Sensor Fusion-Based Machine Learning Algorithms for Meteorological Conditions Nowcasting in Port Scenarios

Modern port operations face increasing challenges from rapidly changing weather and environmental conditions, requiring accurate short-term forecasting to support safe and efficient maritime activities. This study presents a sensor fusion–based machine learning framework for real-time nowcasting of meteorological variables in port environments. Using an IoT architecture compliant with the oneM2M standard and deployed at the Port of Livorno, heterogeneous data from environmental sensors (meteorological stations, anemometers) and vessel-mounted LiDAR systems were integrated through feature-level fusion to enhance situational awareness. A comparative performance analysis of Random Forest, XGBoost, LSTM, ConvLSTM, Bayesian neural networks, and transformer models was performed on a four-month dataset (April–August 2025). Results show that LSTM achieved the lowest single-step prediction error, while Bayesian and ensemble models offered superior robustness and uncertainty quantification during cross-validation. These findings indicate that probabilistic and ensemble methods are better suited for deployment in dynamic maritime contexts. The proposed fusion-based machine learning framework demonstrates the potential for integrating real-time environmental intelligence into future Maritime Autonomous Surface Ship (MASS) operations, enabling safer maneuvering and decision-making under variable weather conditions.