Filtering Low Information Content Optical Pulses from Dual Geostationary Lightning Mapper Data

Currently, the Geostationary Lightning Mapper (GLM) Lightning Cluster Filter Algorithm removes all flashes that contain a single group. A less strict filter would only remove flashes that have a single event. Neither option is ideal as there are single group flashes that are not associated with lightning and there are single event flashes that are associated with lightning. It is difficult to determine which of these low information content flashes are actually associated with lightning without resorting to external ground truth systems, all of which have their own issues with detection efficiency, false alarm rates, and areal coverage. One unique way to test the data for false alarms is to compare lightning in the regions where both GLM-16 and GLM-17 overlap. In the overlap region, flashes detected by both systems, regardless of size, are unlikely to be from non-lightning noise sources. Accordingly the GLM-16 and GLM-17 flashes are separated into two groups of either flashes seen by both instruments or flashes seen by only one of the two instruments. With this technique, the characteristics of the low information content lightning and non-lightning flashes are determined and used to design appropriate filters. The filters are not for real-time processing, but for post-data collection analysis. The filters will ultimately be part of a Common Scientific Algorithm for processing orbital lightning data.