Russian Scientists Use Neural Networks to Map the Arctic Tundra

AI-powered image analysis has drastically accelerated the study of Siberia’s fragile ecosystems, revealing rapid shrub expansion with far-reaching climate consequences

Russian researchers from the Institute of Ecology and Evolution of the Russian Academy of Sciences and the Higher School of Economics, working with international partners, have developed a neural network–based approach to studying the Siberian tundra.

The system processes high-resolution satellite images to track the spread of alder shrubs across Russia’s Arctic. In a recent study, the algorithm analyzed nearly a million landscape fragments from the past 10 to 15 years, covering three key sites in the Siberian subarctic. According to senior researcher Ksenia Ermokhina, the neural network automatically classifies land into four categories, ranging from open spaces to dense shrub thickets.

The findings show a dramatic shift in tundra landscapes. Shrub expansion rates vary from 2.4 to 26.1 percent per decade, depending on local conditions. Scientists warn that these changes could have major implications for the Arctic’s climate system, as vegetation growth alters carbon balance, permafrost dynamics, and regional ecosystems.