Large animals provide crucial seed dispersal services, yet face continued threats and are susceptible to changes in landscape composition and configuration. Thus, there is a growing imperative to improve understanding of animal‐generated seed dispersal using models that incorporate spatial complexity in a realistic, yet tractable, way.
We developed a spatially explicit agent‐based seed dispersal model, with disperser movements informed by biotelemetry data, to evaluate how landscape composition and configuration affect seed dispersal patterns. We illustrated this approach for the world's second largest ratite, the emu (Dromaius novaehollandiae), a highly mobile generalist frugivore considered an important long‐distance disperser for many plant species across Australia.
When animal movement is unrestricted, model parameters related to seed gut passage largely determine seed dispersal kernels. However, as habitat loss and fragmentation increase, the extent of long‐distance dispersal events is reduced and seed shadows became progressively more aggregated. This effect is due to the emu not being able to move between disconnected parts of the landscape, with small changes in habitat structure causing decreased long‐distance dispersal.
We simulated seed dispersal patterns generated by three commonly used generic models of animal movement – unbiased and biased correlated random walks and Lévy walks – to evaluate how different representations of movement affect estimations of animal movements and emergent seed dispersal patterns. Simulated movements informed by the emu biotelemetry data resulted in longer median seed dispersal distances than do the three generic models.
Synthesis. Changes in landscape composition and configuration can dramatically alter patterns of zoochorous seed dispersal as they influence animal movement. However, when models are used to simulate the patterns of seed dispersal, decisions about how animal movement is represented also affect estimates of seed dispersal.