Livestock production is a major driver of global environmental change, contributing to greenhouse gas emissions and biodiversity loss, primarily through the conversion of tropical forests to pasture.
Yet, livestock occur in diverse production systems, each with unique impacts depending on the social-ecological context in which they are embedded. While understanding the spatial distribution of different livestock systems is crucial for sustainability planning, this geography remains unmapped for many regions, leaving policymakers and researchers with a critical data gap. Here, we use largely untapped spatial datasets on livestock, including vaccination, registers, and transaction data, and apply active learning and decision trees to classify and map major livestock systems at scale.
We demonstrate our approach for the 4.2 million square kilometres Dry Diagonal covering the Caatinga, Cerrado, Chiquitano, and Chaco ecoregions across parts of Argentina, Bolivia, and Brazil, which are global hotspots for livestock production and deforestation. Three main findings emerge.
Together, the diverse patterns of livestock production we uncover highlight the need for targeted, context-sensitive land and livestock management strategies in tropical dry woodlands.