Distinct responses of antagonistic and mutualistic networks to agricultural intensification

Species interaction networks, which govern the maintenance of biodiversity and ecosystem processes within ecological communities, are being rapidly altered by anthropogenic activities worldwide. Studies on the response of species interaction networks to anthropogenic disturbance have almost exclusively focused on one interaction type at a time, such as mutualistic or antagonistic interactions, making it challenging to decipher how networks of different interaction types respond to the same anthropogenic disturbance. Moreover, few studies have simultaneously focused on the two main components of network structure: network topology (i.e., architecture) and network ecology (i.e., species identities and interaction turnover), thereby limiting our understanding of the ecological drivers underlying changes in network topology in response to anthropogenic disturbance. Here, we used 16,400 plant-pollinator and plant-herbivore interaction observations from 16 sites along an agricultural intensification gradient to compare changes in network topology and ecology between mutualistic and antagonistic networks. We measured two aspects of network topology-nestedness and modularity-and found that although the mutualistic networks were consistently more nested than antagonistic networks and antagonistic networks were consistently more modular, the rate of change in nestedness and modularity along the gradient was comparable between the two network types. Change in network ecology, however, was distinct between mutualistic and antagonistic networks, with partner switching making a significantly larger contribution to interaction turnover in the mutualistic networks than in the antagonistic networks, and species turnover being a strong contributor to interaction turnover in the antagonistic networks. The ecological and topological changes we observed in the antagonistic and mutualistic networks have different implications for pollinator and herbivore communities in agricultural landscapes, and support the idea that pollinators are more labile in their interaction partner choice, whereas herbivores form more reciprocally specialized, and therefore more vulnerable, interactions. Our results also demonstrate that studying both topological and ecological network structure can help to elucidate the effects of anthropogenic disturbance on ecological communities, with applications for conservation and restoration of species interactions and the ecosystem processes they maintain.