SPATIO-TEMPORAL VARIATION OF SEA FOG IN THE YELLOW SEA AND COMPARISON WITH IN-SITU DATA

Abstract

This study introduces monthly and yearly changes of sea fog distribution during the mission period of Geostationary Ocean Color Imager (GOCI) and investigates spatial-temporal variations of sea fog in the Yellow Sea. Next step is to find a relation between satellite-based sea fog distribution and in-situ data of visibility by KMA and KIOST. We have developed several algorithms including Convolution Neural Network Transfer Learning (CNN-TL) model. Training data of CNN-TL showed up to 96.3% accuracy in matching, both with VGG19 and ResNet50, identically. Thus, it is revealed that CNN-TL is effective for the detection of sea fog from GOCI imagery. Additionally, a semi-automatic detection method was used here to get a high accuracy. As for in-situ data of visibility, KMA data and three ocean research stations were used from 2014 to 2022. There was a big difference of sea fog occurrence season and time (diurnal range) between Socheongcho Ocean Research Stations (SORS) and near island (Baengnyeongdo). In this work, we will compare GOCI-based sea fog and in-situ data tendency to find a combining method.