no doubt, but the metadata for the 10 trees, which as I understand it we don’t have for Briffa’s cores, might give us some environmental clue about the living trees that were sampled.
Re: Anthony Watts (#151),
Absolutely. I have spent enough time in northern larch to guess as to the site types – but seeing is believing. If the larch are growing in fens, not bogs, then the whole issue of oxygenation and root respiration comes into play. Slight drainage could have a huge impact.
Re: bender (#149),
Perhaps you have a too-limited experience of treeline? Time to come to Colorado for a visit!
Please take a look at this panorama, taken at treeline.
Certainly, SOME treeline trees are not light limited. But not far away you can find very dense forest. The demarcation between trees and no-trees can be very very sharp.
One of the challenges in finding Graybill’s BCP’s was the dense forest.
Couldn’t see the trees for the forest 🙂
[sorry if that doesn’t translate well for non-native English readers.]
Re: MrPete (#164),
I’ve been to the Rockies. You must agree the trees in the foreground are clearly not light-limited. Sure, forests in the distance always appear to be dense. The ones in the background … I’d like to see ground-level measurements of light intensity to make a proper comparison. By definition, trees at treeline are not light-limited. If the trees are growing in dense clumps then by definition you must be warmer than treeline.
Re: bender (#166),
I guess we need to find a good definition of “treeline.” I’m assuming we’re dealing with whatever definitions dendro’s use.
Suppose there’s a steep slope. Do they only sample the top row of trees (assuming a perfectly even line)?
Suppose the slope is not so steep. Or even that the trees are at the top of a gentle hill (a “park” in that sense). Then all the trees at/near the top may be treeline trees…and some of them may well grow in clumps.
The panorama I showed you was taken at treeline. Hard to show in a single panorama (esp one not taken for the purpose.)
At the exact same altitude there were both isolated trees, tree clumps, and even denser clumps.
The hard part is treeline is not only defined by altitude, although that seems to be major. With a complex topology, you get all kinds of interesting groupings.
What I know for sure: Graybill’s BCP’s were found in everything from open rock fields to dense forest, all within a narrow altitude range.
Re: MrPete (#169),
I don’t know how dendros define “treeline” in practice. But if they are sampling trees growing in dense forest then those tress are not at treeline, even though they might be generally working at treeline-ish altitudes. The trees they are coring can not be light-limited or it jeapordizes their hypothesis. In the case of Graybill, treeline has probably shifted upslope tens to hundreds of metres since the 1960s.
Re: bender (#179),
AFAIK, Graybill’s work was in the 1980’s. In any case, you can see from our matching photos that while the trees have grown, the tree line is similar. It was actually amazing to me how little had changed in 20 years. I honestly did not expect to be able to match photos. Especially the one-tree photo. We were shocked when Leslie suddenly noticed that a tree I was working on was similar to one of the photos we’d been half-attempting to match for a few days.
I recognize these photos don’t make it easy to diagnose our question. There’s been a lot of growth, and the spread of the trees is not easy to discern.
However, it may help to remember that this is an extreme environment we’re discussing here. Trees don’t crop up quickly like they do elsewhere. I’ve lived here for 15+ years myself (at 7050 feet), with a few acres of trees and meadow. Even here, the trees do not quickly expand into the open space.
Here are matching trees. The trees in the background are now larger but there really aren’t any more.
Perhaps even better, here’s a view from the road. Ignore the foreground trees, which have grown significantly. On the hill in the background you’ll see that the “strip” of tress climing the hill has not spread appreciably. Yes, this is all at treeline. (Sorry, I had limited ability to find the exact same shooting location and parameters… a huge thunderstorm was approaching! 🙂 )
And finally, here’s a view from above.
Bender, if you really think they should not be sampling from shaded trees, chalk up another anomaly.
Here’s another panorama including tree 047.
The crown is in the sun but it has quite a lot of shade. Here’s a full view from the Picasaweb album:
Re: MrPete (#191),
Val La Marche was ca. 1968. He knew about the bcp uptick.
and here’s another interesting reference to six studies (h/t Douglas Hoyt @ WUWT) from fairly nearby areas, published by Oceanology, Phil Trans of Roy Soc, Russian Journal of Ecology etc, all using similar science to the Yamal study and other HS studies, all showing a warmer MWP etc
Don’t know about larches but the reference to the maples perking up prompted a google to this one on maple root systems:
Click to access paper10.pdf
Here is a series of clips from the U-Tube exploration of Yamal – clips which highlight extremely local variations in tree growth, and in density of tree cover.
How much is strongly correlated?
Also, would winter temperatures be a limiting factor on tree growth, so that a warmer winter would show substantially more growth?
There is a difference between the summer and winter temperature trends at Yamal.
Re: MikeN (#155), the limiting factor is number of frost-free days and nights, I think, rather than minimum temperature. However I’d like to see more work on trees’ ability to regulate their temperature – eg like what makes brussel sprouts work in the winter.
Re: Dave Dardinger [No. 142] Larch , like other Deciduous Conifers , propagate via seed cones . As far as spures from roots go , I’ve planted several different types of DC’s over the years including Golden Larch , Dawn Redwood , and Bald Cypress . Though I’m not going to say it’s impossible , in the 20 years I have worked with them , I’ve never seen a failed DC send out Shooters like a broad leaf . In fact , neither have I seen a healthy one do that.
I’m getting a bit confused now.
Is the sample for these plots the Schweingruber Variation or what ?
Re: David (#161),
If people are going to post graphics they should be putting proper axis labels, legends and captions on them. This is the best way to pre-empt any such confusion.
Or if this is the extent of only the Briffa data, can we see a plot for the Schweingruber Variation too ?
I’d like to see graphics showing correlation between these ten trees, especially between 1900 (start of supposed HS blade) and 1990. Eyeballing, it looks like these trees are all outliers, responding to factors other than temperature. Can someone with the skill please do this?
If I have it right, 29 of the CRU cores come from 13 trees.
When this happens is it normal to treat them as just additional cores, or do you average the recovered signals for each set and give them a weight of Sqrt(n).
Also are they being double counted when it comes to calculating samples for a year.
Not as though I suppose it would make much difference.
Alex
Steve: No one knows what CRU does at that level of detail – ARSTAN is well documented and journal articles will sometimes specify which options were used, but they used their own methods. As to core duplication, my own experience in other contexts is that downweighting replicate cores doesn’t usually have a material impact.
Re: Alexander Harvey (#165),
Cheers Steve, I do think that they should not be treated as separate samples when the counts are done.
Alex
A comment on larch propogation at treeline(I posted this reference at WUWT):
from “The Effect of Permafrost on Northern Treeline” by V.V. Kryuchkov published about 1973:
“fallen trunks die and decompose while branches that become transformed into saplings will develop their own rooting system”. Sounds like they can propagate by cuttings as well as cones as long as they are in a peaty mossy environment. But then I’m not a biologist.
I’m afraid I lost my patience and have kludged Steve McIntyre’s code to do my sensitivity analysis (code will be posted after this).
First, here is the chronology without YAD061:
Click to access oyad06.pdf
Whatever that tree was drinking, it looked like it shared the round.
Now for the sensitivity analysis for the CRU archive:
1) Removing the cores less than 72 years old – the drooping tail at the end of the distribution I posted above:
http://img406.imageshack.us/i/cru72.pdf/
As I suspected, these cores don’t contribute much to the chronology.
2) Removing the cores less than 100 years old:
http://img25.imageshack.us/i/cru100.pdf/
Not much difference.
3) Removing the cores less than 150 years old:
http://img156.imageshack.us/i/cru150.pdf/
Still not much of a shift.
4) Removing the cores less than 200 years old:
http://img59.imageshack.us/i/cru200.pdf/
This has removed YAD06 amongst other cores, but the profile remains the same. The noise is increasing but the profile is till clear. There are now 64 cores left, with an average age of 262 years, or an average of 8 cores at any one time.
5) Removing the cores less than 250 years old:
http://img202.imageshack.us/i/cru250.pdf/
Now there are just 32 cores left with an average age of 303 years, or just four cores at any one time. The hockeystick has finally been broken, but only by removing so many cores that the noise has finally overcome the signal.
Briffa’s result appears robust to a very demanding test. I’d appreciate Steve McIntyre’s response to this.
Re: Tom P (#171),
The first link shifted from under me. The comparison with and without YAD06 is here:
Click to access oyad06.pdf
Re: Tom P (#171),
Arghh! it’s the invisible missing backslash at the end of the link that’s causing the problem – this is tougher than R!
Chronology with and without YAD06:
http://img515.imageshack.us/i/oyad06.pdf/
Re: Tom P (#173),
The end of the red curve is higher than the black, contradicting your conclusion. Post a plot covering 1950-2000.
Re: bender (#178),
tom p: you missed this one
Re: bender (#179),
Tom P: I asked you for this plot, and now your failure yo produce it is leading your friend, Eli Rabbett, astray (see Ben Hale’s blog). Please produce the plot. Please just revise the x axis so it covers 1950-1990. Why won’t you do it? I have asked you twice to do so over the last 5 days.
Re: Tom P (#173),
Actually it’s not too hard to see what is happening in your “experiment” simply from the graphs of the Yamal 10 and the plot of tree ages by conard.
Re: conard (#159),
The 5 YAD trees are the ones you eliminated via your age variations. The 5 POR trees weren’t touched. Looking at the charts, all of the Yamal 10 have some degree of hockeystickness. YAD06 has by far the most, but the POR trees each have quite a bit and would likely produce the 2-3 SDs in your graphs. This is why Steve treated the Yamal 10 as one group. Changing them out like with Schweingruber shows how what a particular group of trees contributes to a final result. this was what he was trying to show, not that any group of 5 or so of the 10 Yamal trees would produce a HS. If you want to do a useful variation I’d suggest working on Steve’s green variation, Schweingruber plus Yamel and then compare that with Schweingruber + each one of the Yamal trees individually. There I’d guess you’d find a much lower ending but perhaps the one with YAD06 might be rather higher than the other 9 cases.
Re: Dave Dardinger (#190),
The graph I posted is a concatenation of YamalADring.raw and russ035w.rwl, highlighting the [red] Yamal 10 per Roman’s graph and the [green] Khadyta per Steve’s average age graph at Tom P’s request [#121 request 1]. No trees were intentionally removed. Unfortunately I cannot check for the error you claim because I did not save the script– the graph did not appear to reveal anything new or interesting that Steve and Roman did not capture but if you find it has relevance Roman’s script is easily modified and takes only a few minutes.
