Crowley and Lowery [2000] is a staple multiproxy study: it was cited by IPCC [2001] and is used in spaghetti diagrams in Mann et al. [Eos 2003], Briffa et al. [2001], Jones and Mann [2004]. It is one of the "Hockey Team" supposedly supporting MBH98-99. In October 2004, I obtained smoothed and transformed versions used in Crowley and Lowery [2000] and this information is used here.
The email dataset contains 15 series (the same number as listed in Crowley and Lowery [2000]) and a 15-site composite. CL2000 reported that they transformed data as follows in order to calculate a composite:
they first smoothed the data using 25-year smoothing;
they then scaled the smoothed data to a [0,1] scale on the minimum and maximum of the smoothed data;
they then took the mean of the available proxies.
The method (shall we call it the “Crowley transformation”) seems a little primitive, but I've actually have found this transformation to be quite a useful means of visualizing the effect of individual proxies in the various multiproxy studies, because it provides an easy metric for showing the contribution of each proxy to the total in which values remain positive. (The results are probably pretty similar to what you would get with non-parametric quantiles - which would be a worthwhile exercise). Figure 1 below shows the contribution of the individual proxies to the overall Crowley index.
FIGURE 1. Color-coded Crowley and Lowery [2000] Composite Index. Source: Crowley, pers. comm., Oct. 2004
The color coding strikingly illustrates a number of important points.
Although Crowley and Lowery [2000] argued based on their Figure 1, that there is relatively little synchroneity between proxies from different regions (and thus no MWP or LIA), the color-coded graphic of their actual data could also be construed as showing a certain amount of coherence between the proxies.
There is a breakdown in information in the latter part of the record (post 1965) as many proxy records cease to be available. The breakdown is almost total in the 1980s, when the Crowley index consists only of the Polar Urals and Dunde (and then Dunde) proxies. Some of this is from odd choices of proxy sources: the Central England version used by Crowley ends in the early 20th century for inexplicable reasons; the bristlecone versions are very dated series which end in the 1960s. Crowley's answer is to splice the instrumental record after 1965 to the prior proxy series in one version (with an 1870 splice in version CL2). The change in texture seems particularly marked after the termination of the two bristlecone pine series (lowest 2 series) cease to be available around 1965.
one would not naturally term this index as "hockey stick" shaped. In the portion where there is at least 3 proxies, the 20th century values have simply regained MWP values and, without the Dunde singleton in 1987, do not even end on a peak.
A distinctive hockey-stick shape can be discerned in the 4 lowest records. Indeed, whatever hockey stickness exists in this study is entirely due to these 4 series, which consist of 2 bristlecone pine series, Briffa's Polar Urals series and Thompson's Dunde series: all of which are sterotyped proxies. The bristlecone pine series are prominent in the MBH99 reconstruction and the Polar Urals series in the Jones et al. [1998] series. Both series have problems as discussed here and elsewhere.
Without the contribution of the bristlecones and Polar Urals, the MWP peak would be stronger than the 20th century peak as shown in Figure 2 below, which adopts a different editing approach to CL2000, as opposed to the one adopted by Crowley. Instead of deleting the Sargasso Sea and Central Michigan pollen proxies – both of which are rather well-linked to temperature – the two bristlecone pine series are excluded (as not being good temperature proxies) and the first century of the Polar Urals series is excluded on quality control grounds. The post-1965 portion is not illustrated following Crowley’s own admonitions on quality. While the 20th century portion of Figure 1 was not exceptional on its face, the 20th century portion in Figure 4 is even less exceptional.
Figure 4. Color-coded Crowley and Lowery [2000] Composite, excluding the bristlecone pine series and the 11th century portion of the Polar Urals.
Some of these individual proxies will be considered elsewhere. Crowley's White Mountains proxy is Sheep Mountain [Lamarche, 1974], which we have discussed in connection with MBH98. "C. Colorado" is Almagre, also a bristlecone pine site. We have already discussed problems with bristlecone pines. The 11th century portion of the Polar Urals series consists of as few as 3 cores; some of the cores are only 75 years long and are very weakly dated. This series does not meet quality control criteria prior to 1100 at the earliest. The Dunde series is a high-altitude series at exactly the same latitude (38N) as White Mountains. There is specialist concern regarding whether O18 values in tropical glaciers are measuring precipitation amount (rather than temperature). The pattern of O18 values from monsoon precipitation is reversed from the polar types: in the summer, owing to monsoon rainout, O18 is the lowest; in the winter, O18 is higher. There has been a trend of increasing annual O18 values at Dunde, which Thompson attributes to rising temperatures, but an increased proportion of winter precipitation would have the same effect. I have not seen any analysis by Thompson of this matter which rises above arm-waving. Complicating the matter is the fact that Thompson has never archived the original Dunde data (from 1987). Different grey versions have floated around over the years. Most recently, Thompson has archived decadal averages, but these differ from versions used in MBH98 and by Bao Yang. At this point, I merely draw attention to the prominent role of this series (and it will recur elsewhere) in creating stereotyped patterns.
