Scientists deploy neural networks to forecast storms.

The National Center for Atmospheric Research (NCAR) developed an artificial intelligence system to forecast hail, tornadoes and strong winds during storms that are difficult to predict with traditional weather models. The information is updated twice daily and is freely available online.

According to the developers, the neural network takes into account about 40 different factors, including the spirality of the updraft, the storm’s location, time, dew point, wind speed and surface pressure, and determines their interrelations. The forecasts show the probability of thunderstorm hazards within 40 or 120 kilometres of individual grid points in the model.

The algorithm was trained on a dataset of 500 previous forecasts together with accompanying reports of real storms.

As a result of testing the system, the scientists found that the neural network significantly improved the accuracy of forecasts based on traditional models.

“The neural network not only predicts more deftly when and where serious thunderstorm hazards are likely, but also better forecasts whether hail or wind will prevail,” said NCAR scientist Ryan Sobash.

The developers also reported that, in addition to the rotation of hurricanes, the neural network better predicts hazards associated with straight-line winds than forecasts based solely on the spirality of the updraft.

The researchers launched the experiment in spring 2020. A year later it was redeployed with some refinements.

In May 2021, scientists using artificial intelligence accelerated the modeling of the Universe by 1,000 times.

In March, the United States registered the first AI device for detecting hidden signs of COVID-19 in asymptomatic carriers of the infection.

Subscribe to ForkLog news on Telegram: ForkLog AI — all the news from the world of AI!