More than anything I was curious to see if a matlab/octave user could get a handle on R and spent some time scanning Steve’s code to get oriented with his use of R. It was a fun diversion while waiting for my body to decide to sleep 😉
Re: conard (#193),
You misunderstand what I was telling Tom P. There’s nothing wrong with your graph. I was pointing out that his eliminating trees in several steps by their ages results in the elimination of 5 of the Yamal trees, in particular the 5 trees with the YAD prefix. The other 5 living Yamal trees would still be present in his last step and thus cause the HS which rises about 2-3 SD. IOW, his process, while interesting, proves nothing. I cited your graph so people could see that 5 of the Yamal trees which were with the green section, were younger than 250 years and the other 5 were older. This could also be figured out using Steve’s plating of the trees above, but it’s easier to see with your graphic.
Re: Tom P (#171),
Well, you’re not going to get just one response from Steve, so …
Can you construct a coherent supporting argument to go along with the graphs?
Re: Tom P (#171),
In due course. Grandkids were over. I’m only online for a few minutes.
MBH was exceptionally robust to a number of supposedly “demanding tests” as long as they kept the bristlecones in. But it wasn’t robust to the bristlecones – an unfortunate attribute since even the NAS panel opined that strip bark trees should be avoided in reconstructions. (Not that that’s stopped the Team from continuing to use them.)
I’ll try to look at what you’ve done tomorrow.
########################
##DOWNLOAD DATA AND FUNCTIONS
#############################
source(“http://data.climateaudit.org/scripts/utilities.txt”) #
source(“http://data.climateaudit.org/scripts/tree/utilities.treering.txt”)
f=function(x) filter.combine.pad(x,truncated.gauss.weights(21))[,2]
#utility smooth function
#Hantemirov at NCDC
loc=”ftp://ftp.ncdc.noaa.gov/pub/data/paleo/treering/reconstructions/asia/russia/yamal_2002.txt”
hant=read.table(loc,skip=57,fill=TRUE,header=TRUE) # Year Recon Chron Samples
dim(hant) #4064 4
hant=window( ts(hant[,2:4],start=hant[1,1]),start= -202 ) #minimum in measurement data
#Briffa Chronology from CRU
loc=”http://www.cru.uea.ac.uk/cru/people/melvin/PhilTrans2008/Column.prn”
briffa=read.table(loc,skip=1,fill=TRUE)
name0=scan(loc,n=8,what=””)
name0=outer(name0,c(“”,”count”),function(x,y) paste(x,y,sep=”.”) )
n=nchar(name0[,1])
name0[,1]=substr(name0[,1],1,n-1)
names(briffa)=c(“year”, c(t(name0) ) )
briffa[briffa== -9999]=NA
briffa=briffa[,c(“year”,”Yamal.RCS”,”Yamal.RCS.count”)]
briffa=briffa[!is.na(briffa[,2]),]
briffa=window(ts(briffa[,2:3],start=briffa[1,1]),start=-202)
yamal.crn=briffa[,1]/1000
#Yamal measurement data
loc=”http://www.cru.uea.ac.uk/cru/people/melvin/PhilTrans2008/YamalADring.raw”
download.file(loc,”temp.dat”)
tree=make.rwl_new(“temp.dat”)
tree$id=factor(tree$id) #252
tree=agef(tree)
#save(tree,file=”d:/climate/data/yamal/yamal_cru.rwl.tab”)
#load(“d:/climate/data/yamal/yamal_cru.rwl.tab”) #tree
range(tree$year) #202 1996
yamal=tree
dim(yamal) # [1] 40892 4
yamal$rw=yamal$rw/10 # Sep 28
mean(yamal$rw,na.rm=T) # 61.52668
#################
## INFO COLLATION
#########################
Info=data.frame(id=as.character(levels(tree$id)) )
Info$start= tapply(tree$year,tree$id,min)
Info$end= tapply(tree$year,tree$id,max)
count=tapply(!is.na(tree$rw),tree$year,sum)
Info$max= Info$end-Info$start+1
Info$id=as.character(Info$id)
n=nchar(Info$id);temp=n>3
Info$core=””; Info$core[temp]=substr(Info$id[temp],n[temp],n[temp]);
Info$test=Info$id;
Info$test[temp]=substr(Info$id[temp],1,n[temp]-1)
Info$id=gsub(“_”,”L”,Info$id) #guess
x=substr(Info$id,1,3)
temp=!is.na(match(x,c(“JAH”,”POR”,”YAD”)));sum(temp)#12
Info$site=NA; Info$site[temp]=x[temp]
Info$site[!temp]=substr(Info$id[!temp],1,1);
Info$n=1;Info$n[temp]=3
Info$tree=as.numeric(substr(Info$test,Info$n+1,nchar(Info$test)))
#Info[order(Info$n,Info$tree),]
# Info[order(Info$n,Info$end),]
(index3<-c(“L05721″,”L01031″,”L09211″,”_02341″,”M331″,”L08631″,”L04551″,”L12311″,”L12451″,”L12452″,”L02391″,”L20741″,”L06171″,”L08191″,”L01001″,”L08001″,”L08831″,”L13391″,”P08821″,”L19311″,”L00134″,”L09081″,”L06051″,”L04221″,”P08781″,”L04371″,”L00133″,”L11521″,”L00131″,”L00132″,”L16841″,”L01291″,”L01061″,”L04411″,”L18991″,”L07981″,”YAD041″,”YAD061″,”YAD121″,”L13611″,”L00191″,”L13311″,”L00151″,”L00152″,”L19001″,”L00081″,”P08801″,”_04351″,”L22571″,”L02331″,”P15031″,”L01381″,”L08111″,”L19231″,”P21741″,”L13071″,”L01051″,”L11641″,”L00082″,”L05621″,”L04381″,”L13551″,”L11531″,”L22581″,”L00811″,”L19211″,”L01401″,”L00681″,”L08811″,”L02311″,”L04421″,”L15511″,”L04591″,”L08641″,”L19411″,”L20681″,”L12581″,”L12582″,”L13571″,”L09071″,”L00751″,”L13351″,”L01421″,”L20931″,”L15581″,”L04201″,”L20631″,”L17011″,”_04481″,”L00701″,”L00941″,”L19251″,”L04401″,”YAD071″,”L13421″,”L13261″,”L07511″,”L12601″,”11741″,”L13331″,”L08271″,”L01071″,”L08281″,”L19011″,”L13661″,”L13081″,”L01341″,”L12602″,”L13651″,”L15661″,”L05791″,”L01281″,”L22011″,”L13401″,”L13271″,”L01391″,”L01461″,”L06011″,”_05631″,”L08841″,”L00083″,”L01361″,”L05901″,”P17811″,”L04121″,”L15471″,”L13001″,”L11761″,”L00891″,”L01201″,”L13621″,”L09101″,”YAD081″,”L05771″,”L15571″,”L18921″,”L05921″,”L04521″,”L05481″,”L13441″,”L17651″,”L03881″,”L08701″,”L07571″,”L02281″,”L02231″,”L07441″,”L13581″,”L13191″,”L20881″,”L16791″,”L14331″,”L08791″,”L08681″,”L13321″,”L05911″,”P01231″,”L08661″,”L02321″,”L15851″,”L16611″,”L00661″,”L13251″,”L15662″,”L08711″,”L07431″,”L01471″,”L01241″,”L12661″,”L12662″,”L12561″,”L08691″,”L00961″,”L21991″,”L02371″,”L12562″,”L08671″,”L21981″,”L08721″,”L12111″,”L12112″,”L06021″,”L15831″,”L08131″,”L21841″,”L08741″,”L11571″,”L18881″,”L01411″,”L22001″,”L01171″,”_02361″,”_02351″,”L08431″,”_02271″,”L00771″,”L11551″,”L12421″,”L16861″,”L04431″,”L17181″,”L16071″,”L12422″,”L01331″,”L11572″,”L08621″,”L01151″,”L20731″,”_04491″,”L21821″,”L16081″,”L21781″,”L21831″,”L11831″,”L13561″,”L08751″,”L08651″,”_04451″,”L15071”))#this forces removal of the listed trees
################
# ANALYSIS 1: COMPARE COUNTS: http://www.climateaudit.org/?p=7142
######################
#1. clip the count image from Hantemirov 2002 CA/pdf/tree/hantemirov.2002.holocene.pdf
#2. Compare three versions
#plot count in CRU archive
count.yamal=countf(yamal)
year=c(time(count.yamal));N=length(count.yamal)
par(mar=c(3,3,2,1))
plot(year,count.yamal,col=”grey80″,type=”l”,ylim=c(0,45),yaxs=”i”)
polygon(xy.coords(x=c( year,rev(year)),y=c(count.yamal,rep(0,N))),col=”grey80″,border=1)
title(“Yamal Count: CRU Archive”)
#plot count in H and S version
par(mar=c(3,3,2,1))
year=c(time(hant));N=nrow(hant)
plot(c(time(hant)),hant[,”Samples”],col=”grey80″,type=”l”,ylim=c(0,45),yaxs=”i”)
polygon(xy.coords(x=c( year,rev(year)),y=c(hant[,”Samples”],rep(0,N))),col=”grey80″,border=1)
title(“Yamal Count: Hantemirov NCDC”)
#clip other illustration from HAntemirov and Shiyatov 2002
################
# ANALYSIS 2: COMPARE CHRONOLOGIES
######################
chron=ts.union(hant=hant[,2],briffa=yamal.crn)
Yamal=data.frame(year=c(time(chron)),chron)
fm=lm(briffa~hant,data=Yamal[Yamal$year 200 years”,”Full CRU Archive”))
title(“Yamal RCS Chronologies”)
########################
##DOWNLOAD DATA AND FUNCTIONS
#############################
source(“http://data.climateaudit.org/scripts/utilities.txt”) #
source(“http://data.climateaudit.org/scripts/tree/utilities.treering.txt”)
f=function(x) filter.combine.pad(x,truncated.gauss.weights(21))[,2]
#utility smooth function
#Hantemirov at NCDC
loc=”ftp://ftp.ncdc.noaa.gov/pub/data/paleo/treering/reconstructions/asia/russia/yamal_2002.txt”
hant=read.table(loc,skip=57,fill=TRUE,header=TRUE) # Year Recon Chron Samples
dim(hant) #4064 4
hant=window( ts(hant[,2:4],start=hant[1,1]),start= -202 ) #minimum in measurement data
#Briffa Chronology from CRU
loc=”http://www.cru.uea.ac.uk/cru/people/melvin/PhilTrans2008/Column.prn”
briffa=read.table(loc,skip=1,fill=TRUE)
name0=scan(loc,n=8,what=””)
name0=outer(name0,c(“”,”count”),function(x,y) paste(x,y,sep=”.”) )
n=nchar(name0[,1])
name0[,1]=substr(name0[,1],1,n-1)
names(briffa)=c(“year”, c(t(name0) ) )
briffa[briffa== -9999]=NA
briffa=briffa[,c(“year”,”Yamal.RCS”,”Yamal.RCS.count”)]
briffa=briffa[!is.na(briffa[,2]),]
briffa=window(ts(briffa[,2:3],start=briffa[1,1]),start=-202)
yamal.crn=briffa[,1]/1000
#Yamal measurement data
loc=”http://www.cru.uea.ac.uk/cru/people/melvin/PhilTrans2008/YamalADring.raw”
download.file(loc,”temp.dat”)
tree=make.rwl_new(“temp.dat”)
tree$id=factor(tree$id) #252
tree=agef(tree)
#save(tree,file=”d:/climate/data/yamal/yamal_cru.rwl.tab”)
#load(“d:/climate/data/yamal/yamal_cru.rwl.tab”) #tree
range(tree$year) #202 1996
yamal=tree
dim(yamal) # [1] 40892 4
yamal$rw=yamal$rw/10 # Sep 28
mean(yamal$rw,na.rm=T) # 61.52668
#################
