In today’s post, I’m going to re-examine (or more accurately, examine de novo) Ed Cook’s Mt Read (Tasmania) chronology, a chronology recently used in Gergis et al 2016, Esper et al 2016, as well as numerous multiproxy reconstructions over the past 20 years.
Gergis et al 2016 said that they used freshly-calculated “signal-free” RCS chronologies for tree ring sites except Mt Read (and Oroko). For these two sites, they chose older versions of the chronology, purporting to justify the use of old versions “for consistency with published results” – a criterion that they disregarded for other tree ring sites. The inconsistent practice immediately caught my attention. I therefore calculated an RCS chronology for Mt Read from measurement data archived with Esper et al 2016. Readers will probably not be astonished that the chronology disdained by Gergis et al had very elevated values in the early second millennium and late first millennium relative to the late 20th century.
I cannot help but observe that Gergis’ decision to use the older flatter chronology was almost certainly made only after peeking at results from the new Mt Read chronology, yet another example of data torture (Wagenmakers 2011, 2012) by Gergis et al. At this point, readers are probably de-sensitized to criticism of yet more data torture. In this case, it appears probable that the decision impacts the medieval period of their reconstruction where they only used two proxies, especially when combined with their arbitrary exclusion of Law Dome, which also had elevated early values.
Further curious puzzles emerged when I looked more closely at the older chronology favored by Gergis (and Esper). This chronology originated with Cook et al 2000 (Clim Dyn), which clearly stated that they had calculated an RCS chronology and even provided a succinct description of the technique (citing Briffa et al 1991, 1992) as authority. However, their reported chronology (both as illustrated in Cook et al 2000 and as archived at NOAA in 1998), though it has a very high correlation to my calculation, has negligible long-period variability. In this post, I present the case that the chronology presented by Cook as an RCS chronology was actually (and erroneously) calculated using a “traditional” standardization method that did not preserve low-frequency variance.
Although the Cook chronology has been used over and over, I seriously wonder whether any climate scientist has ever closely examined it in the past 20 years. Supporting this surmise are defects and errors in the Cook measurement dataset, which have remained unrepaired for over 20 years. Cleaning the measurement dataset to be usable was very laborious and one wonders why these defects have been allowed to persist for so long.