*2.4. Community Diversity Indices*

We compared species richness and two phylogenetic diversity metrics for the plants and herbivores across sites. Phylogenetic diversity metrics included the standard effect sizes of mean pairwise distance (*ses*MPD) and mean nearest-taxon distance (*ses*MNTD) in the package *picante* in R [31]. The *ses*MPD metric is sensitive to the phylogenetic diversity of deep-branching lineages, whereas *ses*MNTD is sensitive to diversity patterns close to the tips of the phylogeny [32]. These metrics use null models to determine whether observed communities contain taxa that are phylogenetically clustered or overdispersed [33]. For each site, we randomly generated 999 communities that shuffled the names of taxa across the phylogeny while holding the richness of the community constant.

To characterize turnover in plant and animal communities, we quantified dissimilarity across sites. We considered both the taxonomic and phylogenetic composition of communities using phylogenetic community dissimilarity (PCD) in *picante* [34]. The PCD metric accounts for taxonomic dissimilarity (i.e., PCDc; 'community' species differences) and evolutionary dissimilarity (PCDp; 'phylogenetic' differences between non-shared species), where PCD = PCDc × PCDp. If PCD = 1, two sites are no more or less similar than communities assembled at random from the species pool. If PCDp = 1, then PCD = PCDc, and communities reflect only taxonomic differences. However, if PCDc = 1, then differences are due entirely to the phylogenetic distances between nonshared species (i.e., PCDp). If all nonshared species are closely related, PCDp is low; if two sites have nonshared species from very divergent clades, then PCDp is high. Because the PCD metric is based on species presence/absence patterns, and because PCDp is undefined in comparisons of sites with identical species composition, we focused our analyses on PCD and PCDc and infer the relative importance of phylogeny by comparison.
