Jacoby and D'Arrigo Archive Data!

During the past 20 years, Jacoby and D’Arrigo had obtained over 45 RW series and at least 35 MXD series in North America. To the end of 2005, despite receiivng millions in grants, he had only archived a couple of North American RW series and no North American MXD series. I discussed this before here, listing information on unarchived sites and on grants.

Remarkably, in February 2006, after years of virtual silence, Jacoby and D’Arrigo archived 14 RW measurement data sets from the Seward Pensinsula and 2 Labrador RW measurement data sets (Eyeglass Lake, Medusa Bay), and, stop the presses, on April 12, 2006, archived information on 8 more Alaska sites: 8 RW chronologies; 7 of 8 RW measurement data sets (ak072 – Windy Wolf AK missing for some reason) and 5 MXD chronology and measurement data sets (ak071 – Silvertip; ak073 – Nazina Treeline and ak076 – Chisana Air missing). Here’s a quick survey of the new data.

The 14 Seward Peninsula sites were previously discussed in D’Arrigo et al 2005 and a composite of these 14 sites, together with data from Graumlich and Giddings in the area, is shown in D’Arrigo et al 2006 as one of the 19 contributing sites (see Figure 1 below). It does not show values for the 11th century portion of the MWP; it has elevated late 19th and early 20th century values, with slightly elevated late 20th century values.


Figure 1. RW chronology from DWJ06.

The next figure shows my emulation of the RCS ring width chronology using the 14 recently archived sites. This only goes back to the late 14th century. The visual match to Figure 1 above looks pretty good as these things go – the next time that I visit this data, I’ll re-do the graph in a more similar format.


Figure 2. Emulated RCS Chronology for 14 Seward Pensinsula sites only.

Next here are some working graphs. I realize that journal space is short, but, with the advent of online Supplementary Information, it would be nice to see much more systematic presentation of details and results. The first figure is a simple barplot of the mean ring widths of the 14 sites, indicating a significant inter-site difference. I didn’t plot the differences against map location, but that would be a useful exercise.


Figure 3. 14 Seward Peninsula, Alaska sites. Barplot of mean ring width by site,

Next is a figure showing the count of cores for each site, peaking in the 19th century, showing the very low counts in the 15th century from the Jacoby data. The earlier data obviously relies on the Graumlich and Giddings data.


Figure 4. 14 Seward Peninsula, Alaska sites. Number of cores by site.

The next figure shows the average ring width for each site by age. Note that some very old trees (and the population is very small) are pretty frisky in their growth. This is a surprisingly common phenomenon. This interacts with the RCS methodology disproprotionately as the RCS curves are negative exponential; and the "adjusted" width attributed to the old trees is the observed ring width divided by the age-profile value which is down at low values – thus the "age-adjusted" ring width for these old codgers is very large. The populations here look large enough that this probably doesn’t matter, but many populations are only about 10-20 cores and one or two cores can make a big difference with non-normal distributions.


Figure 5. 14 Seward Peninsula, Alaska sites. Average ring width by age by site.

Here is the spline-fitted age profiles used in my RCS calculations:


Figure 6. Spline smoothed age profile by site.

There is an important non-homogeneity in the age distribution which is characteristic of EVERY ring width site – the 20th century tree sample is much older than sites in the past. Does this matter? Well, typically ring widths in the 20th century are not particularly large even at sites which show profound hockey sticks; it’s usually only the age-adjusted ring widths that cause a hockey stick, making the age adjustment process pretty fundamental. (These age adjustment processes have been the topic of considerable specialist debate in the 1990s.) I can see some areas of potential bias, where I see little evidence of much attention. Trees typically germinate in warm times and die in cold times; so some portion of the age-decline curve may be attributable to temperature i.e. the age adjustment should be flatter. To the extent that this bias exists, it would usually lower 20th century values. Also treelines are rising in the 20th century; presumably higher trees growing at lower temperatures will have narrower widths; these higher trees are not sampled in 20th century collections but are sampled from MWP subfossils. So this also creates a bias between modern and medieval levels which exists in most long chronologies.


Figure 7. Average Age by Year and Site.

The next figure shows the mean ring width by site without age adjustment.


Figure 7. Mean Ring Width by Year and Site.

Finally, the next figure shows the effect of age adjustment on the mean ring width plots and is the RCS chronology for each of the 14 Seward Pensinsula sites.


Figure 8. 14 Seward Peninsula, Alaska sites. RCS standardized chronology by site.

References:
D’Arrigo, R., E. Mashig, D. Frank, R. Wilson and G. Jacoby. 2005. Temperature Variability over the past millennium inferred from northwestern Alaska tree rings. Climate Dynamics 24: 227-236. doi:10.1007/s00382-004-0502-1.
D’Arrigo, R., R. Wilson and G. Jacoby. 2006. JGR.

4 Comments

  1. John Lish
    Posted Apr 12, 2006 at 3:47 PM | Permalink

    and the “adjusted” width attributed to the old trees is the observed ring width divided by the age-profile value which is down at low values – thus the “age-adjusted” ring width for these old codgers is very large.

    This may seem like a simple question but how does one arrive at an “age-profile value” for a sample population?

  2. TCO
    Posted Apr 12, 2006 at 4:51 PM | Permalink

    Note that some very old trees (and the population is very small) are pretty frisky in their growth. This is a surprisingly common phenomenon. This interacts with the RCS methodology disproprotionately as the RCS curves are negative exponential

    If you can do even a preliminary quantification of this effect, it is publishable (of course if preliminary, you must make the caveat that this is so).

  3. TCO
    Posted Apr 12, 2006 at 4:55 PM | Permalink

    Nice analyses, Steve.

  4. Paul Gosling
    Posted Apr 13, 2006 at 4:42 AM | Permalink

    Go on Steve. Take some gredit for the archiving, I am sure it is due.

Follow

Get every new post delivered to your Inbox.

Join 3,381 other followers

%d bloggers like this: