Comments on: More on Screening in Gergis et al 2012

By: Karoly and Gergis vs Journal of Climate « Climate Audit

Karoly and Gergis vs Journal of Climate « Climate Audit — Tue, 30 Oct 2012 16:03:56 +0000

[...] June 10, a few days after the Gergis-Karoly-Neukom error had been identified, I speculated that they would try to re-submit the same results, glossing over the fact that they had changed the [...]

By: Skiphil

Skiphil — Tue, 18 Sep 2012 10:05:29 +0000

link for the PDF document with summary of PAGES AUS2k workshop as quoted above: Univ. of Western Australia hosted PAGES Aus2k working group in April 2011

By: Skiphil

Skiphil — Tue, 18 Sep 2012 10:04:10 +0000

Steve and all, I found this interesting although there’s not much detail — but it does seem that Gergis et al (2012) grew out of this process, and it is said here that they were aiming at a special issue for AUS2k in the Journal of Climate with up to 15 papers to be submitted. Vice Chancellor of UWA welcomed the conference with opening remarks, so he and UWA admin are not exactly unaware of climate issues, although what they may think about Lewandowsky at this point is another question.

Anyone know if the eventual Gergis et al (2012) was to have been part of such a special issue or not? I see that the Volume 25, Issue 14 (July 2012) has approx. 8 papers which deal with tropical Pacific, southern hemisphere, and/or AUS although there are various other papers so the issue is not devoted to SH — still it could be they were trying to have as large a group of papers as possible for that issue. Wonder what current status is for Gergis et al (2012), they seem to be very quiet down under….

2nd Australasia2k Network workshop: Data synthesis and research planning

Meeting of the PAGES Aus2k Working Group – Perth, Australia, 27–29 April 2011

Joëlle Gergis1, P. Grierson2, A. Lorrey3, J. Palmer4 and S.J. Phipps5

“The Vice-Chancellor of the University of Western Australia, Professor Alan Robson, opened the workshop….”

…. [and in conclusion] …..

“The workshop wrapped up on a very
productive note with the development of
a proposed list of 15 papers for consideration
in the Journal of Climate Aus2k special
issue, and a clear direction forward to
deliver Australasia’s best available science
for the Regional 2k global synthesis.”

By: KnR

KnR — Sun, 08 Jul 2012 20:51:59 +0000

NO this paper is abut AR5 , once in there no matter how bad it becomes unchallengeable as accepted ‘wisdom ‘
This is political game not a science one , what has been has been clear since the first ‘hockey stick ‘ , the quality of the science is a very poor second to the political usefulness of the papers claims .

By: KnR

KnR — Sun, 08 Jul 2012 20:45:01 +0000

‘I remain mystified by Thompson’s intransigence in establishing a comprehensive and meticulous archive of his measurement data’

Information is power , never forget that the control of the data gives those controlling it power over how its used .
The professional becomes the personal , in some cases the researcher start to feel more like a ‘parent’ to what they see has theirs, especial if then have had it for long time and invested a great deal in it , and who gives their children away .
I will do it later, this happens when the researcher does not bother to clean there data or organize in such a way that others can understand it, they mean to short it out but they never get around to it and has the data gets bigger the task gets harder so keep getting put off.

They are merely human after all and there are a number of reason why they may fail to archive data that are not becasue they are being tricky but becasue they just being human . The issue is none of these have a roll to play in the scientific approach that demands a behavior which some fail to match .

By: Kenneth Fritsch

Kenneth Fritsch — Mon, 02 Jul 2012 22:49:23 +0000

I have made some progress in determining the chronologies used for the Standard or .crn result in NOAA for the 33 not-Gergis TRW series from the Australasia region. The Spline method in the R function detrend replicates the .crn series almost exactly when applied to the individual trees and then combining those tree chronologies into a series. Using the Modified Negative Exponential (MNE) method under the same conditions produces more of a series ending upward trend not unlike that produced by the rcs function in R. I thought I might be on to something until I realized that the rcs function in R using a Spline method and not the MNE method.

I have parsed the R code so that I can do both the Spline and MNE methods with the rcs function in R.

By: Kenneth Fritsch

Kenneth Fritsch — Sat, 30 Jun 2012 18:02:11 +0000

It is a little late in the day and thread, but I finally used the R library (dplR) to download the raw TRW series (33 not-Gergis from the Australasia region) from the NOAA repository (linked below) and produced chronologies with the rsc function in R. A good summary of the R program is linked below.

http://www.ncdc.noaa.gov/paleo/treering.html
http://www.sciencedirect.com/science/article/pii/S1125786508000350

I was curious whether the RCS (Regional Curve Standardization) would produce the TRW series with series ending upward trends from which Gergis was able to select. Gergis used RCS and combined that with an iterative method that was described by Melvin and Briffa here.

http://www.sciencedirect.com/science/article/pii/S1125786508000374

I have contacted Melvin (with no reply to date) about an outline that would allow me to combine the iterative process with RCS in an R program.

