A movement prediction method of CG flashes based on lightning and convective clouds clustering and tracking technology

Movement prediction of CG flashes is one of the most important lightning prediction techniques in severe weather. Since the randomness and instability of lightning occurrence, very limited approaches have been proposed for this objective. In this paper, a movement prediction method of CG flashes was presented, based on both CG flashes in the past and the clouds in satellite nephogram(cloud image). An improved lightning spatial clustering method was developed and introduced on time-series flashes over a certain period of time, and thus independent, random-happen flashes are aggregated and classified into different thunderstorm groups similar to thunderclouds, meanwhile based on fuzzy C-means algorithm convective clouds were identified and segmented from the cloud images, and then with comparison between the thunderstorm groups and convective clouds in different time sections, the track of CG flashes were extracted. Therefore, on the set of extracted track point, the movement prediction can be calculated by any regression functions. Experiments were made and the result shows that the probability of detection (POD) is around 75%, and false alarm rate (FAR) is around 30%.