MAAP #235: AI-powered detection of Amazon gold mining deforestation in real-time
November 1, 2025
As gold prices continue to skyrocket and shatter historical records, small-scale, but widespread, gold mining activity also continues to be one of the major deforestation drivers across the Amazon.
It often targets remote areas, and in many cases, we presume that this mining is illegal based on its location within conservation areas (such as protected areas and Indigenous territories) and outside mining concessions.
Given the vastness of the Amazon, however, it has been a challenge to accurately and regularly monitor mining deforestation, in order to better inform related policies in a timely manner.
In a recent report (MAAP #226) we presented the AI-based dashboard known as Amazon Mining Watch. This online tool analyzes satellite imagery archives to estimate annual mining deforestation footprints across the entire Amazon, from 2018 to 2024.
In a major advance, Amazon Mining Watch (AMW) just released its first real-time quarterly update, for quarters 2 and 3 of 2025 (April-September 2025).
Here, we present our first quarterly update report based on this fresh data.
The Base Map shows the locations of confirmed recent (Q3 of 2025) mining deforestation across the Amazon, in relation to the cumulative mining impact area. Note that we have confirmed cases in all eight countries plus French Guiana.
Many of the confirmed cases are in protected areas and Indigenous territories, and therefore likely to represent illegal mining. Overall, the data indicates that 96 protected areas and 151 Indigenous territories were impacted, although we have not yet confirmed all of these cases.
Below, we present a series of satellite images confirming the recent expansion of mining deforestation.
Base Map. Amazon gold mining deforestation, highlighting recent confirmed cases. Data: AMW, ACA/MAAP.
Confirmed Cases of Recent Gold Mining Deforestation in the Amazon
In each set of panels displayed below, we show high-resolution satellite imagery before (left panel) vs after (right panel) of the recent gold mining deforestation. The red polygons indicate the pixels highlighted by the AI model as a new quarter 3 mining deforestation event. Yellow polygons are previously detected pixels.
Ecuador
Figure A shows the recent mining deforestation outside of the Cofán Bermejo Ecological Reserve in the northern Ecuadorian Amazon. This area has experienced increasing mining impact in recent years, as further described in MAAP #227. Note that this mining activity is within the Kuri Yaku Indigenous territory.
Figure B shows the recent mining deforestation in the Punino zone in the northern Ecuadorian Amazon, located in the area south of Sumaco-Napo Galeras National Park and north of the Punino River. This area has experienced increasing mining impact in recent years, as further described in MAAP #219.
Figure C shows the recent mining deforestation around the Yacuambi River in the southern Ecuadorian Amazon.
Peru
Figure D shows the recent mining deforestation in the Puerto Inca province in the central Peruvian Amazon. This area has experienced increasing mining impact in recent years, as further described in MAAP #233 (see Central Peru section).
Figure E shows the recent mining deforestation in the San José de Karene Indigenous Territory in the southern Peruvian Amazon. As detailed in MAAP #208, much of the illegal mining deforestation in the southern Peruvian Amazon is occurring within the territory of Native Communities.
Brazil
Figure F shows the recent mining deforestation in the Tapajós Environmental Protection Area in the central Brazilian Amazon. As noted in MAAP #226, Tapajós EPA is one of the Amazon’s top ten most impacted protected areas (in terms of accumulated footprint), including high levels of new mining deforestation in 2024.
Figure G shows the recent mining deforestation in Jamanxim National Forest in the central Brazilian Amazon. As noted in MAAP #226, Jamanxim NF is one of the Amazon’s top ten most impacted protected areas (in terms of accumulated footprint).
Figure H shows the recent mining deforestation in Rio Novo National Park in the central Brazilian Amazon. As noted in MAAP #226, Rio Novo NP is one of the Amazon’s top ten most impacted protected areas (in terms of accumulated footprint), including high levels of new mining deforestation in 2024.
Figures I & J show the recent mining deforestation in Amapá State Forest in the northeastern Brazilian Amazon. This marks the first time this area has been highlighted by MAAP.
Figure K shows the recent mining deforestation in the Grão-Pará Ecological Station in the northeast Brazilian Amazon, along the border with Guyana. This marks the first time this area has been highlighted by MAAP.
Bolivia
Figure L shows the recent mining deforestation in the Apolobamba National Integrated Management Natural Area in the western Bolivian Amazon. This marks the first time this area has been highlighted by MAAP.
Venezuela
Figure M shows the recent mining deforestation along the Orinoco River in southern Venezuela, just across the border from northern Brazil. This marks the first time this area has been highlighted by MAAP.
Suriname
Figures N & O show recent mining deforestation in northeast Suriname. This marks the first time this specific area has been highlighted by MAAP.
French Guiana
Figure P shows the recent mining deforestation in the Parc Amazonien de Guyane National Park in French Guiana. Importantly, in MAAP #225, we reported that this park is a significant carbon sink and has among the highest levels of carbon relative to other protected areas across the Amazon.
Guyana
Figure Q shows the recent mining deforestation in northwestern Guyana, near the border vith Venezuela. This marks the first time this specific area has been highlighted by MAAP.
Colombia
Finally, Figure R shows the recent mining deforestation in the southwest Colombian Amazon. This marks the first time this specific area has been highlighted by MAAP.
Notes/Methodology
Mapped out Q3 differential data. Zoomed in on areas of concentrated mining forest loss alerts and investigated with high resolution imagery (Planet) to confirm or deny recent gold mining deforestation.
Note that in this report, we focus on mining activity that causes deforestation. The vast majority is artisanal or small-scale gold mining, but other mining activities have also been detected, such as iron, aluminum, and nickel mines in Brazil and Colombia. Additional critical gold mining areas in rivers that are not yet causing deforestation (such as in northern Peru, southeast Colombia, and northwest Brazil; see MAAP #197), are not included in this report. This information is not yet displayed in Amazon Mining Watch, but future updates will include river-based mining hotspots.
Our data source for protected areas and Indigenous territories is from RAISG (Amazon Network of Georeferenced Socio-Environmental Information), a consortium of civil society organizations in the Amazon countries. This source (accessed in December 2024) contains spatial data for 5,943 protected areas and Indigenous territories, covering 414.9 million hectares across the Amazon.
Amazon Mining Watch (AMW) is a partnership between Earth Genome, the Pulitzer Center’s Rainforest Investigations Network, and Amazon Conservation. The algorithm is based on 10-meter resolution imagery from the European Space Agency’s Sentinel-2 satellite and produces 480-meter resolution pixelated mining deforestation alerts. Although the data is not designed for precise area measurements, it can be used to give timely estimates needed for management and conservation purposes.
The Amazon Mining Watch partnership is currently working to enhance the functionality and conservation impact of the dashboard, AMW will be a one-stop shop platform including real-time visualization of: 1) AI-based detection of mining deforestation across all nine Amazonian countries, with quarterly updates; 2) Hotspots of urgent mining cases, including river-based mining; and 3) the socio-environmental costs of illegal gold mining with the Conservation Strategy Fund (CSF) Mining Impacts Calculator.
Acknowledgments
This report was made possible by the generous support of the Gordon and Betty Moore Foundation.
