Google's WeatherNext AI model outperforms existing systems in forecasting cyclones, according to a paper published in Nature. The AI-based technology provides an extra day of warning, delivering a three-day forecast for hurricanes or typhoons that matches the accuracy of previous two-day predictions.
AI model provides earlier cyclone warnings
The study, featured in the journal Nature, highlights that WeatherNext's three-day forecasts are as reliable as the two-day forecasts produced by conventional methods. This additional lead time could prove critical for residents in cyclone-prone regions, such as Toamasina in Madagascar, which experienced the aftermath of Cyclone Gezani in February.
This development is part of a broader shift in weather forecasting, with AI tools gaining momentum over the past year. These systems are being integrated alongside traditional methods to enhance prediction capabilities.
From hand calculations to AI pattern recognition
Numerical weather forecasting has a history spanning over a century, beginning with Lewis Fry Richardson's book "Weather Prediction By Numerical Process." Richardson, a pioneering mathematician at the Met Office, proposed calculations that initially had to be done by hand. At that time, generating a daily forecast took more than six weeks, making the approach impractical until computers were introduced. Even modern supercomputers face processing power limitations.
AI sidesteps these constraints by learning patterns from training data rather than performing detailed physics-based calculations. This allows AI systems to rival conventional models on several forecasting scales, particularly for longer-term predictions. While not necessarily superior in all aspects, AI can produce comparable forecasts faster and at lower cost.
AI welcomed in weather forecasting
Despite resistance to AI in many fields, the weather forecasting community generally welcomes it as a useful new capability. The integration of AI tools like WeatherNext represents a gradual enhancement of forecasting accuracy and efficiency, benefiting both meteorologists and the public.
