As you can see from the plot of the Juckes’ proxies, the Yang composite is a very important contributor to the 20th century blade. The Yang Composite is a mainstay of recent Hockey Team reconstructions – its use in Team reconstructions began in Mann and Jones 2003 and was then “randomly selected” into Osborn and Briffa 2006, Hegerl et al 2006 and now Juckes et al 2006.
The two series in the Yang composite that drive its 20th century blade are versions of two Thompson series from Dunde and Guliya. The Juckes “evaluation” did not evaluate the data versions. I’ve posted up on Dunde before, pointing out inconsistencies between the different versions (the Yang version has values only at 50 -year intervals!) Today I checked out the Guliya version. In addition to the Yang version (emailed to me by Dr Yang a couple of years ago), Thompson archived the version used in his Climatic Change 2004 article (only after my intervention) and has just archived a new version in connection with his PNAS 2006 article. The three versions are plotted below – the first 2 are at 10-year resolution; the last one at 5-year resolution:
Figure 1. Three versions of Thompson’s Guliya dO18 ice core series.
There is obviously a strong visual inconsistency between the different versions. This is reflected in the correlations. The correlation between the Yang version and the Climatic Change version is only 0.06. Remarkably, the correlation between the PNAS version (converted to 10-year intervals) and the Climatic Change version from only 2 years ago is -0.005. The data was collected nearly 15 years ago.
I’ve written to Science repeatedly asking them to require Thompson to archive and reconcile these conflicting results, but have been blown off completely.
Should Juckes have evaluated the conflicting versions of Thompson data used in the Yang composite? I think so.
Climate Audit 
57 Comments
You know, it’s hard not to despair at the quality of work in this field when you look at something like this graph with 3 different inconsistent versions of the Guliya series which have ended up in the literature.
There’s obviously a simple, rational explanation why these data series are so very unlike each other. I’m just trying to think of one…
It is surprising to me that these series could drive a twentieth century blade in the HS. How did they manage to do that?
Re #3
Because you are looking at only a sliver of the multiproxy dataset, whereas (in an ideal situation) the blade comes from the common signal shared among all proxies. i.e. PCA picks out commonalities that the human pea-brain can not see.
As a PC, the HS blade is supposedly a signal that is common to many different proxies. IOW you don’t need a HS shape in any one proxy in order to generate a HS shaped PC1. Indeed, that would be damning to the PCA if any one local proxy dominated the “global” PC1*. Anyone who does not understand this does not understand what PCA does.
*Sounding familiar?
This board is attracting a lot of traffic, and not all who come here post. You can be sure that the poor quality of the Hockey Team work, which they appear to be blithely continuing to produce, is increasingly evident to thousands of scientists and others observing. I would think that the reputation of Mann et al, Dr Juckes, and the cheerleaders (you know who you are) blindly supporting the team is in a perilous nose-dive!
It is amazing that supposedly reputable journals (such as Nature, Science etc) publish this c..p!
#3. the blue series is the one used in the Yang composite and it is more HS than the subsequent ones.
bender – this particular series does not enter into a PC series. I’m not saying anything new here for bender, but just distinguishing a nuance: pick-and-average does something very similar to Mannian PC. If you pick and orient the 10 most HS-shaped shaped series out of say 35 or 50 red noise series with persistence, scale and average them; you get a HS reconstruction. The common trend of the picking criterion reinforces and the noise in the shaft cancels out.
#5. Clearly only a fraction of visitors post. However, the visitors must like the discussion since they keep coming back – nearly 50,000 hits/day in November. It’s amazing.
BTW if Juckes wishes to contribute a thread to clarify his paper and even to put his own spin on things, I’ll post it up without changing a comma even where he describes us as “idiots”, our criticisms as “drivel” and other similar genial comments that characterize his recent contributions.
Re#6
I noticed that the blue series was a bit more of a HS, but it is not as dramatic as I recall the bristlecones being. It is true that its trend in the 20th century would help, but by itself I don’t think this one series would give a very dramatic HS. Intuitively, I am thinking of the final “thermometer” as just a weighted average of proxy series. The more HS ones get a big weight when the composite is calibrated to the 20th century ground station records (with as much UHI effect in them as possible!). Still, whatever weight the blue series got, by itself I think it would tend to give higher 15th century temperatures (for example) than would other series you have discussed.
