Comparative Analysis of Ice Datasets to Inform Decision Support Models for Winter Navigation

Abstract The Baltic Sea can be covered by ice during the winter season, with up to 40% of its area affected, significantly impacting ship route planning and traffic management. Accurate ice information is critical for safe and efficient decision-making for winter navigation. For example, severe ice conditions may require many merchant vessels to seek icebreaker assistance to navigate safely and avoid hull damage or becoming trapped. With the emergence of extensive maritime data and advanced machine learning models, there is a growing focus on developing data-driven decision support tools for winter navigation. The effectiveness of these tools heavily depends on the ice data they utilize. Ice data sources can be broadly classified into forecast models such as HELMI and NEMO, and satellite-derived data like IceAnalysis, BALFI, and IceChart. While each of these datasets has its own strengths and weaknesses, a comprehensive assessment comparing these sources and evaluating their suitability for various decision support tools in winter navigation is lacking. This paper aims to fit this gap by comparing five ice datasets and identifying the contexts in which each is most valuable for decision-makers, using both qualitative metrics (data acquisition complexity, resolution, and comprehensiveness) and quantitative metrics (ice concentration, thickness distributions, and ice extent). Based on this comparison, IceChart and BALFI are recommended for localized navigation decisions due to their similar patterns, high resolution, detailed metadata, and uncertainty estimations. Among the forecast models, HELMI currently stands out for its extensive set of variables, NEMO also has the potential for dynamic use cases, provided more variables are available in the product.