Ecological Archives E091-259-A1

Koert G. van Geffen, Lourens Poorter, Ute Sass-Klaassen, Richard S. P. van Logtestijn, and Johannes H. C. Cornelissen. 2010. The trait contribution to wood decomposition rates of 15 Neotropical tree species. Ecology 91:3686–3697.

Appendix A. Validating assumptions.

We calculated average decomposition rates for the stem disks assuming a heartwood : sapwood ratio of 3:1, and 30% of the sapwood being in direct contact with the soil. This assumption was based on field observations, but there is large variation in heartwood : sapwood ratios, both inter-specific and intra-specific. We performed several statistical tests to investigate whether potential deviations from these assumptions would lead to different results. We did this by first testing whether there are differences in wood decomposition rates between sapwood with direct soil contact (SSC), sapwood with no direct soil contact (SNSC) and heartwood (HW). Secondly, we varied the heartwood : sapwood ratios and percentages of sapwood in direct soil contact to see whether this leads to significant changes in wood decomposition rates. Finally, we performed separate regression analyses to test the trait contribution to wood decomposition rates that were calculated using different heartwood : sapwood ratios and different percentages of sapwood in direct soil contact.

By doing so, we were able to statistically test (1) the influence of our assumption on wood decomposition rates, and (2) the influence of potential artefacts on the trait contribution to wood decomposition rates.


(1) Average decomposition rates
First, we compared the average decomposition rates of the three wood types (SSC, SNSC, HW). There were no differences in average decomposition rates between the three wood types (One-Way ANOVA, F = 0.324, P = 0.724; figure 3 in article). This suggests that potential artefacts in our assumptions would not lead to incorrect wood decomposition rates.

Furthermore, there were strong correlations between decomposition rates of SSC and SNSC (Pearsons R = 0.759, P = 0.01), SSC and HW (Pearsons R = 0.773, P = 0.01) and SNSC and HW (Pearsons R = 0.821, P = 0.01).


(2) Varying assumptions
To quantitatively test whether potential artefacts in our assumptions would have major consequences for the average decomposition rates, we varied the heartwood : sapwood ratio

(9 : 1, 4 : 1, 3 : 1, 2 : 1) and the amount of sapwood in soil contact (20%, 30%, 40%), resulting in a total of twelve different assumptions, including our original assumption (Table B1). For each species, a One-Way ANOVA was used to test whether changes in our assumption significantly changed the species average decomposition rate constants. The results (Table B2) show that there is only minor variation in the minimum and maximum K values calculated using the nine different assumptions. The One-Way ANOVA results show that these variations are non-significant. This indicates that wood decomposition rate calculations are robust to artefacts in our original assumption.


(3) Trait contribution
We tested the influence of our assumption on the trait contribution to wood decomposition in two ways: (1) regression analyses were performed for decomposition rates of all three wood types separately to see if the trait contribution to wood decomposition varied between different wood types, and (2) separate regression analyses were performed for each trait using the decomposition rates that were calculated under the nine different assumptions.

The results of the multiple regression analyses that were performed for each wood type separately (Table B3) all point in the same direction: log10-transformed DBH is the only trait that significantly explains inter-specific variation in wood decomposition rates, independent of the wood type. However, the amount of variation explained by log10-transformed DBH was lower for SSC (R2 = 0.29, P = 0.038) than for SNSC (R2 = 0.42, P = 0.009) and HW (R2 = 0.49, P = 0.004).

The results of the separate regression analyses that were run for each trait, using the decomposition rates that were calculated under the different assumptions, are shown in Table B4. The results indicate irrespective of the assumption, log10-transformed DBH is the only significant predictor of inter-specific variation in wood decomposition rates. There is only minor variation in the predictive power of log10-transformed DBH (R2max = 0.46 (assumption 4), R2min = 0.38 (assumption 10, 11 and 12)). This indicates that potential artefacts in our assumption do not have important consequences for the final outcome on the trait contribution to wood decomposition.


Conclusion
From the above-shown tests it appears that potential artefacts in our assumptions for stem disk decomposition calculation (1) do not lead to significant changes in decomposition rates and (2) do not result in large changes in the trait contribution to explaining inter-specific differences in wood decomposition rates. Therefore, we conclude that our results are robust.


TABLE A1. The twelve different combinations of the assumptions that we used for decomposition rate determination. Assumption 5 is the original assumption.

Assumption Heartwood : Sapwood ratio Sapwood with direct soil contact
1 2 : 1 20%
2 2 : 1 30%
3 2 : 1 40%
4 3 : 1 20%
5 3 : 1 30%
6 3 : 1 40%
7 4 : 1 20%
8 4 : 1 30%
9 4 : 1 40%
10 9 : 1 20%
11 9 : 1 30%
12 9 : 1 40%

TABLE A2. The minimum (min) and maximum (max) k-values for each species that were obtained by calculating the k-values with the twelve different assumptions. One-Way ANOVAs were used for each species to test for differences between decomposition rates that were calculated using the twelve different assumptions.

  k-values One-Way ANOVA
Species Min Max F df P
Pseaudolmedia laevis 0.04 0.05 0.464 11 0.921
Hura crepitans 0.01 0.02 0.01 11 1
Ocotea sp. 0.26 0.29 0.046 11 1
Pourouma cecropiifolia 0.15 0.16 0.026 11 1
Pouteria nemorosa 0.08 0.1 0.086 11 1
Sapindus saponaria 0.2 0.22 0.097 11 1
Terminalia oblonga 0.05 0.06 0.019 11 1
Acacia bonariensis 0.26 0.3 0.043 11 1
Carinaria estrellensis 0.09 0.1 0.164 11 0.999
Ficus boliviana 0.06 0.07 0.023 11 1
Ocotea guianensis 0.27 0.27 0.001 11 1
Schizolobium parahyba 0.07 0.07 0.006 11 1
Cecropia concolor 0.22 0.23 0.004 11 1
Heliocarpus americanus 0.3 0.31 0.03 11 1
Trema micrantha 0.07 0.09 0.034 11 1

TABLE A3. Results of the regression analyses that were run for each trait, using the decomposition rates of the three different wood types separately.

