In order to extract thunderstorms from existing lightning phenomena and study thunderstorm activity law

based on the lightning data collected by the lightning location system

this paper develops an improved DBScan algorithm based on Spark framework

carries out lightning clustering analysis

extracts thunderstorms

dynamically analyzes the whole process of thunderstorms ’ occurrence

outbreak

movement and extinction in ArcGIS software

and finally checks them by computer. Thunderstorms are analyzed statistically to find thunderstorm activity rules. According to the historical thunderstorm data

the thunderstorm is excavated in reverse

and the thunderstorm activity law is mastered

which lay a foundation for thunderstorm early warning in the future.