Esper et al 2016 and the Oroko Swamp

Jan Esper, prominent in early Climate Audit posts as an adamant serial non-archiver, has joined with 17 other tree ring specialists, to publish “Ranking of tree-ring based temperature reconstructions of the past millennium” (pdf). This assesses 39 long tree ring temperature reconstructions. The assessment is accompanied by an archive containing 39 reconstruction versions, together with the underlying measurement data for 33 of 39 reconstructions. (It seems odd that measurement data would continue to be withheld for six sites, but, hey, it’s climate science.)
Because I’ve been recently looking at data used in Gergis et al, I looked first at Esper’s consideration of Oroko, one of two long proxies retained in Gergis screening.  I’ve long sought Oroko measurement data, first requesting it from Ed Cook in 2003.  Cook refused. Though Oroko reconstructions have been used over the years in multiproxy studies and by IPCC, the underlying measurement data has never made archived  The archive for Esper et al 2016 is thus the very first archive of Oroko measurement data (though unfortunately it seems that even the present archive is incomplete and not up-to-date).
Despite claims to use the most recent reconstruction, Esper’s Oroko temperature reconstruction is decidedly out of date.  Worse, it uses a n Oroko “reconstruction” in which Cook replaced proxy data (which went down after 1960) with instrumental data (which went up) – in a contemporary variation of what is popularly known as “Mike’s Nature trick”, though Mike’s Nature trick, as discussed at CA here, was a little different.
In today’s post, I’ll look at the “new” Oroko data, which, needless to say,  has some surprises.

My 2003 Attempt to Obtain Oroko Data
To give an idea of just how long Oroko measurement data has been withheld, I originally suggested to Cook in October 2003 that it be archived.  Although ATTP and others occasionally claim that I was insufficiently polite in such requests (as a supposed justification for data obstruction), this was not the case, either in this instance or others.

Dear Dr. Cook,
While you have contributed consistently to WDCP, I notice that the Oroko Swamp data has not yet been contributed to WDCP. I presume that this is an oversight and hope that you plan to continue your previous support of a commendable program.
I notice that the data from the Taimir, Yamal and some of the older Tornetrask sites have likewise not been contributed to WDCP, although these sites have been referred to in well-publicized articles. Perhaps you could encourage your collegues to contribute this data to WDCP.
Yours truly,
Stephen McIntyre

On October 19, 2003 at 11:49 a.m., Cook refused the request, expressing some astonishment that I would “apparently expect [him] to continually update the WDCP as the data are generated” and suspicion at my apparent interest in the long data.

Hi Stephen,
As you say, I have contributed consistently to WDCP and will continue to do so, but not always at the pace that you would apparently like. It took me ten years to finish the 4,000 year long Mt Read, Tasmania chronology, which is now fully available on the WDCP. This work is not trivial to do, nor can it be rushed. I have been going down to New Zealand once a year to get more wood related to the Oroko Swamp project and continue to do so in an effort to solidify the chronology and push it back in time. From what I gather from your email, you apparently expect me to continually update the WDCP as the data are generated. I don’t regard that as a very practical way to go since it would lead to a very confusing collection of overlapping, but different, data sets. So, it was not an oversight on my part. Rather, the project is still in progress. I don’t think that I need to apologize for that.
I have no control over other people’s data. I suggest that you contact them directly. Do I dare ask why you appear to be so interested in these long tree-ring data sets?
Regards,
Ed

When I first collated these old emails for this post, I presumed that I would then have been unknown to Cook.  At the time, McIntyre and McKitrick  2003 had not been published.  Out of interest, I checked contemporary CG2 emails and  noticed that a half-hour before Cook’s refusal, Mann had sent out an email (CG2 thread 1566 – rhis email wasn’t in CG1) to the climate community, ever vigilant for dissidents, warning them that the previously unknown “McIntyre” was “yet another shill for industry”, also accusing me of “trying to break into” the FTP site that he would claim a couple of weeks later to have been “public” all along.

At 11:14 19/10/2003 -0400, Michael E. Mann wrote:
FYI–thought you guys should have this (below). This guy “McIntyre” appears to be yet another shill for industry–he appears to be the one who forwarded the the scurrilous “climateskeptic” criticisms of the recent Bradley et al Science paper.
Here is an email I sent him a few weeks ago in response to an inquiry. It appears, by the way, that he has been trying to break into our machine (“multiproxy”). Obviously, this character is looking for any little thing he can get ahold of.

p.s. I’m setting up my email server so that it automatically rejects emails from the “usual suspects”. You might want to do the same. As they increasingly get automatic reject messages from the scientists, they’ll start to get the picture…

