Developed by a team from the Electromagnetic Compatibility Laboratory at Switzerland's EPFL research institute, the system was "trained" using a database of readings of four basic weather parameters: atmospheric pressure, air temperature, relative humidity and wind speed.
Gathered over a 10-year period from 12 Swiss weather stations in urban and mountainous regions, these readings were cross-referenced with recordings from lightning detection and location systems. This allowed the AI algorithms to learn which weather conditions were associated with lightning strikes in given areas.
The resulting computer system is claimed to be almost 80-percent accurate at predicting where and when lighting will strike to the nearest 10 to 30 minutes, within a 30-km (19-mile) radius. As the technology is developed further, those numbers should improve.