Satellites and AI Team Up to Track Plastic Waste in Rivers

Scientists have developed a new method using satellite imagery and artificial intelligence to detect “hotspots,” or areas where plastic accumulates in rivers, from space. The breakthrough is expected to support global efforts to monitor and reduce plastic pollution in freshwater environments. The method has been tested on three rivers across three continents, including the Citarum River in Indonesia, the Motagua River in Guatemala, and the Odaw River in Ghana, showing highly promising accuracy under a wide range of environmental conditions.

The study, published online in December 2025 in the journal iScience, introduces a semi-automated workflow that combines remote sensing technology with machine learning to map plastic accumulation in rivers at large scales. Until now, such mapping has largely depended on labor-intensive field surveys that are limited in coverage and difficult to replicate globally.

In the research, high-resolution PlanetScope satellite imagery was used to define training areas, known as regions of interest. These areas were then transferred to multispectral Sentinel-2 imagery to train a Random Forest classification model operating on the Google Earth Engine platform. Using this approach, the researchers achieved detection accuracy of up to 99.5 percent when the model was applied within the same river system. When tested across different rivers, the model reached an F1 score of about 79 percent, outperforming results reported in several earlier studies.

Rivers as Plastic Pathways

“Rivers are the arteries of the landscape when it comes to plastic transport, carrying waste from inland sources toward the ocean,” explain Tim H.M. van Emmerik, one of the study’s lead researchers in the report. By combining satellite data with machine learning, he explained, scientists can now identify where plastics accumulate in river systems on a global scale, something that has long been difficult using conventional field-based methods alone.

Researchers emphasized that rivers play a critical role in moving plastic waste from land to sea. Once plastics enter marine environments, they break down into smaller fragments that threaten marine life, coastal ecosystems, and potentially human health through the food chain.

The ability to map river plastic hotspots is expected to provide valuable guidance for policymakers, local governments, and environmental organizations. With more precise information on where plastic accumulates, authorities can design targeted interventions, such as installing waste capture systems, improving local waste management infrastructure, or prioritizing cleanup operations in the most affected river sections.

The method relies on the complementary strengths of two satellite systems. PlanetScope imagery offers the fine spatial detail needed to train detection algorithms, while Sentinel-2 imagery provides broader coverage, allowing the trained model to be applied consistently over large regions. All data processing was carried out using Google Earth Engine, making the workflow accessible and reproducible for researchers worldwide.

Open Access for Global Monitoring

Ámbar Pérez-García, another member of the research team, highlighted the importance of making the tool openly accessible. According to her, the Google Earth Engine–based application allows users anywhere to monitor river plastic hotspots, significantly expanding global observation capacity without the high costs and logistical challenges associated with manual data collection.

The scientists also pointed out that one of the main obstacles in addressing plastic pollution is the lack of reliable data on where plastics actually accumulate along river systems. Traditional field surveys cover limited areas and require substantial time and resources. In contrast, satellite-based remote sensing enables faster, standardized monitoring across wide geographic areas.

Environmental experts not directly involved in the study have welcomed the findings, describing the ability to detect plastic concentrations in rivers from space as a major step forward. They note that such technology could help identify priority areas for intervention and track changes over time, particularly when combined with on-the-ground observations.

One of the test sites, the Citarum River, illustrates how the method can be applied in complex tropical environments. Long regarded as one of Indonesia’s most polluted rivers, the Citarum has been the focus of numerous cleanup initiatives and scientific studies. More accurate satellite-based data could help authorities and non-governmental organizations optimize waste management strategies and river restoration programs.

Looking ahead, the researchers plan to further refine the model to detect smaller plastic items and to improve performance under varying weather conditions and water characteristics. While further development is still needed, the technology is already seen as a promising tool to support global efforts to reduce plastic pollution in rivers and oceans, while fostering cross-border and interdisciplinary cooperation in satellite-based environmental monitoring. (Sulung Prasetyo)