Enhancing night-time sea fog detection with a multi-feature approach optimized by ERA5 meteorological constraints

Frequent advection-cooling sea fog events in the Yellow-Bohai Sea pose a considerable threat to maritime operations and navigational safety, with risks being particularly pronounced during night-time. The objective of this study is to develop a meteorology-remote sensing synergistic model that enhances night-time sea fog detection. We first established a comprehensive sea fog sample dataset from Himawari-8 imagery, utilizing CALIOP vertical profiles as objective anchor points. Based on textural and spectral analyses, three distinct satellite-derived probabilistic indices were constructed and integrated into an initial multi-feature detection model (yielding a mean Probability of Detection of 93.6%, False Alarm Rate of 11.3%, and Critical Success Index of 83.5%). To address the high False Alarm Rate caused by confusing fog with low-level clouds, the core novelty of this study lies in utilizing ERA5 meteorological reanalysis data (relative humidity, wind speed, air-sea temperature difference, sea surface temperature, and temperature inversion) as active physical constraints integrated via a decision tree. Under marginal threshold conditions, the optimized model demonstrates robust performance, yielding a mean POD of 92.1%, a substantially reduced mean FAR of 6.2%, and an improved mean CSI of 86.9%. These results demonstrate that incorporating boundary-layer meteorological constraints effectively suppresses false alarms in mixed cloud-fog regions, offering a reliable approach for regional sea fog monitoring.