AI Identifies Arctic Ocean Plastics

Researchers have unveiled Microscan, the most extensive open database of infrared spectra for microplastics found in the Eastern Arctic waters. This groundbreaking resource allows artificial intelligence to distinguish between polyethylene and polypropylene with over 96% accuracy, directly within the marine environment. It provides environmentalists with a vital instrument for the precise enumeration of plastic debris.

The database contains spectral information from 2,010 actual microplastic particles, ranging from 0.5 to 5 millimeters in size. Initially, it includes spectral data for polyethylene, polypropylene, and polystyrene, the most prevalent synthetic polymers. The foundational material for Microscan was gathered from microplastic particles collected from the surface waters of the White, Barents, Kara, East Siberian, and Laptev Seas between 2020 and 2021.

Microplastics are minute fragments of conventional plastic that persist in ecosystems for extended periods. Their presence in water results from the breakdown of larger plastic items, tire wear, laundry, and discarded fishing gear. Different polymer densities cause them to behave uniquely in water; for instance, polyethylene floats, while other types may submerge.

Previously, scientists faced challenges identifying aged, environmentally altered microplastics and lacked comprehensive spectral databases for training machine learning models. Microscan addresses these issues, serving as a universal training tool that significantly accelerates identification processes during expeditions by enabling AI to learn from real, naturally occurring particles without chemical purification.