I had a concern about using the rcs function in R in that it asked for the pith offset. That piece of meta data is not available with most of the TRW series in the NOAA repository and thus I had only the option of setting it to 1 for all the TRW series. The value of the offset would appear to be downplayed by Esper in the article linked below.

http://www.buentgen.com/uploads/2/2/0/9/2209799/esper_etal.09_trace.pdf

Before continuing with my analysis of the RCS results here I should link to a very good post by RomanM at CA explaining the basis for RCS and his analysis of the Yamal data with it.

http://climateaudit.org/2009/10/21/rcs-one-size-fits-all/

What I found in a direct comparison chronologies of these not-Gergis TRW series was that, indeed, RCS tended to produce the series ending trends more than the Standard or ARSTAN chronologies. The comparison can be viewed in the links below.

http://img838.imageshack.us/img838/1384/notgergistrwrcsstand1.png
http://img96.imageshack.us/img96/5503/notgergistrwrcsstand2.png

While even with RCS, the TRW series appear to be more or less random editions of red noise, the RCS chronologies produced visually perhaps as many as 5 candidates for the Gergis selection process. It is these 5 series that I studied further. I found that grouping the trees by age produced different series ending structure with the older trees producing the upward trends and the younger trees producing little or no trends. I thought that the effect might be the result of the mixing of a tree ages in the RCS process to obtain the regional growth curve. When I obtained much the same differential results when I either used only the older and younger trees to produce 2 separate chronologies or used all the trees in chronology and then extracted the older and younger tree series from that chronology, I think I can conclude that it was not the mixing of tree ages but the trees in the two age groups that gave the differences.

I link both the 5 series below with the grouped tree ages and the growth curves with the tree ring widths plotted by cambial tree age for the entire series. There appears to me to be less of the expected fall off in the growth curve that is modeled by a negative exponential, but that might be because I have not viewed that many TRW series growth curves.

http://img685.imageshack.us/img685/1825/notgergisoldyoung.png
http://img72.imageshack.us/img72/7053/notgergiscambialagegrow.png

I need to look further into exactly what the rcs function is doing in R and determine whether I can produce the Standard and ARSTAN chronologies in R as these data appear in the NOAA repository. The Standard and ARSTAN processes deal with the TRW series individually in detrended the natural growth while the RCS process uses all the data arranged by cambial age to produce a growth curve for detrending. My initial look, given the TRW data, did not reveal an obvious cause of the difference that RCS method produces compared to the Standard and ARSTAN methods.

If anyone is interested in looking further at these TRW series I analyzed here I give the urls below to the raw tree ring data from the NOAA repository:
urls to raw TRW data and for rcs in R:

By: Kenneth Fritsch

Kenneth Fritsch — Mon, 25 Jun 2012 00:54:27 +0000

SteveM, I sent you my Excel spreadsheets of the 33 Australasia TRW series with data from the both the standard and ARSTAN chronologies from the NOAA repository, but I suspect what you wanted was the raw data so that you could obtain the RCS chronologies for the raw TRW data. I plan to obtain all the necessary urls for these raw TRW data which I can send to you. I planned to obtain and run these data with the R program in library(dplR) and obtain RCS chronologies on the 33 series. We could compare results if you use your nls program. The dplR looks like it will be simple to run and helpful in putting the NOAA data into proper form.

The program dplR requires a pith offset value to run and after reading the Esper article linked below I suspect I can default that offset to 1 and not worry about it affecting the results. I have read a CA thread which discussed the pith offset. Do you have any thoughts on the matter?

http://www.buentgen.com/uploads/2/2/0/9/2209799/esper_etal.09_trace.pdf

Using RCS would put me one step closer to the process that Gergis used for her TRW proxies. What is left is the Melvin-Briffa iteration steps. I have emailed Melvin about providing sufficient information about the iterations such that I could code it in R.

Steve: I have collations of all measurement data at NOAA.

By: Kenneth Fritsch

Kenneth Fritsch — Sat, 23 Jun 2012 02:58:00 +0000

AMac, I did not actually realize how much infilling was required in Mann (08) until I really looked at the proxies and then read what was in the SI. Actually if you read carefully and in detail it is all there in Mann (08) and the SI. In-filled data was used in screening test also. It was actually 105 MXD series that were looped off at 1960 and then in-filled.

The Mann (08) SI at some point summarizes were the proxies end and thus you can conclude how much in-filling was done. Or you can go to the data through an SI link and make that determination yourself.

By: amac78

amac78 — Sat, 23 Jun 2012 00:47:25 +0000

Re: Kenneth Fritsch (Jun 22 13:42), > Mann (08) used many TRW series from New Zealand that ended in the 1970s. For the reconstructions Mann (08) these series (amongst many others) were in-filled for 20 plus series ending years using other proxy series. Off topic but perhaps interesting: if I am reading this correctly, Kenneth is stating that many of the NZ TRW series used in Mann08 ended in the 1970s. Thus, for these series, the final 15 to 24 years culminating in 1995 (the terminus of the screening period) were generated by Mann08's authors through infilling (via the RegEM procedure). As best I can tell -- I welcome correction on this point -- "infilling via RegEM" simply generates numbers that plausibly fit the pattern of the time series, past the point where actual data do not exist. There was no need to stop in 1995, as the technique could have supplied values through 2010, or 2035. Mann08's authors used two sets of calibration and validation times. An early (1850–1949) calibration/late (1950–1995) validation, and a late (1896–1995) calibration/early (1850–1895) validation. (Mann08, PNAS, pg. 13254). Assume a NZ TRW series was "infilled" from 1976 on. That would mean that for that proxy, imaginary data was generated for the final 25 years of that series. This would have accounted for 44% of the validation time period under the early calibration/late validation scheme, and for 20% of the calibration time period under the late calibration/early validation methodology. Say it ain't so! Steve: For the 76 or so Briffa MXD series in Mann et al 2008, they didn't just infill missing values, they deleted post-1960 decline values, infilled post-1960 values and used this for calibration/validation.