Evaluation of a Bispectral Fog Detection Technique with a Low Earth Orbiting Satellite for Fog Events in Florida

According to the United States Department of Transportation (US DOT), an average of over 28,000 crashes and almost 500 deaths annually occurred as a result of fog-related vehicular accidents. In Florida, the January 2008 and January 2012 fog-related multi-car accidents claimed the lives of four and eleven people, respectively. A more effective fog warning system could include the use of remote sensing. The ground observation sites used to detect fog statewide are both widely and unevenly dispersed. Many high-traffic areas affected by fog are not monitored by ground equipment, leading to poor forecasting and detection of fog in these areas. A combination of both ground observations and remote sensing may lead to better statewide fog detection and forecasting. A bispectral nighttime fog detection technique is used to determine the presence of fog across the state of Florida. This technique uses brightness temperature differences (BTD) between two infrared (IR) channels. The performance of the technique is validated through the use of six months of observation data from AWOS/ASOS sites across the state. An optimum fog detection threshold is found based on the BTD values. Both the optimum threshold and the skill of the optimum threshold are compared to a previous study which used a geostationary satellite for fog detection. The bispectral technique shows little skill, with a large amount of misses and false detections of fog. The low skill can be attributed to the fact that MODIS makes only one nighttime pass which may not necessarily be when fog has formed. The increased spatial resolution of the MODIS sensor over the previous generation GOES Imager does not make up for the decreased number of nighttime satellite passes in a given day.