Although ecologists have long recognized that species interaction networks may arise as a consequence of both ecological and evolutionary processes, only recently have the effects of the latter been evaluated in community studies. Discriminating the relative effect of these processes has become a major challenge in theoretical and quantitative ecology. Here, we present an analytical framework that integrates phylogenetic and ecological information from bipartite networks. The method uses information from six matrices: one containing observed interactions between two species sets; two with trait information for each species set; two matrices containing phylogenetic distances; and one matrix containing information of temporal co-occurrence (i.e., the synchrony between bird capture and plant fructification during our sampling period). By a series of matrix multiplications and scaling up of functional and phylogenetic information, we obtain four probabilistic interaction matrices, two from each species set: two weighted by trait information and two weighted by phylogenetic information. The effects of traits and phylogeny on interactions, the phylogenetic signal on traits, and evidence of co-evolution were evaluated via matrix and partial matrix correlations, which were tested against null models. We applied this framework on data from a network of plant and frugivorous birds. We ran two sets of separate analyses: including and not including species co-occurrence data.
Results/Conclusions
When co-occurrence patterns were not considered, the only significant matrix correlation was for phylogenetic signal on bird traits (r=0.92, p≤0.001). When we did consider the co-occurrence patterns we found significant correlation for phylogenetic signal on bird (r=0.77, p≤0.001) and plant traits (r=0.75, p≤0.001). Matrix correlations between probabilistic interaction matrices weighted by plant and animal traits were not significant (p=0.15), as well as the partial correlation removing phylogeny (p=0.17). The results suggest that interactions are not mediated by traits, considering the trait sets we used. There is no evidence of co-evolution, since matrices weighted by phylogeny were not correlated (p=0.66). Our results reveal a strong phylogenetic signal on both plant and animal measured traits. Significant matrix correlation between bird phylogeny and traits without considering temporal co-occurrence suggests that phylogenetic signal for bird traits is stronger than for plant traits. Moreover, our findings show that considering forbidden links determined by temporal co-occurrence may shed light on processes structuring ecological networks. Our next step will be the evaluation of the statistical power of our framework, using simulated networks.