A lightning summary and decision model for thunderstorm prediction at Whiteman Air Force Base, Missouri

Abstract

A cloud-to-ground lightning summary was developed for a 139xl85 kilometer area centered at Whiteman Air Force Base. Spatial and temporal patterns, and first stroke peak currents were analyzed from 1989-1995. Stability indices were examined for thunderstorm and non-thunderstorm periods on a seasonal basis. Regression equations developed using these variables distinguished thunderstorm periods from non thunderstorm periods. Decision models were presented that combined responses from the equations with other meteorological considerations. A preferred track for springtime thunderstorms was located between the base and the Ozark Mountains. No preferred track was found during the other seasons. Although diurnal distributions of lightning flashes showed that thunderstorms were possible at any time, late afternoon and nocturnal maxima were observed during the spring and summer. The nocturnal maximum disappeared during the fall. First flash times for thunderstorm events were spread out for the spring and fall. A summertime peak between 1800-2100 UTC (1200-1500 CST) was detected. Last flash times tended to be random, but preferences were observed for early evening and early morning. The percentage of positive flashes nearly doubled in 1994 and 1995 compared to previous years. An increase was observed in the number of positive flashes with peak currents below 60 kA. An increase of negative flashes with peak currents below 20 kA was seen, but contamination by intracloud flashes could not be disregarded. The four spring and fall regression equations for thunderstorm prediction performed better than the two for summer. When tested on independent data, one spring equation had a critical success index (CSI) of 59% and probability of detection (POD) of 100%. One summer equation had a CSI of 55% and POD of 85%. Two thunderstorm decision models were constructed using the seasonal regression equations and several meteorological conditions such as frontal system location. Skill scores for these models were higher than those for any individual equation. The spring model performed best with a CSI of 79% and POD of 100%.