Microsoft’s new AI approach to weather forecasting, Aurora, can accurately predict weather events and outperform traditional systems.
Aurora is able to generate 10-day weather forecasts and predict extreme weather events more accurately, faster and at lower computational costs than traditional numerical forecasting and previous AI approaches.
It does this by analysing a vast amount of diverse atmospheric data sources. The model does not only offer “greater accuracy in general, but it also means we are better at forecasting extreme events”, said Megan Stanley, a senior researcher at Microsoft Research.
For instance, it predicted that Typhoon Doksuri, which wreaked havoc in the Philippines in July 2023, would reach land four days in advance of the event.
Official predictions at the time had placed the storm off the coast of northern Taiwan.
Aurora also beat the National Hurricane Centre in forecasting five-day tropical cyclone tracks and well as outperforming seven major forecasting centres...