## INFO COLLATION
#########################
Info=data.frame(id=as.character(levels(tree$id)) )
Info$start= tapply(tree$year,tree$id,min)
Info$end= tapply(tree$year,tree$id,max)
count=tapply(!is.na(tree$rw),tree$year,sum)
Info$max= Info$end-Info$start+1
Info$id=as.character(Info$id)
n=nchar(Info$id);temp=n>3
Info$core=””; Info$core[temp]=substr(Info$id[temp],n[temp],n[temp]);
Info$test=Info$id;
Info$test[temp]=substr(Info$id[temp],1,n[temp]-1)
Info$id=gsub(“_”,”L”,Info$id) #guess
x=substr(Info$id,1,3)
temp=!is.na(match(x,c(“JAH”,”POR”,”YAD”)));sum(temp)#12
Info$site=NA; Info$site[temp]=x[temp]
Info$site[!temp]=substr(Info$id[!temp],1,1);
Info$n=1;Info$n[temp]=3
Info$tree=as.numeric(substr(Info$test,Info$n+1,nchar(Info$test)))
#Info[order(Info$n,Info$tree),]
# Info[order(Info$n,Info$end),]
(index3<-c(“L05721″,”L01031″,”L09211″,”_02341″,”M331″,”L08631″,”L04551″,”L12311″,”L12451″,”L12452″,”L02391″,”L20741″,”L06171″,”L08191″,”L01001″,”L08001″,”L08831″,”L13391″,”P08821″,”L19311″,”L00134″,”L09081″,”L06051″,”L04221″,”P08781″,”L04371″,”L00133″,”L11521″,”L00131″,”L00132″,”L16841″,”L01291″,”L01061″,”L04411″,”L18991″,”L07981″,”YAD041″,”YAD061″,”YAD121″,”L13611″,”L00191″,”L13311″,”L00151″,”L00152″,”L19001″,”L00081″,”P08801″,”_04351″,”L22571″,”L02331″,”P15031″,”L01381″,”L08111″,”L19231″,”P21741″,”L13071″,”L01051″,”L11641″,”L00082″,”L05621″,”L04381″,”L13551″,”L11531″,”L22581″,”L00811″,”L19211″,”L01401″,”L00681″,”L08811″,”L02311″,”L04421″,”L15511″,”L04591″,”L08641″,”L19411″,”L20681″,”L12581″,”L12582″,”L13571″,”L09071″,”L00751″,”L13351″,”L01421″,”L20931″,”L15581″,”L04201″,”L20631″,”L17011″,”_04481″,”L00701″,”L00941″,”L19251″,”L04401″,”YAD071″,”L13421″,”L13261″,”L07511″,”L12601″,”11741″,”L13331″,”L08271″,”L01071″,”L08281″,”L19011″,”L13661″,”L13081″,”L01341″,”L12602″,”L13651″,”L15661″,”L05791″,”L01281″,”L22011″,”L13401″,”L13271″,”L01391″,”L01461″,”L06011″,”_05631″,”L08841″,”L00083″,”L01361″,”L05901″,”P17811″,”L04121″,”L15471″,”L13001″,”L11761″,”L00891″,”L01201″,”L13621″,”L09101″,”YAD081″,”L05771″,”L15571″,”L18921″,”L05921″,”L04521″,”L05481″,”L13441″,”L17651″,”L03881″,”L08701″,”L07571″,”L02281″,”L02231″,”L07441″,”L13581″,”L13191″,”L20881″,”L16791″,”L14331″,”L08791″,”L08681″,”L13321″,”L05911″,”P01231″,”L08661″,”L02321″,”L15851″,”L16611″,”L00661″,”L13251″,”L15662″,”L08711″,”L07431″,”L01471″,”L01241″,”L12661″,”L12662″,”L12561″,”L08691″,”L00961″,”L21991″,”L02371″,”L12562″,”L08671″,”L21981″,”L08721″,”L12111″,”L12112″,”L06021″,”L15831″,”L08131″,”L21841″,”L08741″,”L11571″,”L18881″,”L01411″,”L22001″,”L01171″,”_02361″,”_02351″,”L08431″,”_02271″,”L00771″,”L11551″,”L12421″,”L16861″,”L04431″,”L17181″,”L16071″,”L12422″,”L01331″,”L11572″,”L08621″,”L01151″,”L20731″,”_04491″,”L21821″,”L16081″,”L21781″,”L21831″,”L11831″,”L13561″,”L08751″,”L08651″,”_04451″,”L15071”)) #this forces removal of the listed trees
################
# ANALYSIS 1: COMPARE COUNTS: http://www.climateaudit.org/?p=7142
######################
#1. clip the count image from Hantemirov 2002 CA/pdf/tree/hantemirov.2002.holocene.pdf
#2. Compare three versions
#plot count in CRU archive
count.yamal=countf(yamal)
year=c(time(count.yamal));N=length(count.yamal)
par(mar=c(3,3,2,1))
plot(year,count.yamal,col=”grey80″,type=”l”,ylim=c(0,45),yaxs=”i”)
polygon(xy.coords(x=c( year,rev(year)),y=c(count.yamal,rep(0,N))),col=”grey80″,border=1)
title(“Yamal Count: CRU Archive”)
#plot count in H and S version
par(mar=c(3,3,2,1))
year=c(time(hant));N=nrow(hant)
plot(c(time(hant)),hant[,”Samples”],col=”grey80″,type=”l”,ylim=c(0,45),yaxs=”i”)
polygon(xy.coords(x=c( year,rev(year)),y=c(hant[,”Samples”],rep(0,N))),col=”grey80″,border=1)
title(“Yamal Count: Hantemirov NCDC”)
#clip other illustration from HAntemirov and Shiyatov 2002
################
# ANALYSIS 2: COMPARE CHRONOLOGIES
######################
chron=ts.union(hant=hant[,2],briffa=yamal.crn)
Yamal=data.frame(year=c(time(chron)),chron)
fm=lm(briffa~hant,data=Yamal[Yamal$year 250 years”,”Full CRU Archive”))
title(“Yamal RCS Chronologies”)
Tom P, I don’t understand. How does demonstrating that the reconstruction is not sensitive to the age of the trees alter in any way the results we see when the CRU 10 are removed?
Also, does this not negate your earlier complaint that the Schweingruber trees weren’t old enough to detect a signal?
What am I missing here?
Re: Michael Smith (#174),
An introduction, a hypothesis, methods, and discussion. What I see are “results” – and not all the links work. He’s somewhere between 1/10 and 1/5 of the way towards something worth contemplating.
Re: bender (#178),
He’s just trying to follow your standards here. Give him a break.
Re: slownewsday (#268),
At your urging, I shall give him a break. He is heartily invited to write this up and explain it to us. My reading of the whole, errr, piece was posted earlier. Perhaps he can clarify where I’ve gone wrong. I’m feeling very patient tonight.
Re: Michael Smith (#174),
I suggest reading my comments above, starting with Steve’s update at the top of the post.
Re: bender (#176),
As you have obviously been following this thread from the start, I don’t know what to suggest. I’ll put all my comments together in one post, though, if Steve allows.
Re: Tom P (#180),
1. You have three readers now complaining that you need to write a report. You think Steve is going to follow what you did when we plebes can’t?
2. I have asked you to revise the x axis on one plot so that you can cure yourself of your confirmation bias.
Tom P., after looking at the progression of your graphs from a reconstruction containing portions of higher aged trees, I would not conjecture what you did.
I thought your initial concern was that the shorter aged trees would not yield a good proxy response and your analysis shows that as you remove the shorter aged trees (around 150 years) the series starts to deviate from the HS and at the final stage has no HS. Do not these results confirm your initial concerns and indicate that the higher aged trees are better responders – and better responders appear to break the HS.
Also please provide your code.
Re: Kenneth Fritsch (#179),
I’ve posted my code twice and it must have got stuck in the spam filter. The important modifications are to replace the line,
(index3= Info$id[Info$n==3])
by
(index3<-c(“ListExcludedTreeIDs”)
and
tree=rbind(yamal[!temp,],russ035)
by
tree=yamal[!temp,]
The conjecture is that the shorter cores suppress the trend, potentially hiding a higher medieval index. This doesn’t seem to happen – no larger underlying trend becomes apparent during that period as the shorter cores are removed from the chronology.
Re: Tom P (#184),
The code that you’ve provided doesn’t generate your figures. It just seems to be an excerpt of my code except for an inelegant list of id numbers.
Re: Steve McIntyre (#228),
The posting left off the second part of my code for some reason, but the important changes are in: (#184).
The list of core IDs longer than a certain age has been extracted from the series by an external rank ordering by core age of the output produced by:
loc=”http://www.cru.uea.ac.uk/cru/people/melvin/PhilTrans2008/YamalADring.raw”
download.file(loc,”temp.dat”)
dpl.br = read.rwl(“temp.dat”)
I agree not the most elegant approach but at least straightforward to follow – I’ve only had a couple of hours experience with R.
I look forward to your comments.
Re: Kenneth Fritsch (#179),
I purposefully kept on removing cores until I could break the hockeystick, but even then it is only because there is a 9th century sudden warming that pops up despite there being little sign of it in the more complete datasets.
The last curve with 32 cores in total has little statistical validity, but I showed it for completeness. If you wish to cling to it for comfort, I must warn you it can bear very little weight.
Re: Tom P (#183),
Your analysis is goofy. My 183 was a crosspost with your 182.
Re: Tom P (#184),
Tom P it would appear that you are conjecturing backwards by assuming that the complete series with all the trees regardless of age is the correct one. Your original conjecture was that short aged trees are poorer responders. Thus we continue to see the same series shape until the higher aged trees are removed from the noise of the sharter ones (your conjecture not mine) and we see them stand alone.
You would have to show the statistical consequences of the reduced samples with the higher aged trees to argue as you have. It might be more the trending away from the HS with higher aged trees that should be tested (by your original conjecture – not mine) than any exact shape of the resulting series with the better responders (your conjecture – not mine).
Re: Kenneth Fritsch (#188),
Tom P, on the other hand, if you have changed your mind about the response of shorter aged trees, then I guess you would say what about the Schweingruber series?
