The apparent inconsistency of ring width chronologies with analysis of changes in treeline elevation has long been an issue that I’ve urged specialists to address. Unfortunately Briffa et al 2013 failed to address the inconsistency at Polar Urals, even though they took note of an obsolete Shiyatov article on treeline changes. (More recent work by Shiyatov’s group has reported that medieval treelines at Polar Urals were even higher than previously thought.)
Worse, there have been significant changes in treeline at Polar Urals over the past 1500 years, an inhomogeneity that needs to be considered in RCS standardization. With the indecisiveness that is so characteristic of Briffa’s work, Briffa et al 2013 noted the possibility of altitude inhomogeneity, but then failed to investigate the problem or demonstrate that they could ignore the inhomogeneity.
For a number of years, I’ve been very interested in treeline movements as an index of centennial-scale climate changes. A survey of such changes is long overdue. Shiyatov’s work at Polar Urals is one of the most detailed and remarkable analyses and it is very disappointing that Shiyatov’s work on altitude changes is not clearly analysed in Briffa et al 2013, an article purporting to be a definitive treatment.
I noted Shiyatov’s work on treeline changes at Polar Urals in an early Climate Audit post here, which relied on Shiyatov 1995 and Shiyatov 2003, quoting Shiyatov (1995) which stated:
The 12th and 13th centuries were most favorable for larch growth. At this time the altitudinal position of the timberline was the highest, stand density the biggest, longevity of trees the longest, size of trees the largest, increment in diameter and height the most intensive as compared with other periods under review.
Reconciling this conclusion with ring width chronologies that record little variation between the medieval and LIA periods is something that seems to me to be a prerequisite for using ring width chronologies as useful climate proxies.
Shiyatov’s Altitudinal Transects
Shiyatov’s recent work (see Mazepa et al 2011) has resulted in substantial revision to his diagram of upper treeline at Polar Urals since the earlier work discussed at CA here, which was based on the relatively inaccessible Shiyatov (1995) which I placed online.
Shiyatov’s earlier work reported results from transects that went above present treelines, but only to 340 m. Newer transects have extended the sampling to 420 m and show that medieval trees grew substantially higher than reported in Shiyatov 1993, 1995 or 2003. The revision is shown in the graphic below, which plots the treeline values of Mazepa et al 2011 (Figure 11) on the previous graphic from Shiyatov 2003.
The dramatic change is also evident in Shiyatov’s transect diagrams in which he plotted the floruit of each tree against time (vertical axis) and transect (horizontal). The diagram below shows results for the original two transects (up to 340 m). The high medieval treeline is evident by the trees on the left side of the diagram. The decline in treeline is shown by the developing gaps, with the complete absence of trees in the 19th century in the higher transect evident by the white space. Modern growth at about 250-260 meters is evident, with saplings and trees beginning to grow in the higher transect. Shiyatov observed that the modern treeline is rising rapidly and is limited by the rate at which seeds move uphill; he stated that the potential treeline under modern temperatures is higher than the present treeline. (Thus, to use treelines to compare medieval and modern temperatures, one needs to allow for the lag and allow for future increases in treeline. ) The left block of the diagram shows the higher 1983 transect (elsewhere Transect 2), while the right block is the lower 1968 transect. In the top margin, I’ve annotated an approximate altitude – only approximate to 10-20 meters since the slope is not uniform.
Figure 3. Tree floruits in two altitudinal transects from Shiyatov et al 2011 Figure 3. The y-axis is time. The x-axis shows the horizontal distance along each transect. The slope lines show the altitude (right axis) along each transect. The two transects go from about 340 m to 180 m. In the top margin, I’ve shown an approximation of the vertical altitude along the combined transects. It’s an indication as it presumes uniform slope. Shiyatov explains the gap at certain altitudes as due to local topography which results in high snow accumulation, slow melt and non-germination.
Although Mazepa et al 2011 was published in the same year as Shiyatov et al 2011, it includes information from new and higher transects and reports very different results. In the diagram below, I’ve spliced its Figure 10 – a much revised version of the upper transect – as the left panel below. (Visually the plot under 340 m looks similar to me to the earlier version though this is not stated in the article itself.) The new data shows that the medieval forest extended to the upper limit of the new transects (420 m), higher than previously reported. It also revises the time on which the great lowering of treeline took place.
From this data, Mazepa et al 2011 reported:
On the base of life-span for dead trees (more than 1200 pieces) which grew above current tree-line ecotone the estimation of the most high-altitude position of tree-line for the last 1500 years was received [sic – revised ?](Fig. 11). The most high-altitude position of tree-line during the medieval climatic optimum was estimated at altitude 400-420 m a.s.l. This border occupied the highest position during 13th and in the beginning of 14th centuries. After that the huge dying off of trees, decrease of sparse tree growth and light forests up to the beginning of 20th century has begun. Intensive decrease of this border and thinning of forest stands occurred in 15th and 16th centuries and especially in 19th century.
