Originally tweeted by Stephen McIntyre (@ClimateAudit) ABOVE April 30, 2023.
Note from Anthony: For those of you who remember the yeoman work Steve McIntyre did at the Climate Audit exposing Michael Mann’s hockey stick and flawed methodology, this should come as no surprise. Again, hockey sticks are generated when the original data is not visible. The only conclusion you can draw is that the data and methods have been adapted to the desired and predetermined results. This series of Tweets has been compiled here for easier access and readability. – Anthony
[M]Most readers are familiar with Climategate’s famous “hidden decline” phrase. Here are the 1850-2000 parts of the 5 series calculated from the Asian tree ring data, explained below. I recently got some great PAGES2k reverse engineering from @detgodehab and visiting again.
The data illustrated below comes from
(1) Asian-origin 2001 Briffa series with late 20th century decline (cut out of Mann’s IPCC scheme);
(2) average of Asian series in MXD series with reticle sent by Briffa/Osborn to Rutherford and Mann, superficial input in Mann 2008
(3) the average of (45) meshed MXD as used in Mann 2008. As discussed long ago at the Climate Audit, Mann removed the breach reductions and replaced them with thermal data. degree. This is a different incident from the IPCC chart or the 1999 WMO “hidden decline” chart.
(4) PAGES2K (2013) introduced a novel Asian reconstruction (Cook et al) from tree rings where the late 20th century C decline observed in the Schweingruber data was non-existent. . The close value is similar to the mid-20th century high. The difference is not adjusted for PAGES2K.
(5) the current PAGES reconstruction – “Woke Reconstruction” for short and used in IPCC AR6 – contains a subset of the Asian PAGES2K subset, the average of which produces a bar monster sword. “Reject” is in Woke’s rearview mirror.
A few years ago, I noticed that some tree ring chronologies located in the Woke Reconstruction had giant closed blades that seemed impossible to reproduce by standard methods.
I also made some threads on Twitter:
I asked the two lead authors of PAGES 2019 about the origins of the Asian tree ring series, but got nowhere. They do not assume that they have any responsibility as lead author of a Nature paper to answer questions about their data.
The Basic Reference (Cook et al. 2013) contains only a single sentence as a purported description of chronological methodology: that they took “careful care” to avoid the ‘curse of fractional length’ paragraph’, with “partial use” of a new technique recently introduced by UEA’s Tom Melvin
the keepers of these chronologies at Columbia U, which adamantly denied the data as I was trying to figure out the mysteries of the Hockey Stick. Jacoby: “Fifteen years is not a delay. The time has come for poorer quality data to be forgotten and not archived.”
Either way, @detodehab readers were intrigued by the puzzling chronology of the Asian tree ring and their computational reverse engineering. He reproduced the results down to the last detail. No one can imagine the actual calculation from the details in PAGES2K or Cook et al 2013.
It’s hard for a statistical method to be “wrong”. Mann’s principal component method is a seemingly unique example. PAGES2K’s Asian tree ring chronology is something else. It’s worse than anyone can imagine.
unfortunately, the interpretation of the faulty calculation is technical and will take some time. But for now, the monster blade in the Asian tree ring data “Woke” PAGES 2019 is bogus. PAGES2019 has selectively handpicked the biggest cards, nearly all of which come from bogus chronology.
I have a question for the readers in order of presentation. Which should come first:
1) narration of detective work where the calculation is reversed
2) pathology of the Asian PAGES2K tree ring method
Not all Asian chronologies are pathological, but the largest blades are.
on the left is a chronological description of PAGES Asia2K and on the right is my brief description of their actual algorithm inferred by @detgodehab and verified by me. I’ll come back to it later on how he figured this out. Currently, the pathology of the method.
I will describe more or less the pathologies of the Asian PAGES2K algorithm as we discuss them in the DM over the past few weeks. @detgodehab started with analyzing paki033, the series that I featured in a blog post and theme for 2021.
@detgodehab ported the algorithm to Octave. The paki033 repeat was stopped after 20 repetitions. So let’s take a look at the evolution of chronology. It opens as an unremarkable series on the left and closes with a large blade. On the right is the sequence of steps showing the appearance of the closing blade
what happened to the individual cores? The “date” of the tree ring is calculated as the difference between the measure and the smoothness (pseudo-model). Between the start and the end, the ‘model’ gets closer to zero at the end, so the contribution to the chronology (shown right) increases dramatically
this looks very suspicious as a procedure. An obvious question is whether the monster blades of certain PAGES2K chronologies are some sort of artifact, as opposed to “climate”. Therefore, I suggest that @detgodehab see what happens when the data from the past 50 years is not used? As a test.
bingo. Except for the data of the last 50 years, paki033 has an even bigger blade // 50 years ago //. For good measure, @detgodehab tested for 25 years and got the same big blade // 25 years ago //.
so it is very clear that the large blade produced by the chronology of PAGES2K Asia is bogus and some kind of artifact according to their method and NOT the climate. This does not rule out global warming. It is only related to PAGES2K.
In addition, PAGES2K uses many other data. Nor are all 2K chronologies of Asia calculated using this bogus method. But PAGES2019 has picked the worst and most bogus chronologies (claiming that they are the best) and that’s why there’s a monster blade in the opening tweet.
while obviously the Asian PAGES2K method is sick, when @detgodehab cut and slash for 100 years without a tongue. I didn’t parse this to see why. There are many mysteries in the algorithm that I will continue to describe.
In an iteration, convergence is assumed. But the PAGES2K algorithm does NOT converge for paki033. It stopped at the 21st iteration. I ran D’s algorithm for 100 iterations and noticed that Melvin’s “test statistic” (which is odd) went up to ~68 iterations, then dropped all of a sudden , then increase
this is NOT acceptable behavior in a valid algorithm for the purpose of convergence
what happened to the chronology of these wild changes in ‘converging’ statistics? The blade (stopped at ~4) continued to grow to ~37 at iteration 50, then dropped to ~15.3 at iteration 100. Obviously not climate-appropriate.
blade max by repeating for 500 repetitions is shown below. It did indeed converge according to Melvin’s statistics – but with an unbelievably large blade of 15,338.
@detgodehab observed “such a clear convergence does not mean that the date is valid”. Clear.
recall the diagram shown previously where i extracted the smooth ‘model’ for an individual core. On the right are the ‘models’ for each core at convergence: they all progress towards the end point.
a technical point: there are two main approaches to “decreasing” each core to allow the subtree to grow: 1) a separate curve for each tree/core; 2) a curve (usually negative exponential plus constant) for the site.
I did a test applying Melvin’s iteration and smoothing supsmu as follows: 1) for individual cores as done in Asia PAGES2K; 2) a curve for the entire site (“RCS”). The monster blade only appears with a combination of PAGES of Melvin iteration and supsmu core smoothing.
so far, i have discussed a paki033 website.
@detgodehab verified that the same flawed algorithm was used for at least 8 other Pakistani sites. Note that these sites (along with Columbia U’s Mongolian chronology) dominate the list of major contributors to blade closing.
Please also note that all of these sites are prioritized for selection under the PAGES2019 screening process (which I have criticized elsewhere.)
Originally tweeted by Stephen McIntyre (@ClimateAudit) ABOVE April 30, 2023.
Related
