2023 Shortlist

Airstrips of Destruction in Brazil and Venezuela by the Pulitzer Center´s Rainforest Investigations Network (RIN) –

Entry type: Portfolio

Country/area: United States

Publishing organisation: Armando Info, El País, Earthrise Media, The Intercept Brasil, The New York Times, Pulitzer Center´s Rainforest Investigations Network

Organisation size: Big

Cover letter:

I write this letter on behalf of an alliance of journalists and outlets that collaborated to reveal the criminal infrastructure fueling the destruction of the Amazon rainforest. Their stories were published by Armando Info, El País, The New York Times and The Intercept Brasil – a collaboration coordinated by the Pulitzer Center’s Rainforest Investigations Network (RIN).

The RIN was launched in 2020 with the goal of promoting collaborative investigations to unveil the causes of environmental destruction in the Amazon, Congo Basin and Southeast Asia. As the editor of RIN, I have worked with some of the most talented investigative journalists of our time and have been inspired by their embrace of teamwork and collaboration.

The project I am submitting represents an important breakthrough because of its innovative methodology to prove a crucial, underreported issue: the connection between criminality and deforestation in the Amazon. Brazil and Venezuela share a sad history of land invasions, violence against Indigenous people, and pollution. A great part of this is caused by the rush for minerals, especially gold.

The investigation broke new ground as the first journalistic endeavor to use artificial intelligence and satellite imagery to map mines and runways in a tropical rainforest. These airstrips are the most blatant element of the growing illegal economy in the Venezuelan and the Brazilian Amazon. Driven by the high profits from mining and drug trafficking, clandestine groups, such as militias and guerrillas, conquer territories by tearing up the forest to allow them to advance into more remote regions. Many of these areas are Indigenous territories and represent the last barriers to the total destruction of the planet’s largest tropical rainforest.

Bringing together journalists from local and global media organizations with civic technologists required diligence and sensitivity given the different languages, cultures, and audiences. The dedicated team from the Rainforest Investigations Network had to manage the expectations, workflows, and dynamics among these partners through countless meetings, one-on-one conversations, and a customized collaborative platform called Confluence.

Through a collaboration with Earthrise Media, a non-governmental organization that uses geospatial data in journalistic investigations, the team developed a machine-learning algorithm that analyzes satellite images to identify the covert runways. In addition to publishing major stories in The New York Times, El País, Armando Info and The Intercept, we made the geospatial data of gold mines and runways detected by the algorithm publicly available and downloadable, to assist other journalists and investigators in their work.

By publishing methodology and data used, we believe the project will have an enduring resonance beyond this first batch of compelling stories. In partnership with Earthrise Media, the Rainforest Investigations Network has transformed the monitoring of airstrips and mining areas in the Amazon into an interactive tool that other journalists as well as civil society can use.

We published a platform called Amazon Mining Watch (AMW), which will continue to update the data and inspire new investigations. AMW has since been featured by journalism publications including the Global Investigative Journalism Network (GIJN), the Online Journalism Blog, and the Reuters Institute for the Study of Journalism.

The reports drew immediate global attention. They were republished and quoted in Venezuelan and Brazilian media. Prominent journalists and media organizations in the United States and Europe also highlighted them. An example of the investigation’s lasting impact is the fact that, with the transition of government after Bolsonaro lost reelection in 2022, legislators in the Brazilian parliament are using the number of airstrips as an indicator of the urgency to reverse the destruction in the Amazon. The new minister of the environment, Marina Silva, cited the investigation’s findings in her speeches and positions.

Description of portfolio:

The principal authors AIrstrips of Destruction are Hyury Potter, for The Intercept Brasil, Manuela Andreoni, Blacki Migliozzi, Pablo Robles, Victor Moriyama and Denise Lu, all for The New York Times, and Joseph Poliszuk, María Antonieta Segovia and María de los Ángeles Ramírez, from Armando Info/El Pais. The project was done with the coordinatination of the Pulitzer Center’s Rainforest Investigations Network editorial team: Environmental Investigations Editor: Gustavo Faleiros, Executive Editor: Marina Waker Guevara, Data Editor: Kuek Ser Kuang Keng and Reseach Editor: Jelter Meers. Partnership with Earthrise Media developers Caleb Kruse and Edward Boyda.

The products of this collaboration are
– 1 multimedia (front page) article published at The New York Times about airstrips in Brazil
– A series of six articles called “Corredor Furtivo” with multimedia elements published simultaneously at El País and Armando Info, about the relation between the airstrips and illegal armed groups in Venezuela
– 1 featured article at the Intercept Brasil about the relation between the airstrips and illegal mining in Brazil
– 1 documentary launched by The Intercept about the life and work of the pilots working in the illegal runaways.
– The website Amazon Mining Watch and GitHub page managed by the Earthrise Media with all the documentation and code for the geospatial analysis

Since its inception, the investigation had the ambition of generating strong stories and groundbreaking data analysis that would serve as a blueprint for other journalists and civil society actors investigating criminal activities in the Amazon.

From their previous reporting, Potter, Poliszuk, Ramírez and Andreoni knew that unregulated landing strips were key to the criminal infrastructure that facilitates gold mining and other illegal activities. But until now no one had been able to systematically research, identify, and map the gold mines and airstrips, in part because of the impenetrable geography of the Amazon.

As Rainforest Investigations Fellows, the three journalists decided to join forces to combine advanced data analysis and technology with traditional shoe-leather reporting to reveal for the first time the true scope of the problem. The team partnered with Earthrise Media to build a machine-learning algorithm and the public database that underpins the project.

The Amazon’s illegal economy is an extremely dangerous beat. To complement and verify the computational findings, our reporters did extensive field reporting and had to take many safety precautions. They talked to the people affected by illegal mining and to the pilots flying to and from the mines. They visited the different parts of the illegal mining supply chain, from dig sites to clandestine gold shops.

Different parts of the analysis included painstaking manual work to improve the machine learning algorithm and verify the results. Lining up satellite imagery with a computer program also created coding problems and technical delays.

The development of the algorithm consisted of manually selecting information from historical satellite imagery and then using this baseline for programs to find similar patterns in recent images. This machine-learning process allowed the team to scan thousands of high-resolution Sentinel 2 satellite images dating back to 2016, covering an enormous area in the Amazon.

Such a statistical and computational model is known as an artificial neural network. It was trained to look at patches the size of 44 by 44 10-meter pixels, an equivalent of 440m by 440m on the ground. The AI program found roughly 1269 airstrips in the Brazilian Amazon and 47 in Venezuelan Amazon . Many were found by cross-referencing information in Open Street Map, a community-built global map of roads and other infrastructure.

Project links: