Fog is defined as a phenomena with horizontal visibility less than 1 km by water droplets and supercooled water in the atmosphere. In Korea, fog occurs most frequently at 00:00 ~ 08:00 KST, and it is dissipated at 07:00 ~ 10:00 KST. At daytime, the fog has been detected using the reflectance of the visible channel and brightness temperature of infrared channel. And the dual channel difference (DCD) method based on the difference in emissivity between the shortwave infrared channel and the thermal infrared channel has been mainly performed at nighttime. However, it is difficult to detect fog at dawn because the reflectance of the visible channel and DCD are significantly affected by the temporal change of solar zenith angle (SZA). Therefore, I tried to optimize the definition of the dawning period and improve the fog detection algorithm of dawn developed by KNU(Kongju National University) based on Himawari-8/AHI (Advanced Himawari Imager) data. The channels used in this study are 3.9, 8.6, 10.4, 11.2 and 12.3 μm.
The definition of dawning period was optimized by 79? ~ 87? of SZA through the sensitivity test according to the various SZA. Four fog detection methods, Ngt_Fog, Ngt_Dwn_ST, Ngt_Dwn_DT, and Dwn_Fog_DT were designed and evaluated for the 12 training and 10 validation cases. Ngt_Fog uses the night fog detection results without further detection. Ngt_Dwn_ST detect dawn fog through the sum of night fog detection results and dawn time fog detection results with a single threshold. Ngt_Dwn_DT is the same as Ngt_Dwn_ST except for the dynamic threshold. And Dwn_Fog_DT detects fog with a dawn time algorithm without using the night fog detection results. These algorithms were validated using the 10 minute average data of 280 visibility data operated by the Korea Meteorological Administration. The initial threshold values set by frequency analysis for each evaluation factor, which were optimized through 12 training cases. The qualitative and quantitative evaluation of the dawn fog detection were performed by geographic location and fog type. The detection skill of NGT_DWN_DT is best with the average POD (probability of detection) and FAR (false alarm ratio) for the validation cases are 0.92 and 0.48, respectively, and Bias is 1.78. But the skill of DWN_FOG_DT is worst with the average POD(0.75) and FAR(0.62), and Bias(2.00). And the POD tended to be higher on land than on coast. Therefore, it is necessary to improve the detection level of coastal fog through the more study.