Actually, thinking about this just gave me another idea. The other way you could get a good HS is to have two series each with “mild” positive 20th century trends but in centuries prior to the 20th one of them is below mean when this blue series is above mean and vice versa. Then the negative correlations would make the “fitted” pre-20th century temperatures low while the 20th century ones remain high. Is there another series like that in this composite?
#8. Peter, without the Guliya and Dunde series, the Yang composite is not a HS. The blue version has a “cold” MWP and affects the MWP-modern relationship.
There are three ice cores from the Guliya Ice Cap. I don’t know which of them have been analysed for d18O, but if more than one has been worked on this could explain the discrepancy. Some error in the age model could yield the poor correlations.
Please can you plot ice core d180 plots using type=”s”.
Thompson et al. 1995 A 1000 year ice core climate record from the Guliya Ice Cap, China and its relationship to global climate variability. Annals of Glaciology, 21, 175-181.
Steve-
Is it possible that you might run MBH 98, 99 (w/o stripbark, bristlecones, Gaspe, etc.), but, particularly, w/o the “instrumental splice” and do a regression on the means. This would give us all some idea of how much of the “blade” is “instrumental”. Yes, only to 1980 – we need up-to-date proxies. I realize that this will have no absolute reference, since the “instrumental” has been removed, but it seems to me that it should provide anomalies (referenced to the mean), that can shed some light on how much of the hockeystick regression is tree-based vs. “instrumental”.
Steve, on reading your post I initially thought, like RichardT, that the Guliya data must have been obtained from different cores. However, I’ve just scanned over the 2006 PNAS paper, checked the WDC data archive, and also looked at the Thompson et al. 1997, Science article which is referred to in the recent PNAS paper.
It’s late here and way past my bed time but my reading is that all the data were derived from a core drilled on Guliya in 1992. This core is 300+m long and in excess of 12000 samples were cut for d18O measurement. The ice at the base of the core is believed to be very old (>500k years). Am I wrong in this interpretation?
Given that the data are supposedly from the same core it is difficult to think of a reason for the difference between the archived WDC (Climatic Change 2004) and the PNAS data. I’m intrigued that in the PNAS article they only give data for the past 400 years when they have a much longer record to hand. Also that they have a 5 year resolution whilst the WDC archive has a 10 year resolution.
It’s possible that the core was revisited and further measurements taken. If this were the case then it poses serious questions about the analyses and data quality control. For those who don’t know it’s possible to measure ice/water d18O to a precision of about +/-0.05 per mille based on replicated analyses.
This is a very curious observation you have uncovered Steve and I wish I knew what was going on!
The following is first line in Frank Scammell’s post #11. Anyone know what causes the non-ASCII chars? I get them in a lot of posts.
Is it possible that you might run MBH 98, 99 (w/o stripbark, bristlecones, Gaspe, etc.), but, particularly, w/o the “instrumental splice” and do a regression on the means. This would give us all some idea of how much of the “blade” is “instrumental”.
Bob
Funny but the non-ASCII chars are don’t showup in my post. I’ll try one more using blockquote.
Bob
#12. Paul, I raised an almost identical issue regarding the Dunde core last year with Science, reporting to them the inconsistency between Dunde versions and got blown off. I’ve raised the same question with Cicerone of NAS and likewise got blown off. Here’s the same information. The WDCP information was only archived in 2004 about 17 years after the core was drilled, when Climatic Change adopted a new policy requiring authors to provide data (in response to my review of an MBH submission, which was eventually rejected because they refused to provide data.) However, they kited the academic check by citing the rejecte study in Jones and Mann 2004 and, having kited the check, didn’t care any more about the original rejection. So whatever is going on at Guliya, something similar is going on at Dunde and perhaps elsewhere.
Here’s my speculation on what’s going on with the Dunde data: the estimated age in the core is decreasing. It looks to me like the black Clim Chg version is dated about 100 years earlier than the red version comparing 1500s (blck) with 1600s(red). One can scarcely guess what’s going on with the Yang version, but maybe the blue uptick centered in the 1300s corresponds to the black in the 1500s.
I think that the PNAS version of Guliya is also dated quite a bit younger than the Clim Change version in 2004.
I wonder which versions were used in the Al Gore hockey stick.
No wonder Thompson doesn’t want to talk about this or archive his results. It looks like the dating might be a mess. Interested parties might look at my notes on his dating on Kilimanjaro, which looks highly problematic.
