Predicting frugivore generated seed rain in different environmental contexts: a modelling approach applied to a forest specialist

Frugivorous primates play a key role in the regeneration of tropical forests, ecosystems that have been heavily deforested and altered, by dispersing viable seeds via endozoochory. Due to their highly contingent and site-specific behavioural and ranging responses to habitat change, simulation models can help to better understand the consequences of different environmental contexts on their seed dispersal services. Here, we extended a previously developed agent-based model (ABM), to reproduce, in silico, the spatially explicit seed dispersal pattern (seed rain) of an endangered Neotropical primate, the black lion tamarin (Leontopithecus chrysopygus). The model focused on behavioural modes, thus integrating responses to resources (fruiting trees) with motion capacity and internal energy state. We validated the model with data from four lion tamarin groups inhabiting forest fragments of different sizes: one small (100 ha), one medium-sized (515 ha), one continuous (33800 ha) and one riparian forest. In all four forest types, our model produced realistic movement patterns as well as seed dispersal patterns generated by these lion tamarin groups, suggesting it captured most of the processes resulting in the seed rain. We compared the simulated patterns with observed patterns by comparing the used area (hectares), daily path length, activity budget, movement rate, path twisting and the defendability index (all related to movement). Simulated used areas only matched with empirical observations after the implementation of resource monitoring behaviour at the territory border. This underlines the need of specific behavioural rules based on the pattern-based modelling approach to successfully generate the observed patterns. The model slightly overestimated the seed dispersal distance (SDD) and the pattern of seed deposition was less aggregated than observed empirically, but still congruent with sampling limitations. The model can run on a small set of parameters that can be estimated within a few days of data collection (feeding and sleeping tree coordinates and group behaviour sampling), thus having the potential to predict the seed rain from multiple groups of lion tamarins. Finally, we discussed potential improvements and encouraged replication of such modelling efforts to other forest dependent species, enabling a spatially explicit evaluation of the processes guiding the seed rain by forest specialists in different environmental contexts.