2011-03-25 08:38:55Lunacy continues at WUWT and Climate Audit
Rob Honeycutt

robhon@mac...
98.207.62.223

WTFUW...Those guys?  Now McIntyre found where, I guess it's Briffa in 1999, deleted some data.  Data that obviously doesn't agree with any of the other data.  Here's the write up on WUWT.

I think anything other than just plotting raw data is a form of corruption to these guys.  It's just utterly ridiculous.

2011-03-25 08:59:13
Albatross
Julian Brimelow
stomatalaperture@gmail...
199.126.232.206

Rob, without knowing the details, I think that this is probably just another attempt to fabricating controversy and to libel/slander climate scientists.  They have to do so to keep the rabid deniers of AGW frothing at the mouth.  Sad but true.

McIntyre is losing it and with each day and passing is showing his true (and scary) colours, not to mention his incredible desperation. I have noted that WUWT and CA are working a lot more closely now, probably in an effort to brain wash as many people as possible, and to keep the converted convinced that this is all a conspiracy-- despite the shit hitting the fan all around them.

I have no idea how one deals with this-- to be candid, McIntyre or Watts in handcuffs is probably the only thing that will slow things down.  Note that i did not say "stop".  These guys are relentless, and have many faithful followers.

 

 

 

 

2011-03-25 09:06:23
grypo

gryposaurus@gmail...
173.69.56.151

Yes, this is going to go viral.  I'm hearing bombshell all over the threads at CA.  Watts said acedemic misconduct.  Isn't he ususally one of the ones who refuses to cry fraud, unless proven?  He just lets others do it for him after he dog-whistles? Or is that just McIntyre?

2011-03-25 09:07:51Briffa
dana1981
Dana Nuccitelli
dana1981@yahoo...
64.129.227.4

I'm trying to figure out what paper McIntyre is referring to.  It looks like this one.  The 1550-1960 tree ring data is referenced from other papers though.  Briffa 1998 from Nature is one of them.  The data goes back to about 1400, and doesn't look like McIntyre's plot pre-1550.  Same with another of the references, Briffa et al. (1998)Briffa 1999 has a plot which goes all the way back to 0 AD, and again, doesn't look freakish between 1400 and 1550, unlike in McIntyre's plot.

So I don't know what to make of McIntyre's claims.  Seems to me like the don't jive with the rest of Briffa's publications.  I wish these deniers would stop obsessing over papers from 12 years ago.

2011-03-25 09:15:17
grypo

gryposaurus@gmail...
173.69.56.151

Could there be some serious de-weighting of the 1402-1550 data.  Is this the age-banding method?  PC?

 

page 2938 from this has a spaghetti graph, plate 3.  Read the description.  It mentions the 1402 data and error bands at the end of the description.  Need a paleo/stats person.

 

EDIT:

It appears as tho the low error bar runs along where McIntyre has his data plotted.  So the data was used to represent unlikely data and error?

2011-03-25 09:22:00
grypo

gryposaurus@gmail...
173.69.56.151

2011-03-25 09:27:42different
dana1981
Dana Nuccitelli
dana1981@yahoo...
64.129.227.4

Nah, that looks different grypo.  Look at McIntyre's plot - it absolutely plummets between 1550 and 1400 (going backwards in time).  Plus you're looking at a 2001 paper from Briffa.  McIntyre is looking at 1999.

2011-03-25 09:42:17
grypo

gryposaurus@gmail...
173.69.56.151

yeah, around 1480 it goes down to .9.  wow.  He discusses briffa 2001 in an earlier thread, wondering why it changed from 1999.  possibly to blame Mann, surprise.  It looks like whoever delves into this will need to look at all 4 threads.  and all the papers.  yay!  what a waste of time!

2011-03-25 09:44:25
Rob Honeycutt

robhon@mac...
98.207.62.223

Okay...  I'm about to be banned from WUWT in 3...2...1....

I keep messing up my html tags over there.  Totally by accident.  :-)

2011-03-25 11:00:41Comment
Robert Way

robert_way19@hotmail...
134.153.163.105

I'm not sure what my opinion on this subject is at this point. If Steve Mc is correct then I do think that Briffa has some explaining to do. Personally I'm not a fan of the deletion of data for a figure either way... I would get blitzed by my supervisor if I did it in any document so I don't know why the same standards shouldn't apply.

Either way though, it just goes to show that some tree ring datasets are probably too difficult to use and other proxies like ice cores...etc... will be better for reconstructions.

2011-03-25 11:14:27
nealjking

nealjking@gmail...
91.33.110.91

Has anyone done a full-on explanation of how the proxy picture has been put together?

2011-03-25 11:29:07
grypo

gryposaurus@gmail...
173.69.56.151

Dana, let's rewind

Here is something from one of the previous CA threads.

Briffa et al 2001 uses virtually the same population of sites as Briffa and Osborn 1999. The B2001 population was 387 sites, while the Briffa et al 1998 (Nature 393) population (cited in BO99) was 383 sites – immaterially different. The Briffa et al 2001 site count was 19 sites in 1550, 8 in 1500 and only 2 in 1402, but there were enough for Briffa to report a reconstruction. (Readers should bear in mind that the Jones reconstruction, for example, was based on only 3 proxies in the 11th century, one of which was a Briffa tree ring site with only 3-4 cores, well under standard requirements.)

How did Briffa get from the 1999 reconstruction version (magenta) to the 2001 version (limegreen)?

So what McI's asking is, even though the same population is used, why did briffa change the reconstruction from 99 to 01.But oddly, he claims to know.

This is an interesting exercise that I’ll describe in more detail on another occasion. But I can’t resist a quick preview. The Briffa-Osborn 1999 version was based on averages of all available sites – a sensible enough procedure. Changes in the Briffa 2001 methodology include the calculation of regional averages followed by stepwise principal components. (The methodological description in Briffa et al 2001 is very sketchy and, unfortunately, the Climategate computer dossier didn’t include Briffa’s programs.)The effect of using principal components on regional averages is to change the weights for individual sites, including the possibility of negative weights i.e. flipping the regional MXD series. In particular, the closing uptick in the Briffa 2001 reconstruction may well depend on the flipping of data – a point that I’ll try to examine in the future.

The next thread only looks at the code to see how the data was clipped, so, there's nothing there of real interest in looking for fraud.  Is this accurate so far?

2011-03-25 11:54:21Accurate
dana1981
Dana Nuccitelli
dana1981@yahoo...
69.230.97.203
Yep, that's what I've gotten so far too, grypo.
2011-03-25 12:13:49
Alex C

coultera@umich...
67.149.101.148

Yes, they're talking about the NH tree-ring density plot in Briffa and Osborn 1999, which Dana linked to.  That paper, though, discusses a reconstruction from Mann et al, and it does not present the findings of a new construction from Briffa himself.  Note the lack of site maps, and any real discussion or graphs regarding the data in question.  The density data, which was plotted with other reconstructions, comes from "

  1. K. R. Briffa

Nature 391678 (1998K. R. Briffa, P. D. Jones, F. H. Schweingruber, T. J. Osborn"

Which is NOT, though, the Briffa 1998 that Dana linked to (twice - mis-copypaste?).  It is this, Reduced Sensitivity of Recent Tree-Growth to Temperatures at High Northern Latitudes, abstract here (note the Nature page and issue number).

The data that Steven shows is not plotted in this paper at all - this paper only plots data from the twentieth century.  Of course, the citation given in the 1999 paper is Briffa's.  I do not know where Briffa got the data to plot in Briffa and Osborn 1999.  Does someone have any more insight?

2011-03-25 12:23:55
grypo

gryposaurus@gmail...
173.69.56.151

No hyperlink.  So I can't find the data and i can only assume the Hughes data in the Dana paper and isn't what I'm looking for, even tho it appears to be it, just minus McI's pre-1550 decline.  The data that changed in B2001 that McI is concerned with is made up of 2 data sets that go back to 1400 and 1402 labeled NSIB and ESIB.  No idea which is raw, but neither look like McI's plot at all when plotted in simple excel.  It doesn't tell me anything because I don't know what being used, and I'm no good with code so I'm at an impasse.

2011-03-25 12:26:20
Alex C

coultera@umich...
67.149.101.148

Uh oh.  I clicked on the (3) in the paper Dana linked to.  It went not to the third source but the FOURTH, this one:

"P. D. Jones et al., Holocene 8, 455 (1998)"

Was this an error in publication that was never addressed?  Which paper is that?

Edit: Here (full here).

Edit 2: Well, data extending back that far is shown, but no drop off is.  Perhaps the citations are correct.

McI is referring to data from an email.  Did he provide a direct citation (better yet, hyperlink) to which one it is?

He seems to be complaining about the usage of the data in that one paper, even though it was shown in full in another (mysteriously hidden) paper by itself.  Was the cutoff even of any consequence toward the discussion in Briffa and Osborn 1999?

2011-03-25 13:18:31
grypo

gryposaurus@gmail...
173.69.56.151

crap, I accidentally edited another post.  Here's that relevent mcI comment

 

Steve – the derivation of the deleted data is described in meticulous detail in yesterday’s post,http://climateaudit.org/2011/03/17/hide-the-decline-sciencemag/ the first part of Hide the Decline- Sciencemag. The post includes scripts generating the figures, with the scripts proving the provenance of the data, and thereby establishing what Briffa didn’t show.
For Briffa et al 2001, the additional data is obtained from a Climategate email – see discussion in Dec 2009.
As I’ve noted elsewhere on a number of occasions, Briffa showed the decline only in standalone graphics. However, the decline was deleted in spaghetti graphs comparing to other reconstructions, thereby giving false rhetorical coherence – which in turn was used to justify untrue statements in the running text, as shown in these posts.

2011-03-25 13:59:42Well I feel stupid.
Alex C

coultera@umich...
67.149.101.148

Disregard my little mishap with the citation numbers, I must have highlighted the source earlier, forgot, and was shocked when I clicked on the (3) to have it scroll down and (gasp!) show a highlighted number 4.

Still no closer to the data.

2011-03-25 18:21:42I haven't looked in depth at this yet
John Cook

john@skepticalscience...
121.222.20.55

But I've been investigating Briffa's data for entirely different reasons. I just downloaded the data from:

ftp://ftp.ncdc.noaa.gov/pub/data/paleo/treering/reconstructions/n_hem_temp/briffa2001jgr3.txt

Apparently either Briffa 2000 or Briffa 2001 was used in the WMO graph (I'm currently not sure which) but just to see for my own eyes, I plotted both curves together to have a look as this also is relevant to this new issue:

It's unsmoothed, yearly data, I make no comments on it as of yet. I've uploaded the excel spreadsheet of the data.