Briffa et al 2013 on Treeline Changes
Briffa et al 2013’s discussion of Shiyatov’s work on treelines is very inadequate. Here is a full quotation:
We note earlier work on tree-elevation changes on Rai-Iz Massiv, Polar Urals (Shiyatov, 1993) involving a survey of tree remnants extending over an elevational range from 280 to 340 m.a.s.l. during the period 850 to 1990 CE. This remains one of the most detailed studies of tree-line position and composition over time undertaken to date. This work showed an early phase of relatively high larch germination at approximately 1100-1250, a notable phase of tree death at 1300-1350 and a complete absence of germination along the whole elevational transect from 1650 to 1900. The strong phase of tree germination at high elevation at the start of the twelfth century revealed in Shiyatov (1993) implies a distinct warming at that time. While some high growth-rate intervals are evident in our chronologies during the 1100-1250 period (Fig. 9), they are neither continuous nor notably higher than earlier or later intervals e. g. in the 11th and 15th centuries. It should also be noted that while the study area was virtually devoid of living trees at the start of the 20th century, it is now covered in larch forest (Shiyatov, 2009).
With the perspective of the above review, one can see errors and omissions in the Briffa et al 2013 account arising in part from their surprising reliance on the obsolete Shiyatov 1993 (making no citation of Mazepa et al 2011, Shiyatov et al 2011 or even Shiyatov 2003. Briffa et al 2013 report medieval treelines only up to 340 m, though these results were superceded by the new transects to 420 m in Mazepa et al 2011. The die-off phase previously dated to 1300-1350 is dated later in Mazepa et al 2011.
They noted that high-growth intervals in their chronology were “neither continuous nor notably higher than earlier or later interval”, they did not reconcile the inconsistency between the chronology and the changes in treeline elevation. To more fully show the inconsistency, in the graphic below, I’ve overplotted a smoothed version of the Briffa et al 2013 Polar Urals RW chronology against Shiyatov’s diagram of treeline changes (showing the Shiyatov 2003 and Mazepa et al 2011 versions as above). I’ve centered the RW chronology to coincide with the reference altitude in Shiyatov et al 2003 and scaled the RW chronology to show the comparison of the RW chronology to the elevation changes. It is evident that the RW chronology has negligible centennial-scale coherence with the altitude changes. The failure of Briffa et al 2013 to grasp this particular nettle shows typical Briffa indecisiveness: the veiled allusion to the problem indicates their awareness and would be used by them as evidence of their awareness of the issue, but they failed to reconcile or even clearly report the inconsistency.
It is evident that there have been substantial changes in treeline elevation on a centennial scale, presumably responding to temperature changes. On the basis that ring widths (and MXD) are proxies for temperature, altitude changes are an important inhomogeneity that needs to be allowed for: indeed, the problem of altitude inhomogeneity was raised in an early CA post here.
Unfortunately, Briffa et al 2013 completely dropped the ball on altitude inhomogeneity. They referred to the problem as a sort of CYA, but neglected to do anything about it. Here’s their only statement:
We have not investigated the influence of sample elevation on the absolute magnitude of tree growth or made any allowance for such differences in our analysis, but this may not be a very significant factor (Briffa et al., 1996).
The purported citation (Briffa et al 1996) stated the following as their “statistical” analysis purporting to discount the influence of altitude:
We examined the evidence for a time-dependent elevation bias in the reconstruction by regressing mean MXD against mean sample elevation for different age classes of trees … These results (shown in [their] Table 2) indicate no significant elevational influence on mean density, at least over the range of elevation involved in these calculations”
As noted at CA here, as too often, Briffa is totally at sea when it comes to even elementary statistical analysis. We know from first principles that temperature declines with altitude. Therefore, if ring widths (or MXD) are temperature proxies, then there must necessarily be a relationship with altitude. If, as Briffa et al 1996 claim, there is no (inverse) correlation between MXD (ring width) and altitude, then this, on its own, indicates a defect in the proxy.
On the other hand, since we know that the treeline has changed elevation over time, the supposed lack of correlation is probably nothing more than a failure on Briffa’s part to disentangle centennial climate changes from altitude. The original caption in Briffa et al 1996 stated:
The density data in each age class are averaged over different time periods so that effects of climate variability should hopefully largely cancel out.
This statement is no better than Briffa’s original attribution of the “Decline” to a cargo cult (see for example here). The high-elevation samples occur in the medieval period and were not averaged with Little Ice Age samples, because trees did not grow at high altitude in the Little Ice Age. In addition, there has been negligible recovery of medieval trees at lower altitudes because moss and other weathering have mostly destroyed lower altitude medieval trees. Low altitude trees are primarily from the Little Ice Age. The dismissal of altitude inhomogeneity in Briffa et al 1996 was based on worthless analysis, making the failure of Briffa et al 2013 to consider and analyse altitude inhomogeneity completely unwarranted.
In the new Polar Urals chronology presented by Briffa et al 2013, altitude inhomogeneity appears to be an extremely important issue, as the average altitude of living trees is notably lower than the average altitude of subfossil trees in the original Schweingruber sample. Briffa appears to have made the matter worse by incorporating the apparently inhomogeneous purlasi_sc dataset.
Briffa’s RCS methodology is extremely sensitive to inhomogeneity. Needless to say, CRU are very alert to inhomogeneities that go in their favor (e.g. root collar inhomogeneities), but obtuse to inhomogeneities that go the “wrong” way e.g. altitude at Polar Urals, purlasi_sc. Given Shiyatov’s unequivocal and convincing analysis of elevation changes at Polar Urals, it is unacceptable that Briffa et al neglected to even investigate the impact of sample elevation.