Post-1980 Proxies #2: Alaskan Tree Rings

D’Arrigo et al. [GRL 2004] reports on tree rings taken from the Seward Peninsula in northwestern Alaska from white spruce from 14 sites near elevational treeline in summer 2002. We show here their figures, demonstrating that the "proxy" show no evidence of the warm 1990s and hot 1998, raising questions about the ability of this type of commonly used proxy to pick up similar warm periods in the past.

D'Arrigo Figure 3 excerpt

The original study calculated site "chronologies" were calculated for both ring width and maximum density (MXD), but, contrary to AGU policy requiring data archiving for its publications (including GRL), d’Arrigo et al. have not archived their data.

The bottom panel of their Figure 3 below shows the ring width chronology through to 2002 (showing the same graph as in the preview). There is obviously no loud 1998 or loud 1990s.

They do not show the actual MXD site chronology. Instead, the top panel of their Figure 3 is a "temperature reconstruction" from the MXD chronology (and is presumably a linear transformation of the MXD chronology); thus the top panel can be construed as an image of the MXD chronology up to 1970 and has the the same features (in terms of maxima and minima) as the MXD chronology itself. Note the truncation at about 1970. This is explained as due to a "weakening" of the relationship between temperature and MXD.
Darrigo et al 2004 Figure 3

Original Caption: Figure 3. Top line: May–August NTR extending from 1389–1970 (recent decades truncated, see text), based on 1909–1950 model. Thick lines shows smoothed values. Horizontal lines indicate mean values. Chronology based on 3 SP sites: Death Valley (65.19N, 162.27W, 239m), Alpine View (65.11N, 162.18W, 282m) and Hey Bear (65.22N, 162.22W, 229m) [D’Arrigo et al., submitted to Holocene, 2004]. Gray shading shows changing sample size over time, which ranges from 1–7 radii prior to 1640. Inferences of past climate prior to ca. 1640 should be interpreted with caution. Low value in 1783 is labeled. Middle line: May–August NAR [Jacoby et al., 1999]. Bottom line: SRW series [D’Arrigo et al., submitted to Holocene, 2004]. Vertical lines in SP records indicate acceptable EPS cutoff of 0.85 [Cook and Kairiukstis, 1990].

Their Figure 2 (shown below with original caption) appears to contain the information left out of their Figure 3. Here the dotted line presumably shows the shape of the MXD chronology in the period after 1980. As with the ring width chronology, there is no increase through the 1990s.

Darrigo 2004 Figure 2

Original Caption. Figure 2. Actual (solid line) and estimated (dashed line) Nome May–August temperatures from 1909–2001 based on SP-MXD chronology, including both calibration (1909–1950) and verification (1951–1970) periods. Note that estimates after ca. 1970 tend to underpredict actual values (calibration ar2 declines from 42% for 1909–1950 to 38% for 1909–70 and 23.2% for 1909–2001). The ar2 for the verification period is 26%, with a positive RE of 0.20 in verification. Additional tests: calibration 1909–1920: ar2 67%, verification 1921–1950: RE = 0.31; calibration 1921–1950: ar2 38%; verification 1909–1920: RE = 0.40.

Reference: Rosanne D’Arrigo, Erika Mashig, David Frank, Gordon Jacoby, and Rob Wilson, 2004. Reconstructed warm season temperatures for Nome, Seward Peninsula, Alaska, GRL 31, L09202, doi:10.1029/2004GL019756.


  1. Steve Funk
    Posted Feb 16, 2005 at 4:26 PM | Permalink

    Are there any actual local temperature records for the last few decades included in this study? You would think that in this location, there would be a significant positive correllation between warmth and ring width. In interior northern California, where I practice forestry, there is probably a negative correllation, since tree growth is highest in years with more spring and summer precipitation.

    Steve’s comment: The bolder line in their Figure 2 is Nome temperature; the dotted line is the “reconstruction” which here is just a projection of the MXD chronology. So you see that the MXD values do not pick up warm temperatures very well. I guess the bristlecones are a little to the south of you. The theory that there is a positive linear relationship between ring width and temperature looks very insubstantial to me. What are your thoughts on this?

  2. John A.
    Posted Feb 17, 2005 at 3:40 AM | Permalink

    What’s apparent to me about the reconstructions, especially “Figure 3”, is that the smaller the number of tree rings used, the larger is the deviation in “temperature”

  3. Steve Funk
    Posted Feb 17, 2005 at 2:40 PM | Permalink

    A forest service research paper on the rocky mountain bristlecone pine populations, says that radial growth is higher in years following heavy monsoon precipitation, typically in El niño years. Following La Niña years, with high late-season temperatures and low precipitation, radial growth is minimal.
    The expert, both on bristlecone pine and dendroclimatology, seems to be Dr Hal Fritts, University of Arizona emeritus. I haven’t read anything except short summaries and a brief presentation by his colleague, M.A. Stokes, so take this for what it’s worth. Stokes says the limiting factor for tree ring growth in the southwest is precipitation, acting in concert with temperature. Fritts says his model incorporates temperature/precipitation interactions, and includes lag periods. He claims good results in showing climate cycles in the western United States, including El niño and la niña cycles, and a good correllation with known temperature changes resulting from the Krakatoa eruption in the 1880’s. His model doesn’t appear to be a simple linear relationship. He says his model is useless in the eastern and southern United States.
    Conceptually, it is hard for me to grasp how one can solve for temperature, the least significant influence, without having some independent knowledge of precipitation, the more significant variable. This is like an algebra problem with two unknowns and one equation. But if this problem can be solved, there are other issues. Most tree cores I have seen show a pattern of consistently decreasing ring widths over time, due to increasing inter-tree competition. With a large enough sample size, growth trends due to climate could probably be isolated from limitations due to competition. With a smaller sample, this would be problematical. Bristlecone pine is uniquely suited to measuring climatic influences because it tends to grow in very open stands, as well as being very old. White spruce, on the other hand, grows in dense stands. The onset of the active growth season in white spruce is triggered by photoperiod rather than temperature. A late frost after growth begins would severely limit that season’s growth, even if average temperatures for the rest of the season were high (which in Alaska we would presume to be optimum.) White spruce is highly susceptible to damage from snow, ice and wind, and non-fatal attacks from spruce budworm. A severely damaged tree would show virtually no radial growth, and probably be taken out of the sample. But moderately damaged trees might easily be kept in the sample. These non-climatic influences could be particularly significant when the sample size is small. Between 1780 and 1700, the sample size drops from app. 20 to less than 10. The data probably become worthless sometime during that period. I wonder how you could test the strength of a relationship when the sample size varies from one to 45. My own background, unfortunately, is one semester of statistics in an era when the teachable concepts were limited by the numbers you could crunch.

  4. what
    Posted Nov 1, 2005 at 8:34 AM | Permalink

    what the heck are tree rings

%d bloggers like this: