Russian Scientists Are Rewriting the Rules of Arctic Environmental Monitoring
Although the Arctic is still widely seen as one of the most pristine places on the planet, the region is increasingly feeling the effects of human activity. Microplastics pose a major threat to Arctic ecosystems. Researchers writing in the scientific journal FMS report that 90% of birds are carrying these harmful particles.
Novgorod State University, together with the Shirshov Institute of Oceanology of the Russian Academy of Sciences, has unveiled MICROSCAN, a digital project designed to take the analysis of Arctic waters to a new level.
Arctic Under AI Surveillance
The developers have created the world’s largest open database of infrared spectra from real microplastic particles collected in the waters of the eastern Arctic. It contains 2,010 particles ranging from 0.5 to 5 millimeters in size, collected from the White, Barents, Kara, East Siberian and Laptev seas.
The main weakness of previous recognition systems was that AI models were trained on “sterile” industrial plastic. In the real ocean, however, plastic changes. Saltwater, intense ultraviolet radiation and microorganisms alter its properties. Identifying this kind of “aged” debris using standard reference libraries is practically impossible. The MICROSCAN database brings together fragments collected specifically from the natural environment.
During the experiment, each sample was tested three times: through conventional spectral matching, expert assessment by chemists and a neural network model. The CNN1D convolutional neural network can identify up to 120 types of polymer materials with an accuracy of 96–97%, approaching absolute accuracy. Scientists can now hand over the routine but important task of identifying pollutants to digital intelligence, freeing up their time for more complex analytical work.
Standardized Identification Rules
More than 25 million tons of plastic are estimated to be floating in the world’s oceans. According to the United Nations, by 2050, the total mass of plastic is expected to exceed the weight of all fish. Each year, pollution kills roughly one million seabirds and 100,000 marine mammals.
In 2021, the Shirshov Institute of Oceanology published the landmark monograph Microplastics in the Marine Environment, bringing together established research methods. In 2022, scientists at Tomsk State University studied how waste enters the Arctic through Siberia’s Yenisei and Ob rivers, considered the main “arteries” carrying human-generated waste into the Arctic. By 2023, researchers were also addressing the need to standardize methods so that data from across the Eurasian Arctic could be compared accurately.
In 2024, the global scientific community, including experts from the International Atomic Energy Agency, reinforced the trend toward using AI to analyze complex spectra. The Novgorod researchers’ work fits into that broader movement while accounting for the Arctic’s unique conditions. Previously, microplastic research methods varied, making it difficult to combine results into a coherent picture. MICROSCAN offers a common language for identifying plastic debris. That, in turn, should speed up the analysis of samples brought back from field expeditions.
A Universal Training Resource
Last year, the Far Eastern Federal University and Yandex’s School of Data Analysis unveiled a neural network that detects water pollution in hard-to-reach regions. Using satellite imagery and drone footage, the model marks the coordinates of waste on a map, identifies its composition and weight, and can even predict where it will move based on currents and weather conditions. The system was used to organize a cleanup along the Sea of Okhotsk coast of the South Kamchatka Sanctuary. About 3 tons of marine debris were collected there, 85% of it plastic, including containers, fragments of fishing gear and other household waste.
MICROSCAN is designed to serve as a “universal training resource for machine learning” for similar models. Researchers in other countries can use the open Russian dataset to train and validate their own models.
In the coming years, the developers may expand the database to cover additional polymers, integrate neural networks directly into expedition laboratories and incorporate the technology into a national environmental monitoring system for the waters along the Northern Sea Route. That integrated system is now taking shape, and MICROSCAN is expected to play a central role in it.