  SSC SNSC HW
Trait R2 P R2 P R2 P
WD 0.02 0.653 0.00 0.891 0.00 0.905
DBH* 0.29 0.038 0.42 0.009 0.49 0.004
VA 0.06 0.387 0.08 0.302 0.10 0.259
FA 0.17 0.130 0.30 0.297 0.07 0.358
PA 0.07 0.341 0.02 0.631 0.01 0.756
CN 0.06 0.387 0.01 0.737 0.01 0.807
N** 0.06 0.383 0.01 0.676 0.01 0.745
P** 0.04 0.488 0.03 0.552 0.00 0.910
L*** 0.19 0.105 0.23 0.074 0.14 0.175
PE** 0.13 0.193 0.07 0.331 0.05 0.418

WD=Wood density (g cm-3), DBH = Living tree diameter at breast height (cm), VA=Vessel area (%), FA=Fibre area (%), PA=Parenchyma area (%), CN=C:N ratio, N=N concentration (mg / g DW), P=P concentration (mg / g DW), L=Lignin concentration (mg / g DW), PE=Phenolic extractives (mg / g DW)

* = log10-transformed prior to analyses

** = ln-transformed prior to analyses

*** = power (x2) transformed prior to analyses


TABLE A4. Results of the regression analyses using decomposition rates calculated under the twelve different assumptions. For assumptions associating the numbers, refer to Table B1.

  Assumption number
  1 2 3 4 5 6 7 8 9 10 11 12
Trait R2 P R2 P R2 P R2 P R2 P R2 P R2 P R2 P R2 P R2 P R2 P R2 P
WD 0.00 0.919 0.00 0.900 0.00 0.864 0.00 0.852 0.00 0.861 0.00 0.861 0.00 0.862 0.00 0.844 0.00 0.844 0.00 0.848 0.00 0.848 0.00 0.848
DBH 0.44 0.007 0.45 0.006 0.45 0.006 0.46 0.006 0.41 0.011 0.43 0.008 0.44 0.007 0.44 0.007 0.44 0.007 0.38 0.014 0.38 0.014 0.38 0.014
VA 0.08 0.314 0.07 0.343 0.07 0.346 0.09 0.276 0.02 0.290 0.09 0.290 0.10 0.252 0.10 0.261 0.10 0.261 0.08 0.293 0.08 0.293 0.08 0.293
FA 0.03 0.530 0.03 0.540 0.04 0.482 0.03 0.521 0.04 0.506 0.04 0.506 0.03 0.536 0.03 0.557 0.03 0.557 0.02 0.602 0.02 0.602 0.02 0.602
PA 0.00 0.990 0.00 0.985 0.00 0.982 0.00 0.960 0.00 0.990 0.00 0.990 0.00 0.933 0.00 0.912 0.00 0.912 0.00 0.884 0.00 0.884 0.00 0.884
CN 0.06 0.382 0.06 0.383 0.06 0.401 0.05 0.430 0.05 0.405 0.05 0.405 0.05 0.427 0.05 0.408 0.05 0.408 0.06 0.381 0.06 0.381 0.06 0.381
N 0.08 0.318 0.08 0.324 0.07 0.340 0.07 0.361 0.07 0.343 0.07 0.343 0.07 0.356 0.07 0.340 0.07 0.340 0.08 0.316 0.08 0.316 0.08 0.316
P 0.00 0.893 0.00 0.912 0.00 0.951 0.00 0.969 0.00 0.935 0.00 0.935 0.00 0.977 0.00 0.992 0.00 0.992 0.00 0.961 0.00 0.961 0.00 0.961
L 0.24 0.066 0.23 0.072 0.23 0.074 0.21 0.085 0.21 0.086 0.21 0.086 0.20 0.098 0.19 0.110 0.19 0.110 0.18 0.116 0.18 0.116 0.18 0.116
PE 0.06 0.369 0.07 0.343 0.09 0.285 0.06 0.365 0.07 0.352 0.07 0.352 0.06 0.402 0.06 0.390 0.06 0.390 0.05 0.418 0.05 0.418 0.05 0.418

WD=Wood density (g cm-3), DBH = Living tree diameter at breast height (cm), VA=Vessel area (%), FA=Fibre area (%), PA=Parenchyma area (%), CN=C:N ratio, N=N concentration (mg / g DW), P=P concentration (mg / g DW), L=Lignin concentration (mg / g DW), PE=Phenolic extractives (mg / g DW)

* = log10-transformed prior to analyses

** = ln-transformed prior to analyses

*** = power (x2) transformed prior to analyses


[Back to E091-259]