Machine Learning Helps Journalists Map Illegal Amazon Mining
Satellite imagery combined with machine learning is enabling journalists to identify thousands of illegal gold mines across the Amazon Basin. Venezuelan journalist Joseph Poliszuk used the technology to map 3,718 mining sites in his country, some located within protected indigenous lands and national parks.
Poliszuk left Venezuela in 2018 after exposing government corruption. When illegal gold mining surged during the COVID-19 pandemic, he wanted to document the expansion, but manually reviewing satellite images covering 123 million acres of rainforest was impractical. Working with the Pulitzer Center and the nonprofit Earth Genome, he trained a machine learning model to recognize mining pits and airstrips carved into the forest.
From Detection to Enforcement
In January 2022, Poliszuk published his findings in El País under the series "Corredor Furtivo." By cross-referencing the mapped mines with crime data, he identified which operations were run by Venezuelan syndicates, Colombian guerrilla groups, or Brazilian miners. The Venezuelan military announced bombing raids on illegal airstrips the week after one article appeared.
Brazilian journalist Hyury Potter developed a parallel project the same year, identifying hundreds of previously unreported airstrips in the Brazilian Amazon. The New York Times later published its own investigation using Potter's analysis.
Expanding to Nine Countries and Beyond
The Pulitzer Center, Earth Genome, and Amazon Conservation launched the Amazon Mining Watch platform in 2022. It now covers all nine countries in the Amazon Basin using machine learning to detect mining activity.
Earth Genome is moving toward larger foundation models trained on satellite imagery, radar, land cover data, and elevation data. According to physicist and co-founder Edward Boyda, these models can be adapted with only a few examples to recognize different objects on Earth's surface.
A similar platform for Africa launches publicly in July. Africa Mining Watch will map mining activity across the tropical belt, including the Congo Basin. Twenty-five African journalists tested the platform during a virtual mapathon on Earth Day in April.
Tools for Investigative Work
Earth Genome is developing Earth Index, a tool that lets journalists and researchers select a geographic area and mark examples of what they want to identify. The platform then locates similar sites in the region.
In beta testing, Earth Index has been used to investigate illegal logging in Albania, commercial flower farms in Uganda, and palm oil production in Brazil.
For writers learning to work with satellite data and machine learning outputs, AI Data Analysis Courses and AI Research Courses can build skills in analyzing complex datasets for investigative reporting.
Your membership also unlocks:
