Google’s WeatherNext AI model is better than existing systems at forecasting cyclones, according to a paper published in Nature.
The AI-based technology gives an extra day of warning, providing a three-day forecast for hurricanes or typhoons that is as accurate as previous two-day predictions.
But this is just one AI tool transforming weather forecasting, part of a gradual shift that has been gaining momentum over the past year.
Numerical forecasting dates back more than a century to Weather Prediction By Numerical Process, a book by Lewis Fry Richardson, who was a pioneering Met Office mathematician. At that time, the calculations had to be carried out by hand. It took more than six weeks to generate a daily forecast, making the method impractical until the introduction of computers. Even with modern supercomputers, processing power can still be a limiting factor.
AI sidesteps this, because rather than having to make detailed calculations like physics models, it develops predictions based on patterns learned from training data.
AI-based systems now rival conventional models for forecasting on several scales, especially longer-term predictions. And while they are not necessarily superior, they can produce comparable forecasts faster and at lower cost.
There is resistance to AI in many areas. In weather forecasting though, it is generally being welcomed as a useful new capability.
