AI for Arctic Sea Ice Classification and Navigation

ESA's Earth Observation initiative, led by the Technical University of Denmark, applies AI Convolutional Neural Networks to automate Arctic sea ice concentration mapping from Sentinel-1 radar imagery. The project addresses a critical gap: despite advances in digital technology, Arctic sea ice maps used for maritime navigation are still produced manually by analysts interpreting radar images—a slow process with only 20% map release frequency and uncertainties reaching 20%, with discrepancies up to 60% between analysts. The AI system processes dual-polarized radar data alongside radiometer and wind speed measurements, generating automated ice concentration maps with significantly improved accuracy, resolution, release frequency, and spatial coverage. This directly improves Arctic navigation safety and supports climate change monitoring by enabling more frequent and reliable sea ice thickness assessments. The project falls under ESA's Digital Engineering strategy (Chapter 6.9) addressing Arctic operational challenges through AI-powered Earth observation.