Plant-flower visitor networks in a serpentine metacommunity: assessing traits associated with keystone plant species

Consistent topology of plant-pollinator networks across space may be due to substitutability of the plant species most important for community function (keystone species). It is unclear, however, whether keystone species identity varies within a community type and what traits underlie this variation. Using a network biology approach, we assess whether keystone plant species vary across a metacommunity of five serpentine seeps in California and determine the features that predict their identity. We define keystone species as those with high strength, low node specialization index (NSI), and/or low d' and determine whether these parameters are predicted by floral traits (flower biomass, number of open flowers per plant, symmetry, or stamen number) and/or ecological features (variation in local floral abundance, endemism) within seeps and across the metacommunity. Keystone species identity varied among seeps and was associated with local flower abundance: mean floral abundance correlated positively with strength but negatively with NSI within most seeps as well as across the metacommunity. For the metacommunity, flower biomass correlated negatively with NSI while variation in flower abundance correlated negatively with strength. Across the metacommunity, the d' metric was associated with flower biomass, whereby plants with smaller flowers interacted with the most abundant pollinators across the metacommunity. Results suggest that connectance and interaction evenness may not be greatly influenced by community composition turnover due to substitution of keystone plant species across space. Keystone species can be predicted by functional traits but which trait (flower abundance or size) depended on the metric used and the level observed.