Satellites to detect plastic waste on Earth

Researchers unveil in PLOS ONE a new algorithm capable of detecting plastic waste on earth.

Every year, about 8 million tons of plastic waste ends up in the oceans. Plastic pollution is wreaking havoc at sea, but also on land. In May 2020, scientists from the English University of Plymouth unveiled an algorithm for detect plastics floating on the surface of the sea by satellite using data from the Sentinel-2 satellites of the European Space Agency (ESA). Now, American and Australian scientists are revealing in the review PLOS ONE : a new computer system that uses data from these same satellites to identify plastic waste sites on earth.

The European Space Agency’s Copernicus Sentinel-2 program provides free access to Earth observation data by satellite, with a moderately high spatial resolution (10, 20 or 60 meters/pixel), a wide spectral range, and a temporal frequency of 5 days. The researchers use all of this data to track down plastic waste.

A new algorithm to track waste

At sea, the spectral signature of floating plastics can be characterized. The spectral data from Sentinel-2 is therefore sufficient to detect this waste by satellite. On land, the spectral diversity of litter and land cover requires more information.

The researchers were thus interested in neural networks, a system ofmachine learning algorithms, which revolutionize the classification of images. They thus developed a new system based on a convolutional neural network, capable of analyzing spectral, structural and temporal information from Sentinel-2 satellite data. The system thus makes it possible to identify the aggregations of plastic waste on Earth throughout Southeast Asia.

Identifying Plastic Waste on Earth in Southeast Asia

To combat plastic pollution at sea, it is necessary to better understand the hotspots of waste on land, whether they are uncontrolled or official dumps. To assess the system’s performance, the researchers first trained it at 184 known sites in Indonesia. He then detected 374 waste sites there, “more than double the number of sites reported in public records”, says the study. Then applying it to 12 Southeast Asian countries, the system identified a total of 996 potential waste aggregating waste sites. That is “nearly three times the number of publicly registered sites”, note the authors. 53% of candidate locations were subsequently validated by human evaluators as litter aggregators.

The researchers demonstrated that their new system can be used to monitor discharges over time. In addition, they showed that 19% of the waste sites they detected are within 200 meters of a watercourse. Some visibly flow into rivers that eventually reach the ocean.

The researchers plan to refine and expand their new system for monitoring waste sites globally. They hope to help inform waste management policies. The data is publicly available at Global Plastic Watch to facilitate actions against this pollution.

Leave a Comment

Your email address will not be published. Required fields are marked *