Here’s a link http://www.climateaudit.org/?p=327 to my correspondence with Science regarding Thompson. It’s pretty frustrating to look back at it.
Steve, I’m just about to take the kids to school, then will be in my lab. I want to look at these data a little more closely. I think you are right about the inconsistencies in the dating. However there is a clear remark in one of the Thompson group papers that Guliya has annually resolved layers down to a depth equivalent to 400 years. This being the case there should be little or no issue with the dating, at least for the most recent 400 years of the record.
Dr Juckes???
Steve, I’ve now had a chance to read again the relevant Thompson papers (Science, 1997; Climate Change 2003: PNAS, 2006). I’m in no doubt that they are referring to data obtained on the 1992 Guliya core that is described in Thompson et al. 1997.
I’ve looked through the additional data lodged with the WDC (Climate Change 2003) and the PNAS (2006). The first comment to make and one which you have yourself made is that the files are not really data. They simply give the values of the numbers plotted in the associated figures.
The second point is that there are material differences between the WDC and PNAS archived data. To try and bring the two sets to a common base of decadally averaged data I have simply averaged the 1980-84 and 1985-89 data to get a figure for 1990 and so on. I think that Thompson and co-workers have used simple arithmetic means of data in constructing their multi-year averages and not weighted them according to the mean annual accumulation rate. Thus averaging two 5 year periods should give the decadal average.
The difference in the data sets is significant. I don’t think there is a discernible modern trend in the Guliya data however it is presented. None the less the difference between the sets is worrying as it implies poor data management and quality control if one is to put this in a generous light.
What is needed is a full archiving of the complete Guliya data set for 18-O and all the other measured parameters….Cl, SO4, dust etc. Over 6000 sample s were cut for measurement. These should be readily archived. Only then can we truly judge what the situation is with respect to Guliya.
It’s very worrying that different versions of these data are being propagated around the various proxy-climate reconstructions. I’ll write to Lonnie Thompson and see if I can get hold of the full data set. Don’t hold your breath!
From what you say similar problems are found with Dunde and possibly the Andean glaciers as well.
Paul, would you write to Science as well? They are familiar with the file.
I’ve had a look at Thompson et al. 1995, his first paper on the Guliya ice cap. There is a little more information on the methods here than in the later paper. The chronology for the core is based on annual dust layer counting, at least in the upper part. This must extend for at least 1000 years, as there is an accumulation record (based on the thickness of annual layers, corrected for iceflow), so the chronology should be reasonably secure. The work in Thompson et al. 1995 seems all to be based on the long core, though the other two cores are long enough to cover several hundred years (one perhaps a thousand). This paper includes a low resolution isotopic record, presumably initial data, that appears to be different from all three curves shown above.
Will do Steve.
Richard, I agree that the dating should not be an issue. Certainly the first 400 years can be counted and as it is only this period with which we can compare the PNAS and Climate Change profiles the age model is certainly not the issue.
Bob W,
I’ve had the same problem with the characters displayed. On your browser’s View menu changing the character encoding to UTF-8 seems to take care of it.
Re #23
Something comes to mind regarding Steve’s speculation in #16. Does Thompson 1995 address possible loss of top layers of ice? If the last X number of years is missing, it would be possible to misdate the core by shifting it to the right in the plot above. How do we know that the first layer counted is from 1992? It is an elementary issue that should be addressed by correlating with volcanic ash or some other indicator.
Earle, this is a good point that you raise. I don’t know very much about the Tibetan Plateau ice cores, the depth of the Firn layer etc. I’m presuming that the top several 10’s of metres is actually firn and not ice. The longest Gulyia ice core is ca. 300m with the top 100m representing the last 1000 years. On the basis of cosmogenic isotopes (36-Cl) it has been suggested that the bottom 20m may be more than 500 thousand years old. Thompson says that the first 400 years shows marked annual layering and that this is particularly evident because accumulation is markedly seasonal and dominated by the monsoon (80% plus annual accumulation during this period).
However, what I’m more concerned about is the apparent mismatch between the different archived versions of this data set. I don’t know how many analyses were completed on Gulyia and whether or not the core was stored and re-visited for further analyses. According to the Thompson et al 1997 Science paper more than 12,000 samples for oxygen isotope analyses were cut from the core. Reading both the Climate Change (2003) and PNAS (2006) papwers we are referred back to the 1997 article. I can only surmise that the data for these later papers represent a sub-set of the original 12,000 analyses. Given this then it is difficult to explain the differences in the archived data as they would appear to represent the same data sets.