Re: Kenneth Fritsch (#188), Re: Kenneth Fritsch (#189),
Tom P you have forgotten to comment on my posts noted above.
Re: Craig Loehle (#201),
Tom P, as I have noted before, you may have broken the Briffa 2002 HS. Say it is not so, i.e. please explain your graphs again with Craig L’s post in mind and some estimation of CLs for the trending away from the HS with older trees and the smaller sample. I think you can do it if you put your mind to it.
Remember a sensitivity study such as that applied by Steve M with the Schweingruber series is not saying that Schweingruber is correct, but just different. If the younger trees in that series are a problem than the younger ones in the Briffa series must be also and we must look at the oldest trees in Briffa for a sensitivity test. That is in effect what you have done. I do not know how robustly you have broken Briffa’s 2002 HS but broke it you may have.
The opening post states that YAD06 is the most influential tree in the world. That it’s contribution to one chronology or one climate reconstruction might be small in absolute terms does not mean it is not the most influential trees in the world, in relative terms. Other trees simply have less influence.
.
I’m not sure what Tom P is trying to prove, but I know that his “sensitivity test” ain’t gonna answer the question. If he would like to pinpoint a tree that is MORE influential, I’d love to see that.
Tom P: fess up. How old are you?
As someone who reads often and comments almost never, I think Tom P’s participation is a good thing, and he should not be discouraged. He is using Steve’s code and giving out his modifications. This is how the ‘team’ should have been interacting over the years. One comment on the plot with and without YAD06 (http://img515.imageshack.us/i/oyad06.pdf/). The fact that one tree makes such a big difference in the results means that the authors of the papers did a poor job of analysis. Looking at your plot, there is roughly a .25 (out of 2.5) difference height of the red vs black curves (along with other differences harder to discern). To me, this is an appallingly big effect from one tree. The YAD06 should either be tossed or modified to be less of an outlier. It should have been explicitly discussed and its handling made clear. The inclusion w/o discussion of such an outlying data point does not give one confidence in the scientific correctness of the paper, and should lead to a serious examination of its conclusions and its methods by the community beyond Steve M.
Re: anon (#194),
Not at all – the Yamal chronology shows an historically unprecedented recent increase in the tree index independent of any tree’s inclusion. Its exact value is not important. Far more important is that there is no evidence of any earlier value of the index that comes close.
Certainly not the tossed out – that would be cherrypicking! But even if YAD061 were thrown out, the conclusion of an unprecedented recent increase would be unchanged. That’s one reason why Briffa’s analysis is robust.
Re: Tom P (#196),
I have been watching your manipulations and I am amazed with how quickly you seem to have arrived at the level of making expert pronouncements on data analysis with such confidence.
Just for the record, what exactly do you think the reconstruction chronology represents and what relationship does it have with global and/or regional temperature or climate?
Re: romanm (#198),
Don’t be amazed by me, but you might be amazed as to how well Briffa’s analysis has held up to a completely unbiased sensitivity analysis. I didn’t know what the result might be when I suggested this approach. Indeed Steve McIntyre had asked if I was disturbed by the much larger fraction of shorter fossil cores that made up the earlier part of the record. I’d said it might indicate a problem.
But after this sensitivity analysis, it is very difficult to now see how there could have been any manipulation, or even inadvertant biasing of the records.
The issue at stake here is not any recent warming – that can be measured in the instrumental record. It is whether trees have responded in the past in a similar way as they are doing now. At Yumal the clear answer is that fossil record shows only a fraction of the growth spurt seen in the live cores of the twentieth century.
Yumal is only one area, but the strong signal there has attracted considerable attention. That is why Steve McIntyre has put the work into this audit in the first place.
Steve can of course reserve his option to criticise the methodology of RCS. However, as far as the core record itself is concerned, I’d like to hear if he has any grounds left to question its validity for such an analysis.
I think Briffa can be criticised for not releasing the data. But as someone who has had to work for months myself on preparing data for public release, I understand the burden it puts on any active research scientist. The resulting innuendo concerning this data has partly been a result of this delay, though.
Now the CRU archive has been given a thorough going over, I hope now Steve can help dispel any remaining doubts concerning the robustness of this dataset.
Re: Tom P (#204),
1. There are just too few cores to produce useful error bounds.
2. As Steve pointed out several posts ago, there are strong indications that the particular living trees in Yamel were selected (by whom unknown, but probably the Russians) from a larger or even much larger population, which would destroy the value of the remaining sample in a climate reconstruction.
3. Having other tree series replace the Yamel live trees destroys the HS.
4. Replacing the entire Yamel proxy with other equivalent proxies like the Polar Ural Update also eliminates the HS in multiproxy climate reconstructions.
Is that enough grounds?
Re: Dave Dardinger (#206),
Read http://www.treeringsociety.org/TRBTRR/TRRvol60_2_77-90.pdf . Once below five or six trees overlapping at a time there is a problem, hence the obvious breakdown in the statistics in my last plot of just 32 cores. More than that, though and the signal is consistent.
Are you referring to breaks in the numbering? There were breaks in every core series, including Schweingruber. There are quality control issues irrespective of any signal that will mean a core is not included in an analysis.
There is only one other tree series, Schweingruber, and that just injects noise through introducing much shorter cores than the CRU archive – read Craig Loehle’s reference just above.
You really are determined to throw away data you don’t like! You’re right though, if you throw away all the hockeysticks there will be none left. There is a name for this.
Re: Tom P (#208),
1.) I skimmed through the article and I don’t think it particularly encourages small populations. The small groupings in figure 7 were for 15+ trees and while they were able to identify low frequency trends, they mention that it’s if the trend is strong enough. I’d have to read closer to see if that could possibly apply to fewer trees and what I think is a low-amplitude trend.
2.) You were asking for “any grounds”. Until there’s an accounting for the cores they constitute a ground for doubt. And if such a small count of living trees is actually all there is, an explanation for why this is the case needs to be made.
3.) I don’t know that Schweingruber is the only other series of trees in the area. It was just one Steve is familiar with.
4.)
This has a very familiar ring to it. It’s the same lament Mann used against Steve and has the same flaw. Steve is not trying to create a new multiproxy climate reconstruction. He’s trying to show that ones ones at issue (most that don’t have Bristlecomes have Yamal) are dependent for their hockeystick shape on a proxy which consists of a small number of trees with no metadata, limited providence, and a liklihood of being pre-selected at some point.
Let me add an additional ground for doubt.
5) Looking at the graphs for the individual living trees, there doesn’t appear to be much coherence among them. There’s a general hockeystickishness in all of them, but where the uptick comes varies a lot and makes one suspicious that we’re seeing a growth-spurting mechanism which happened to each of these trees but at various points in time and not as a result of a climate change which just happened to affect small areas at any given time.
Re: Tom P (#204),
You don’t seem to appreciate the subtleties in my first sentence.
You didn’t answer my question. Nobody doubts that larger tree rings formed on some of the trees. The question was why these differences exist [Hint: There are MANY possible reasons].
Your single-minded insistence that the result is “robust” must mean (since you imply that you understand what that term means in statistics – it’s not the same as “I get the same result every time regardless of what I leave out of the data set”) that some parameter is actually being estimated robustly. So, in that context, I will repeat the question again:
Re: RomanM (#209),
If you want a discussion of the basis of dendroclimatology, please ask someone else. Steve McIntyre obviously thinks there is enough basis to warrant this analysis, and he knows much more about the subject than I do.
Re: Tom P (#211),
I don’t want a discussion.
You don’t seem to understand that the statement “the process is robust” is meaningless unless there are parameters being estimated by the procedure. I simply want to know what you think those parameters are (even roughly).
Re: Tom P (#211),
more, more! i can’t get enough.
Re: Tom P (#196),
Please correct me if I am missing something here. Your studies (http://www.climateaudit.org/?p=7241#comment-359086) look at the effect of removing various trees from plots of years (horizontal axis) vs some quantity (related linearly to temperature, I assume) on the vertical axis. You are using Steve’s code that replicated what was done in the Briffa paper (with some modifications).
I am basing my comments precisely on your figure http://img515.imageshack.us/i/oyad06.pdf
The inclusion/no inclusion of ONE tree makes a 10% difference in the final uptick (roughly .25 out of 2.5). In fact, I am being generous here–the effect of the one tree is better stated as a 20% effect–it brings the final uptick down about 1/5 of the way to the local peak of ~1.6 seen at year ~300 on your plot. When I see one tree making a big difference, I wonder what else makes a big difference. My next step might be to look at other outlier trees and ask the question:”How many trees need to be removed before the hockey stick vanishes?”. There could be outlier trees giving the opposite sensitivity (excluding them gives a more pronounced hockey stick). To maintain symmetry, you could ask: “How many trees do I remove to double the size of the hockey stick?” Having made the effort to understand R and Steve’s code, these questions should be easy to answer. You have looked at one appoach to removing trees–there are others–and the robustness of the results have yet to be demonstrated, IMO. When confronted with an outlier point like this tree, they needed to presented as such, to discuss how they handle it (include/exclude/reduce its significance) and give reasoning for any decision. What one should not do in a solid scientific work is leave include the outlier and not say anything about it. BTW: Certainly it can be left out if you decide that its response is not related to the temperature (which is what you are trying to infer). It is not cherry picking if one has some legitimate justification for exclusion. But, first and foremost, they need to be upfront and explain it in the paper.
Bender
What is fess up. I ask because in french fess is quite rude. Your phrase would mean arse up is that what you mean to say?
Re: stephen richards (#197), In colloquial American English it means to confess, “fess” being short for “confess” or “come clean”. Hope that helps.
Tom, you are misusing the word robust, but yes removing the one tree doesn’t change things too much, just the magnitude of recent ‘warming.’ The results are not robust whether Yad06 is included or excluded.
I found it. Tree response to climate depends on age (not stationary) in Larch.
Marco Carrer, Carlo Urbinati (2004) AGE-DEPENDENT TREE-RING GROWTH RESPONSES TO CLIMATE IN LARIX DECIDUA AND PINUS CEMBRA. Ecology: Vol. 85, No. 3, pp. 730-740.
Dendrochronology generally operates under the assumption that climate–growth relationships are age independent, once growth trends and/or disturbance pulses have been accounted for. However, several studies have demonstrated that tree physiology undergoes changes with age. This may cause growth-related climate signals to vary over time. Using chronology statistics and response functions, we tested the consistency of climate–growth responses in tree-ring series from Larix decidua and Pinus cembra trees of four age classes. Tree-ring statistics (mean sensitivity, standard deviation, correlation between trees, and first principal component) did not change significantly with age in P. cembra, whereas in L. decidua they appeared to be correlated with age classes. Response function analysis indicated that climate accounts for a high amount of variance in tree-ring widths in both species. The older the trees are, the higher the variance explained by climate, the significance of the models, and the percentage of trees with significant responses.
