Computational protocol: Functional Trait Strategies of Trees in Dry and Wet Tropical Forests Are Similar but Differ in Their Consequences for Succession

Similar protocols

Protocol publication

[…] We used principal component analysis to quantify spectra of trait-based multivariate plant strategies for each forest type separately. The PCA biplots show the main trade-offs across (standardized) functional traits based on principal axes of variation, where binary variables are treated as dummy variables. Trait spectra for dry and wet forest species were compared by correlating the correlation coefficients of all pairwise trait combinations; in each site 11 traits were measured, resulting in 55 pairwise trait correlations per site. Subsequently the pairwise trait correlation coefficients derived from dry forest species were correlated with the pairwise trait correlation coefficients derived from wet forest species. Spearman correlation coefficients were used, since not all traits are normally distributed, except for relating the binary variables [deciduousness (De), leaf compoundness (LC) and biotic dispersal (Di)] when we used the Phi coefficient, a measure of association between binary variables whose interpretation is similar to correlation coefficients.We also examined whether the trait associations found were influenced by evolutionary histories. To this end, we recovered phylogenetic trees for the dry forest species and the wet forest species using Phylomatic [], scaling branch lengths to one. For all traits and each forest type we explored phylogenetic signal (Blomberg’s K []) and compared this to random trait distributions over the phylogenetic tree, using the package “Picante” []. Phylogenetically independent contrasts were computed as the difference in the mean trait values for pairs of sister species and nodes, using the package “Ape” [] and we compared whether trait associations were similar with and without considering phylogeny [].Species scores on the first two principal components of the PCA were scaled up to community level using the Community Weighted Mean (CWM) [,], which is calculated as follows: CWM= ∑i=1Swi× xi where S is the total number of species, w i is the relative basal area of the ith species and x i is the score on the PCA axis of the ith species. Relative basal area is a measure of species’ relative contributions to the total basal area represented by functional trait measurements in each plot (which is in turn at least 80% of total basal area in a plot). The relative basal area was used for weighting, rather than the abundance, because it reflects the species’ biomass, an indicator of plant performance and adaptation to local conditions. These community weighted mean scores on the PCA axes reflect the average multivariate plant strategy in the community, and were regressed against stand basal area (m2/ha) (including cacti in the case of dry forest). Stand basal area is a structural variable of succession and logarithmically relates to forest age in both forest types [see supplementary material in ]. Stand basal area was used, and not age, because it better reflects aboveground biomass, understory light interception and environmental conditions [] as well as competitive interactions []. All statistical analyses were carried out using R v. 2.13.1 []; for multivariate analysis we used the package ‘Vegan’ []. […]

Pipeline specifications

Software tools Phylomatic, PHYSIG, Picante, APE
Application Phylogenetics