The structure of flower visitor networks in relation to pollination across an agricultural to urban gradient

1. Pollination is a major ecosystem service in which insects, particularly bees, play an important role for the reproduction of most angiosperms. Currently, this service is considered under threat due to reported bee declines. Moderately urbanised areas could be important for pollinators and pollination; however, compared to agricultural and natural systems, they are poorly studied. 2. Here, we investigated the relative effects of local habitat quality and anthropogenic land use across an agricultural to urban gradient for local plant and flying insect communities. We quantified local flower visitor networks and related network architecture to these local and landscape factors using structural equationmodelling. Flower visitor network architecture is often assumed to act as a surrogate for the ecosystem service of pollination. To test this idea, we related network metrics to pollination of four experimental, insect pollinator-dependent plant species. 3. Overall, local land use markedly influenced plant and flying insect communities. Flower richness and bee richness were higher in urban compared to agricultural areas. Flower visitor network metrics (e.g. linkage density) increased with the proportion of urban area surrounding a site. Also, relative to agricultural areas, urban flower visitors were more generalised and foraged from a higher number of plant species, likely a consequence of higher urban flowering plant richness. However, urban bees also visited a lower proportion of the available flowering plants (higher specialisation). Surprisingly, linkage density, network specialisation and flower visitor generality were not related to pollination of our four experimental plants per se. Rather, it was the proportion of urban cover, flying insect abundance and bee richness that were positively related to pollination. 4. Our findings show strong effects of local land use on plant and flying insect communities and flower visitor interaction networks. We observed increased overall visitation rates and pollination services to our experimental plants in urban compared to agricultural areas, despite increased urban flower visitor generality. Indeed, flower visitor network metrics were a poor proxy of provision of the ecosystem service of pollination. Nevertheless, our results point to potential facilitating effects of diverse urban floral and bee communities for pollination.