Good for when you’ve got a time machine that needs charging.
Lightning has often been one of nature’s most intimidating forces. A mighty bolt packing massive voltage striking seemingly at random; you can’t prepare for it, you can only hope it doesn’t strike you or something you care about. But notice how I said “seemingly at random.” That’s because, thanks to a new AI system, we might just be able to predict where and when a bolt will strike.
A research team from the École Polytechnique Fédérale de Lausanne in Switzerland has developed a surprisingly simple machine learning-based system that can pinpoint a bolt of lightning within 10 to 30 minutes in a 30-kilometer radius (that’s about 18.6 miles). “We have used machine learning techniques to successfully hindcast nearby and distant lightning hazards by looking at single-site observations of meteorological parameters,” one of the researchers explained in their published paper.
Hindcasting, as the name may imply, involves utilizing past weather data to make an educated guess about what’s coming. The researchers fed their AI a swath of meteorological data that would allow it to determine the atmospheric conditions that usually lead to lightning. This included data about things like humidity, wind speed, and air pressure, among other variables. Armed with this knowledge, the AI had a fairly easy time pinpointing lightning bolts before they actually strike, correctly guessing the strike point 80% of the time.
This system, in addition to being efficient, is also pretty cheap to replicate. As long as there’s weather data lying around to feed it, any country could have their own lightning-predicting AI. Before that, though, the researchers are hoping their AI could lend a hand in Europe’s ongoing Laser Lightning Rod project, which would utilize laser light to completely redirect lightning bolts into harmless locations.