New AI Platform Monitors Mining in the Amazon Rainforest
Featured
  • Facebook

21 نيسان/أبريل 2022 Author :   Pamacc reporter
Illegal mining in the Amazon

PAMACC News: Mining, one of the main causes of the degradation of rivers and forests in the Amazon, can now be monitored remotely by journalists, scientists, and other concerned citizens. Today, the Pulitzer Center in partnership with Earthrise Media launches the Amazon Mining Watch, a platform powered by an algorithm that analyzes satellite imagery to detect gold mines and other open-pit mining activities in the world's largest rainforest.

In its beta version, the platform performs 326 million analyses of high-resolution images every 4 months, covering the entire Amazon region, which encompasses nine countries or 6.7 million square kilometers. The algorithm has identified an area with characteristics of mining activity the size of 6.8 thousand square kilometers. To get an idea of the size of the environmental impact, this area is equivalent to about four times the entire city of New York.

The Amazon Mining Watch platform is a partnership between the Pulitzer Center's Rainforest Investigations Network (RIN) and Earthrise Media. The two nonprofit organizations had already collaborated to investigate the alarming criminal expansion of gold mining operations in Brazil and Venezuela and, from this experience, they decided to expand their methodology to the entire Amazon region.

The AI model recognizes mining zones using topographical features, and therefore it is not possible to determine when using the platform which mining sites are legal and which are illegal. It is also important to point out that because it is an automatic detection method using satellite images, there are false positives. The main goal of Amazon Mining Watch is to encourage  journalists, researchers, and activists to use the data as a springboard to further investigate the results, thus contributing to contextualizing the AI findings

How the algorithm was developed

Earthrise Media, an organization dedicated to supporting communicators and organizations in the use of geospatial analysis, enlisted high school students in the United States for "identification marathons" of looking at mines through satellite images of the Amazon. This human effort eventually "trained" the AI model to recognize the characteristics of an open-pit mine, making it possible for computers to distinguish between mining areas and other land uses, such as agriculture.

This statistical and computational model, known as an artificial neural network, was trained to look at patches the size of 44 by 44 pixels, an equivalent of 440m by 440m on the ground. It did this by looking at thousands of Sentinel 2 satellite images from the last four months.

In this way, every four months, Amazon Mining Watch will be able to update its analysis of the status of mining on Amazon.

The code behind the platform as well as the data generated is open to the public and can be downloaded for other uses.

Reporting from the analysis

The Rainforest Investigations Network (RIN) was created in 2020 by the Pulitzer Center to support investigative journalists in the three main rainforest regions: the Amazon, the Congo Basin, and Southeast Asia. In its two years of operation, the network has awarded 25 fellowships to reporters investigating environmental crimes, corruption, and the supply chains that drive forest destruction.

The Amazon Mining Watch platform originated from a series of collaborations with journalists seeking to expose illegal mining activity and document its impacts on the environment and Indigenous communities in Brazil and Venezuela.

One of these collaborations is the investigation “Illegal Mining Set Air Bases in the Jungle” (Las pistas illegales que bullen en la selva Venezolana)” published together by El País and ArmandoInfo in early 2022: The first story of the series Corredor Furtivo (The Stealthy Corridor)  by RIN Fellows Joseph Poliszuk and María de Los Ángeles Ramirez.

--------- --------- --------- ---------
Top
We use cookies to improve our website. By continuing to use this website, you are giving consent to cookies being used. More details…