2011-03-25 20:55:57
Ari Jokimäki

arijmaki@yahoo...
192.100.112.211

One Briffa & Osborn 1998 data is here. Doesn't look like anything in McIntyre's graph.

2011-03-25 22:05:56
Ari Jokimäki

arijmaki@yahoo...
192.100.112.202

I forgot to mention that McIntyre's graph shows that he's plotting different dataset. See the dotted line at the left of where McIntyre's addition starts (at about 1550 to 1575). It turns in different direction than where McIntyre's graph is going. Even with such a short piece of that line you can see clearly that it's different dataset. That perhaps explains why McIntyre has plotted only the beginning and the end of the data he's using - I bet it doesn't follow the dotted line. It also might explain why he doesn't explain exactly what data he's plotting.

By the way, the dotted line deviates from McIntyre's line also in the end (after 1950's).

2011-03-25 22:26:38
grypo

gryposaurus@gmail...
173.69.56.151

Somebody finally provided a link to the data.

ftp://ftp.ncdc.noaa.gov/pub/data/paleo/contributions_by_author/jones1998/jonesdata.xls

It is the 2nd sheet on that excel doc, briffa.  I was able to reproduce the curve, with a 50 yr smoothing.

 

 The plot points are off from the zero point but that doesn't matter for this purpose as we only need to validate what was done (perhaps the difference between the email data and online data?).  As stated before, the sample size for briffa, osborn 99 was very smaill and didn't concur with the other proxies and that was likely reason it was cut off from 1400-1550.  The next step is to figure out how Briffa got to 2001.  I plotted tje data and it looks just like it should, as the paper had it.

http://www.cru.uea.ac.uk/cru/people/briffa/jgr2001/jgrdata2.xls

So how did Briffa get there, McI already claims to know, but here's the description on the data from Briffa.

 The accompanying Table contains reconstructions, from Plate 2, of April-September mean temperature (as degree C anomalies from the 1961-1990 mean) for each of the regions defined in Figure 1. These are based on tree-ring density, processed using the age-band decomposition (ABD) standardisation procedure that aims to maintain low frequencies in the reconstructed data. The interannual values are provided here and must be smoothed with  a 10-year low-pass filter to correspond with curves shown in Plate 2. The skill of the reconstructions generally deteriorates back in  time, due to fewer chronologies available within each region and  the early sections of the series (as delineated by the vertical  red bars in Plate 2) could be considered not sufficiently reliable. 

Some statisitical analysis happened or the proxies changed dramatically (increase in number?) to allow for Briffa to have confidence in those to write what he wrote in the 2001 paper.  That's as far as I can go with it.  I don't want to make any more assumptions because I don't know enough about the techniques or the proxies.  

John, can you ask one of your contacts to look at McI's threads?

2011-03-26 06:09:03
grypo

gryposaurus@gmail...
173.69.56.151

So I've gone through a few things, I think.  Possibly not, it was fun for me anyway.  The data used in Briffa, Osborn 1999 is Eastern Siberia data with 150 cut off.  You can find the full set labeled ESIB in the Briffa 2001 data.  The other data that goes back to 1400 is Jones' Northern Siberia data (NSIB).

I don't think NSib data was used in B99, as it's own proxy anyway, possibly included with a full proxy going back to year 1.  But if the data is plotted from B2001 on JUST THOSE 2 proxies, it looks like this.

 

So you can see where the ‘problem’ starts.  So the 1st 150 years was deleted from the B99 as it diverged so much from all the other data used in that paper just as it diverges from this NSIB data.

B2001. 

http://www.cru.uea.ac.uk/cru/people/briffa/jgr2001/Briffa2001.pdf

The graph is pictured here and it shows what I graphed, so the data and picture match curves, pretty much, gin=ving me confidence I'm using the right data.  Notice the red line.  That is explained in the caption:

The vertical red lines in some reconstructions indicate how far back the age-banded reconstructions are considered to be reliable.

So in this study (B01) the ESIB data was only reliable to 1460 as opposed to 1550 in B99.

But why?  Perhaps a different method was used?  McI says yes, and he’s right, but watch his wording in:

http://climateaudit.org/2011/03/21/hide-the-decline-the-other-deletion/

“This is an interesting exercise that I’ll describe in more detail on another occasion. But I can’t resist a quick preview. The Briffa-Osborn 1999 version was based on averages of all available sites – a sensible enough procedure. Changes in the Briffa 2001 methodology include the calculation of regional averages followed by stepwise principal components. (The methodological description in Briffa et al 2001 is very sketchy and, unfortunately, the Climategate computer dossier didn’t include Briffa’s programs.)

The effect of using principal components on regional averages is to change the weights for individual sites, including the possibility of negative weights i.e. flipping the regional MXD series. In particular, the closing uptick in the Briffa 2001 reconstruction may well depend on the flipping of data – a point that I’ll try to examine in the future.”

But when I read the 2001 paper.  It introduces the method and goes to tell a more specific story.  Whether or not this is 'sketchy' is unknown to me or if 'flipping' is a good word to use or not.

http://www.cru.uea.ac.uk/cru/people/briffa/jgr2001/Briffa2001.pdf

 

And then he describes what happened with the ESIB data that is causing all the fuss.

 

But watch closely how McI’s assumptions lead his readers to think conspiracy in http://climateaudit.org/2011/03/17/hide-the-decline-sciencemag/.

One can reasonably wonder whether the key conclusion of Briffa and Osborn 1999 – “[despite] the problems associated with interpreting many of them as unambiguous measures of hemispheric temperature change, this conclusion [MBH] must surely be accepted” – would have stood up if the decline had been shown.”

 

“Briffa and Osborn 1999 contains a very sly reference to the divergence problem:

A number of tree-ring chronologies have displayed anomalous growth or changed responses to climate forcing on different time scales in very recent decades (3 – Briffa et al 1998 (Nature), 9 – Jacoby and D’Arrigo 1995). Understanding the reasons for these changes is important for understanding the causes and limits on past tree growth. Paradoxically, therefore, more work in the recent period is required to better interpret the early proxies. Few of the proxy series run up to the present, however, and updating these will involve considerable effort.”

 

Climategate scientists were well aware of the importance of figures. Briffa and Osborn knew that the graphic with the deletion of the decline would leave a different impression than one that disclosed the decline. The sly wording of the running text compounds the problem. Yes, there are proxies that need updating, but the MXD data used in the Briffa reconstruction came right up to the early 1990s. Unavailability of data is not the reason why the Briffa reconstruction ends in 1960.”

 

I don't know what’s more interesting, Briffa looking for good temperature signal in dead trees from Siberia or McIntyre’s dog-whistle.  He doesn’t even really say anything terribly important, or even something that people wouldn’t have known by looking at an excel spread sheet graph.

Perhaps someone can make a case against them for not being more explicit in what happened to the 150 years of bad data in Briffa 99, but I bet the reason they didn't is because they didn’t think anyone would give a shit or because this is referrenced in another paper, possibly.  I hardly believe that they thought that the 'wording' would fool anyone within the field.   It really doesn’t matter, just as the 1960 data doesn’t matter.  It only matters when studying the divergence problem.

2011-03-26 10:46:06Question regarding "flipping,"
Alex C

coultera@umich...
67.149.101.148

Is McI implying that the data set itself is actually flipped?  As if you could plot a horizontal line through that set and reflect it over, and it would be correct?  If so, why do the peaks still match up, even though the absolute variance in the Northern Siberian data is obviously greater?

2011-03-26 14:03:59
grypo

gryposaurus@gmail...
173.69.56.151

The 'flip' appears to just be wordplay, and bad wordplay.  He actually used a similar word in a peer-reviewed response to Mann where he said data was used 'upside-down'.  This is the whole Tiljander debate over the other data.  Had he used a better phraseology that argument may have been settled one way or another, but Mann, apparently, had no idea what he was saying and called it 'bizarre'.

From what I see for this situation, it looks like filtering against highly correlative data from other proxies causes the bad data to not be as weighted as much when the reconstruction is presented in graph form, but instead causes high error bars.  This is my impression from what I've read, but I need verification on that.

2011-03-26 16:06:45
Ari Jokimäki

arijmaki@yahoo...
91.154.107.62

I don't understand why we are looking at Jones et al. data when the dataset in question in Briffa & Osborn (1999) is:

K. R. Briffa et al., Nature 391, 678 (1998) [GEOREF]; K. R. Briffa, P. D. Jones, F. H. Schweingruber, T. J. Osborn, ibid. 393, 450 (1998) [GEOREF]; K. R. Briffa et al., Philos. Trans. R. Soc. Lond. B Biol. Sci. 353, 65 (1998).

There is one Jones et al. dataset in Briffa & Osborn (1999) but that one is between 1000 and 1992 and has not been cut by Briffa & Osborn.

2011-03-26 17:42:58comment
Robert Way

robert_way19@hotmail...
142.162.22.136

The Tiljander debate showed that Mc was right on that issue. Kaufmann had to fix his series because he also used it upside down. Didn't make too much of a difference but Mann's response of "Bizarre" was pretty lazy if you ask me. The original Tiljander series people even said Mann and Kaufmann used it wrong. That being said Mc is a conspiracy wackjob...

2011-03-27 00:14:41
Ari Jokimäki

arijmaki@yahoo...
91.154.107.62

McIntyre was not correct on Tiljander. There was indeed a mistake in Mann et al. when using Tiljander, but McIntyre's claims about it were false also. One of the most non-sensical of McIntyre's comments was the official comment in PNAS. It's no wonder Mann et al. called it bizarre because that's what it was.

2011-03-27 00:30:51
grypo

gryposaurus@gmail...
173.69.56.151

Robert, it probably was 'lazy' but the way in which McI worded his response, was confusing.  Using words like 'upside-down' or 'twisted' or 'flipped' is more like word play than good argumentation, especially in a peer-reviewed literature. Perhaps that's why Mann used the word 'bizarre', who knows.  My critism is that this would be settled had that original comment not used the buzz words he uses to dog-whistle on CA.

 

Ari, the reason I showed the other data, is that I wanted to chart a comaparison between the other siberian data that goes back to 1400, to show the divergence and why the Briffa data was cut off.  But, further investigation tells me that the ESIB is either not raw ot different from B98.  So I'll chart the data straight from your source later.  i missed your previous post and that would have saved me some time.

 

ThefordPrefect has a piece on it.  He's the one I asked about the B01 data.

2011-03-27 05:05:45
Ari Jokimäki

arijmaki@yahoo...
91.154.107.62

But do we know that McIntyre used the Jones' data in that graph? I thought you meant that when you said that someone provided the link. I haven't read ClimateAudit comment threads, so I don't know if McIntyre has offered any further information of the data he used. I doubt he will, though, as it's quite obvious that he's using wrong dataset there.

Current status seems to be this:

- Details in McIntyre's graph suggest that he is using different dataset than the original graph.

- None of the references to the reconstruction in question show similar decline at 1500 as McIntyre's addition.

- It seems that an Excel-sheet containing Jones et al. (1998) reconstruction also has data labeled as "Briffa" and that data is a candidate for McIntyre's source.

We need to know if Jones' Excel-sheet really is the source. McIntyre probably won't tell it but we can try to recreate the McIntyre's graph with that data. If it turns out to be the source, then we need to track down what Briffa data that is. That is probably found in Jones et al. (1998) reference list.

By the way. I checked the text version of Jones et al. (1998) data in NOAA's paleoclimatology website - it doesn't have the extra data the Excel-sheet has.

2011-03-27 05:51:03
grypo

gryposaurus@gmail...
173.69.56.151

Ari,

I found where McI decribes his data:

http://climateaudit.org/2011/03/17/hide-the-decline-sciencemag/

The Briffa and Osborn 1999 version of the Briffa MXD reconstruction doesn’t match the version of Briffa et al 1998 or the subsequent version of Briffa et al 2001, both of which were archived. Oddly enough, it does match (after truncation) a version archived at NCDC in December 1998 in connection with Jones et al 1998 (though not used in that article), where it occurs in the second sheet of an Excel file here. To my knowledge, this particular version of the Briffa reconstruction was not otherwise published. (The Briffa reconstruction seems to have been very fluid in this period, as the versions in IPCC TAR Zero Order Draft and First Order Draft appear to be different again and still unaccounted for.)

Also, moving on to the pre-1550 decline, McI says:

http://climateaudit.org/2011/03/21/hide-the-decline-the-other-deletion/

As noted previously, Briffa data was deleted after 1960. (Smoothing was done after the deletion further accentuating the impact of the deletion of post-1960 data.)In addition, and this point has not been previously discussed, Briffa and Osborn did not show data prior to 1550 for the Briffa MXD reconstruction. I’d previously noticed that an archive for Jones et al 1998 (surprisingly) contained a Briffa version that is linearly related to the Science graphic – the match is shown as a dotted line. (The basis of the linear relationship is not reported and not known to me at present, but can nonetheless be used empirically to show the extensions.)In the graphic below, I’ve shown (in magenta) not just the hide-the-decline extension, but the deleted data prior to 1550. Take a look. (Update note- see yesterday’s post for provenance.)

So it looks like he uses the excel spread sheet data.  The way he posts makes it hard to follow.

 

edit:

As far as where Briffa got the data, I can only guess it is the East Siberia he used again in 2001, in one form or another.  It is the only data set I can find that goes from 1400 to current time with diverges at both ends.

2011-03-27 23:01:37
grypo

gryposaurus@gmail...
173.69.56.151

EDITED FOR IMPORTANT CORRECTIONS

 

Rewind, collecting data from throughout this thread.

 

1.  The data used in Briffa Osborn 99 that McI shows is cut at 1550 [pale blue] is referenced as [3] in the chart desription.

 

2.  The paper that is refererenced is "Influence of volcanic eruptionson Northern Hemispheresummer temperatureover the past 600 years"

 

3.  That paper used data that Ari first cited a few posts back.  The paper decribes the data in detail in the methods section.

exerpt, but the whole section is educational:

Series NHD1 is preferred, because the two-stage averaging reduces any biasthat NHD2 might have towards a region with a large number of chronologies.Nevertheless, it would be incorrect to average the regional time-series togetheron an equal basis when one series is based on very few chronologies, so theregional series were first weighted by their expressed population signal, EPS(ref. 38). EPS measures the extent to which an average time-series of a finitenumber of sample series approximates the theoretical population average (thatis, infinitely replicated) time-series (EPS ¼ n9¯r). If the time-dependent sampleis very small, the time-dependent EPS will be low and the region will be given alower weighting. Both NHD1 and NHD2 were then normalized with respect totheir means and standard deviations over the 1881–1960 period

 

4.  Briffa Osborn (BO99) truncate that data at 1550.  An brief explanation (not directly mentioning truncation) is in the description:

northern NH tree-ring densities [1550-1960, from (3), processed to retain low-frequency signals] are in pale blue

 

5.  Then McI uses data from an excel spread sheet here to complete his graph of the data that was truncated out of BO99. The data he uses is only labeled "briffa et al' and is on the second sheet called 'science3'.  All this data, including the link to the excel file, comes from a directory (Index of /pub/data/paleo/contributions_by_author/jones1998/) with text data and explanations of the data in a read me.  

SUGGESTED DATA CITATION: P.D. Jones, K.R. Briffa, T.P. Barnett, 
and S.F.B Tett, 1998, Millennial Temperature Reconstructions.  
IGBP PAGES/World Data Center-A for Paleoclimatology 
Data Contribution Series #1998-039.  
NOAA/NGDC Paleoclimatology Program, Boulder CO, USA.

ORIGINAL REFERENCE: P.D. Jones, K.R. Briffa, T.P. Barnett, and S.F.B. Tett, 1998, 
"High-resolution Palaeoclimatic Records for the last Millennium: Interpretation, 
Integration and Comparison with General Circulation Model Control-run Temperatures",
The Holocene 8, 1998, 455-471.

6.  So why did McI use this data, which is not labeled, on a second sheet, in a directory for Jones' papers?  These do not directly lead to the BO99, so I don't know.  Does he?  Yes, he's go the data here that best matched the uncut curve.

7. The green line is 1400-1550 that McIntyre spliced onto the end of that data.  This isn't a splice of different data.  That Briffa data is correct from 98 and the excel file, not the nhd1 file.  The nhd1 is adjusted differently. But the nhd1 data used in the B98 (volcano) is important in finding out why Briffa had to use the unadjusted data in the excel.  The nhd1 data would not work in a reconstruction.  The excel data was processed, but only to retain low frequency signals (then truncated)

 

8.  Here is McIntyre's

2011-03-28 02:36:22No.
Alex C

coultera@umich...
67.149.101.148

The paper that Briffa 1999 references is not the volcano one, but the one I brought up in one of my first comments, here.

Again, note the Nature issue and page numbers.  The volcano one does NOT match those numbers; this one does.

The "Seeing the Wood from the Trees" paper can be found in full at Science; cross-referencing the citation in question leads to here.

2011-03-28 03:35:57
grypo

gryposaurus@gmail...
173.69.56.151

It's actually both.  The data was used in both those papers.  Both 391 and 393 in nature.  Had the same 4 out of six authors in same magazine.  The citation in BO99 for [3] has both there.  The volcano one has data going back to the time discussed.

K. R. Briffa et al., Nature 391, 678 (1998) [GEOREF];

K. R. Briffa, P. D. Jones, F. H. Schweingruber, T. J. Osborn, ibid. 393, 450 (1998) [GEOREF]

; K. R. Briffa et al., Philos. Trans. R. Soc. Lond. B Biol. Sci. 353, 65 (1998).

I haven't tracked downt the 3rd one

2011-03-28 03:41:34Apologies
Alex C

coultera@umich...
67.149.101.148

I had only payed attention to the first citation and the citation of the reprint by the RS.

I actually had found the RS paper, let me see if I can get it again.

2011-03-28 03:43:09
Alex C

coultera@umich...
67.149.101.148

Here you go (here).

2011-03-28 11:12:49
James Wight

jameswight@southernphone.com...
112.213.158.190

grypo, what's up with the red line in the 20th century?!

2011-03-28 12:19:50
grypo

gryposaurus@gmail...
173.69.56.151

No need to apologize.  It fooled me too.  

 

I'm glad I caught it because reading the 'methods' section of the volcano paper was enlightening as to the data we are discussing.  I'm pretty sure that I have a good idea what happened and a couple other people are pretty much saying the same as what I'm thinking.

See Nick Stokes and thefordprefect comments in this CA thread.

 

Especially see Nick Stokes in this AV thread.  Notice what he says about sample size.

 

I think we can really hit back hard on this one with an extensive blog post citing the differences between real science and blogscience.  Some high profile 'skeptics' jumped the gun and said some incediary things and the whole episode went viral.  I'm thinking 'team' effort (pun intended) on this one.  Anonymous grypo on the interwebs won't really attract any attention.  Plus, I'm lacking the scientific background to be very specific.  Let me know what you all think.

 

James, the end points aren't smoothed, so it looks a little messy, but the main curve in the 20th century is the famous 1960 divergence in the original 'hide the decline'.  And the reason the it goes higher than the BO99 graph is the entire graph was normalized to a different zero point by about -.25    That is only an assumption.

SEE ABOVE FOR CORRECTIONS TO REWIND POST

Posted Mar 24, 2011 at 12:24 PM | Permalink | Reply

Anthony Watts

From my perspective, with this example it appears they have crossed the threshold from simple incompetence and confirmation bias into the realm of academic misconduct.

2011-03-28 15:30:05comment
Robert Way

robert_way19@hotmail...
142.162.22.136

sample distribution is also important here. maybe pre 1600s the spaitial distribution was low

2011-03-29 00:03:40
grypo

gryposaurus@gmail...
173.69.56.151

Yes, the volcano paper touches on that.

Two methods of using the site density series to construct a meanNH record have been explored. In each, time-dependent changes insite spatial coverage are accounted for by adjusting the variance, inaccord with the diminishing number of sites, especially in the earlydata (see Methods section). The two derived records are, nevertheless, highly correlated (r ¼ 0:92, AD 1400–1990) and similarextreme values are indicated in the lowest 5% of the distributionof each of them (Table 1). The (preferred) NH density record,NHD1 (Fig. 1), formed as the average of 5 subcontinental-scalemean density series (Fig. 2), correlates with a series of instrumentalsummer (April–September) average temperatures9combined overthe areas corresponding to the chronology sites2at r ¼ 0:76 (AD1881–1960; p p 0:01 after correcting degrees of freedom for autocorrelation).

