Quantifying phylogenetic and nonphylogenetic patterns in the richness, frequency, and identity of links in a herbivore-parasitoid interaction network

Uncovering the patterns and structure in species interactions is central to understanding community assembly and dynamics. Species interact via their phenotypes, but identifying and quantifying the traits that structure species-specific interactions (links) can be challenging. Where these traits show phylogenetic signal, link properties (such as which species interact, and how often) may be predictable using models that incorporate phylogenies in place of trait data. However, quantification of phylogenetic patterns in link properties is conceptually and methodologically challenging because it requires co-estimation of multiple phylogenetic and nonphylogenetic pattern types in interaction data for multiple sites, while controlling for confounding effects and making biologically plausible assumptions about which species can interact. Here we show how this can be done in a Bayesian mixed modelling framework, using data for trophic nteractions between oak cynipid galls and parasitoid natural enemies. We find strong signatures of cophylogeny (i.e. related arasitoids attack related host galls) in both link incidence (presence/absence) and link frequency data, alongside patterns in link incidence/richness and identity across sites that are independent of either parasitoid or gall wasp phylogeny. Our results are robust to substantially reduced sample completeness, and are consistent with structuring of trophic interactions by a combination of phylogenetically conserved and phylogenetically labile traits in both trophic levels. We show that incorporation of phylogenetic relationships into analyses of species interactions has substantial explanatory power even in the absence of trait data, with potential applied use in prediction of natural enemies of invading pests and non-target hosts of biocontrol agents.