Towards Risk-Informed Evaluation of Baltic Sea Winter Navigability Using Machine Learning

This paper addresses the critical challenges of winter navigation in the Baltic Sea, a region known for its complex and rapidly changing ice conditions. These challenges increase the risk of maritime accidents, operational delays, and safety hazards for various vessels. To mitigate these risks, the paper presents novel tool for navigability assessment through route planning. These are based on the Winter Navigation Risk Index (WiNRI), an innovative data-driven model that combines a rich dataset of 329 historical accident reports with expert insights and machine learning techniques. WiNRI assesses navigability by evaluating multiple factors, such as vessel type, ice conditions, and environmental parameters. Using a modified Dijkstra algorithm, the safe route predictions are generated and hazardous areas through dynamic risk maps highlighted. These visual tools enable real-time assessment and better decision-making for ship operators, icebreakers, and maritime authorities, ultimately enhancing maritime safety in the Baltic Sea.

Keywords: ice routing, risk index, WiNRI, winter navigation, decision support systems, machine learning.