Age influence on climate sensitivity is likely to be non-monotonic. In L. decidua, the most important response function variables changed with age according to a twofold pattern: increasing for trees younger than 200 years and decreasing or constant for older trees. A similar pattern was observed in both species for the relationship between tree height and age. It is hypothesized that an endogenous parameter linked to hydraulic status becomes increasingly limiting as trees grow and age, inducing more stressful conditions and a higher climate sensitivity in older individuals.
The results of this study confirm that the climate signal is maximized in older trees, but also that a sampling procedure non-stratified by age (especially in multi-aged forests) could lead to biased mean chronologies due to the higher amount of noise present in younger trees. The issue requires more extensive research as there are important ecological implications both at small and large geographic scales. Predictive modeling of forest dynamics and paleo-climate reconstructions may be less robust if the age effect is not accounted for.
Re: Craig Loehle (#201),
Good reference – now what was I saying about the Schweingruber series?
Re: Craig Loehle (#201),
Good find. What this tells us is that the climate effect will be understated by the shorter-lived trees. Using longer lived trees to calibrate to the modern era results in apple-orange comparisons. In fact using trees of a similar age distribution to those in the past will result in responses which are more similar for the two times.
As well, the age “corrections” applied to the tree ring sequences could very well have a magnified effect on the RCS standardization process for these longer lived trees.
Re: RomanM (#213),
Sorry, it’s hard to read.
Snip away.
Re: Craig Loehle (#201), In the paper I cite above, they found that older LARCH trees are more responsive to climate. In the Briffa study, the trees in the recent period are all old trees. Most of the cross-dated subfossil wood are younger. This in itself will create a hockey stick, which is the amplified by the RCS standardization. When there is an artifact, it is necessary to factor it out or control for it, not just ignore it. If you did a cholesterol study and forgot to control for subject age, would that be “robust”?
Re: Craig Loehle (#218),
This is the essence of the question that I asked of TomP yesterday. He did not address the question?
Re: Craig Loehle (#218),
It’s not that short cores will necessarily create a hockey stick, it’s that they can suppress any trend present in the longer cores. It’s only by taking out the shorter cores that such a trend, if present, can be seen. Of course if too many cores are removed, the statistics become poor.
Hence when Steve McIntyre replaced the original CRU cores with the much shorter cores of the Schweingruber series, he was removing the cores that were more responsive to climate and therefore removing the recent trend. His sensitivity test was confirming exactly what your paper says – if you want to detect a climactic signal, try to use longer cores.
There are indeed plenty of shorter fossil cores in the CRU archive, so a reasonable question to ask is might these be obscuring a climactic signal in the earlier record. That is precisely what my sensitivity test set out to answer: by successively removing the shorter cores it allowed any climactic signal to emerge. None did.
Hence when the CRU archive is made more sensitive to a climate signal by removing the shorter cores, the hockey stick shape is preserved. That is why the Briffa analysis is robust.
Re: Tom P (#226),
None did? What do you call the peak that shows up around 700 – 800 on your “CRU>250” chart, a peak that is higher than the peak at the end of the original reconstruction?
Granted, you are down to so few cores that one can have little confidence in drawing any conclusions about that 700-800 peak — but then that same lack of confidence would apply to any conclusion you attempt to draw from this analysis, wouldn’t it?
Re: Tom P (#226),
Tom, Tom, Tom, by your hypothesis and Craig Loehle’s excerpt (They found that older LARCH trees are more responsive to climate)taking out the shorter cores leaves those most likely to respond to climate in the current time and past. That would explain the Briffa versus Schweingruberseries difference. You have not, nor anyone else here, examined the statistics of the smaller sample size that shows that as one goes back in time with higher aged trees the HS deterioates. So if I add Steve M’s sensitivity test to yours, on tree age, I get a distinct feeling that the Briffa 2002 HS is broken or at least now under serious doubt
Yes, Tom, yes and the basis of your sensitivity test with age – that needs to be followed to its limits, i.e. at what age would we expect the response to climate to kick in. It would appear that there might be some effective threshhold age.
Tom P, you are not only ignoring your own data you are contradicting your view of why the short cored Scweingruber series fails to respond to modern warming.
Tom P, based on what I see in your sensitivity test there is every indication that the longer/longest lived trees do in fact break the 2002 Briffa HS and also explain the age differing reactions to the modern warm period.
If you wanted to start with the premise that the Briffa analysis is robust (actually the tentative conclusion of the analysis) and then attempt to show that, I would suppose now that you would have to make a distinction between short lived trees in Schweingruber and those in Briffa and totally ignore the results using the longer/longest lived tree responses in the Briffa series.
Re: Craig Loehle (#202),
This is the excerpt from Craig Loehle which I think is one of the most important ones in this thread and needs repeating and further discussion.
Re: Kenneth Fritsch (#276),
Yes, it needs discussion, Ken. And BEFORE people start jumping to categorical conclusions about the likely consequence of age-biased samples. Such dicey propositions should be left to those willing to take on quantitative analysis and audit.
.
I see the gatekeepers at RC have taken to calling analysts at CA “frauditors”. How pleasant.
I mentioned in the first thread that perhaps the LP&_ were based on the scientific names, with _ should be B.
Tom you are hunting for graphs to support your conclusion without thinking through what you want to show. People mentioned early on that there were hockey sticks in the other trees as well. Everyone already knew the results you posted, just by looking at Steve’s graphs above. Yad04 is the only tree that doesn’t appear to have a hockey stick.
Tom P: runnin down a dream.
Nice Tom.
I don’t suppose that you might consider that the analysis of the trees is a separate issue from what they represent. Or is that way beyond a PhD.
RE TomP #174,
Thanks, Tom, for plotting Yamal with and without YAD06. It clearly makes a difference, but not as big as I would have thought given Steve’s initial graphs.
Re: Hu McCulloch (#222),
Hu,
I’ve looked at this quickly and it’s late. In a quick calculation, the kaufmanized Yamal chronology for the 1990s decade is reduced from 7 sigma to 5.61 sigma merely by removing YAD061. Yamal contributes about 39% of Kaufman’s 1990s value – thus tree YAD061 is singlehandedly contributing about 8% of Kaufman’s 1990s temperature.
I submit that this makes YAD061 the most influential tree in the world. Stay thirsty, my friends.
Re: Steve McIntyre (#223),
That’s a bit premature. Tom P has all night to come up with his next act. No guessing what it’s going to be.
Tom P and the Jesus stick.
Firstly, I’ll point out that I am just a scientist, not a statistical expert or dendroclimate expert, so this is my scientific take on events:
.
Statistical methods and sample sizes aside, the point that still seems to be being missed by some is that the reason for the reconstruction is *not* to simulate modern day temperatures, it is to determine *past* temperatures. The modern day reconstruction period is used as a tool to assess the amount of error and uncertainty that can be expected to be found in the “rest” of the samples. Therefore, if the methods used to choose the trees suitable for the temperature calibration period are different from those used to choose past samples, the amount of error and uncertainty in the past samples cannot be accurately determined, inserting unavoidable bias and rendering the reconstruction inconclusive at best. It is vitally important that no selection bias has taken place in this reconstruction because none of the calibration trees stretch back further than 1/10th of the total reconstruction, leaving no continuous record throughout the period (and that doesn’t even take into account arguments about whether trees lose temperature sensitivity).
.
Briffa has so far failed to answer this important question and the type of within sample testing by TomP will not shed any light on whether the reconstruction is robust or not. If the methods were bad, all the within sample sensitivity testing is simply garbage in, garbage out. Steve attempted to answer this question by supplimenting the data with other samples from nearby areas. This is not perfect, but, ironically, appears to have precidents in Briffa’s own work. This suggests that the selection criteria used by Briffa is not representative of the overall sample and, therefore, that the past reconstruction is tainted by calibration period sample bias.
Re: DaveJR (#232),
Precisely! Couldn’t have put it better myself.
It wasn’t until a few days ago that I had this lightbulb moment. Maybe I’m dumb. It hadn’t dawned on me that the deep-past reconstruction uses different trees to the recent-past reconstruction. Doh!
More constructively, I think this is a good example of why debate on this site and the use of R scripts is useful. I’m a trained scientist but not a statistician. Things that might seem obvious to some of you guys take a while to sink in with non-specialists. Following a civil online debate really brings out some of these issues.
Keep it constructive, guys. $20 in the tip jar.
Tom P
Your graphs do not show robustness, and here’s why.
As far as I can see, approximately half of the Briffa cores for the late 20C segment are 200-250 years old, with the rest being older.
Hence, your graphs where you remove cores based on age, do not actually remove many until the final 250 year cutoff you plot.
At that point, half the Briffa 10 selection disappears, and suddenly there’s a warming spike around the year 800AD which is 0.5 degrees warmer than the 20 C spike.
Also having one tree responsible for 15-20% of the warming spike can hardly be considered robust either.
Finally, the Briffa reconstruction in not robust because it’s late 20th C segment is only dependent on 10 cores, and can not be said to be statistically valid (particularly when the cores do not even correlate well with local temperatures). Removing cores REMOVES statistical validity and robustness, adding cores (which is what Steve did) adds statistical validity and robustness, and as Steve showed, removed the 20th C spike.
Sorry, David, I didn’t see your post pointing out the peak at 800AD or I wouldn’t have wasted bandwidth pointing it out a second time.
Re: Michael Smith (#234),
There’s nothing wrong with independent verification of observation through a crosspost.
Actually I just looked at the data. The Briff 10 are the following:
Tree Life Age
POR011 1803-1990 187
POR031 1683-1990 307
POR051 1610-1990 380
POR081 1626-1990 364
POR111 1580-1990 410
YAD041 1803-1990 187
YAD061 1803-1990 187
YAD071 1849-1990 149
YAD081 1875-1990 115
YAD121 1805-1990 185
Obviously none of these trees are removed from the “sensitivity test” between TomP’s 200 and 250 cutoffs because they’re all either under 200 or over 250 years old.
So the spike in the 800AD region is due to OTHER trees being removed by his algorithm, ie he’s not testing the sensitivity of Briffa 10, but the sensitivity of graph based on age of the cores. So i’m not really sure exactly what he’s trying to prove. There’s clearly no relation to temperature and changing the cutoff from 200 to 250 probably leaves a few freak trees covering the 800AD area (in much the same way as Briffas 10 freaks cover the late 20th C area).
It would be interesting to know whether after the 250 year old cut off, there are more or fewer cores covering 800AD than cover late 20C (ie more than 4 or less than 4).
Once you get down to looking at the data you realise that this whole “low number of tree core” type of analysis is goverened by some outliers and has basically no chance of having some kind of recoverable temperature signal (even assuming there were not loads of other factors affecting growth).
It’s simply stunning that this stuff is not discussed more openly, and taken more seriously, by the proxy reconstruction scientists themselves.
But Phil. will be along any second, I’m sure, to compare Tom Petard’s little stunt to Gerald Browning’s posting at RC, and to set the record straight. Cause even-handed Phil’s like that.
