Sang-Won Lee

Ship-generated NO_x emissions are concentrated along major shipping routes and port approaches, highlighting the need for spatially explicit assessment of emission-control measures. This study developed a high-resolution framework to quantify ship NO_x emissions and evaluate Tier III scenarios in the territorial waters of the Republic of Korea. One-minute Automatic Identification System (AIS) data and vessel particulars for container ships operating in 2023 were used to estimate main-engine NO_x emissions. Consecutive AIS positions were connected into trajectory segments, and emissions were allocated to 1 km × 1 km grid cells according to segment–grid intersection length. Annual NO_x emissions were estimated at 16,297.5 t and were concentrated along major shipping routes and port approach waters. Six Tier III scenarios combined two vessel construction-year criteria, 2016 and 2021, with three application areas: Territorial Waters (TW), Air Quality Management Area (AQMA), and Emission Control Area (ECA). The 2016-based scenarios achieved greater reductions than the corresponding 2021-based scenarios, while overall reductions decreased as the regulated area narrowed from TW to AQMA and ECA. These results indicate that Tier III effectiveness depends on both the regulated vessel population and the spatial distribution of baseline emissions.

Accurately identifying fishing vessel behavior is essential for preventing collisions with merchant vessels and ensuring maritime safety. However, this task remains challenging due to diverse fishing operations, ambiguous boundaries between fishing and navigating states, and limited availability of labeled data. We aimed to develop a scalable and data-efficient framework for classifying fishing vessel behavior using unlabeled Automatic Identification System (AIS) trajectory data. Raw AIS data from fishing vessels operating along the South Korean coast were preprocessed, and Hermite interpolation was applied to generate uniform 1-s interval trajectories. Trajectories were segmented into 10-min windows, and 11 domain-based features were extracted from each trajectory, including Menger curvature, Steaming Intention, Directionality, speed statistics, and point proximity. A SimCLR-based self-supervised learning model with tabular data augmentations (Gaussian jittering and feature dropout) was employed to learn embeddings. Optimal hyperparameters (τ = 0.2, batch size N = 1,024) were identified via ablation studies. The embeddings were then classified using an AutoML framework (AutoGluon) for final behavior classification, minimizing manual intervention in model selection and hyperparameter tuning. The proposed method achieved an F1-score of 0.870 and accuracy of 0.926 on the test set. Validation across three sea areas (East, Yellow, and South Seas) yielded F1-scores up to 0.974. A 1 km × 1 km grid-based spatial density analysis further identified high-intensity fishing zones across the South Korean Exclusive Economic Zone. The framework provides an effective and scalable solution for fishing vessel behavior classification and maritime traffic management.

Coastal ferry has the main purpose to transport cargo such as vehicles and trailers, as well as to transport passengers. For safe cargo transportation, the cargo securing safety of coastal ferry must be ensured to prevent the maritime accident. However, the standards for the cargo securing safety are not properly presented as the rules for coastal ferry in accordance with actual operations. Therefore, coastal ferry is exposed to the risk of accidents due to not sufficient standards for the cargo securing safety. In this study, the accident of coastal ferry that occurred in Korea and Japan are analyzed from the perspective of cargo securing safety. First, the wave conditions at the time of the accident were reproduced using a wave simulation model. The reproduced wave spectrum was applied to the ship motion models to estimate rolling and pitching of the coastal ferry. Finally, the acceleration of the trailer loaded on the ferry was calculated based on the ship motions. The simulated results were compared and evaluated with the standard value of IMO's Cargo Securing Manual. As a result, the exact accident time was predicted, and the wave height and ship motions were additionally presented as evaluation criteria for cargo securing safety. The results of this study are expected to improve the safety standards of coastal ferry and could reduce the maritime accident cause by the insufficient cargo securing.