It’s possible that the timing and form of Cook’s refusal was a coincidence, but it’s also possible that it wasn’t.  In any event,  far from Cook being inconvenienced by “continually” updating the archive as new data became available, Oroko data remained unavailable.
The Oroko Reconstruction in Esper et al 2016
In the first figure (below), I’ve compared the Esper-2016 temperature reconstruction to a chronology that I calculated from the measurement data, both series being scaled to facilitate comparison.  Prior to 1958, the temperature reconstruction and my chronology estimate track very closely, with the chronology dated one year earlier than the reconstruction.  In a SH context, this means nothing more than the tree ring measurement data has been dated to the calendar year opening the austral summer (consistent with a convention reported at NOAA Paleo), while the reconstruction has been dated to the calendar year closing the austral summer (consistent with Jan-Mar calibration reported in early Cook articles on Oroko).
However, after 1958, the chronology from measurement data and the Esper-2016 reconstruction diverge dramatically.  The Esper reconstruction ends on a high note (more than 2 sd units), while the chronology from measurement data ends on a low note. The divergence between temperature “reconstruction” and chronology arose because Cook spliced instrumental data after 1958 into his reconstruction.
Figure 1.  Comparison of Esper et al temperature reconstruction and chronology calculated from archived measurement data, both scaled.
Obviously the incident is reminiscent of “Mike’s Nature trick” “to hide the decline”, though the details and nuance differ.  These incidents cannot help but remind one of Mann’s vehement denial in 2004 that any researchers had ever engaged in such practices:

No researchers in this field have ever, to our knowledge, “grafted the thermometer record onto” any reconstruction. It is somewhat disappointing to find this specious claim (which we usually find originating from industry-funded climate disinformation websites) appearing in this forum.

Cook purported to justify the exclusion of post-1957 proxy data on the grounds that the “catastrophic loss of climate correlation” was due to disturbance, most likely due to logging:

This catastrophic loss of climate correlation in the surviving trees confirms that a stand-level disturbance occurred. It is likely that the cause is the known logging activity described earlier.

In the first article on the site, Cook noted that there was no actual “evidence of cut stumps” along the transect and posited other possible methods of disturbance, though later articles simply cited logging.

Although there was no evidence of cut stumps along the transect line, the cutting nearby would have opened up the forest and perhaps made it more susceptible to wind-throw. L. colensoi is also known to form root grafts between individuals, producing an interlinked network (Moar, 1955). Alternatively, the hydrology of the swamp may have been altered when drains associated with road construction and light rail access for milling were put into the area.

By the time of Esper et al 2016, the authors (including coauthor Cook) seem to have lost track of the earlier splice of instrumental and proxy data.  One cannot help but think that the splice of temperature data into the reconstruction will give an inappropriate boost to the resulting correlation to instrumental data. It’s therefore odd that Oroko was one of the lower ranking temperature reconstructions in terms of correlation to instrumental temperature.
The “Most Recent” Reconstruction
In cases where there were multiple versions of a chronology, Esper et al said that they used the “most recently published” version:

In those instances where there are multiple versions of a reconstruction, we cite the (GPmost recently published account as it contains references to all previous work

However, this practice wasn’t followed for Oroko. Esper et al used an antique Oroko version which Cook had emailed to Briffa in 2005.  They cited Cook et al 2002 (GPC), the earliest article on the site.  A more recent version is used in numerous recent articles including PAGES2K (Gergis et al, 2012; PAGES2K 2013; Neukom et al 2014; Gergis et al 2016).  These use of Oroko in these studies is complicated somewhat by quixotic and inconsistent lags in various versions (generally undocumented), but the same data underpins all versions.
The most recent version of the Oroko reconstruction (in Neukom et al 2014 etc) does not contain the splice of instrumental data. It goes down after 1960. The underlying measurement data used for the more recent versions of the chronology appears to be different (and presumably larger) than the data set archived with Esper et al 2016.  For example, the recent chronology version contains three additional years past the most recent measurement (1999) in Esper’s Oroko archive.

Figure 2: Oroko versions: comparison of Neukom et al 2014 unscreened to chronology calculated from measurement data in the Esper 2016 archive. The Neukom et al 2014 unscreened version is the most complete of the various recent versions.
 
 
The long Oroko chronology (in the unscreened Neukom 2014 data) is shown below. To my eye, its modern values are unexceptional over the past two millennia.
 

Figure 3. Oroko temperature reconstruction (from Neukom et al 2014 data)
Conclusion
Esper et al 2016 is an interesting article and the authors have made a reasonably diligent effort to ensure that the relevant data is available.  I haven’t examined other datasets yet. Data issues tend to be idiosyncratic and the issues impacting Oroko are unlikely to impact other datasets.  I’m glad to see that some Oroko measurement data has finally been archived  – thirteen years after my original request – but it’s all too typical that, when finally archived, it appears to be a version that could have been made available in 2003. So despite Cook’s original sarcasm at the idea that he  should be expected “to continually update the WDCP as the data are generated”, the data, when finally archived, appears not to incorporate data from the past decade. And worse, the reconstruction in which instrumental data was spliced with proxy data continued in use without warning labels, the authors of Esper et al 2016 seemingly unaware of or forgetful of the original splice.
 
 

Source