I hope this is not too far off-topic.
I find it very interesting the reaction at RC and Rabett to Tom p.’s rapid-reaction, shot-in-the-dark, zeroth-order cut at throwing mud on Steve’s analyses. Blind acceptance without having any, ( as in none, no, null, zilch, nada, not a single clue ) of Tom P.’s methods and only a rough idea of the data he used. The same situation very likely is happening at BCL, and tamino.
All calculations, even Tom P,’s require Verification prior to acceptance. Apparently the simple, straightforward concept of Verification has yet to sink in whenever the roughest, research-grade numbers produce the expected results.
Wonderful theatre. Highly addictive. Go Tom go!
Obviously another effect has to be put into these equations and that is that old trees were once young and during that phase did not hypothetically respond as sensitively to climate as they would when older. Another factor involved, given this hypothesis, is that an older tree that responds more sensitively to climate will do so only when the climate is actually changing, and therefore we have a correspondence effect to look at, i. e. the time periods when the tree age is most effective in sensing a climate change.
I do not know how much data could be applied, but a sensitivity test where tree data beyond certain threshold ages for individual trees were used in the reconstruction – in contrast to what I think Tom P used: tree age, but all of that tree’s data.
Perhaps the petard has petered out?
I presume every one knows of this document?
Click to access Briffa_HB_2008.pdf
discussing faults and improvements on RCS
Do the calculations use the correct version as Brifa used?
From McIntyres code:
# this calculates an RCS chronology for a tree ring data set: the script is one of my earlier scripts
# it works, but I’d write it more cleanly now.
# it permits a couple of different ‘methods’ – in this case, variant styles of RCS
#my preferred method is a nls fit
I’ll be commenting later on things but have family obligations today. In the meantime, I’ve placed Briffa et al 2001 online in CA/pdf/tree/ as it contains discusses age decomposition and has an interesting comment on RCS.
I didn’t pay much attention to Tom’s charts before. But when Gavin Schmidt claimed that he thought Tom’s charts showed that Briffa’s Yamal curves were robust I had to take a closer look.
Here is what Gavin wrote on RC in response to another commenter.
Gavin:
[Response: If you look at the fossil trees you will see similar things, and of course there is a systematic issue related to sampling living trees that might well be younger on average than the fossil ones. 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]
Here is what I wrote in response to Gavin:
“Actually, Gavin, Tom P’s data shows that the Briffa results are not robust. As Craig Loehle has posted, Larches seem to get more sensitive to temperature variation as they get older. Maybe that’s because their roots are deeper and they have less issues with water – who knows. But what we do know is this – the trees that Briffa has to represent the modern era are on average much older than the trees that Briffa has to represent older history. So when Tom begins to pull out trees, most of the trees that come out first are the trees in the older version. What we can see from his graphs is that the now older data begins to climb as compared to the contemporary data. Finally, when all of the younger trees have been removed from the older history we see Tom’s last chart. Now the trees from the earlier history are as old as the trees in the current era. The result shows that there is a peak around 800 that is larger than that around 2000. And there are peaks around 200 and 1450 that compete with 2000.”
Gavin hasn’t posted this yet. It will be interesting to see if he does.
Re: Tilo Reber (#247),
No it doesn’t – the noise increases but the average remains the same at around an index of 1 until the 20th century – the hockeystick is maintained.
If you look more carefully you can discern a slightly warmer period around 1100 (index of 1.2) and a cooler period from around 1500 to the 1800’s (index of 0.7). These would be the MWP and LIA.
The last chronology excluding all but the cores less than 250 years is statistically invalid. I posted it for completeness to show that the only way to break the hockeystick was to take out so many cores that the all we are seeing for most of the record is noise.
I was concerned that there might be a few people with a weak grasp of statistics might want to claim sudden very warm periods of a few decades scattered around the last two thousand years on the basis of this invalid chronology – at least you are not alone in making this error here.
Re: MikeN (#248),
No, I wrote on RealClimate:
I received only encouragement.
Re: Tom P (#250),
That it is noise, says you Tom P – and without any corroborating statistics. It is not something you can simply posit.
Consider the following when talking about noise and sample sizes:
We assume that these trees show reasonable climate sensitivity after a thresh hold age. Let us assume that age to be 125 years. Now when comparing the entire Briffa series to that where the lower aged trees are removed, the difference should not be measured in number of tree samples, but in the tree years that are sensitive to climate change. I’ll need to do the coding to obtain a comparison, but I know that a sensitive-tree-years comparison will not show the difference that tree samples would.
If this hypothesis is correct, we have another problem that would cloud the interpretation of the multiple age tree proxy of Briffa’s. We need to relate the sensitive tree years to the various periods of the proxy. Determining the variations in sensitive tree years over the period of the proxy would be critical to determining how well the various time periods were covered with reasonably expected sensitive tree responses.
Re: Tom P (#250),
I was also concerned that there might be at least one, so this comment caused me to finally go back and take a look at Tom’s “test”.
Look at the “contributions” made by Tom P. He came in pretty much cold to the entire process. No problem with that , we all start at some time. However, all of a sudden he was making blanket proclamations about the entire situation:
No, the selection of proxies that were quite different by their time span from the earlier proxies were not a problem. I can’t think of anyone with statistical experience who would make such a pronouncement without further information and analysis to back up their contention.
Steve’s substitution as a sensitivity test did what it was intended to do. It showed that using another legitimate sample removed the sharp rise in recent times – clearly a rise that was due only to the live proxies which were selected in some determined fashion. Without further examination of the selection process used, there is a question as to the result of using it.
Tom’s criticism was that Steve replaced the long lived proxies with shorter ones. He proclaimed that this was simply replacing “signal” with “noise”. Maybe somebody should inform the dendros about that so that they wouldn’t waste their time collecting useless information. On what basis would such a proclamation be made? Presumably on the paper which supposedly modeled the behavior of artificially generated autoregressive generic proxies. Counts which should serve as general guidelines were being interpreted as gospel truth.
So Tom came up with his “test”. What did that consist of? Adapting a program previously written by Steve with a very minor alteration, he removed proxies by size in stages. The hockey stick (flat until recntly and then a sharp rise) remained until the ones selected for recent times started leaving . What a surprise. What did this show? Well, supposedly it was to see if as the “noise” of short term proxies was removed, a signal would emerge. It didn’t. Well, not really (until the end) because the hockey stick (which is unitless) is not fomed by the short proxies depressing the early record, but by the late extremes producing a large uptick. However, it seems that his analysis was acclaimed and praised by the “scientists” and “statisticians” doing independent evaluation at RC so it is OK.
What Tom P. seems to have ignored in his defense of Briffa’s reconstruction is the value of such a large recent uptick. I asked him a question earlier about what he thought the reconstruction was estimating. He did not seem to understand that because this reconstruction was being used constantly in paleoclimate temperature analyses, the answer was clearly that it was a temperature (unitless, but the shape was important because the fitting was usually linear). So Tom, what temperature is being estimated and how good is the fit by this “robust” result?
Bender, I think you have it right.
Re: romanm (#259),
RomanM,
It was clear from the get-go that this guy was a nutter. Now he is over at RC claiming vindication of the “robustness of Briffa’s data”. Let’s review. No matter what he removes from the Briffa sample, a hockey stick remains. Umm. This proves that Briffa’s sample is internally consistent and thus massively non-representative of any sample that includes the Schweingruber data. Tom P doesn’t udnerstand that it’s not what you take away from the Briffa sample that tests its “robustness” vis a vis representativeness, it’s what you add to it! Leave it to Tom P to confuse addition and subtraction.
.
And it’s been that kind of a week, folks. I’m getting a raw lorax.
Re: romanm (#257),
RomanM, my take on Tom P’s sensitivity test is that it leads either to the opposite of his “Briffa is robust” proclamation or what he proclaims means nothing and we are back to the Schweinberger series being different than Briffa’s.
Remember that in a Craig Loehle post above he had an excerpt from a paper stating that older larch trees are more sensitive to climate changes and young ones relatively insensitive. If one follows that conjecture or hypothesis (I am not a dendro) then, with some exceptions, the Schweingruber series, the CRU data in the modern era and the change in shape of the Yamal series when younger trees are removed starts to fit.
Let us wait until tomorrow when Steve M promises to give more details on age dependence of larches as stated by the experts. I’ll be baby sitting my grandkids all week, but should have access to this blog and thread. I am looking forward to how this all turns out.
I agree that it is a little difficult to comprehend that dendros would use samples that are merely noise in a proxy series.
What I do like about these developments is that it gets the discussion deeper into the science – and it is a lot more fun than replying to Lorax.
Re: Tom P (#249),
Tom, I have three questions for you about your work.
First, you say:
“The last chronology excluding all but the cores less than 250 years is statistically invalid. I posted it for completeness to show that the only way to break the hockeystick was to take out so many cores that the all we are seeing for most of the record is noise.”
And in an earlier posting where you first show your results you say this about the trees that are over 250 years old.
“Now there are just 32 cores left with an average age of 303 years, or just four cores at any one time.”
So if the small number of cores produce a statistically invalid result, then why doesn’t Briffa have statistically invalid results when the last years of his chronology have only 5 trees.
My next question is about tree overlap. Let’s use your sample where trees are over 200 years old, since you seem to believe that is still statistically valid. Take the 50 year time slice going from 1940 to 1990. This time slice is made up of live trees that reached their maximum age within that time slice. Another words, all the data that you are getting for that period comes from tree rings that are old tree rings greater than 150 years. Now, take a time slice from 940 to 990. Your selection method insures that all of those trees lived to be over 200, but it does not insure that they were over 150 in that time slice. Trees in that time slice could have been in their first 50 years of life – or their second – or their third. So while your method insures that all the tree rings used in the second half of the twentieth century are over 150 years old, the tree rings used will, on average, be much younger for all of the earlier periods.
My third question has to do with this. The sorting based on age is valid because, as Craig Lohele has found, Larches become more sensitive to climate as they get older. But, as we all know, elements such as water, nutrition, sunlight, etc. still play a part in tree growth as well. Briffa has stated that chronologies that do not match the surface temperature record of the region are thrown out. But that only serves to bias the chronologies to the ones that have trees that are sensitive to climate in the 20th century. Earlier trees in the same chronology receive no selection bias benefit. It seems to me that this would give you a better representation of 20th century climate, but it would make comparisons to warming in earlier periods invalid.
Tom, you posted on RealClimate this sensitivity test, asking for comments. Did anyone respond with comments on whether the then proposed test was valid?
If larch reproduce asexually as I stated in #168, and if all of these 10 trees started their lives as branches on old mature trees which then fell over in the boggy soil, and if they then began their lives as independently rooted saplings, would they not then reflect muted growth in their early rings when they were still branches on old trees? And a growth spurt when they put down their own roots and were fed by the trunk of their rotting parent tree? In effect showing the growth of a mature tree and a sapling on the same set of rings? Or is this impossible?
