RESUMEN
BACKGROUND: The occurrence and frequency of plant-pollinator interactions are acknowledged to be a function of multiple factors, including the spatio-temporal distribution of species. The study of pollination specialization by examining network properties and more recently incorporating predictors of pairwise interactions is emerging as a useful framework, yet integrated datasets combining network structure, habitat disturbance, and phylogenetic information are still scarce. RESULTS: We found that plant-pollinator interactions in a grassland ecosystem in the foothills of the Rocky Mountains are not randomly distributed and that high levels of reciprocal specialization are generated by biological constraints, such as floral symmetry, pollinator size and pollinator sociality, because these traits lead to morphological or phenological mismatching between interacting species. We also detected that landscape degradation was associated with differences in the network topology, but the interaction webs still maintained a consistently higher number of reciprocal specialization cases than expected. Evidence for the reciprocal evolutionary dependence in visitors (e.g., related pollinators visiting related plants) were weak in this study system, however we identified key species joining clustered units. CONCLUSIONS: Our results indicate that the conserved links with keystone species may provide the foundation for generating local reciprocal specialization. From the general topology of the networks, plant-pollinators interactions in sites with disturbance consisted of generalized nodes connecting modules (i.e., hub and numerous connectors). Vice versa, interactions in less disturbed sites consisted of more specialized and symmetrical connections.