Trait convergence in a plant-pollinator network

Background/Question/Methods:

To understand biodiversity patterns in a community, and the mechanisms by which biodiversity is generated and maintained, it is essential to study the patterns of interactions among the species within the community. Interaction patterns can be complex, but recent methods in network ecology allow us to identify certain interaction rules that may help to predict how species are affected by the interactions they are involved in. For example, interactions in a community seem to be organized in modules, which are groups of species that tend to interact more frequently with each other than with those included in other modules. It has been hypothesized that species within a module will tend converge in their traits. Here we attempt to understand how pollinators affect the evolution of their resource plant in the context of a plant-pollinator interaction network, by searching for patterns of convergence in the plant traits that are assumed to mediate their interaction. The studied network consists of 40 plant species, 88 pollinating insect species, and 19,166 total interactions. The study was done in the Monte desert of Mendoza, Argentina, between 2006 and 2010. Plant traits include diameter, depth, color, symmetry, and orientation of the corolla; average flower height; and flowering length and timing. Pollinator species were grouped according to their potential selection pressure on the plants with which they interact, based on the literature. We used an Evolutionary Principal Components Analysis (EPCA) to evaluate two hypotheses: 1. species pollinated by pollinators in the same functional group will show trait convergence and 2. species that belong to the same module will show trait convergence.

Results/Conclusions

We found that plant species tend to converge in flowering length and timing, corolla symmetry and orientation, and flower height. A Generalized Linear Model analysis indicated that pollination by large bees and hovering flies explains the position of plant species in the EPCA axes. Lastly, modules did not reflect the convergence of traits suggested by the EPCA, as had been predicted in previous studies. Our results suggest that, in order to understand patterns of selection and coevolution at the community level, it is important to focus on functional groups that potentially exert similar selection pressure. Because modules are formed based on interactions between species (not functional groups) they might not reflect coevolutionary patterns.