One problem is that the 2003 Climate Change article uses decadally averaged isotope ratios and the PNAS 5-year averages. I’m pretty sure these are just arithemetic averages without any weighting for the amount of accumulated snow in any particular year. If this is the case then simply averaging adjacent 5 year periods should give the 10 year average. You can introduce a phase shift between the two to account for a mis-identification of the 5 year period and decade but it is still not possible to reconcile the archives any better.
The Yang version of the Guliya 18-O isotope record has appeared in an 1800 year continuous record in the following papaer:
Yang, Yao, Gou, Wang and Neumann, 2006 Wavelet analysis reveals periodic oscillations in a 1700 year ice-core record from Guliya, China. Annals of Glaciology, V43, 132-136.
The paper is explicit that the record is based on the 308.7m ice core reported in Thompson et al., 1997 and Yao et al, 1997.
So here we have two papers, both published in 2006, purporting to use results from the same measurements on the same ice core yet with different data!
Steve, can you post a copy of the Yang Guliya data set. I tried one of your previous links but the URL no longer seemed active.
Steve: http://data.climateaudit.org/data/Yangbao.data.txt
Paul, for some cores, the PNAS version is substantially the same/ For example, here is Dasuopu with PNAS converted to a simple decadal average:
Figure 1 – PNAS and Clim Chg versions of Dasuopu
However, for Guliya and Dunde, there are substantial differences using the same methodology: Here is Guliya –
Figure 2 – PNAS and Clim Chg versions of Guliya – decadalHere is Dunde (on a different scale) –
Figure 3 – PNAS and Clim Chg versions of Dunde – decadal
Paul, the Yang Yao version truncates the early part of the Yang Bao version:
From Yang Yao et al – top panel truncates elevated early values of Yang digital data below.
Steve, the fact that the Dasuope data are coincident would appear to confirm that they have simply averaged the annual results to obtain 5, 10 or whatever period averages.
The differences for Guliya and Dunde are significant and require an explanation. I’m still trying to be charitable over the reasons, but it’s becoming increasingly more difficult to hold this line.
Then there is the Yang data set which is different again. I think it possible that the TP drilling was probably part of a Sino-American programme and that the data was to be shared amongst the groups. It is interesting to note in the most recent Yang paper that no acknowledgement, or reference (other than Thompson, 1997) is given for the Guliya profile.
Whoa…there’s a 7 per mille shift in delta 18-O. By the reckoning of Yang, Yao and others that’s a 11.2 degree drop in temperature between about 200 and 500 AD.
I think it more likely that what we are seeing here are changes in atmospheric circulation and disruption/enhancement of the monsoon.
A neat truncation though!
I wonder if we can get to the bottom of this with an open scientific comment submitted to Science/PNAS/Annals of Glaciology. Here are 3 different journals publishing refereed papers based on supposedly the same data sets but which are actually very different.
It’s rather nice that the Yang version of the Guliya data also surfaced this year. HEre are some interesting comments on the need to examine Thompson’s withheld sample data from Seimon 2003.
Anton Seimon GRL 2003 CIRES, University of Colorado, Improving climatic signal representation in tropical ice cores: A case study from the Quelccaya Ice Cap, Peru
I wonder what Seimon would think of the Guliya situation. In contrast to Quelccaya where results at least had been archived, not a single thing had been archived from Thompson’s Himalayan cores before I started shaking the tree.
Another thing that adds a little interest to the inquiry: the hockey stick illustrated by Al Gore on page 64-65 is taken somehow from Thompson’s data. They label a little bump circa 1370 as the “Medieval Warm Period”. This is on the same page as the one in which Gore describes the “fierce attack” by skeptics against the “hockey stick a graphic representing the research of climate scientist Michael Mann and his colleagues”, going on to say that Thompson’s ice core record is one of the “most definitive” confirmations of Mann.
Speaking of Gore, his Thompson ice core version on page 64-65 is denominated in deg C. I wonder where that came from.
I don’t think we can better the comment by Seimon:
“This investigation indicates that individual sample measurements would more effectively serve as the basis for paleoclimate analysis with this data series, rather than subjectively determined layer averages made widely available by the original investigators.”
