Phylogenetic and functional
structure
The phylogenetic and functional structure measures were taken to test
the hypothesis that Caatinga vegetation near roads will be more
phylogenetically and functionally clustered than Caatinga vegetation far
from roads. For the phylogeny of the woody component, we used all the
tree, shrub and liana species of the phytosociological survey identified
at least to the family level (87 species). The phylogeny of herbaceous
and sub-shrub species (non-woody component) included all non-woody
species identified at least to the family level for a total of 86
species. Except for Selaginella convoluta (see phylogenetic
signal section), all the morphospecies sampled in this survey were
inserted into mega-tree R20160415.new (Gastauer & Meira-Neto, 2017) via
phylomatic function, which was dated (Bell et al. , 2010) using
the algorithm bladj in combination with the file ‘ages’ in Phylocom
(Gastauer & Meira-Neto, 2017; Webb et al. , 2011). That mega-tree
was chosen because it is an updated mega-tree using the updated APG IV
classification (The Angiosperm Phylogeny Group, 2016).
With incidence data, we calculated phylogenetic structure for plots near
roads and for plots further from roads, considering woody species or
non-woody species, using the following phylogenetic measures with
Phylocom 4.2 software (Webb et al. , 2011): mean pairwise distance
(MPD), mean nearest taxon distance (MNTD), net relatedness index (NRI)
and nearest taxon index (NTI). Phylogenetic analysis was done using the
unconstrained model of Phylocom (Gastauer & Meira-Neto, 2015; Webbet al. , 2011) that maintains the species richness of each plot,
but all species of the metacommunity (pool of species) have the same
chance of being included (Webb et al. , 2002) within each of the
10,000 randomizations. We defined the metacommunity as all the species
sampled in the collection area (8,000m2), identified
at species, genus or family level. We used the ’comstruct’ function to
calculate the phylogenetic metrics. We calculated these measures at
different spatial scales: 10-m x 10-m and 20-m x 50-m.
In order to evaluate the functional structure of communities, we
computed Gower distances among species based on the trait matrix (see
above) using the ‘gowdis’ command from ‘FD’ package in R to perform the
rarefaction and extrapolation of functional diversity (Villéger et al.
2008). From this we built a functional dendrogram in Newick format using
‘hclust2phylog’ from the ‘ade4’ package in R (Dray et al. , 2018).
We then computed traitNRI and traitNTI values for each plot using
“ses.mpd” and “ses.mntd” commands and the unconstrained null model
that assumes the same probability for all species to enter randomized
communities. This works similar to the computation of phylogenetic
community structure, but instead of phylogenetic distances among species
of a plot, functional distances among species with the focal trait are
calculated. The functional tree was built with all woody and non-woody
species (n=179).
We tested statistical significance through a bilateral t-test for one
sample, using the generated NRI and NTI metrics (Gastauer & Meira-Neto,
2015).