Comparison of Airport Lightning Warning Skill Using NLDN, GLD360, and Single-Point Sensor Lightning Data

Cloud-to-ground (CG) lightning discharges pose significant hazards to exposed assets and personnel at airports. As a result, flight and ground operations typically halt when there is a risk for a CG strike. During these warning periods, airports incur significant direct and indirect costs. Hence an effective CG warning system must balance the probability of detection (POD) of CG strokes with sufficient lead time with both the warning duration (WD) and false alarm ratio (FAR). The POD defines the safety of the system, whereas the WD and FAR characterize the efficiency.

This study compares the warning skill, defined as the trade-off between the POD and the WD and FAR, of lightning warning systems based on two types of single-point sensor, local Electric Field mill (EFM) measurements and flash measurements from a single-point lightning detection sensor, as well as total lightning (TL) data from both the National Lightning Detection Network (NLDN) and the Global Lightning Dataset GLD360.  Using EFM and single point lightning sensor data collected from a single site in Louisville, CO, this study uses warning data from the summer lightning seasons of 2015 and 2016 to compare the warning skill of all four systems using the same reference storms.

While the warning skill of each system in part depends on the development and propagation patterns of lightning activity over the area of interest, this paper seeks to tie the relative warning skill of each system to the performance and detection methodology of each system.  Among the single-point lightning sensor and network data solutions, improved spatial resolution and detection efficiency are shown to improve the overall warning skill. In this case study, the warning system based on EFM data demonstrates the lowest overall skill. The performance of warning systems using a combination of measurement technologies, such as merging alerts from network data with alerts from EFM data, is also analyzed.