The John F. Kennedy Space Center (KSC) and Cape Canaveral Air Force Station (CCAFS) jointly operate a network of 31 surface-based electric-field mills distributed across the two facilities for lightning hazard-warning decision support. We have been engaged in an effort to understand the spatial and temporal evolution of patterns in contours of electric field in order eventually to try to develop improved lightning hazard-warning techniques. To the best of our knowledge, no current lightning hazard-warning criteria incorporate objective application and interpretation of the temporal and spatial evolution of contours of the electric field at the surface before, during, and after active lightning periods in storms. We are interested in two problems of considerable importance at KSC/CCAFS, and, for that matter, at any location where outdoor operations and activities take place during thunderstorm seasons. The problems are 1) estimation of when and where the first flashes (both in-cloud (IC) and cloud-to-ground (CG)) in a storm might occur and 2) assessment of the potential for flashes (both IC and CG) to occur late in the lifetimes of storms, after flash rates have decreased drastically. Initially we have restricted our attention to CG flashes.

We have been studying isolated air-mass convective storms that developed over the KSC/CCAFS area from early May through late September, 2004-2006. We have restricted our attention initially to this simplest class of thunderstorm for a number of reasons, not least of which is the fact that they are more likely to produce a lightning flash at a given location with little or no warning. As shown by Lengyel (2004), the majority of lightning casualties are a result of one of the first few CG flashes in a storm or of CG flashes occurring at the end of storms. When long-lived moving storms have begun to produce lightning, it is relatively easy to project where they will go and what they will do. It is much more difficult to determine when and where a growing storm will make the first lightning flash and whether a dissipating storm has produced its last flash.

We identified thunderstorms that fit the air-mass or “pop-up” criteria by examining rainfall data, CG lightning data, and archived WSR-88D reflectivity data from the Melbourne, Florida (KMLB) radar. For the storms of interest we performed a two-pass Barnes objective analysis on the electric-field data. Each analysis cycle resulted in one filled-contour plot of averaged data, superimposed on a map of the sites of the operational field mills, on which the ground-strike point of the CG flash of interest was marked. We then used the contour plots to create animations of the spatial variation of the field contours for 58 first CG flashes and 62 last CG flashes.

We previously reported on the results of the analysis of contours leading up to first CG flashes. We noted that on the basis of this relatively limited data set we have not yet seen any consistent patterns that would lead us to suggest that every time we see it we can say when and where a CG flash is likely to strike. We were however able to show that it is possible with 70% to 80% probability of detection to predict that a lightning flash will occur within 10 km and 10 minutes of the location of fields exceeding magnitude o 1 kV/m.

Since then we have been working on the issue of flase-alarm rate (FAR). We have calculated the FAR. in several different ways. We created a grid encompassing the KSC/CCAFS area based on the latitude/longitude readings of each of the 31 operational electric-field mills. Hits and misses were determined based on spatial, temporal, and electric potential gradient thresholds. CGLSS data were looped over the grid points for the 30 minute period before the first CG lightning strike. If a strike occurred within the domain a hit was recorded. Otherwise, a miss would be assigned. From these findings, the FAR was calculated. We have also investigated a second method, in which we separated the KSC/CCAFS area into regions based on latitude. Our preliminary results suggest that false-alarm rates comparable to those of severe thunderstorm/tornado warnings are achievable, at least under some circumstances.