Real-Time Forecasting of Marine Visibility with Self-Adaptive Deep Learning

Marine visibility is a crucial factor for safe navigation and operational efficiency at sea, as poor visibility can lead to costly accidents and pose significant safety risks. While deep learning models, such as artificial neural networks (ANNs), have shown effectiveness in forecasting land visibility, predicting marine visibility presents unique challenges due to its complexity and variability. In this paper, we introduce selfadaptive deep learning (SADL) models to enhance visibility prediction across varying time periods and distances from ships. Our approach employs real-time training models that continuously learn and adapt based on incoming data, enabling dynamic and comprehensive visibility forecasts. We present case studies to validate the effectiveness of the SADL models and demonstrate their accuracy across diverse locations, time frames, and scenarios. The results show the models' capability to predict visibility accurately and in a timely manner, thereby effectively improving marine safety.