Abstract
Plant-pollinator networks have been widely used to understand the ecology of mutualistic interactions between plants and animals. While a number of general patterns have been identified, the mechanisms underlying the structure of plant-pollinator networks are poorly understood. Here we present an agent based model (ABM) that simulates the movement of bees over heterogeneous landscapes and captures pollination events, enabling the influence of landscape pattern on pollination networks to be explored. Using the model, we conducted a series of experiments using virtual landscapes representing a gradient of forest loss and fragmentation. The ABM was able to produce expected trends in network structure, from simulations of interactions between individual plants and pollinators. For example, results indicated an increase in the index of complementary specialization (H2') and a decline in network connectance with increasing forest cover. Furthermore, network nestedness was not associated with the degree of forest cover, but was positively related to forest patch size, further supporting results obtained in the field. This illustrates the potential value of ABMs for exploring the structure and dynamics of plant-pollinator networks, and for understanding the mechanisms that underlie them. We attribute the results obtained primarily to a shift from specialist to generalist pollinators with increasing forest loss, a trend that has been observed in some field situations.