One sentence summary: Low interaction frequency and high specialization between interacting partners additively increase the vulnerability of mutualistic interactions to disruption in a fragmented habitat.

Interactions between species form complex networks that vary across space and time. Even without spatial or temporal constraints mutualistic pairwise interactions may vary, or rewire, across space but this variability is not well understood. Here, we quantify the beta diversity of species and interactions and test factors influencing the probability of turnover of pairwise interactions across space. We ask: 1) whether beta diversity of plants, pollinators, and interactions follow a similar trend across space, and 2) which interaction properties and site characteristics are related to the probability of turnover of pairwise interactions. Geographical distance was positively correlated with plant and interaction beta diversity. We find that locally frequent interactions are more consistent across space and that local flower abundance is important for the realization of pairwise interactions. While the identity of pairwise interactions is highly variable across space, some species-pairs form interactions that are locally frequent and spatially consistent. Such interactions represent cornerstones of interacting communities and deserve special attention from ecologists and conservation planners alike.

Abstract. Island biogeography theory predicts that species richness increases with habitat area and declines with isolation. We expand this framework to address changes in the number of links and species in pollination webs from 12 isolated hills, ranging in area from tens to thousands of hectares, immersed in the agriculture matrix of the Argentine Pampas. We also studied whether total interaction frequency is partitioned more evenly among individual links in richer webs. Our results reveal a direct effect of area on the number of links and species present in each pollination web. However, link richness increased twofold faster than species richness with area. These area effects were not confounded by sampling effort or correlated incidence of exotic species, despite widespread habitat disturbance. Habitat proximity, an inverse measure of isolation, had a marginally significant influence on link but not on species richness. Increased link number was associated with decreasing dominance by any particular interaction and increasing interaction evenness. Despite the strong area effect, a rich pollination web sampled from a small, protected sierra suggests that simple conservation measures, such as reduced grazing and fire suppression, may effectively preserve much local interaction diversity.

Mutualistic interaction networks have been shown to be structurally conserved over space and time while pairwise interactions show high variability. In such networks, modularity is the division of species into compartments, or modules, where species within modules share more interactions with each other than they do with species from other modules. Such a modular structure is common in mutualistic networks and several evolutionary and ecological mechanisms have been proposed as underlying drivers. One prominent explanation is the existence of pollination syndromes where flowers tend to attract certain pollinators as determined by a set of traits. We investigate the modularity of seven community level plant-pollinator networks sampled in rupestrian grasslands, or campos rupestres, in SE Brazil. Defining pollination systems as corresponding groups of flower syndromes and pollinator functional groups, we test the two hypotheses that (1) interacting species from the same pollination system are more often assigned to the same module than interacting species from different pollination systems and; that (2) interactions between species from the same pollination system are more consistent across space than interactions between species from different pollination systems. Specifically we ask (1) whether networks are consistently modular across space; (2) whether interactions among species of the same pollination system occur more often inside modules, compared to interactions among species of different pollination systems, and finally; (3) whether the spatial variation in interaction identity, i.e., spatial interaction rewiring, is affected by trait complementarity among species as indicated by pollination systems. We confirm that networks are consistently modular across space and that interactions within pollination systems principally occur inside modules. Despite a strong tendency, we did not find a significant effect of pollination systems on the spatial consistency of pairwise interactions. These results indicate that the spatial rewiring of interactions could be constrained by pollination systems, resulting in conserved network structures in spite of high variation in pairwise interactions. Our findings suggest a relevant role of pollination systems in structuring plant-pollinator networks and we argue that structural patterns at the sub-network level can help us to fully understand how and why interactions vary across space and time.