Detection of Arctic sea fog using FY-3D/MERSI-II multispectral imagery during the summer

Arctic sea ice melting enhances surface evaporation over open waters, thereby intensifying sea fog occurrence and posing serious risks to navigation safety. However, accurate satellite-based detection of Arctic sea fog remains a persistent challenge due to weak spectral contrasts among fog, low cloud, sea ice, and snow, especially under low solar elevation in high-latitude regions. To address these issues, we propose the Surface-Adaptive Multispectral Threshold Segmentation (SA-MSTS) method tailored to the Medium-Resolution Spectral Imager-II (MERSI-II) aboard the FY-3D satellite. SA-MSTS dynamically adjusts spectral thresholds based on the underlying surface type (sea ice or open water), enabling robust fog discrimination under diverse backgrounds. High-quality ground-truth labels are derived from the Aerosol and Carbon Detection Lidar (ACDL) onboard the atmospheric environment monitoring satellite, DQ-1. Through systematic spectral analysis across all MERSI-II channels, nine critical bands (bands 1, 3, 5, 6, 7, 16, 20, 24, 25) and four optimized spectral indices for sea fog detection are identified. The method is validated using in-situ shipborne meteorological observations collected by the research vessel Xuelong 2 during Arctic expeditions. Validation shows that SA-MSTS achieves a probability of detection (POD) of 87.5% over open water and 88.6% over sea ice, with corresponding false alarm ratios (FAR) of 23.7% and 29.5%, and critical success indices (CSI) of 68.8% and 64.7%, respectively. These results demonstrate that SA-MSTS can achieve high accuracy in Arctic sea fog detection across heterogeneous Arctic surface conditions and highlight the potential of MERSI-II data for polar atmospheric applications.