Inter-species feedbacks drive emergent productivity in agroforestry systems - an agent-based analysis

Introduction Conventional agricultural intensification has increased global food production but has also accelerated deforestation, soil degradation, and biodiversity loss. Agroforestry offers a sustainable alternative by integrating trees into farming systems to enhance ecosystem functions. However, predicting how reciprocal interactions between a focal crop species and multiple associated tree species shape long-term productivity under adaptive management remains a major scientific challenge. Methods We developed an empirically calibrated Agent-Based Model (ABM) based on a decade of measurements from a 25-year old multispecies agroforestry experiment integrating Ilex paraguariensis (yerba mate) with nine tree species. The model simulates species-specific growth, canopy shading, harvest, pruning, soil organic matter (SOM) feedbacks, and management interventions. It represents 778 interacting perennial individuals and enables quantitative exploration of reciprocal inter-species feedbacks under fixed and adaptive management strategies. Because the model is deterministic, statistical replication of simulation runs is not applicable. Results Simulations reproduce key field-observed patterns with quantitative agreement. Starting from degraded soil conditions, both management strategies show an initial fertility decline followed by recovery driven by endogenous SOM accumulation. Adaptive management yields a ~56% higher net productivity than fixed management, shortens the recovery time of soil fertility from ~260 weeks to ~88 weeks, and produces nearly threefold higher total biomass. Across species, the model reproduces observed relative Ilex paraguariensis yield differences, correctly predicting that Toona , Cañafístola , Petiribi , Anchico , and Kiri support higher harvest yields than the control (no trees), consistent with experimental field observations over a 10-year period. This quantitative agreement strengthens the model’s validity in capturing beneficial inter-species synergies. Conclusion The simulations reveal that the focal crop responds to tree-mediated shade and nutrient inputs while actively reorganizing soil fertility gradients through biomass extraction and residue return, thereby reshaping tree regrowth and competitive structure. Together, these dynamics define a mechanistically transparent and predictive framework linking empirical field data with long-term system forecasting.