The purpose of using a PCA was to generate only two variables
(principal components) that would explain the most variance. The
species would be distinguished mainly on the first axis in a
gradient from low (negative) to high (positive) values (Gauch,
1982). Since the variance explained by the first axiswas high (54%),
we consider that this objectivewas fulfilled, and therefore selected
the species that appeared at the positive end of the ordination axis;
those which have a combination of high values for the five
variables used. Before applying the PCA, we combined the data
obtained from samples taken at the 17 sites. This is because not all
species were present at all sites, so the complete matrix had too
many zeros. Because of this, we did not have replicates for all
46 species of shrubs. The PCA was performed using the PC-ORD
2.0 statistical package.