Sources: Pexels

A new artificial intelligence model has improved the prediction of tropical cyclone tracks, intensity and wind structure, according to research published in Nature. Google DeepMind says its WeatherNext Cyclones system can provide about one additional day of useful forecast lead time compared with earlier models.

The advance could have implications for countries regularly exposed to hurricanes, typhoons and cyclones. The World Meteorological Organization estimates that tropical cyclones caused more than 779,000 deaths and about $1.4 trillion in economic losses globally over the past 50 years.

WeatherNext Combines Global And Local Storm Data

Cyclone forecasting requires accurate predictions of both storm movement and changes in strength. WeatherNext Cyclones combines global atmospheric conditions with historical cyclone observations.

Google DeepMind says the model was trained on nearly 20 terabytes of atmospheric data and information from almost 5,000 historical storms in the International Best Track Archive for Climate Stewardship, or IBTrACS.

NOAA describes IBTrACS as the most complete global collection of tropical cyclone data. The database brings together observations from meteorological agencies worldwide.

WeatherNext can also generate 1,000 possible forecast scenarios for a cyclone. Such ensemble forecasts are designed to represent uncertainty and help forecasters assess a wider range of potential outcomes.