Scaling up behavior: modeling the interaction between agent-based seed dispersal and forest structure

Background/Question/Methods

Seed dispersal is vital to a forest ecosystem. In tropical forests, frugivorous vertebrates disperse the vast majority of seeds from fruiting canopy trees. As these agents forage, they deposit seeds aong their search paths, which are driven by the spatial distribution of resources. The resulting pattern of seed deposition sets the template for subsequent distributions of seedlings and adults. This implies second order dynamics resulting from a feedback between movement and landscape processes: the movement of dispersers alters the landscape upon which they base their subsequent movement decisions. Despite the complexity of seed dispersal and forest demography, can landscape and movement processes work in concert towards a self-organizing and self-reinforcing equilibrium of forest structure? To address this question, we developed a spatially explicit, individual-based model which simulates foraging, seed dispersal and forest growth in a single interconnected cycle. The objectives of this study were to 1) outline the fundamental dynamics of this novel simulated system and 2) identify realistic parameter values that induce spatial structure in the landscape.

Results/Conclusions

We found the model was sensitive to three parameters: spatial scale of foraging behavior, F, seed retention time, R, and tree lifespan. Shorter lifespans mean higher turnover of the forest, allowing patterns the agents have established in the seedling bank to be quickly realized in the adult stage. When R, also equivalent to maximum dispersal distance, is smaller than F, seeds are rarely taken beyond the local search area. This produces significant clustering of the landscape (according to a join count statistic), but is unrealistic for many vertebrate dispersal systems in which seeds are routinely taken out of local patches. In that case, where R is greater than F, and lifespan is on the order of decades, we found no significant clustering of trees, but instead strong seed deposition patterns. Compared to a null model, two seeds from the same mother were significantly more likely to be found under a pair of conspecific than heterospecific trees. Agents, while unable to shape tree distributions, were able to create patterns of genetic diversity in the seed pool. This suggests that for realistic forest turnover rates, foraging behaviors over short time scales can indeed affect forest structure over large spatial and temporal scales.