Probabilistic Guidance for Lightning Cessation Within Isolated Convection Near Cape Canaveral, Florida

Almost daily summer time thunderstorms in central Florida frequently halt outdoor operations, requiring that one wait some prescribed time after an observed lightning flash to safely resume activities. This is an especially important problem for the U.S. Air Force’s 45^th Weather Squadron (45WS). Prior research suggests that these wait times might be safely shortened by observing reflectivity values and hydrometeor type with radar to safely predict that lightning has ended for a particular isolated thunderstorm. This paper describes the creation and testing of a usable operational tool that creates probabilistic guidance for the 45WS to use for determining total lightning cessation for isolated storms. The study analyzed dual-polarized radar data from isolated thunderstorms to develop probabilistic lightning cessation guidance for the 45WS. We tracked 184 isolated storms in central Florida at 1 min intervals using radar and lightning detection systems including radar reflectivity and hydrometeor classification at isothermal levels. For each isolated storm we investigated its maximum reflectivity and graupel presence at the 0, -5, -10, -15, and -20°C levels and composite (maximum) reflectivity. A random sample of all the 1 min interval data was used to train a generalized linear model (GLM) to make a probabilistic prediction that cessation had occurred. The GLM revealed that the most statistically significant predictors for lightning cessation were maximum reflectivity at the composite and 0 °C levels along with graupel presence at the -5, -10, -15, and -20°C levels. The GLM was trained with 1000 random samples of minutes to bootstrap the results, with the median values of the final set of predictor coefficients used to calculate probabilities that cessation had occurred at that minute. Forecast verification statistics from another random sample of tracked minutes then were used to analyze the performance of the GLM with different probability thresholds (95.0%, 97.5%, and 99.0%) for determining lightning cessation. Applying this cessation guidance from our GLM as though the storms were occurring in real time revealed that only about 1% of the 184 storms in our data set had observed lightning after the GLM suggested cessation had already occurred, an event which would threaten life and property. Even the median of the most conservative probability threshold (99.0%) improved on the guidance currently being used by the 45WS, while the 95.0% probability guidance had a median wait time of just 9 min after cessation. These excellent results exceeded our expectations.