There are over 12000 oxygen isotope analyses for the Guliya core. Assuming they are equally spaced, samples would have been taken every 2.5 cms. For the recent past this represents a very high resolution during the main accumulation period, probably on the order of 1 to 2 months.
#35. That’s not true for Guliya since we don’t even have subjectively determined layer averages. WE have only subjectively determined decadal averages for dO18 for part of the core, without dust or chemistry. It’s outrageous.
re #26
The firn-ice transition in Guliya is very shallow – just a few metres (Thompson et al. 1995). This is very different from the polar glaciers which have equivalent temperatures. They measured tritium concentrations in the ice – a global marker horizon for 1963 courtesy of US and Soviet atmospheric testing of nuclear weapons. The tritium peak is concordant with the age model.
The results of the Seimon paper on Quelccaya are less impressive than its abstract. The chronological errors he finds are small – no greater than the ~1% error you should expect with layer-counting chronologies. The expectation that earthquake-derived dust layers can be used to date the core is speculative – at least it should have been supported by a figure. The hypothesis that widespread tephra from the Huaynaputina eruption was only deposited on the ice after being remobilised by an earthquake is scarcely credible without supporting information.
I realise that Seimon was hindered by the lack of accessibility to the raw data, and his paper is useful, but the real conclusion is that the Quelccaya chronology is pretty good. This doesn’t help the Guliya case. At Guliya, unlike Quelccaya, the long core was taken to the lab frozen (and presumably an archive section remains).
RichardT an age model is hardly required when you can count annual layers in the ice. It’s reassuring that the early 60’s spike in atmospheriuc tritium is recorded and is in agreement with the layer counting.
Given this it is hard to conceive how several different profiles for Guliya can exist in the extant literature and that these different profiles are still being propagated in new and upcoming papers by different groups. The problems can be resolved quickly and easily if the original 12,600+ isotope analyses are archived in full with the other chemical and physical data recorded for this core.
Yang et al 2000 (ENSO Events Recorded in the Guliya Ice Core
Climatic Change 47, 401-409 DOI 10.1023/A:1005696702385) Have a figure (no data) showing annually resolved d180 and accumulation for the last 300 years. There’s a lot on interannual variability. Does anyone have a program read in the data from the graph, then we could compare offset decadal means?
#40. Hans Erren and Willis know how to do this from pdf’s. Hans tried to explain it to me. Why don’t you email the article to Hans. (Hans/Willis, why don’t you post up your instructions on how to extract times series from pdf’s?)
Well spotted Richard.
Steve I think Willis digitised some previously published data using VectorWorks. This is a CAD program and fortunately I have a copy of it. I’ll try and extract some annual data from Figure 2 of the Yang et al paper.
There are two ways in digitising a graph from a pdf.
If the graph is a true vector (i.e. you dont see pixels when zooming in to extreme magnification),
then you can use a postscript driver to print the pdf to eps format, somewhere in the file the graph is found as an x,y coordinate list (unit points), the x,y data can then be rescaled using the point values of the axes, eg in excel.
If the graph is a bitmap (you’ll see pixels at extreme magnification) then you need digitising software, eg CAD, Rockworks or a GIS package.
RE #42
Paul,
I don’t know if Willis has a VectorScript for collecting the placed loci x,y to text data, but if not, I found a tip that might help. The idea is to move the placed loci to a clean file and then export the file as a VectorScript file. The VectorScript file, which is in text format, will have all the loci x,y data that can copied out using a text editor.
Cliff, thanks for the tip. I’m going to try and digitise the profile tonight. I’m not too optimistic as the picture is a bitmap file and quite coarse. None the less the quality should be good enough to allow us to differentiate between the various grey versions of the Guliya data that are appearing and re appearing in different publications.
Paul,
Another paper by Yang et al. (2006), contains the same graph but without the smoother. Its still low resolutions though. I suspect getting data out of this group is going to be difficult since, almost ten years after they published the main record in Science, they are still publishing bits of it afresh.
I described the manual method for digitizing here. I use Vectorworks, but any good CAD program should work as well. For the manual method, the key is to establish a grid of lines representing the times that the observations represent, and to turn on the “Snap to Objects” option so that your times are exactly correct. I work at pretty high resolution.
The method used by Hans for vector based graphics (convert to PostScript, use the text description) works well, and is the method of choice. Unfortunately, not many graphics are vector based.
