Climate Audit

One of my first blog postings was on Briffa’s very first fudging (2011 update: the “Briffa bodge”) of a temperature reconstruction – his adjustment of the Tornetrask reconstruction – a reconstruction that is used in virtually every study. This was one of the first encounters with the Divergence Problem. Tornetrask MXD went down in the 20th century. Briffa resolved this by simply “straightening” out the reconstruction (although, unlike in IPCC TAR, at least he reported it in the original article.) At the time, per, an early reader of the blog, remarked that he was “gobsmacked” at this. This has recently come up again in two contexts and this time even I am “gobsmacked”.

If you fell like looking at the earlier post, it will provide some context. In the recent D’Arrigo, Wilson article on the Divergence Problem, they refer to Briffa’s brute force elimination of the downward MXD as an “empirical correction”. (I don’t think that it’s unfair to use the term “fudge factor” for Wilson’s “empirical correction”.)

As a result of the divergence problem, attempts to directly estimate large-scale temperatures for the recent period in dendroclimatic reconstructions have generally not been successful (Briffa et al., 1998a,b; Briffa, 2000; Briffa et al., 2001; Esper et al., 2002; D’Arrigo et al., 2006, see Fig. 3). The inability of many reconstruction models to verify in the recent period has compelled a number of researchers to eliminate recent decades from their calibration modeling, effectively shortening the available periods for direct calibration and verification testing between tree rings and climate (e.g., Briffa et al., 2001; Cook et al., 2004a; Rutherford et al., 2005; D’Arrigo et al., 2006). Another alternative is to use an empirical correction for the divergence effect (e.g., Briffa, 1992; Osborn et al., submitted for publication, Glob. Planet. Change).

I was thinking about posting on this, when I stumbled on Hakan Grudd’s PhD thesis online. Grudd was the first author in Grudd et al 2002, which is cited as authority for the Tornetrask version that Briffa has been using. Grudd has many interesting things to say about the Divergence Problem and has a new reconstruction for Tornetrask. D’Arrigo and Wilson do not mention Grudd’s new work.

Here are some important comments from the introductory part of the thesis:

‘€¢ Over the last two centuries, tree-ring width shows a diverging trend with respect to maximum latewood density and temperature, implying an increasing productivity in response to temperature. If not attended to, this phenomenon introduces a bias in the climatic reconstruction, resulting in depressed temperatures in the pre-instrumental period.

‘€¢ The maximum latewood density data from Torneträsk are continuous from AD 441 to 2004. By including data from young trees in the most recent period, it is demonstrated that the previously reported reduced sensitivity in density to temperature in the late 20th century was an artefact from the age structure in the old data.

‘€¢ A new multi proxy reconstruction for the last 1,500 years, based on maximum latewood density and ring width, shows significantly higher temperatures in the pre-industrial period as compared to previous reconstructions from Torneträsk. Several periods during the past 1,500 years were warmer than the 20th century. The most striking feature is a long and warm medieval period centred on AD 1000. The coldest period occurred around AD 1900 which is consistent with evidence of maximum glacier expansion over this period.

‘€¢ Methodological differences in the development of local dendroclimatological reconstructions sometimes hampers our ability to assess climate variability on a regional scale (Wilson et al., 2005). It is, therefore, imperative to continue to supply open-access databases with raw data, e.g. the International Tree-Ring Data Bank (http://www.ncdc.noaa.gov/paleo/treering.html) and the Dendrochronological Database (http://www.wsl.ch/dendro/dendrodb.html), and to develop public tools to extract and standardize the data in a consistent way, e.g. ARSTAN (Cook & Krusic) (http://www.ldeo.columbia.edu/res/fac/trl/public/publicSoftware.html).

None of Grudd’s new Tornetrask data is archived. I’ve been trying for a couple of years to get the updated Tornetrask data and Briffa has obstructed every attempt. It’s nice that Grudd believes that the data should be archived and I hope that he will do so. I’ve emailed him suggesting that he live up to this.

Paper II (still unpublished) is the salient one. He may not be finding it all that easy to get it published despite Grudd’s competence, since Grudd seems to have left the reservation. Here are some comments:

However, the lack of data after AD 1980 in the Torneträsk records hampers the calibration with instrumental data. This lack of data in the most recent period is a common problem in many tree-ring records throughout the Northern Hemisphere (Briffa et al., 2004). Furthermore, there is increasing evidence for a change in the sensitivity of tree growth to temperature in the most recent period (Jacoby & D’Arrigo, 1995; Briffa et al., 1998b; Briffa et al., 1998c; Briffa et al., 2004; Waterhouse et al., 2004; Wilmking et al., 2004; Driscoll et al., 2005; Wilmking et al., 2005; D’Arrigo et al., 2006). The phenomenon is not fully understood and the evidence is sometimes contradictory (Myneni et al., 1997; Briffa et al., 2004; Waterhouse et al., 2004). A similar change in sensitivity has been reported also for the Torneträsk data (Briffa et al., 1992). Hence, there is an urgent need to update the Torneträsk and other TRW and MXD records and to analyse the data for recent changes in sensitivity to climate.

Our present knowledge about natural climate variability in the last 1,000 years is to a very large extent based on tree-ring proxies. However, many existing tree-ring records have problems with age structure, with changing sample replication, and with a lack of data in the most recent period. Inevitably, sample replication will change through time in long tree-ring records and this will affect the signal-to-noise ratio so that low replication will lead to wide error bars around the reconstructed climate signal. Furthermore, the age of the trees that constitutes a tree-ring chronology will change through time: Some periods in the record will be dominated by relatively young trees while other periods will be dominated by relatively old trees and the climatic growth response may be very different. For example, in nearly all existing records, data for the most recent period is based on tree rings that on average have a high cambial age as compared to the earlier parts of the record. This causes an in-homogeneity in the age structure that may misrepresent the calibrated growth response and lead to spurious reconstruction results. The widely reported change in the sensitivity of recent tree-growth to temperature could, potentially, be explained as an artefact from using inadequate data in the most recent period. Hence, a major task in the near future will be to update and renovate existing tree-ring collections throughout the Northern Hemisphere. The new tree-ring data from Torneträsk isjust one step in this work.

Existing Torneträsk TRW and MXD data are updated to AD 2004 using samples from 35 living trees. The new data are derived from a mixture of young and old trees. This has significantly changed the age structure of the MXD data in the 20th century and the previously reported decline in density after about AD 1950 is no longer present in the data.

The results show that the previously reported loss in sensitivity of MXD to temperature in the late 20th century was an artefact of the age structure in the data. The reconstructed temperatures show higher values in the pre-instrumental period as compared to earlier temperature reconstructions based on Torneträsk TRW and MXD data. The Medieval Warm Period around AD 1000 is notably warmer than temperatures in the 20th century. Other notably warm but shorter periods occur around AD 750, 1400 and 1750. The coldest temperatures occur around AD 1900, which is consistent with the maximum extent of glaciers in Swedish Lapland in the last 1,500 years.

The New Reconstruction

Here’s Grudd’s new reconstruction for Tornetrask, which shows the elevated MWP as advertised:

Original Caption Figure 11. In the lower panel (b), Reconstruction IV is compared with two previously published temperature reconstructions based on tree-ring data from Torneträsk: The thin curve is from Briffa et al. (1992) and based on TRW and MXD. The hatched curve is from Grudd et al. (2002) and based on TRW. All three reconstructions have been smoothed with a 100-year spline filter and have a common base period: AD 1951 — 1970.

Now here’s something funny. Here is my estimate of what Briffa’s original reconstruction would have looked like without his “empirical correction”/fudge (See earlier post) . Look at the bottom row and compare it to the new Grudd version which I’ve shown a second time immediately below for comparison. Although the scales are different, I think the shapes of the two versions are, in the phrase so beloved of the Team, “remarkably similar”. Downstream of the site reconstructions, the scales tend to get standardized, so it’s the shape that matters more than the scale. The interesting thing about this reconstruction is that it has major impact on the Jones et al 1998 reconstruction, which is still a major component of the spaghetti graphs. I’ve asked Grudd for a digital version of his new reconstruction and am looking forward to re-doing Jones et al with the new Grudd version for Tornetrask and the new Esper version for the Polar Urals. Some new light on an old “empirical correction”.