Automatic Fog Detection Algorithm with Applicatuion of Dynamic Thresholds

One of keys to get fog information from satellite imagery is using emissivity difference between infrared (IR) and shortwave-infrared (SWIR), and it can be used in a nighttime algorithm. However, the difference varies according to time because SWIR channel measures both reflectivity and emissivity during daytime. Therefore, the measuring property makes it difficult to estimate fog area continuously after sunrise. Moreover, fog surface temperature is not that differ from the surface since the fog height is very low. In those reasons, current fog detection algorithms using satellite data have a lot of limitations and discontinuity. In this study, dynamic threshold values were applied in order to derive fog area continuously and automatically. First of all, SWIR-IR as the function of the solar zenith angle is basically executed after RTM simulation. Then, some corrections with IR differences varied with IR brightness temperature and clear-sky visible reflectance produced in 15 days are utilized for reducing contaminations from high cloud and surface. Finally, the fog characteristics like temporal consistency and spatial homogeneous are considered in this algorithm. The results show that this algorithm presents fog area similar to that we see in satellite image, and these agree comparatively well with GTS stations reporting fog at that time. Also, we could clearly see the development and declination of fog area from the result images. However, the performance is very dependent on the fog types different drop size, intensity and horizontal size. In the future, more case studies for various fog types are needed for general use, and this algorithm will be finally added for the fog detection module in COMS meteorological data processing system.