Functional groups of communities
We calculated fuzzy-weighted species composition to reflect the functional relatedness among communities (Pillar et al. 2009; Duarte et al. 2016), using the matrix.x function of package SYNCSA (Debastiani & Pillar 2012). For that we considered only the effect traits, thus, not including the life forms. Next, based on Euclidean distances between plots, we submitted the fuzzy-weighted species composition to cluster analysis using Ward’s method. For the derived classifications up to five functional groups of plots, we tested group partition sharpness by using the bootstrap resampling procedure (Pillar 1999). Further, for a synthetic view of the functional patterns across communities we submitted the fuzzy-weighted community composition matrix to covariance-based Principal Components Analysis (PCA) and tested the significance of the ordination axes (Pillar 1999). For these analyses we used the functions Cluster and Ordination in MULTIV software (available at http://ecoqua.ecologia.ufrgs.br).