Deterministic modelling of seed dispersal based on observed behaviours of an endemic primate in Brazil

Plant species models are among the available tools to predict the future of ecosystems threatened by climate change, habitat loss, and degradation. However, they suffer from low to no inclusion of plant dispersal, which is necessary to predict ecosystem evolution. A variety of seed dispersal models have been conceived for anemochorous and zoochorous plant species, but the coupling between vegetation models and seed dispersal processes remains rare. The main challenge in modelling zoochoric dispersal is simulating animal movements in their complex habitat. Recent developments allow straightforward applications of hidden Markov modelling (HMM) to animal movements, which could ease generalizations when modelling zoochoric seed dispersal. We tested the use of HMM to model seed dispersal by an endangered primate in the Brazilian Atlantic forest, to demonstrate its potential simplicity to simulate seed dispersal processes. We also discuss how to adapt it to other species. We collected information on movement, fruit consumption, deposition, and habitat use of Leontopithecus chrysomelas. We analysed daily trajectories using HMM and built a deterministic Model Of Seed Transfer (MOST), which replicated, with good approximation, the primate’s movement and seed deposition patterns as observed in the field. Our results suggest that the dispersal behaviour and short daily-trajectories of L. chrysomelas restrict the species’ role in large-scale forest regeneration, but contribute to the prevalence of resource tree species locally, and potentially maintaining tree diversity by preventing local extinction. However, it may be possible to accurately simulate dispersal in an area, without necessarily quantifying variables that influence movement, if the movement can be broken down to step-length and turning angles, and parametrised along with the distribution of gut transit times. For future objectives, coupling MOST with a DVM could be used to test hypotheses on tree species survival in various scenarios, simulating regeneration and growth at regional scales by including data on main dispersal agents over the area of interest, distribution of tree species, and land use data. The principal advantage of the MOST model is its functionality with data available from the literature as the variables are easy to parametrise.