I've also noticed a marked goalpost shifting on the threads from hysterical giddiness about activist scientists deleting data to the choice of where the deletion began.  So it is not that scientist deleted data, it's they haven't properly informed their future auditors as to what they choose to do for end smoothing and on what year they decided to do it.  Perhaps a blog post here isn't necessary, we can just wait until they are informed, and wait for complaints about the choice of font and how that was really hard to read.  

The ironic part of this is that they keep saying that the deletion was convenient, but if you look at what they deleted, it was data which, given equal weighting, would have diminished the MWP!  Those damned scientists are always trying to make the middle ages look warmer!  Frauds!  Oh, and look, they archived their raw data!  The openess of the scientific process makes it easier to fool ourselves!

2011-03-29 07:12:13Journal of Coastal Research
oslo

borchinfolab@gmail...
90.149.33.182

WUWT has propaganded the following story today:

http://wattsupwiththat.com/2011/03/28/bombshell-conclusion-new-peer-reviewed-analysis-worldwide-temperature-increase-has-not-produced-acceleration-of-global-sea-level-over-the-past-100-years/

What is this journal anyway?

How are they able to provide such crap?

Do they have any review process at all?

The story has been mentioned in various media (forums) in Norway, but thorowly debunked on basis of narrow geographical area of observation.

Is there no decensy at all?

2011-03-29 10:46:04The Great Karnak says...
Daniel Bailey
Daniel Bailey
yooper49855@hotmail...
97.83.150.37

What is this journal anyway?   Crap

 

How are they able to provide such crap?  They try very, very hard

 

Do they have any review process at all?  Only on days that do not end in Y

 

Is there no decensy at all?  None, just in-decensy

 

GIGO - Garbage In, Garbage Out

2011-03-29 16:26:17Discussions at CA
Robert Way

robert_way19@hotmail...
134.153.163.105

Here is my commentary from over at CA

If you wanna call this:

Adequate spatial sampling then that is your choice but anyone can see pretty clearly that pre-1500 two regions would dominate the variability in the reconstruction. So considering 3 regions have a total of 3 series and 2 have a total of 21 series my criticism is quite justified. Would you using temperature compute a regional composite with that distribution of thermometers and call it accurate… not a chance!

To further my point: http://treesfortheforest.files.wordpress.com/2010/05/ghcn_global_temp_area_1961-1990_5x51.png

Is the pre-1880 data in this graph reliable? Well no, but then why isn’t there an accusation of hiding inconvenient data? Well simply because the sample density is too low to include past whatever threshold. Yes the authors should have been more clear but this foray is counterproductive.

And for the record RomanM I know you have considerable statistical expertise but if you call that spatial distribution (pre-1500) as being adequate then it is your credibility on the subject that should be questioned, not mine.
It is obvious from that sample distribution that no full composite could accurately represent all 5 regions pre-1500. How can you defend your position by saying 4/5 regions are included? Yeah they are but 87% of the data comes from 2 regions. Bias much?

Further supporting my point from Briffa 2001 comes:

“Bias might be introduced in cases where the spatial coverage is not uniform (e.g., of the 24 original chronologies with data back to 1500, half are concentrated in eastern Siberia) but this can be reduced by prior averaging of the chronologies into regional series (as was done in the previous section)… Eight different methods have been used… They produce very similar results for the post-1700 period… They exhibit fairly dramatic differences, however, in the magnitude of multidecadal variability prior to 1700… highlighting the sensitivity of the reconstruction to the methodology used, once the number of regions with data, and the reliability of each regional reconstruction, begin to decrease. The selection of a single reconstruction of the ALL temperature series is clearly somewhat arbitrary… The method that produces the best fit in the calibration period is principal component regression…

“…we note that the 1450s were much cooler in all of the other (i.e., not PCA regression) methods of producing this curve…”

2011-03-29 22:48:02
grypo

gryposaurus@gmail...
173.69.56.151

Robert,  very good.

McI's responses are now getting more specific and defensive.

Whether or not they provided adequate justification for PC methods in Briffa et al 2001 is also immaterial to the issue of whether the deletion of data in Briffa and Osborn 1999 was done because of insufficient replication as defenders have alleged. if the replication was insufficient in Briffa and Osborn 1999, then it was insufficient in Briffa et al 2001.
No one has refuted this point.

In my post '26 Mar 2011, 6:09 AM' above, I copied a piece from Briffa 2001 that explained the use of age banding for that data.

 

Also, getting even deeper for a full refutation of McIntyre --

I don't have time this morning (east coast USA) but I found this very important dissertation. It's Long.

http://www.cru.uea.ac.uk/cru/pubs/thesis/2004-melvin/melvin-2004-thesis.pdf

Historical Growth Rates and ChangingClimatic Sensitivity of Boreal Conifers

Thomas Michael Melvin

 

It is beyond very likely that the data Briffa uses is Luosto, Finland and Helldalisen, Norway.  This paper is all about that data.  Page 28 in the dissertaion.  I found 4 sites with that data on NOAA's website.

http://www.ncdc.noaa.gov/paleo/metadata/noaa-tree-3972.html

http://www.ncdc.noaa.gov/paleo/metadata/noaa-tree-3971.html

http://www.ncdc.noaa.gov/paleo/metadata/noaa-tree-3975.html

http://www.ncdc.noaa.gov/paleo/metadata/noaa-tree-3974.html

For someone who is good with data, this could be the treasure trove.

