A Probabilistic Fog Stability Index for Predicting the Occurrence of Radiation Fog at U.S. Airports

Abstract

Fog is a phenomenon that is widely known to affect the aviation industry adversely, as evidenced by economic losses due to the hindrance of fog on airport operations. This is because fog has been a consistent cause for delays, diversions, and cancellations of scheduled commercial airline flights that have subsequently resulted in a substantial negative economic impact on the transportation industry as well as society. This study’s purpose was to determine suitable predictors of the occurrence of radiation fog at U.S. airports. It assessed an extant Fog Stability Index (FSI) and 1-Day Persistence model for their reliability in predicting radiation fog. This research study addressed the overarching question: is it possible to predict the occurrence of radiation fog at airports using a probabilistic methodology? Therefore, the study’s objective was to compare the reliability of using a theoretical or traditional approach to FSI, a probabilistic FSI, and 1-Day persistence as predictors of radiation fog. The research utilized data period spanning the years 1973 through 2020, at six airports in east-central Florida. The study utilized archival data from Iowa State University’s Environmental Mesonet and radiosonde data provided by NASA’s varying weather observation equipment. The study isolated the occurrences of radiation fog as opposed to advection, sea fog, and other types of fog. This was of specific interest because radiation fog is potentially predictable with such measures and could affect airports to a more noticeable degree, comparatively. Thus, observations were limited to 1000Z to 1500Z and METAR weather codes to BR, FG, MIFG, and BCFG for the occurrence of radiation fog. A statistical analysis of the data was performed utilizing logistical regression analyses supporting the use of a dichotomous dependent variable: the occurrence or nonoccurrence of fog. The preliminary analyses found that 1-Day persistence may or may not be a suitable predictor of radiation fog. This was inconclusive due to the rarity of those events within the sample and resulted in a lack of viable data for logistic regression analyses. Further research will be required confirm the suitability of 1-day persistence for predicting radiation fog. The primary analyses found that both the theoretical and probabilistic approaches to using FSI were reliable predictors of fog as evidenced by a contingency analysis of predicted fog events versus actual fog events within the sample. However, the probabilistic approach yielded better results with respect to hits – correctly predicting the occurrence of fog when it occurred, and misses – not predicting that fog would occur when it did, as opposed to the traditional FSI high, medium, low fog-event chance model. The traditional or theoretical model yielded a lower percentage of hits and a greater percentage of misses. Thus, the study concluded that using a probabilistic FSI model to predict radiation fog events could positively impact the air transportation industry by providing accurate, additional information to decision-makers reducing the consequent economic impact of delays, diversions, and cancellations.