A common theme to recent questions has been the relationship of ring width chronologies to temperatures, and, in particular, the relationship of bristlecone chronologies to temperature. Rob Wilson has recently weighed in on this.
While it was nice of Rob to present some new material, people should not lose sight of the fact that the North American reconstruction presented by Rob is unpublished and it is based on unarchived material; his other comparandum, Salzer and Kipfmueller, is a very recent publication, again based on unarchived material.
So yes, the material is interesting and deserves consideration, but don’t you think that the matter should have been dealt with somewhere in the literature prior to the last few minutes?
Rob compared a composite of 3 bristlecone chronologies to a North American composite chronology, based primarily on unarchived data from D’Arrigo et al 2006. I’m not going to make a detailed response today. In order to respond, I started looking back into predecessor D’Arrigo and Jacoby articles since the 1980s and, in doing so, ran into enough interesting issues to make up about 15 columns. As a result, I’m going to respond in rather a piecemeal basis and maybe try to pull a consolidated reply at the end.
Rob’s North American chronology, drawing on Jacoby and D’Arrigo material, shows somewhat of a growth pulse in the last half of the 20th century. We’ve talked about the "Divergence Problem" in the past. If Rob’s chronology is right, then there wouldn’t appear to be a Divergence Problem. In fact, the Divergence Problem originates mainly from Schweingruber’s sampling and it turns out that there seems to be a remarkable discrepancy between Schweingruber and Jacoby-d’Arrigo as to whether ring widths in the last half of the 20th century went up or not – a controversy which originated in 1988-1989. A simple question like this seems to be unanswered nearly 20 years later. This issue inter-relates with the issue of CO2 fertilization – there’s a sort of 3-cornered battle with Graybill in one corner proposing increased growth and CO2 fertilization; Schweingruber in another corner saying no increased growth; and Jacoby and D’Arrigo in the third corner saying increased growth but due to increased temperature.
I’ll try to summarize the history of the debate from this point of view over a few posts.
Lamarche et al 
Let’s start with Lamarche et al , the et al including Fritts, Graybill and Rose, all important dendro figures, which reported enhanced 20th century growth for Sheep Mountain bristlecones, arguing that the increased growth could not be accounted for by climatic factors.
As a follow-up to this paper, Graybill carried out a fairly significant collection of high-altitude sites in the mid-1980s, reported on in Graybill (1987), conference proceedings which I’ve been unable to locate, but which was cited in contemporary literature. This collection led to virtually all of the sites in Mann’s PC1. It was financed by the Carbon Dioxide Information program (CDIAC) of the U.S. Department of Energy. Later, as we’ve pointed out, the Graybill sites morphed into Mann’s PC1. Graybill and Idso  reported on many of these sites (pdf here). While the collection was nearly all bristlecones and foxtails, there was also one high-altitude limber pine site from Colorado, which I’m going to come back to on another occasion.
The concept of CO2 fertilization was strongly criticized by dendrochronologists, for a variety of reasons (to be discussed below or on another occasion), but was more favorably received in botanical research, leading in particular to a number of detailed studies of high altitude conifers by the Swiss [especially studies associated with C. KàÆàner]. Interestingly a Walter Ammann is a coauthor of Hattenschwiler et al, 2004, “Atmospheric CO2 enrichment of alpine treeline conifers” [I wonder if he's related to Caspar?] I’ll get to these articles on another occasion; I’m just noting this line of developments for now.
One line of criticism on the dendro front, starting with Kienast and Luxmoore , was that they were simply unable to replicate anomalous late 20th century growth in high-altitude conifer chronologies. I’ve probably paid insufficient attention to date to this line of evidence, but it needs to be looked at squarely, prior to accepting Rob’s composite at face value. I had noticed some time ago in passing that the HS-ness of Graybill’s high-altitude North American sites was substantially higher than the HS-ness of the other high-altitude North American sites, but didn’t pursue the point.
But the minute that you pause and think about it, it is important: in our discussions of bristlecones, we’re not just talking about bristlecones, but site chronologies collected by one author (Graybill), who was attempting to prove a point about increased growth. One needs to show that Graybill’s results are consistent with those of other authors on the same material. Here one would like to see an update on the classic bristlecone sites, but little has been published in nearly 20 years. Hughes did an update at Sheep Mountain in 2002, but none of the results have seen the light of day.
Salzer and Kipfmueller updated San Francisco Peaks AZ, which was a Graybill site – but curiously about the only bristlecone site with no HS-shape. Unfortunately Salzer and Kipfmueller have not archived their measurement data, so it’s hard to explain the radical differences between the Graybill result and S and K result (maybe I’ll re-visit this on another occasion.)
Kienast and Luxmoore 1988
Kienast and Luxmoore  was one of the first responses to the Lamarche et al  hypothesis of increased ring width due to CO2 fertilization. They denied that there was any significant increase of ring width ( a position later held by Schweingruber).
They studied 34 sites on transects in 4 different regions, one of which was Colorado. Each transect went from the dry lower border to the upper border. They stated that only 8 of 34 sites had even somewhat anomalous post-1950 growth and none of the 8 met the Lamarche et al  conditions. They stated that the increase coincided with favourable climatic conditions in 4 cases and in the remaining 4 sites, the increase in ring width exceeded the expected CO2 fertilizing effect as determined from seedling experiments. Their conclusion was that the CO2 effect was unproven.
In their Colorado transect, only one site had enhanced growth – a lower border site, not an upper border site. Among the Kienast and Luxmoore sites said not to show enhanced growth are two high Colorado Engelman spruce sites (Arapahoe at 3400 m; Niwot Ridge at 3320 m). So an obvious back-tracking question is: did Graybill collect any Engelman spruce and how do they compare with Kienast/Schweingruber? Similarly, did Schweingruber collect any limber pine or bristlecones in Colorado and how do they compare with Graybill? Which sites are used in MBH98?
Jacoby and D’Arrigo 1989
Jacoby and D’Arrigo 1989 articulated a third point of view: that there was increased growth (contra Kienast and Luxmoore), but that the increased growth was due to climatic factors, not CO2 fertilization (contra Lamarche). The controversy was rather a triangle. Jacoby and D’Arrigo 1989 was a seminal multiproxy article and its reconstruction (as extended by D’Arrigo and Jacoby 1992) became a mainstay of the multiproxy world.
I’ve posted about the Jacoby collection from time to time (see right frame category Jacoby). Two points off the top bother me about the Jacoby and d’Arrigo collation: they collected 36 sites and selected only the 10 “most temperature sensitive” and did not even archive the results from the other 26 sites. Some of their collecting appears to overlap with Schweingruber, but no codex has ever been published. Second, the Jacoby 10 sites all have southern exposure, as the northern exposure sites do not share the common “signal”. This bothers me as one wonders whether some sort of insolation effect e.g. cloudiness might be confounding results with only southern exposure sites.
Using their chronologies and local temperature histories, Jacoby and d’Arrigo then carried out a variety of statistical analyses, including the principal components followed by regression (in that sense, anticipating MBH). For all the bellyaching by some blog readers about the right of non-climate scientists to comment on articles by climate scientists, the salient arguments in Jacoby and D’Arrigo 1989 are entirely statistical; one does not need to have drilled cores to comment on that. Discussing the statistical procedures of Jacoby and D’Arrigo 1989 is a long column in itself (and probably worth doing). For now, I merely mention the following, which is a little bit worrying:
“the low-frequency variance and autoregressive properties of the two sets of time series, tree rings and temperatures, do not conform to the theoretical assumptions of standard statistical tests for significance of regression results.”
This observation remains valid 9 years; it obviously remains a problem with MBH98. To cope with this, Jacoby and d’Arrigo also did simulations to provide new benchmarks. I’m not going to examine these simulations today, but it’s obviously something that should be done.
The conclusion of Jacoby and D’Arrigo 1989 was a temperature reconstruction, showing elevated 20th century values (the reconstruction going from 1671 to 1973.) They stated:
“In the eleven individual chronologies the recent increase in growth over the period of increasing atmospheric CO2 is more obvious and universal then the earlier (1700s) increase. Our NH reconstruction through 1973 (Figure 1, with instrumental data added through to the present — another splice) thus supports the hypothesis that the current warming trend has exceeded the level of natural climate variability, at least over the past several hundred years….There is as yet (through 1973) no evidence of any direct effects of CO2 fertilization at these northern sites (Lamarche et al 1984) (Figure 6). This may be due in part to severe nutrient limitations in arctic ecosystems (Oechel and Riechers 1986), (Figure 6)
Now their Figure 6, mentioned above, illustrated the residuals, purporting to show that the residuals were random. I did not find their statistical procedures very meaningful. The data sets used in this article ended in the 1970s and early 1980s. The measurement data has been archived. In most cases, this represents the most recent archived measurement information from Jacoby and d’Arrigo (but stop the presses, 14 Alaska sites from the Seward Peninsula have just been archived) — another topic.
They did not discuss Kienast and Luxmoore 1988.
Schweingruber et al 1993
Next Schweingruber et al , et al including Briffa, reported on a total of 69 sites at the North American treeline – 25 sites collected in B.C., Yukon and Alaska in 1984 and 44 sites collected in Canada from Yukon to Alaska in 1989. Like Kienast and Luxmoore, Schweingruber et al. also concluded that there was little evidence of anomalous growth. In the acknowledgements, it is reported that “Gordon Jacoby jointly selected and sampled many sites during the 1989 field season.” Two increment cores were collected from 12-15 trees per site. 34 sites were Picea glauca, 24 P. mariana, 4 P engelmannii and 3 P sitchensis and one larch.
