Development of Fog Detection Algorithm during Nighttime Using Himawari-8/AHI Satellite and Ground Observation Data

Abstract

In this study, we developed a hybrid nighttime fog detection algorithm based on the optical and textural characteristics of fog from Himawari-8/ advanced Himawari imager (AHI) data and ground temperature data. The Dual Cannel Difference (DCD) caused by the emissivity difference between the 3.9 and 11.2 μm is the main evaluation element for fog detection. And, the local standard deviation of Brightness Temperature (BT) and difference between fog top BT and ground temperature (sea surface temperature) were used to distinguish between fog and low cloud. The thresholds and weights of the three evaluation elements were initially determined by visual inspection of fog cases and optimized through receiver operating characteristics analysis using training cases. Although the level of fog detection differs depending on the fog intensity and weather conditions, the quantitative evaluation of results using ground observed visibility data showed that average probability of detection and false alarm ratio are 0.64 (0.24 ~ 0.89) and 0.56 (0.33 ~ 0.71), respectively. We performed sensitivity tests for fog detection methods because the detection levels can be affected by fog detection method. As a result, the Weighted Sum Method (WSM) showed a slightly lowered detection level compared to that of the Simple Decision Tree (SDT), average differences, hit rate, hanssen-kuiper skill score, threat score, and bias are −0.31, 0.01, −0.12, and 2.07, respectively. And more works are needed for the improvement of fog detection levels through the sophistication of thresholds and weights using more cases, because detection level is sensitive to fog cases.