In 2018, Joseph Polisluck fled Venezuela. That same year, he was the target of a lawsuit by wealthy Maduro supporters after exposing corruption in then-President Nicolas Maduro’s government. He and several colleagues from the independent media Armando.info packed up their lives and fled the country under threat of imprisonment.
For years, Mr. Polischuk has published articles about southern Venezuela, a sparsely populated state that spans a wide swath of the Amazon and Orinoco basins. Through his field reporting, Polischuk exposed illegal gold mines, drug trafficking operations, and crimes against indigenous peoples throughout the region’s rainforests. Now in exile and working first in Colombia and then Mexico, Mr. Polischuk has been forced to rethink how he works thousands of miles away. He began experimenting with satellite surveys.
Satellite imagery has long helped investigative journalists gather information about conflict zones and track changes in landscapes in remote areas. Now, in a new wave of satellite-based investigations, reporters are relying on machine learning models to automate some of this work and scale up analysis to an unprecedented degree.
This innovation is most evident in environmental journalism. Polischuk is just one of a team of South American investigative reporters who have used geospatial data and AI-powered pattern recognition to track illegal mining, large-scale logging operations, and cattle ranching across the Amazon.
As illegal gold mining surged during the COVID-19 pandemic, Polisluck realized there was a story to be told in documenting the growth of these mines across Venezuela’s rainforest. However, manually sifting through satellite images of more than 50 million hectares of rainforest was not practical. Poliszuk wondered if a machine learning model could be trained to detect mine scars in these images, as well as adjacent airstrips cut into dense vegetation and used to transport minerals.
With financial and editorial support from the Pulitzer Center’s first Rainforest Research Network (RIN) Fellowship and technical support from the nonprofit Earth Genome, Polis Owl was able to do just that. In January 2022, he co-published the first article in the series using custom machine learning models. el pais The title is “Corredor Frutivo” [Clandestine Corridor]”
Polischuk was able to identify 3,718 gold mining sites in Venezuela’s Amazonas and Bolívar states. Some of those mines operated on protected indigenous lands and within Canaima National Park, home to Angel Falls, the world’s tallest waterfall. By matching maps identifying mining activities with criminal data from Venezuelan authorities, Polischuk was also able to determine whether a mine was run by a Venezuelan syndicate, a Colombian guerrilla group, or a Brazilian national. Garimpeiro (prospector).
The week after Poliszuk published one of his first novels in a magazine; el pais During the series, the Venezuelan military announced that it had bombed several illegal airstrips operating in the region.
“I have 20 years of reporting experience.” [illegal mining]”Thanks to this technology, we can show people the dimensions of this phenomenon. Thanks to this movement, we now know that we can track things from the air that we can’t prove on foot,” Poliszuk told me.
view from above
Even for journalists who don’t work in exile, reporting on the Amazon poses a number of accessibility issues and security risks. Polischuk said getting from Venezuela’s capital to one of the mines in Amazonas state requires a two-hour flight, six hours by car, four hours by boat and then another four-hour trek through the jungle, often through dangerous areas occupied by armed militias. These same groups often have a monopoly on oil and gas in the region, which can make fuel expensive and difficult to source.
“It’s dangerous. It’s difficult. You can’t go there like you can go from Boston to Washington or from Caracas to Maracaibo,” he said.
The same year that Polischuk pitched his project to the Pulitzer Center, Brazilian journalist Huli Potter pursued a similar investigation with a RIN Fellowship. He also used machine models from Earth Genome, which works with a number of journalists to conduct AI-based environmental and human rights research. Potter went on to publish several key findings in The Intercept Brazil, identifying hundreds of previously unreported airstrips in the Brazilian Amazon and documenting the explosion of illegal gold mining on protected indigenous lands. The New York Times published its own investigation based on analysis of satellite imagery and collaborated with the Potter and Pulitzer Centers in the reporting process.
“It was like a think tank trying to figure out how to do this work,” Polischuk said of his time at the RIN Fellowship. “It was a great time to think about new journalism, a different way of doing things.”
Based on the results of these initial studies, the Pulitzer Center decided to build a dedicated platform that uses machine learning to track mining activity across nine countries that are part of the Amazon basin. Earth Genome built the interface and contributed to the underlying geospatial detection model. Amazon Conservation, a nonprofit advocacy group, contributed to the fundraising efforts and helped develop impact metrics. In 2022, three partner organizations launched Amazon Mining Watch.
“That was the beginning. Inspired by Joseph and Fury’s work, we were able to extrapolate and cover the entire Amazon,” said Gustavo Faleiros, former director of environmental research at the Pulitzer Center.
In the early days of Amazon Mining Watch, we relied on small, task-specific machine learning models. These models were trained by Earth Genome itself and customized solely to identify gold mining sites and airfields in satellite imagery. But more recently, Earth Genome has been experimenting with more powerful geospatial-based models, models pre-trained on vast amounts of data such as radar, land cover, and elevation data, as well as satellite imagery.
Perhaps these larger models will make geospatial research more accessible to journalists, beyond just covering the Amazon and illegal mining.
“In the same way that people discovered how to unsupervised train AI models for text (a technique that has grown into large-scale language models), we’ve been doing the same thing in the geospatial data space,” said Edward Boyda, a physicist and co-founder of Earth Genome. “With little additional input from the user, and perhaps only a few examples, the model can be effectively tuned to detect a variety of objects on the Earth’s surface.”
Beyond Amazon
The Pulitzer Center and Earth Genome are currently partnering with the nonprofit Code for Africa to bring a similar platform to the continent. Earlier this month, the organizations announced the launch of Africa Mining Watch. The platform uses geospatial sensing models to track mining operations across the Tropical Gulf, a region that includes the world’s second-largest tropical rainforest, the Congo Basin. It is scheduled to be opened to the public in July.
Last week on Earth Day, 25 journalists from across Africa took part in a seven-hour virtual mapthon to experiment with the platform and test its ability to identify landmines in areas of interest.
“My hope is that African Mining Watch becomes a platform that connects not just the gold mining story, but the strategic minerals story,” Faleiros said, pointing to the cobalt, copper and coltan mines discovered across the Congo Basin.
Earth Genome is also building a unique platform that leverages these foundational models for journalists. A tool called Earth Index allows reporters, researchers, and policy makers to access a platform and select regions on a world map. Once they select an example of what they want to identify, say an artisanal gold mine in Ghana, the platform will highlight other potential gold mines in the region.
Earth Index is in its invitation-only beta stage and has been used to investigate illegal logging in Albania, commercial flower farms in Uganda, and palm oil production in Brazil. Boyda says he plans to make the Earth Index publicly available in late April.
“The idea of Earth Index is not to give people data, but to give them the tools to create their own data,” Boyda says. “People who work in a particular field will know that context better than we do. This tool will allow them to build the datasets they need.”
This story has been updated to correct Earth Genome strongholds.
