The ordination topology of the sampling locations revealed
distinct clumping patterns according to the seven geographical
regions (Fig. 7A). The random sets of samples showed only modest
scatter around the centers of gravity on the two-dimensional
ordination space. The ordination topology was very similar when
done on species presence-absence data with almost no overlap
between the seven regions (data not shown). Phylogenetic beta
diversity, expressed as the unique (non-shared) fraction of total
phylogenetic diversity (branch-length) between two communities,
also revealed a clear clustering of the geographic regions, bearingmuch similarity with the two-dimensional topology of the
compositional relatedness (Fig. 7B).
For further interpretation we independently fitted environmental
gradients of salinity and temperature (mean values of each
location) on the ordination planes (envfit; vegan library of R software).
The arrows on Fig. 7 show the direction of the most rapid
change of salinity and temperature within the ordination space (the
direction of the gradient), and the strength of the gradient is
expressed as the square of the correlation coefficient between the
ordination and the variable (R2). The significance (p-value) of the
environmental variables was assessed by 999 random permutations
of the data. None of the random permutations had a similar or
better R2 than the original environmental fitting, giving highly
significant p-values