Scientists involved neural networks to predict a storm

The US National Center for Atmospheric Research (NCAR) has developed an artificial intelligence system to predict a city, tornado and strong winds during a storm, which is difficult to predict with the help of traditional weather models. Information is updated twice a day and is freely available on the Internet.

According to the developers, the neural network takes into account about 40 different factors, including the spiralality of the ascending stream, the location of the storm, the time, the dew point, the wind speed and the pressure on the surface, and determines their ratio with each other. Forecasts show the likelihood of a thunderstorm hazard between 40 or 120 kilometers from individual points of the grid in the model.

The algorithm was trained at a set of data from 500 past predictions along with accompanying reports on real storms.

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

“The neural network not only more skillfully predicts when and where serious thunderstorms are likely, but better predicts whether a hail or wind will prevail,” said NCAR researcher Ryan Dogs.

The developers also said that in addition to the rotation of hurricanes, the neural network better predicts the dangers associated with rectilinear winds than forecasts based exclusively on the spilation of the rising stream.

Researchers launched an experiment in the spring of https://gagarin.news/ru/terms-of-use/ 2020. A year later, she was again involved with some improvements.

Recall that in May 2021, scientists with the help of artificial intelligence accelerated the modeling of the universe by 1000 times.

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

Subscribe to FORKLOG news at Telegram: Forklog AI – all news from the world of AI!