“Figure 10a,b shows the MXD and TRW chronologies for each major region plotted from AD1800onwards. ..These curves indicate that no recent anomalous growth increases have occurred. Previous work (Jacoby and d’Arrigo 1989) has highlighted significantly increasing ring-width growth of some trees. This previous work was concerned with severely stressed trees growing the forest-tundra ecotone. Selected sites and trees at these locations, with particular ecology and aspect, may exhibit a sensitivity to temperature variability representative of larger climatic regions (G. Jacoby per comm.). Our results, along with other recent work (D’Arrigo et al 1992) indicate that increasing tree growth through the 20th century is not characteristic of the wider North American boreal forest.
Over recent decades, the regional chronologies described here do not exhibit any clear increasing growth tendencies that might be expected to result as a consequence of the anthropogenic environmental modifications or “natural” climate change.
As to the correlation of ring widths to temperature, they reported:
“By comparison, the ring width-related variables appear to be poorly correlated with temperature. Certainly LRW displays apparently random insignificant associations. TRW (and ERW) display mode consistently positive associations for the months of June through September and more particularly in the Gleichlaufigkeit values, with the June values (barring the WINNIP series) displaying consistently high values.
Graumlich 1991 is a criticism of Lamarche et al  and Graybill  from the Kienast and Luxmoore angle. She looked at 5 subalpine conifer sites in California near the bristlecones: 3 foxtail pine — the bristlecone cousin/brother, one lodgepole pine and one juniper. She reported that, at 3 of the 5 sites, their 20th century growth could be modeled adequately as a function of climatic variation (considering both temperature and precipitation). This is again a statistical argument and a number of questions occur to me as to whether she has overfit the models. Unfortunately, no data is archived or available.
None of the 5 chronologies illustrated in her Figure 5 (West Tyndall Foxtail; West Tyndall Lodgepole; Bighorn Plateau; Crabtree; Kaiser Pass) visually show anomalous 20th century growth, although all 5 end on an uptick. Graumlich also notes problems with the data meeting regression assumptions:
I have chosen to retain the autocorrelation within the data set despite the fact that it could be removed by prewhitening because preliminary analyses indicated that prewhitening reduces the upward trend in many of the tree ring time series. Prewhitening would thus remove that feature of the time series of most interest for this study. As an alternative to prewhitening, the residuals from the growth/climate regression models were tested for autocorrelation as part of a general strategy to test for violations of the assumptions underlying ordinary last squares regression.
Graumlich pointed out the possibility of a temperature-precipitation interaction “in which drought stress limits growth in years of high winter precipitation and cool temperatures limit growth in years of high winter precipitation. While the contribution of precipitation in governing growth of subalpine trees in the southern Sierra Nevada and the nearby White Mountains has been recognized (Lamarche 1974, Scuderi 1987) the important distinction that the effects of temperature and precipitation on growth are nonlinear and multiplicative rather than linear and additive has not been fully appreciated.
Graumlich distinguished her results from LaMarche and Graybill by the following alternatives (1) their sites are less arid, and if CO2 fertilization is due to increased water use efficiency then effects would be more pronounced in more arid environments; (2) strip bark morphology characterized Graybill sites and possible interaction with tree morphology; (3) site differences in soil nutrients may be a factor; (4) Lamarche’s exclusion of climate as an explanation was based on linear analysis, while Graumlich considered interaction term.
Graybill, Donald A. 1987. A network of high elevation conifers in the western US for detection of tree-ring growth response to increasing atmospheric carbon dioxide. in G.C. Jacoby and J. W. Hornbeck, editors, Proceedings of the International Symposium on Ecological Aspects of TRee-ring analysis, U.S. Department of Energy Conference Report DOE/CONF-8608144.
Graumlich 1991, Subalpine tree growth, climate and increasing CO2: an assessment of recent growth trends. Ecology 72, 1-11.
Jacoby, G. and R. D’Arrigo, 1989. Reconstructed Northern Hemisphere annual temperature since 1671 based on high latitude tree ring data from North America, Clim Chg 14, 39-59.
Lamarche, V, Graybill, Fritts and Rose, 1984. Increasing atmospheric and carbon dioxide: tree ring evidence for growth enhancement in natural vegetations. Science 225, 1019-1021.
Kienast, F. and R. J. Luxmoore, 1988. tree ring analysis and conifer growth responses to increased CO2 levels. Oecologia 76, 487-495. Schweingruber, F.H., K.R. Briffa and P. Nogler, 1993. A tree-ring densitometric transect from Alaska to Labrador, Int J Biometeorology 37, 151-169