A Wrinkle in Space-Time: Spatial machine learning uncovers stage-specific drivers and hidden hotspots of illegal wildlife trade

Hannah Murray, Alexander Perigon, Bienvenu M. Takem Mbi, Anne Stephanie Kobla, Roméo Tene, Meredith L Gore, Matthew H Shirley, Bistra Dilkina

The illegal wildlife trade (IWT) persists despite intervention, demanding new approaches to allocate limited prevention and enforcement resources more effectively. Spatial machine learning has been applied to IWT for this purpose, but mainly along one dimension – place. We extend this to simultaneously address place, species, and time, applying machine learning to participatory mapping data analyzed through a crime script framework to identify drivers and undetected hotspots across the preparation, pre-activity, activity, and post-activity stages of the illegal pangolin scale trade in Cameroon. The level of biodiversity, distance to international borders, and tree cover emerged as shared drivers across all stages, while hunting zones, food insecurity, and banking infrastructure distinguished individual stages. Each location was further assigned a relative risk score of trafficking involvement, surfacing high-risk candidate hotspots absent from existing records. These findings demonstrate that machine learning can extend IWT analysis beyond space to include time and species, sharpening targets for interventions like enforcement.