Finally, on the Mac, there’s a program called “GraphClick” which automatically converts a line in a graph to (X,Y) coordinates. It’s pretty smart about it, you set the X,Y scales and the output is in terms of those scales, so no re-scaling is necessary. It does require a clean graph, and is not too smart about colors and crossing lines, so some pre-processing is often required.
CAVEAT!!! – Once I have the data, I plot it in Excel, copy the plot, and paste it back into my CAD program. There I overlay the original graph with my digitized version, to make sure that the digitized version corresponds exactly to the original. DO NOT OMIT THIS STEP ON PAIN OF LOOKING VERY STUPID.
w.
I’ve had a look at the annual data presented in the Yang et al. 2006 papers and had a first pass at digitising the results, and then calculating 5-year and 10-year averages.
My preliminary conclusions are:
1) The recent profiles published by Yang et al. are not in agreement with either the Thompson (PNAS 2006), or Thompson (Climate Change 2003) profiles. The correlation between the Thompson 2006 archive and the recent Yang profile (averaged over 5-yearly intervals) is r^2 = 0.11.
2) There are similarities between the early Yang archive and the recent (Yang et al., 2006) annual profile. However I can’t make a detailed comaprison as I don’t have a digital version of the decadally averaged Yang profile. Do you have a copy Steve?
3) When I return from a meeting tomorrow I’ll post the plots and archive the data for everyone to look at.
Paul, here’s what I think is going on with the Himalayan data. In the last century and half, the decadal versions of Dunde, Dasupopu and Guliya have “significant” positive correlations with one another (0.3-0.6) and a trend. Prior to that, the various series are essentially uncorrelated (0 and negative). I’ll bet that Thompson has used his dating flex to end the 19th century at a low point in each of the three series and yield a positive 20th century trend for each core. I’ll bet that the synchronization would fall apart if anyone got to see the sample data.
#48. Paul, could you post up the digitized data mentioned here? Thx.
#50 Steve, I’ll try to post the graphs and data file tonight. Believe it or not today I am interviewing for a fellowship in Isotopes and climate change studies in my lab!
If I have trouble posting the graphs I’ll email a copy of the Excel file to you.
Steve I have sent you an email and attachments to your Yahoo address!
I’ve digitized the Guliya ‘ˆ’€šO18 values and placed them here. The file also contains the local temperature values.
Also, here’s a graph of the ‘ˆ’€šO18 values vs. the local temperature.
As you can see, the local temp has dropped over the period. However, Yang says that the signal is not from the local area, but represents the temperature of the ocean where the water is coming from. I haven’t checked that yet.
w.
Willis, thanks for posting the data. The correlation between dO18 and HadCRUT is -0.1, not significant. Is that a surprise?
Willis, you are so kind to post the data. I have see another paper by Thompson.
Thompson, L.G., T. Yao, M.E. Davis, E. Mosley-Thompson, P.N. Lin, T.A. Mashiotta, V.N. Mikhalenko and V.S. Zagordonov. 2006. Holocene climate variability archived in the Puruogangri ice cape from the central Tibetan Plateau. Annals of Glaciology, 43, 61-69.
In this paper, Fig. 8 exhibits O-18 variations of different Tibetan ice cores. Would you digitalize them? It is very interesting to discuss sth of all the curves.
Thank you in advance!
Steve: Last time I checked the NGRIP people were not sharing their new Greenland chemistry data either. Please tell me if that’s changed. I have other information to add relative to this thread but would much prefer you contact me offline.
For the record, Paul [F.] Dennis is at the University of East Anglia, School of Environmental Sciences, Norwich, United Kingdom — unless Google has misled me 😉
Cheers — Pete Tillman
3 Trackbacks
[…] the Thompson ice core data, see also “Juckes, Yang, Thompson and PNAS: Guliya” (CA 12/3/06), “Dunde: Will the real Slim Shady please stand up?” (CA 4/12/07), […]
[…] Himalayan ice core information for Dasupo, Dunde, and Gulaya, is detailed (inter alia) here, here, here, here, here, here, and […]
[…] definitive archived versions of his measurements. See, for example, IPCC and the Dunde Variations, Juckes, Yang, Thompson and PNAS: Guliya, Gleanings on Bona Churchill, and Mann on Irreproducible Results in Thompson (PNAS […]