Multivariate flower phenotype and proboscis length shape specialised pollination niches

By analysing multivariate floral traits and the topology of the pollination network in a community of coexisting plant species that share pollinators, we investigated whether plant-pollinator interactions are influenced by multivariate flower phenotype and/or  morphological traits. 

Methods
We used plant-pollinator network topology to objectively characterise pollination niches. We built the quantitative plant-pollinator interaction matrix assigning to each cell the average visitation rate (i.e. number of visited flowers per observation period) across observation periods of each pollinator morphospecies (j columns) to a given plant species (i rows). We first tested whether the interaction network exhibited a nested and/or modular structure. We detected the main pollinators in each module taking into account the individualisation of species with greater importance in the network modules recognition.To characterise the overall variation in floral phenotype among the 11 vervain species, we constructed a matrix containing nine quantitative variables of flower measurements from five individuals per species (N=55). The variables included were: floral display (number of flowers per inflorescence and inflorescence area), flower colour (PC1 and PC2 coefficients of the PCA of the reflectance data), flower shape (PC1 and PC2 coefficients of the geometric morphometric analysis of corolla shape in frontal view and corolla tube length) and nectar reward (nectar volume and amount of sugar per flower). Besides, we also tested whether the specialised pollination niches recognised by the modularity network analysis were associated with the morphological correspondence between flowers and probosces and trait-matching.