Tracking Thunderstorm Cells Using Lightning Density Information

Several systems to track and nowcast thunderstorms are available, most of them using algorithms based on radar and satellite information (among the most common are TITAN, WDSS-II etc.). Most of theses systems are able to assimilate lightning data available from lightning location systems (LLSs) to improve the resulting information. On the other hand, only few tracking systems are based uniquely on lightning data, mainly due the limited amount of spatial information (number of events in a timestamp). With the advance LLSs in the detection of total lightning (intracloud and cloud-to-ground), more information became available, allowing the use of point pattern analysis (PPA) tools to create a suitable spatial/temporal window to track the thunderstorm development. In this work we propose a tracking algorithm based on a kernel density, using it contours (polygons and centroids) and some spatial (Boolean) criteria - overlapping areas, merge and splitting conditions. The use of density information regardless to other cluster methods (DBSCAN, K-means, mean shift etc.) to aggregate lightning belonging to the same thunder cell reduces the noise and allows the use of several statistical information about the density distribution inside each polygon. The idea of our algorithm is similar to the mentioned TITAN, but using only lightning information, which become essential for countries like Brazil were the radar/satellite spatial/temporal coverage is limited. Since the total lightning network is being expanded over the country, the algorithm might be applied for wide areas, and could improve the civil defense warning systems.