Mapping artisanal and small-scale mines at large scale from space with deep learning.

Area of Science:

• Earth Science
• Remote Sensing
• Artificial Intelligence

Background:

• Artisanal and small-scale mining (ASM) is expanding globally, offering economic benefits but lacking regulation due to unknown locations.
• Informal or illegal ASM operations are difficult to monitor, hindering effective governance and research.
• Accurate spatial data on ASM is essential for understanding its impact and developing policy.

Purpose of the Study:

• To develop and validate a method for mapping artisanal and small-scale mining (ASM) locations using satellite imagery.
• To create the first large-scale, comprehensive dataset of ASM activity in Sub-Tropical West Africa.
• To assess the performance of a convolutional neural network for detecting surface mining operations.

Main Methods:

• Utilized a convolutional neural network (CNN) for image segmentation to detect surface mining from satellite data.
• Developed a novel dataset covering 1.75 million km² across 13 countries in Sub-Tropical West Africa.
• Achieved 70% precision and 42% recall in detecting diverse ASM activities ranging from 0.1 ha to 2,000 ha.

Main Results:

• Successfully mapped a significant number of artisanal and small-scale mining (ASM) sites across a vast region.
• Generated a unique dataset quantifying ASM activity, crucial for policy and research.
• Demonstrated the effectiveness of AI-driven satellite image analysis for monitoring informal mining.

Conclusions:

• The developed method provides a robust and scalable approach for mapping artisanal and small-scale mining (ASM) globally.
• The comprehensive dataset facilitates improved research and policy development for the ASM sector.
• This approach can be adapted to other regions, enhancing global monitoring of mining activities.