TomP
Hahah… talk about sly! You removed the 250 year cutoff from your rotating gif, which showed a higher signal in 800AD than the end of the 20th C.
“No it doesn’t – the noise increases but the average remains the same at around an index of 1 until the 20th century – the hockeystick is maintained.”
How do you know what is noise and what isn’t ? Do you have a “real temperature” history over the past 2000 years you are comparing it to ?
Every single cutoff performed DECREASES the “unprecedentedness”, and therefore the hockeystickness, of the late 20th C spike, until with a cutoff of 250, it’s significantly lower than the 800AD spike. Given that the whole purpose of cutting short lived cores out of the survey was to reduce “noise”, what does that say about the robustness ?
So far we have both fewer cores reducing significance of the 20th C spike (which does not correlate with 20th C temperatures there anyway), and more cores resulting in the 20th C spike disappearing altogether. Seems yet again a climate study is perched on the balance point of “maximum 20th C spike compared to historical signals” and they didn’t even have to pick PC’s to do it, just the particular cores. Amazing.
Even a homeopath would blush.
Tom P.
“No it doesn’t – the noise increases but the average remains the same at around an index of 1 until the 20th century – the hockeystick is maintained.”
No, it’s not. The blade is maintained. You are clearly getting more variation around the shaft that you were before. And you are reaching higher points around the shaft than you were before. Remember, the AGW motto is that nothing was happening until the 20th century. Go look at Al Gore’s chart.
“The last chronology excluding all but the cores less than 250 years is statistically invalid.”
Why?
“I posted it for completeness to show that the only way to break the hockeystick was to take out so many cores that the all we are seeing for most of the record is noise. ”
And having only 5 cores at the end is not statistically invalid. Are you saying that you don’t have enough old cores for the earlier periods? And if you don’t, then do you have a set that is good enough to compare to he modern era.
“the noise increases but the average remains the same at around an index of 1 until the 20th century – the hockeystick is maintained.”
The 20th century is where the chronologies were sorted to remove those that didn’t match the surface temperature. So where is the suprise in that. Do you have all 2000 cores, or just Briffa’s subset.
The 20th century is where the chronologies were sorted to remove those that didn’t match the surface temperature.
Any proof of that, Tilo? Stephen McIntyre has explicitly stated that he does not believe that to be the case. You know better, do you?
Re: Simon Evans (#255),
Ummm, No.
Steve has explicitly said that he see no sign that Briffa deliberately cherry picked his data. That is completely different than not thinking he did something statistically ill advised which had a side effect of essentially cherry picking.
Tilo’s point is made by Briffa himself as Jeff Id has pointed out earlier:
There you go. Now the real test of your integrity is whether you are still appalled now that you know that Briffa not only screened by temperature, but thought that it was a good idea, or is your tune going to change so that instead of it being a slur by Tilo, this is all just fine and dandy and Briffa “Did the right thing” ?
Re: Artifex (#257),
You have made the very mistake that Steve asked us to be careful not to make, and which I pointed out very early on: core samples vs. chronoligeis. Simon Evans is correct. There is no evidence that sorting happens at the level of cores in a chronology. This is what Briffa stated in his reply, and there is no evidence to the contrary. Thank you, Simon, for pointing out the error.
Re: bender (#258),
Bender I obviously need to reread. Link ?
Re: Artifex (#261),
Mine was #80 in “Briffa’s reply”. Steve picked up on it a little later in the same thread. I don’t mean to be harsh, Artifex. Let’s just make sure we’re correct about the things we say before we say them. Simon is not aften wrong.
Artifex,
Tilo’s point is made by Briffa himself as Jeff Id has pointed out earlier
You quote Jeff Id quoting from Osborn and Briffa’s SOM relating to their 2006 paper (“The Spatial Extent of 20th Century Warmth in the Context of the Past 1200 Years”), where they are explicitly explaining the modifications they have made for that specific paper (Yamal was, of course, featured in Briffa’s 2000 paper).
So what was your point about integrity?
But really, Artifex, never mind. It’s possible that I misunderstood Tilo’s post, and I think we are now clear enough as to what is being said.
–It’s possible that I misunderstood Tilo’s post, and I think we are now clear enough as to what is being said.–
Yeah, I’m a little perplexed by bender’s comment to Artifex in #258.
Tilo seemed to clearly be saying that chronologies were sorted not cores which is what I thought Steve M and bender were both saying earlier, so I didn’t get Simon’s response to Tilo. And bender’s assertion that Simon was correct confused me a bit more.
On top of all that I didn’t see where Artifex implied screening of cores; only screening by agreement with temp, which is what Briffa did, correct? Only he did so by cherry picking certain chronologies not individual cores within chronologies or am I not getting this?
What am I not understanding?
I do think Artifex’s comments regarding Simon’s integrity were over the top and snippable.
Re: Brian B (#264),
Good work, Brian B. Seems I did not go back to Tilo’s original comment. I assumed Simon’s interpretation was correct, because normally he’s quite careful. His read was incorrect. It was Simon that did the conflation and that misled Artifex. My apologies to Artifex. Everyone’s thanks to Brian B.
.
Now, that being said, Simon may have a different bone to pick with Tilo. Briffa has done some chronology substitution in the past. That’s not exactly cherry-picking in the sense of trolling through dozens of possibilities looking for a fit. Similar, but not as egregious.
Re: bender (#266),
Has Steve M written on Briffa’s now-documented Athabaska substitution?
Brian B,
I’ve reread, and I think I misread. I was actually confused by the latter part of Tilo’s comment which I didn’t quote – “Do you have all 2000 cores, or just Briffa’s subset.” Anyway, this is probably a confusion of my making, so let’s move on 🙂
Thanks for the explanation, Simon.
Nice to be confused by something other than my own brain for a change.
See how nicely everything goes when we search for truth and work from facts? Good work everyone.
I really enjoyed Tom P’s suggestion that those who are questioning him (I’m thinking RomanM in particular) have “a weak grasp of statistics”.
Anyone know a statistics teacher who can straighten us out? 😉
Simon:
“Any proof of that, Tilo? Stephen McIntyre has explicitly stated that he does not believe that to be the case. You know better, do you?”
Simon, Steve said that he does not think that Briffa picks individual trees. He did agree that he picks chronologies. Briffa himself has said that he picked chronologies according to how they mached the surface temperature record. He said it in an earlier paper and he said it again in his response to Steve’s work. But it seems obvious to me that this is going to introduce a huge bias in the relative warming that you get between the 20th century and earlier periods. It seems to me that using such a selection method would basically invalidate all tree ring chronologies that use the method. I tried to explain it as simply as I was able to do so here:
http://reallyrealclimate.blogspot.com/2009/10/tree-ring-chronology-selection-problem.html
Re: Tilo Reber (#275),
My apologies for messing that exchange up, Tilo. I trusted Simon to get it right, but didn’t verify. Tsk.
.
But you must be careful. The magnitude of the warm-trend bias introduced by Briffa’s biased substitutions will depend on how frequently and egregiously he does this. There are not many documented cases of this. The push hsould be on to compile a list of all such substitutions, and not just by Briffa.
Re: Tilo Reber (#274),
Tilo,
Agreed, that’s what dendros do!
MrPete, (and Craig Loehle)
Consider carefully that the variance used in crodssdating comes from a different part of the spectrum than the variance used in climate reconstruction. The former is interannual to sub-decadal. The latter is secular. To suggest that selecting samples for crossdating MUST lead to a bias vis a vis climate signal is to assume that the variance in the two parts of the spectrum is non-independent. That leap in logic is not a small one.
Kenneth:
“This is the excerpt from Craig Loehle which I think is one of the most important ones in this thread and needs repeating and further discussion.”
Definitely important Kenneth. So it occurs to me that a selection of chronologies that reflect 20th century warming then inadvertently also becomes a selection for older trees. But again, this selection only effects the trees in the 20th century. Earlier times do not reflect the same selection, and so earlier times will have a mix of older and younger trees. And those times will have less climate sensitivity.
Re: Tilo Reber (#280),
Tilo, please look at Tom P charts at
Re: Tom P (#171),
You will see a significant departure of the Briffa series away from the HS when only trees older than 200 years are used and no HS when trees 250 years and older are used. See my post below for more on this.
Re: Kenneth Fritsch (#284),
Kenneth, I’m with you on that. In fact that is what I was trying to point out to Tom P in #246 when I said,
“So when Tom begins to pull out trees, most of the trees that come out first are the trees in the older version. What we can see from his graphs is that the now older data begins to climb as compared to the contemporary data. Finally, when all of the younger trees have been removed from the older history we see Tom’s last chart.”
I also like your bins. Looks like there should be plenty of trees left > 250 years. But if sparsity of numbers is a statistical problem, why isn’t it one when there are only 5 trees left in the last years of Briffa’s chronology?
It seems to me that there is another problem that I’m calling the overlap problem. I’m going to describe it below in another response to Tom P. Tell me what you think?
Bender:
“But you must be careful. The magnitude of the warm-trend bias introduced by Briffa’s biased substitutions will depend on how frequently and egregiously he does this.”
Of course you are right. I was trying to get my hands around that issue – even in rough terms. I don’t remember where I picked this up now, but it’s from Briffa:
“Description: Palaeoclimate reconstructions extend our knowledge of how climate varied in times before expansive networks of measuring instruments became available. These reconstructions are founded on an understanding of theoretical and statistically-derived associations acquired by comparing the parallel behaviour of palaeoclimate proxies and measurements of varying climate. Inferences about variations in past climate, based on this understanding, necessarily assume that the associations we observe now hold true throughout the period for which reconstructions are made. This is the essence of the uniformitarian principle. In some northern areas of the world, recent observations of tree growth and measured temperature trends appear to have diverged in recent decades, the so called “divergence” phenomenon. There has been much speculation, and numerous theories proposed, to explain why the previous temperature sensitivity of tree growth in these areas is apparently breaking down. The existence of divergence casts doubt on the uniformitarian assumption that underpins a number of important tree-ring based (dendroclimatic) reconstructions. It suggests that the degree of warmth in certain periods in the past, particularly in medieval times, may be underestimated or at least subject to greater uncertainty than is currently accepted. The lack of a clear overview of this phenomenon and the lack of a generally accepted cause had led some to challenge the current scientific consensus, represented in the 2007 report of the IPCC on the likely unprecedented nature of late 20th century average hemispheric warmth when viewed in the context of proxy evidence (mostly from trees) for the last 1300 years. This project will seek to systematically reassess and quantify the evidence for divergence in many tree-ring data sets around the Northern Hemisphere. It will establish a much clearer understanding of the nature of the divergence phenomenon, characterising the spatial patterns and temporal evolution. Based on recent published and unpublished work by the proposers, it has become apparent that foremost amongst the possible explanations is the need to account for systematic bias potentially inherent in the methods used to build many tree-ring chronologies including many that are believed to exhibit this phenomenon.”
