Since the Millennium Ecosystem Assessment first evaluated the state of Earth’s life-support systems, recognition that nature sustains and enhances human life has been rapidly increasing. As such, many dimensions of environmental decision-making now require quantitative information on ecosystem services, precipitating increased demand for ecosystem-service models. Though widely valued and acknowledged as a biodiversity-driven ecosystem service, biological control of crop pests is rarely considered in land-use decisions because suitable biocontrol models do not exist. Here, we introduce a new tool and database for pest-control modeling at landscape to regional scales. Specifically, we compiled a comprehensive database of predator and pest abundances, predation rates, and crop damage from 132 studies across 6789 sites in 31 countries. We used this database to both determine the extent to which landscape composition alone can predict spatial heterogeneity in biocontrol variables and test one model’s ability to predict results found in another system. We further identified strategies for selecting and applying models to maximize predictive power.
Despite spatial heterogeneity in environmental conditions, farm management, and species assemblages, we found that landscape models explained significant spatial variation in every dimension of biocontrol, from enemy abundances to crop yields. Indeed, correlations between model predictions and observed data were significantly greater than 0 for each biocontrol variable, ranging from 0.37 to 0.44. Pests and enemies, however, exhibited context-dependent responses to landscape composition, in some studies increasing and in others decreasing in landscapes with more natural vegetation. Despite such variation, landscape models predicted variation in entirely independent datasets when models and data shared the same crop and similar landscape features. Our modeling approach thus offers a new path forward to integrate biological control into land-use planning and environmental decision-making. Moreover, we envision the supporting biocontrol data to be a living database— by adding studies focused on different crops and geographies, the reach of our modeling framework will continue extending to novel contexts, increasing predictive power and offering opportunities to answer critical questions about the ecology of biocontrol.