3Published methods for reconstructing tree age from increment cores without any visible pith
4generally follow geometric/graphical methods (e.g. Rozas 2003) or a regression approach (e.g.
5Rohner et al. 2013). Graphical methods require measuring the curvature of the innermost rings or
6direction of parenchyma rays, a type of information not always available, especially in old datasets.
9The estimation method we used to estimate pith location is based on the initial growth rate from an
10increment core, and follows the approach that Rozas (2003) found most effective for beech cores
11lacking the arcs of inner rings. We modified Rozas approach so that it could be applied to tree-ring
12measurements even without having access to the actual cores. The method was based upon three
13conditions: (1) cores from a certain tree should not be excessively short in relation to its stem DBH
14(Norton et al. 1987); (2) initial growth rates were derived using the first 20 years (Rozas 2003); (3)
15the number of years to be added should not exceed 10% of the maximum age observed amongst all
17The first step consisted in identifying, for each forest, the number of increment cores where pith
18was present. Then, to estimate the number of missing rings in the remaining incomplete cores of
21 Reconstructed stem DBH (DBHrec): twice the sum of available ring-widths (RWs).
22 Initial ring-width (irw20) of the increment core: average width of the first 20 rings in a core.
23 Median Initial Ring-Width of the Forest ( IRW20): for each forest, the median width of the
24 first 20 rings, using exclusively cores with the pith. When comparing different forests, the
1 1
2
26 Maximum tolerated Number of Missing Rings (NMRbelt) added to reach the pith: valid for all
27 forests within a belt, set to 10% of the maximum lifespan observed within each belt (Table
28 1).
29 Maximum Tolerated Missing Radius (TMRbelt): valid for all forests within a belt, it was the
30 lowest integer value among all the i forests of a belt obtained when applying the formula:
31 IRW20i*0.1(Maximum Age)I . According to this formula, the tolerated missing radius at each
32 site depends on the growth rates and the maximum lifespan, both connected to site
33 productivity. As before (see NMR calculations), the threshold used for the number of rings
34 was 10% of the maximum lifespan observed. Finally, TMRbelt was 3 cm for high-mountain
36
38 1. First selection step: choose cores according to their length. For each tree, we calculated the
39 missing radius (MR) as: MR = (Measured DBH)-DBHrec. Trees with MR < TMRbelt were
40 excluded.
41 2. Second selection step: choose cores according to the missing number of years to the pith. In
42 the remaining trees, we divided their missing radius (MR) by the core irw20 to estimate the
43 number of missing rings. Cores with an estimated number of missing rings above the
45 3. Calculate tree age by summing the number of visible rings and the estimated number of
46 missing rings.
47For each site, the final dataset contained trees with observed or reconstructed age.
48
3 2
4
49TABLE 1. Summary of cores with/without pith and statistics used for calculating the Maximum
50Tolerated Missing Radius (TMRbelt) and the Maximum tolerated number of Missing Rings to Pith
51(MRPbelt).
52
53
54On average, only 66% cored trees had a visible pith (Table 1). We thus tested our method using only
55cores containing the pith. According to the belt of pertinence, we removed from each core an initial
56portion equal to TMRbelt and, using the irw20 of the remaining radius we estimated the number of
57missing rings to the pith. The absolute and relative errors were generally low (less than 10-20 years;
58Table 2), and comparable to those found on beech by Rozas (2003). For all belts, the reconstruction
59method generally underestimated the effective number of years (Fig. 1), and the overestimation was
60always below 20 years. Often, age in old-growth forests could be underestimated due to periods of
61initial suppression (Rozas 2003). Most trees showed errors within 50 years, and in many cases the
5 3
6
63Considering the objectives of our paper, the selected approach revealed to be fairly simple and
64conservative enough. Thanks to it, we were able to recover an extra 6% trees, so that the final
65usable sample was on average 72% of the original coring (Table 3). Within each belt, managed
66forests had often a higher percentage of used trees, probably in connection to the widespread
67presence of younger trees with simpler growth trajectories (i.e. lower errors in reconstructing pith).
68Age estimation was generally in line with observed data on trees with pith and, generally, the
69maximum age at each site was observed rather than reconstructed (Table 3). The reconstructed age
70was higher than the observed only at the site ORI, but the difference was only 9 years. COP was an
71exception, since the oldest tree didnt reach the pith and cannot be recovered because it didnt pass
72the MRP threshold (estimated number of missing rings 10% of the maximum age observed within
73the belt).
74
75TABLE 2. Absolute and percentage error for each bioclimatic belt, using only trees with pith. On
76each core TMRbelt by each trees irw20. (Left) N* vs. N. Dotted line: slope=1; filled circles are
77trees with N* maximum tolerated number of missing rings to pith (MRP, corresponding to 10% of
79
80
81
7 4
8
82
83FIGURE 1. Comparison of reconstructed (N*) vs. actual (N) number of years on cores containing
84pith within each bioclimatic belt (Alps and Apennines merged together to simplify the analyses). N*
85obtained by dividing TMRbelt by each trees irw20. (Left) N* vs. N. Dotted line: slope=1; filled
86circles are trees with N* maximum tolerated number of missing rings to pith (MRP,
87corresponding to 10% of the maximum age observed within the belt). (Right) Frequency
9 5
10
89TABLE 3. Number of trees recovered by the age estimation method and maximum age
90observed/estimated within each study forest. Discarded trees: cores with estimated missing rings
9150. (Used refers to trees with DBH 37.5 cm, used in the final analyses)
92
93
94
95REFERENCES
96Norton D.A., Palmer J.G., Ogden J. (1987). Dendroecological studies in New Zealand. An
97evaluation of age estimates based on increment cores. New Zealand Journal of Botany 25, 373383.
98Piovesan G., Di Filippo A., Alessandrini A., Biondi F., Schirone B. (2005). Structure, dynamics and
99dendroecology of an old-growth Fagus forest in the Apennines. Journal of Vegetation Science 10,
1001328.
11 6
12
101Rohner B., Bugmann H., Bigler C. (2013). Towards non-destructive estimation of tree age. Forest
103Rozas V. (2003) Tree age estimates in Fagus sylvatica and Quercus robur: testing previous and
13 7
14