Recovery of seed-dispersal interactions and functions in tropical forest ecosystems

Biodiversity includes species and their interactions, which together are fundamental to maintaining ecosystem functioning. At the community level, species interactions form complex networks that drive ecological processes such as seed dispersal, which arises from interactions between fleshy-fruited plants and frugivorous animals. Seed-dispersal is central to the natural regeneration of tropical forests, a process that underpins forest recovery. During natural regeneration, plants recolonize degraded areas via seed dispersal by animals, while animals depend on recovering plant communities for food and shelter. Deforestation and other human disturbances constrain this process at multiple spatial scales. Locally, the recovery stage of a patch determines which species and interactions can occur, and at the landscape scale, surrounding habitats define the species pool capable of recolonizing. Despite its importance, natural forest regeneration remains poorly understood because studies of forest recovery often overlook species interactions. A trait-based approach helps address this gap. Functional traits determine how species respond to disturbance, select interaction partners, and contribute to ecological processes. Using this approach, my thesis investigates how plant and animal communities, and their seed-dispersal interactions and functions, recover after human disturbances in tropical forests. I explored these processes across local and landscape scales in three complementary studies. In Study I, I developed a trait-based simulation model to disentangle how local and landscape-scale factors jointly shape the seed rain reaching recovering forests through animal-mediated dispersal. Using data on species traits from the literature, I tested how local resource diversity, distance from an undisturbed forest, and the degree of network specialization affect the abundance and diversity of seeds dispersed by birds. I found that seed arrival declined with distance but increased with resource diversity in recovering forests. Moreover, birds in less specialized networks dispersed a broader range of seeds, increasing seed diversity. These findings reveal that seed dispersal between forest patches is determined by local resource availability, landscape connectivity, and the structure of interaction networks. Enhancing fruit resource diversity near forest remnants can therefore help attract frugivores to degraded areas and overcome spatial limits to natural regeneration. For studies II and III, I conducted extensive fieldwork in the Chocó lowland forests of north western Ecuador, one of the world’s most biodiverse and threatened regions. Across 62 plots spanning active land use, regenerating forests (1–38 years of recovery), and old-growth forests, I collected data on seed-dispersal interactions, seed rain, and plant and animal traits. In Study II, I examined how local forest structure and patch connectivity influence the functional recovery of seed rain, mediated by local seed-dispersal interactions. I found that local forest structure mainly influenced the functional diversity of the seed rain through its effects on plant–animal interactions, while patch connectivity affected its functional composition directly and indirectly. Structurally complex forests promoted diverse interactions and thus higher functional diversity, while better-connected patches facilitated the arrival of large, late-successional seeds via animal seed dispersers. This demonstrates that forest structure and connectivity shape distinct but complementary aspects of forests' functional recovery. In Study III, I estimated the time needed for the functional diversity of seed-dispersal interactions to recover. Using hierarchical Bayesian models, I separately modelled the recovery of fruiting plants, frugivorous animals, and their interactions, accounting for variation in patch connectivity. I found that plant functional diversity reached values similar to old-growth forests even under active land use due to the presence of remnant trees, whereas animal functional diversity required nearly 40 years to recover. Seed-dispersal interactions, on the other hand, took roughly two decades to return to old-growth levels. Well-connected patches recovered faster, but uncertainty increased with isolation. These results emphasize the importance of long-term conservation policies that allow for local ecological processes to recover. Overall, my thesis provides one of the first comprehensive, trait-based perspectives on natural forest regeneration. By developing, combining, and adapting new and established methods, including simulation modelling, the interaction trait-space, and Bayesian recovery models, I bridge the gap between community responses, the recovery of interactions, and the ecological functions they sustain. The approaches developed here can also be applied to other processes such as pollination, herbivory, or decomposition. My findings underscore that forest recovery depends on both local (resource diversity and forest structure) and landscape-scale (connectivity and distance to old-growth forests) factors. Ultimately, restoration should focus not only on species but also on the interactions that link them, as only through this approach can we strengthen the connection between biodiversity and ecosystem functioning and ensure the long-term maintenance of tropical forests.