Stevia is a satellite-based geospatial intelligence system trained on South American mining and deforestation sites to detect the spectral and morphological signatures of artisanal and illegal gold extraction. Rather than treating forest loss as a single variable, our model analyzes features including exposed sediment, excavation morphology, mine ponds, riparian disturbance, access corridors, and temporal land-cover change to identify candidate operations and measure their expansion. The environmental signal is significant: artisanal and small-scale gold mining (ASGM) produces roughly 20% of global gold and is the world's largest source of mercury emissions to air and freshwater; the WHO estimates it accounts for approximately 37% of global mercury emissions. In the Amazon, mining concessions and illegal operations overlap approximately 450,000 km² of Indigenous territory, affecting 1,131 Indigenous lands, while mercury contamination associated with mining has been documented in at least 30 Amazonian rivers. Indigenous territories affected by mining in Bolivia, Ecuador, and Peru experienced at least three times the forest-loss rate of comparable Indigenous territories without mining. Stevia converts these disturbances into georeferenced detections that can be intersected with watershed topology, protected-area boundaries, Indigenous territories, elevation, and historical satellite observations rather than producing another generic deforestation heat map.
The value of this detection becomes greater when environmental and epidemiological layers are analyzed together. Mining excavation produces disturbed pools and altered hydrology while simultaneously bringing highly mobile populations into remote forest regions, conditions associated with malaria transmission. From 2007–2022, 358,774 mining-related malaria cases were recorded in Brazil, and research in a Brazilian Amazon mining population found gold miners were nearly five times more likely to acquire malaria. More recent spatial analysis of Yanomami territory estimated that every 1% increase in annual mining area was associated with a 24% increase in monthly malaria cases, with an estimated 102,870 excess cases from 2018–2023 associated with increased mining activity. This is precisely where satellite intelligence addresses a surveillance gap: illegal operations are frequently absent from administrative mining databases, inaccessible to health services, and distributed across enormous cross-border territories. Researchers have specifically noted mining areas missing from official information systems in states including Amapá and Roraima. Stevia therefore functions as a spatiotemporal anomaly and risk-layer generator, identifying previously unmapped disturbance, quantifying change through time, and linking candidate sites to downstream watersheds and ecological or epidemiological exposure zones so limited field and enforcement resources can be directed toward locations with measurable evidence of active disturbance.
Log in or sign up for Devpost to join the conversation.