Optimizing size thresholds in a plant-pollinator interaction web: towards a mechanistic understanding of ecological networks.

Using functional traits together with abundance effects strengthens the prediction of interactions between pairs of species in ecological networks. Insights into the way species interact as well as prediction accuracy can be gained when thresholds for trait value combinations that make interactions possible are optimized through model selection. I present novel data of two subalpine plant-pollinator communities and build several stochastic models integrating flower abundance and morphological threshold rules that allow or restrict interactions between species. The number of correctly predicted interactions was highest when thresholds were set so that the insect's proboscis was not shorter than the nectar-holder depth minus 1-1.6 mm, and not wider than the nectar-holder width minus 0.5 mm. In comparison with models based solely on plant abundance effects, the model incorporating optimized size thresholds better predicted the distribution of the trait differences between plants and insects. This indicates that a mechanistic approach of interaction webs based on optimized size thresholds provides valuable information on community structure. The possible implications for community functioning are discussed.