The day before yesterday, I reported that Briffa and Osborn (Science 1999) had not just deleted the post=1960 decline (see also CA here), but had deleted the pre-1550 portion as well – the deletions contributing to an unwarranted rhetorical impression of consistency between the reconstructions, an impression that was capitalized upon in the commentary in the running text of Briffa and Osborn 1999.
Figure 1. Annotated version of Briffa and Osborn 1999 Figure 1. See here and here for derivation.
The dossier of computer programs in the Climategate documents is far from complete – programs from Tim Osborn and Ian Harris are in the dossier, but not programs from either Keith Briffa or Phil Jones.
In the directory osborn-tree6, the program science99_fig1.pro both by its name and contents appears to be the program that was used to produce the figure in Briffa and Osborn (Science 1999) – see here for example, though the program in the Climategate zip file is not dated until Feb 16, 2000, about 9 months after the publication of the article.
But there’s an apparent curious inconsistency between the program in Osborn’s archive and the figure as published in Science (Update Mar 23 pm – reader PaulM has now traced the implementation of the deletion of pre-1550 back further).. Here’s the section of code in which the Briffa reconstruction is retrieved:
2: begin ; Age-banded MXD
alltit=”Age-banded density NH growing-season reconstruction”
; Period to consider
perst=1402
peren=1960
fac=0.0 ; do not smooth it any further!!!
; restore,filename=’../treeharry/densadj_all(330).idlsave’
; timey=x
; ; CONVERSION FACTORS FOR AGE-BANDED MXD, BY REGRESSION ON INSTR.
; ts=densadj*0.156525 ; converts it from density to temperature anom
timey=newagetime
ts=newagets
kl=where((timey ge perst) and (timey le peren),nyr)
timey=timey(kl)
ts=ts(kl)
; ts=ts(kl)-0.140369 ; to convert it oC wrt 1961-90
end
The start period for the reconstruction in the code is 1402 (the start of the magenta portion), rather than 1550 – the start of the Briffa version in the actual graphic.
It’s therefore evident that they had, at one time, plotted the Science 1999 spaghetti graph showing data before 1550, but elected to delete the pre-1550 data as well as the post-1960 data. I presume that there is another version of this program (with the corresponding line reading perst=1550) that was used to generate the figure in the Science article. It’s odd that it isn’t in the Osborn archive.
Update Mar 23 pm: – PaluM observes that the deletion of the pre-1550 portion can be accomplished through the parameter yrmxd, which proves to come from either the directory bandtempNHsm50_calmultipcr.idlsave or bandtempNHsm50_calmultipcr_NSIBhug.idlsave.
There is a parameter timey that is used to shorten the interval in the line
kl=where((timey ge perst) and (timey le peren),nyr)
This parameter timey is set to newagetime which is set to yrmxd in the second line of the code.
So to get the figure truncated to 1550, or any other start date you like, you can just run this code as it is with yrmxd set to whatever number you want, depending on how much data you want to hide from the reader. It is not really a code inconsistency.
The truncated time period newagetime is obtained in the lines:
24] ;restore,filename=’bandtempNHsm50_calmultipcr.idlsave’
[25] ;restore,filename=’bandtempNHsm50_calmultipcr_NSIBhug.idlsave’
[26] ; Gets: nyr,nhtit,yrmxd,prednh,fullnh,predse
The directory bandtempNHsm50_calmultipcr_NSIBhug.idlsave results from a bodge that is mentioned in one of the programs – the age-banded NSIB series did not accord with expectations in some respect and, in some applications, the Hugershoff version was substituted for the age-banded version – as presumably in the directory referred to here.