This divergence problem that Briffa admits to makes me think that there is no small number of chronologies that need to be thrown away. The problems with the polar Urals as well as the Osborn/Briffa quote given again by Artifex further makes me think that chronology rejection may be somewhat common. But the truth is, I don’t know – and I certainly would like to.
Bender, let me say one more thing. Any random sampling of trees that you use to create a chronology will probably contain some sensitive trees and some less sensitive trees. Since you need to fit to a fairly steep temperature line in the 20th century, you probably need for your chronology to contain a very high percentage of sensitive trees. Given the Briffa description, those high percentages might be hard to come by. If that is the case, there might be a lot of chronologies that are thrown out. It may also explain Briffa’s small sample size. It would be easier to get a sample of 10 that is uncontaminated by non-sensitive trees than a sample of 30 that is uncontaminated.
Re: Tilo Reber (#282),
I understand. But suspicion and proof are two different things.
Tom P plays down the trend away from the Briffa HS when the series is limited to older and older trees and calls it, without futher detail, noise that presumably arises from smaller numbers of trees that are older.
I did a simple calculation for the Briffa trees and then in tree age bins I counted the number of trees, tree years and finally useful tree years (by conjecturing that tree ring data in trees younger than 100 years is insensitive to climate change):
Bin = >250 years; Number = 27: Years = 8400; Useful Years = 5700
Bin = >200150100<150 years; Number = 78: Years = 9800; Useful Years = 2000
Bin = <100 years; Number = 59: Years = 4100; Useful Years = 0
Re: Kenneth Fritsch (#286),
WordPress did its thing on greater and lesser than symbols so here are my bins again.
The overlap periods for the various aged trees need to be put on a pretty graph. It will have to wait for me to do it for a few days.
Greater than=gt and lesser than = ls
Bin = gt250 years; Number = 27: Years = 8400; Useful Years = 5700
Bin = ls250gt200; Number = 34: Years = 9900; Useful Years = 6500
Bin = ls200gt150: Number = 54: Years = 8500; Useful Years = 3100
Bin = ls150gt100; Number = 78: Years = 9800; Useful Years = 2000
Bin = ls100 years; Number = 59: Years = 4100; Useful Years = 0
The Carrer and Urbinati paper Craig L. quoted refers to “climate sensitivity” not temperature sensitivity. Did they define climate sensitivity as temp sensitivity and if not, doesn’t this lead right back to the recurring question of what exactly in the climate the larches are sensitive to; CO2, precip, seasonality of precip, temp, etc?
Briffa selected for correlation to temp but the trees didn’t necessarily do so. They may have responded to something caused completely, partially or not at all by the correlated temperatures. And the smaller the number of cores and the more localized their area of collection the more likelihood of a spurious corellation, no?
Or am I again missing something, again?
Re: Brian B (#294),
Did you read the whole paper?
re #295.
Didn’t have a chance, bender.
Just got back online quickly this morning before leaving on a trip in about 30 minutes and the question occurred to me. I did a quick google to try and find the paper but these guys seem pretty prolific so there were pages of hits to search through and they seem to be especially prolific in Italian which I don’t read too well. This thread will probably be dead in a week when I get back so that I’d at least ask it.
If by climate sensitivity they mean temp then my question has been mooted in any event.
The point about spurious correllations I’m assuming is one of the reasons for the concern regarding such small samples.
Anyway, got to go.
Re: Brian B (#297),
The answer to your question is almost certainly contained in the paper.
Still no answers to my questions to Tom at 289 or at RC. His ability to ignore reality when it doesn’t suit his agenda is simply breathtaking.
Re: Tilo Reber (#299),
Breathtaking? But he’s a nobody, so it’s to be expected. To me what is breath-taking is the vacuum of silence coming from Gavin Schmidt’s larynx on the topic of selective use of data. It’s sucking my lungs out. And Eli’s silence on enforcibility of IP rights on climate data and models. That silence is the sound of fear.
Bender:
I’m feeling a little sorry for Gavin. He’s left defending a topic that he knows very little about. Mann is suppose to be part of the team. Where is he? Briffa’s response was short, weak and incomplete. Why isn’t he helping more? Of course some of my pity for Gavin is dispersed by the fact that he is blocking questions from the opposition that he finds embarrassing or difficult. Other than Gavin, Tom P. is the best that they have over there. And that says a lot.
OT: The most influential mammoth in the world? Also from Yamal…
http://news.bbc.co.uk/1/hi/6284214.stm
In the information posted at NCDC for Hantemirov and Shiyatov 2002 regarding their 4000 year chronology, the numbers in the “samples” column on the right for 1995 and 1996 (the last two years of data) are both “5.”
In the data posted for Briffa at CRU, “YAD061” apparently ends in 1995, leaving only 4 samples in 1996 (YAD041, YAD121, YAD071, YAD081).
Here are the last two lines in the CRU data for YAD061. Am I construing them correctly as showing the end in 1995?
“YAD061 1980 1150 1020 1920 2120 2500 1890 1570 1670 1710 2040
YAD061 1990 1610 1580 1710 2870 1800 1720 -9999”
Was YAD061 not one of the samples used in H&S 2002, or is there merely a mistake by me or a typographical error in one place or the other? In other words, did H&S 2002 use some tree other than YAD061 which had a core sample ending in 1996?
Re: Micajah (#303),
Yes, that series ends in 1995. -9999 is the series endmarker, slotted in 1996.
Re: Micajah (#303),
http://www.ncdc.noaa.gov/paleo/text/tucson-format.txt
ftp://ftp.ncdc.noaa.gov/pub/data/paleo/treering/treeinfo.txt
The latter file gives a quick overview. In practice there are several ‘no data’ values and some really interesting deviations– sometimes several uses in the same file.
Thanks. I suppose it’s just a little thing to wonder about with this tree and the data sets used by Briffa.
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 the graphs of the modified RCS series below with each including the archived Briffa chronology for reference. Additionally I have posted 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:
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 the graphs of the modified RCS series below with each including the archived Briffa chronology for reference. Additionally I have posted 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:
.
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:
.
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 with greater than sign replaced with (gt)*:
.
Code for tree ring counts by year graphs:
.
[Jean S: This was caught in Spam filter. There seemed to be several copies of this, so I didn’t release them all. Please repost if there is something essential missing from this one.]
Jeff ID noticed that my series graphs were prematurely cut off for the latest years. In order to correct that I did the following:
.
From Steve M’s RCSchronology function I changed limit for min to max as in:
series124″,xlab=”Years”,ylab=”Tree Ring Width”,type=”l”)
The series graphs are presented below and the observations/conclusions are the same as those from my previous post.
Re: Kenneth Fritsch (#308), Hantemirov and Shiyatov 2002 selected subfossil samples using some criteria that sought to pick those with higher variability as well as length. Is it practical to incorporate variability into your selections too?
Re: Micajah (#309),
The question becomes, I think, do some some tree ring ages better respond to the climate. The Craig Loehle excerpt from Post #202 above deals with age and larches (the trees Briffa used in Yamal). Until someone offers to discuss this literature I think it bears repeating:
I do not see where diversity necessarily works here and it implies that offsets are required for some nonclimatic influences that are age dependent. One could argue that as more younger aged tree rings are removed from the series that the smaller resulting sample size yields more uncertain results, but the trending with older tree rings (and not older trees that also were once young) appears to be showing something in my view.
I should also rewrite what I changed to get the full individual series to be included in the plots:
.
The greater than symbol got me again and thus I will use gt to represent that symbol or equal sign where appropriate.
.
From Steve M’s RCS chronology function I changed limit for start=min(tree$years to end=max(tree$years) as in:
series =ts(series, end=max(tree$year))
.
I then used this code to plot the series separately (for tree rings older than 124 years in this case):
.
ts.plot(chron.yamal$series, main=”Emulated RCS Chrononolgy for TR Ages gt 124″,xlab=”Years”,ylab=”Tree Ring Width”,type=”l”)
Reblogged this on Climate Ponderings.
20 Trackbacks
[…] YAD06 – the Most Influential Tree in the World […]
[…] Steve McIntyre, a Canadian former minerals prospector and climate sceptic who has analysed the data, suggests that one tree, known as YAD06, could be “the most influential tree in the world“. […]
[…] climate sceptic who has analysed the data, suggests that one tree alone, known as YAD06, could be "the most influential tree in the world".In the hacked emails from the Climatic Research Unit at the University of East Anglia, one word […]
[…] read: “We recognize that a few treemometers in Yamal, and particularly tree YAD061, aren’t really representative of the global climate for the past millennium and therefore a […]
[…] base their argument on one tree, […]
[…] base an argument on one tree, […]
[…] YAD06 – the Most Influential Tree in the World, 30 September 2009 […]
[…] features prominently, again, and includes his old tree rings as well as Keith Briffa’s hockey stick shaped Yamal tree rings, especially YAD06: the most influential tree in the […]
[…] and climate sceptic who has analysed the data, suggests that one tree, known as YAD06, could be "the most influential tree in the world".In the hacked emails from the Climatic Research Unit at UEA, one word looms large: Yamal. The […]
[…] https://climateaudit.org/2009/09/30/yamal-the-forest-and-the-trees/ […]
[…] for this story -YAD06 – the Most Influential Tree in the World by Steve McIntyre Sept. 30, 2009 https://climateaudit.org/2009/09/30/yamal-the-forest-and-the-trees/Welkom in de EUSSR! (ik ga dit voortaan standaard onder al mijn artikelen zetten)/*//>*/Aanverwante […]
[…] 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 […]
[…] one greatly exaggerates the importance of a single proxy record – but who would do that? (Oh yes: YAD06 – the most important tree in the world, The global warming industry is based on one MASSIVE lie etc.). Note that premature public access […]
[…] represent global temperature, why wouldn’t he dress up as that famous Yamal tree, YAD06, the Most influential Tree in the World? I’d pay to see […]
[…] noted though that tree ring paleoclimatology is an inexact science, and as we’ve seen, even a single tree can go a long way to distorting the output. On the plus side, it is good to see that this paper defines and corrects biases present in the MXD […]
[…] ultimate example of the use of outliers in climate science has to be the Yamal hockey stick, where one tree came to represent the entire global temperature (see also: […]
[…] ultimate example of the use of outliers in climate science has to be the Yamal hockey stick, where one tree came to represent the entire global temperature (see also: […]
[…] ultimate example of the use of outliers in climate science has to be the Yamal hockey stick, where one tree came to represent the entire global temperature (see also: […]
[…] YAD06 – the Most Influential Tree in the World, 30 September 2009 […]
[…] McIntyre’s objections to the Yamal tree (“the most important tree in the world”) in the global warming hockey stick are being vindicated as the new version of the data series […]