Estimating the Footprint of Artisanal Mining in Africa

Abstract

Artisanal and small-scale mining (ASM) is a significant but often unregulated economic sector in sub-Saharan Africa. The lack of comprehensive monitoring of ASM activity hinders efforts to understand and address its environmental and social impacts or promote its sustainable development. We address this gap by creating the first high-resolution estimate of ASM’s spatial footprint across sub-Saharan Africa. To do so, we manually label a large compilation of ASM sites across five countries — Sierra Leone, Central African Republic, Tanzania, Zimbabwe, and the Democratic Republic of the Congo — and train machine learning models that integrate geographic features with satellite imagery to map ASM on a large spatial scale. We demonstrate that these models outperform surveyors on the ground in areas that have previously been monitored and successfully detect ASM in unmonitored regions. We develop low-cost machine learning methods for domain adaptation in order to train models that accurately extrapolate to new spatial regions, where standard machine learning approaches fail. Within our five countries, we identify $>$231,000 1 km$^2$ [+/- 2 standard errors: 170,153-297,710] grid cells containing ASM activity, more than 40 times that recorded by existing ground-based mapping efforts. We estimate that 44% [37-49%] of people reside within 1km of ASM activity and that 2% [2-3%] of protected areas and 18% [15-20%] of biodiversity hotspots overlap with ASM. Using algorithms adapted for out-of-domain prediction, we map ASM activity across 15 additional countries in sub-Saharan Africa, estimating that 3% [1-8%] of the region’s area and 9% [3-24%] of its population are impacted by ASM. Our findings suggest that ASM encroaches on a far larger share of Africa’s ecosystems and population than previously understood. Our methods for spatial generalization can be applied to many settings where predictions are needed across large regions but labeled data are sparse and unrepresentative..