page 29

Crossdating was achieved using both visual comparison of cores and correlation statisticsobtained from program COFECHA (Holmes 1986). First, the measured ring-widths of allcores for each tree were crossdated against each other and second, all trees within eachregion were crossdated against each other. Sections of cores for which crossdating wasuncertain were removed from the chronologies: generally these were where growth wasslow and there were coincident missing rings in both cores. For trees with a missing ringon one core and a ring on the other core the missing (zero measure) ring was inserted.Missing rings (zero measure) were inserted for trees that were found to have a missingring on both cores in situations where the missing ring could be inserted with confidence.Details of these crossdating activities were recorded. Series of ring-width measurementsrepresenting each tree were created as the arithmetic mean of the series from each core ofthat tree. Due to low replication of trees, some of the earliest segments of the oldest treeswere removed leaving the Luosto and Helldalisen site chronologies both commencing at 1550 and the Rutajarvi site commencing at 1700. With no need to use the earliest data ofthe “modern” chronologies, this action reduces the problems of increased variance anddecreased signal to noise ratio which are dependent on reduced sample depth (Briffa &Jones 1990). Crossdating was subsequently confirmed by correlation of the sitechronologies with the Tornetrask (Grudd et al. 2000) chronology

About 2001 p. 12

Standardisation is the name given to the process by which series of ring measurementsare converted into a series of chronology indices which represent the magnitude of theannually-varying “common” forcing on tree growth (Cook et al. 1990). Standardisation isa large subject area and it is necessary to limit the scope of the description given here.The focus of this project is on long-timescale variability: the effects on its expression inring-width chronologies and the extent to which it reliably represents the influence ofchanging temperatures in northern Fennoscandia. Two types of standardisation methodsare distinguished here by whether they can or cannot represent variance at timescalesbeyond the length of individual tree-ring records to which they are applied. Those thatcannot are generally based on “curve-fitting” techniques. The variance represented bycurves fitted through the measurement series from individual trees is removed. Thosemethods that can represent variance at timescales beyond that of the length of individualtrees in a chronology can be referred to as “Regional Curve Standardisation” methodsbecause a regional growth curve is used on all trees. Fitting a horizontal line, the mean ofa series of measurements, can preserve variance to the length of the series and is a specialcase of curve-fitting while fitting the same horizontal line to all series, which canpreserve variance beyond the lengths of individual series, is a special case of the RCSmethod. The Ageband Decomposition (Briffa et al. 2001) method is a special case of theRCS method in which tree measures are averaged by ring age class and division is usedto create series of indices for each ring age class instead of for each individual tree

2011-03-30 04:34:02blog post
dana1981
Dana Nuccitelli
dana1981@yahoo...
64.129.227.4

There's been a request that we do a blog post on this subject, which I second, once we can get all the facts straightened out.  Maybe grypo or Robert could handle it?

2011-03-30 06:06:02
grypo

gryposaurus@gmail...
173.69.56.151

I'll definately pitch in what I can, but I'm not knowledgeable on dealing complicated data, code, or stats.  What I can do is point out the two comments in McIntyre's "the other deletion" thread that we should deal with.  One is 

(1)

Needless to say, one of the reasons for the reader being “uninformed” is the deletion of adverse data (both before 1550 and after 1960) to give the impression of “corroboration” of the “general validity” of the reconstructions. Note that smoothing after deletion enhances the impact of the deletion – look at the strong divergence pre-1550 as well as post 1960.

It is quite clear what is being inferred without actually saying it.

2ND:

(2)

How did Briffa get from the 1999 reconstruction version (magenta) to the 2001 version (limegreen)?

This is an interesting exercise that I’ll describe in more detail on another occasion. But I can’t resist a quick preview. The Briffa-Osborn 1999 version was based on averages of all available sites – a sensible enough procedure. Changes in the Briffa 2001 methodology include the calculation of regional averages followed by stepwise principal components. (The methodological description in Briffa et al 2001 is very sketchy and, unfortunately, the Climategate computer dossier didn’t include Briffa’s programs.)

The effect of using principal components on regional averages is to change the weights for individual sites, including the possibility of negative weights i.e. flipping the regional MXD series. In particular, the closing uptick in the Briffa 2001 reconstruction may well depend on the flipping of data – a point that I’ll try to examine in the future.

What he does here is makes sure that even if there is an answer to (1), then how did Briffa get (2).  He has a certain style that infers things without actually saying them.  This is what needs to be focused on because this is the dogwhistle that makes the bull go viral.

When those questions are answered (removing any ideas of fraud or misconduct) then there is the problem that the statisical analysis may not be correct, which, I assure you, McIntyre will claim, and sort-of already has (2).  This is where we point out that whatever his problem is with the analysis needs to be put into the literature.  In fact, we should demand it.  If his ideas are correct then he is doing science a disservice by promoting buzzwords like 'upside-down', 'bodge', 'trick' 'flipped' etc instead of getting his important work in front of the commuity that needs to see it.  

In other words, we point out the difference between what McIntyre does and what dendrologists do.  This can be high impact or fall flat depending on presentation.

2011-03-30 20:49:07
Ari Jokimäki

arijmaki@yahoo...
192.100.112.210

Isn't here quite simple to just show that McIntyre is using wrong data (none of the references in B&O99 shows the 1500-decline)?