Switzerland uses artificial intelligence to forecast lightning

According to Climate and Atmospheric Science, scientists from Lausanne, Switzerland, have developed a simple machine learning system that can predict lightning 30 minutes before it occurs, based on normal weather data.

"The current lightning forecast systems are very slow, they need specialized data sources such as radars or radiators, " said Lausanne, a Federal Press agency, Lausanne (EPFL) press agency. The Swiss scientists' method of using information can be collected by any weather station, which means that it is possible to prepare forecasts even for the most remote places on Earth. land".

Lightning arises in very specific conditions that can be predicted and calculated using artificial intelligence - (Image: Shutterstock)

As early as 1749, Benjamin Franklin, the famous politician and natural scientist, shed light on the nature of lightning. His experiments with lightning rods and kites show that lightning bolts are electrical discharges that run between thunder-cloud clouds and the Earth's surface.

At the beginning of this decade, through research by Russian physicists, the picture became clearer. As it turned out, cosmic rays were involved in the occurrence of lightning, acting as a kind of "trigger" to discharge. As recently discovered by Japanese scientists, gamma-ray bursts all herald lightning strikes. In principle, gamma ray bursts can be used to accurately predict when and where subsequent lightning strikes will be generated.

Amirhossein Mostajabi and colleagues have developed a less accurate, but more affordable, and more scalable system to predict lightning strikes by studying whether lightning strikes occur in what kind of cloud and under what meteorological conditions. These observations led them to the idea that lightning arises under very specific conditions, which can be predicted and calculated using artificial intelligence systems.

Encouraged by this idea, physicists have prepared a set of machine learning algorithms with the idea that some relatively inaccurate prediction systems can learn from previous mistakes and gradually. Go to the correct answer combined in a sequence.

To train this program, physicists have prepared a special set of data that dozens of meteorological stations installed in different regions of Switzerland collected for decades. During this period, thousands of lightning strikes occurred in the vicinity. This allows artificial intelligence to learn the identifying characteristics of lightning and learn to predict lightning based on characteristic changes in temperature, humidity and air pressure as well as other weather data.

The current version of the machine learning system, tested, correctly predicted 76% of lightning strikes in half an hour before they actually appeared in the vicinity of weather stations. This, according to the scientists, can greatly increase the safety of the aircraft and ensure the continuous operation of the electrical network and other infrastructure.

1. Successfully researched lightning warning device 30-60 minutes ago