Integrating Earth Observation and Ecological Modeling to Advance Sustainability Assessment of Cocoa Expansion

Abstract Expansion of agri-food crops (e.g., cocoa) into tropical forests threatens biodiversity and ecosystem services, yet crop-specific impacts remain poorly understood and rarely integrated directly from Earth observation to ecological models. We present a scalable framework that fuses ~10 m Sentinel-1/2 satellite imagery with supervised machine learning to generate model-ready, crop-specific maps and quantify ecological impacts. Applied to Peru's Ucayali Agroecological Living Landscape, we mapped 25,659 ha of cocoa in 2020, 68% of which directly replaced forest since 2000. This expansion reduced biodiversity by ~0.5%, eliminated ~3.0 million tonnes of carbon storage, and increased nitrogen export to runoff by ~58 tonnes per year, degrading water quality. Our globally transferable framework delivers spatially explicit, policy-relevant metrics that pinpoint ecological impact hotspots, thereby informing targeted, sustainable land management strategies in tropical agri-food landscapes.