A framework for capturing farm-scale variation in process-based simulations for agroecosystem resilience

• Model-independent framework captures potential farm-scale variation in simulations. • Application at two UK sites captured realistic outcomes and rare extremes. • Farm-scale heterogeneity influences agroecosystem performance and resilience. • Harmonised inputs enable consistent, comparable applications of different models. • Supports scenario testing and resilience analysis under management and climate change. Agroecosystem models typically simulate average conditions in fields or over a gridded landscape. This approach overlooks the spatial variation that shapes resilience at the farm scale. Yet, this variation can strongly influence production and regulating services, and their robustness to stresses. We developed a model-independent five-step framework for capturing farm-scale variation in process-based simulations for agroecosystem resilience. The framework comprises: 1) data acquisition; 2) data processing; 3) simulation; 4) output sampling, and 5) analysis and interpretation. We applied the framework to two contrasting UK case exemplars, differing in climate, soils, and management. We used the Rothamsted Landscape Model to simulate calorific productivity, nitrate runoff and N 2 O emissions. Simulations captured realistic farm-scale variability, including rare outcomes, reflecting the emergent effects of heterogeneous soils, weather, and farm management. This flexible and transferable approach enables the simulation of agroecosystem performance and resilience using process-based models. It can help to explore system dynamics and risks at the farm scale under plausible spatial and temporal heterogeneity. By explicitly capturing these sources of heterogeneity at the scale most relevant for decision making, i.e., the farm, it provides a robust basis for applied research, policy design, and scenario exploration under current or future environmental change.