What if we lose a hub? Experimental testing of pollination network resilience to removal of keystone floral resources

On the basis of theoretical predictions, pollination networks seem to be resilient to random node elimination but sensitive to targeted exclusion. However, such predictions have a very weak empirical basis. In order to test the robustness of the pollination network to short-term disturbances, we removed inflorescences of the most connected species occurring in a lowland meadow network using the before-after approach and compared the result with that obtained by network modelling. The manipulated network showed no significant differences for the most commonly used metrics, but was more generalized than control networks, owing to a change in the preferences of pollinators. Furthermore, no secondary extinctions (emigrations) were found, owing to the considerable natural variation found among insect species assemblages. Following elimination of the most linked plant species, a new hub was detected in the experimental meadow, the hub node being a plant species with a similar inflorescence to that removed, and formerly playing the role of a peripheral node. We conclude that exclusion of the main food source forced insects to change their specialized preferences to other plant species that were available. Mostly, these had inflorescences similar to those that were removed.