Microsoft has created an artificial intelligence (AI) model that outperforms current forecasting methods in tracking air quality, weather patterns, and climate-affected tropical storms, according to studies published last week.
The new model, known as Aurora, provided 10-day weather forecasts and forecasted hurricane courses more precisely and quickly than traditional forecasting, and at a lower cost, according to researchers who published their findings in journal Nature.
"For the first time, an AI system can outperform all operational centers for hurricane forecasting," noted senior author Paris Perdikaris, an associate professor of mechanical engineering at the University of Pennsylvania.
Aurora, trained just on historical data, was able to estimate all hurricanes in 2023 more precisely than operational forecasting centres such as the US National Hurricane Centre.
Traditional weather prediction models are based on fundamental physics principles such as mass, momentum, and energy conservation, and therefore demand significant computing power. The study found that Aurora's computing expenses were several hundred times cheaper.
The trial results come on the heels of the Pangu-meteorological AI model developed and unveiled by Chinese tech giant Huawei in 2023, and might mark a paradigm shift in how the world's leading meteorological agencies predict weather and possibly deadly extreme events caused by global warming.
According to its creators, Aurora is the first AI model to regularly surpass seven forecasting centres in predicting the five-day path of deadly storms.
Aurora's simulation, for example, correctly predicted four days in advance where and when Doksuri, the most expensive typhoon ever recorded in the Pacific, would reach the Philippines. Official forecasts at the time, in 2023, showed it moving north of Taiwan.
Microsoft's AI model also surpassed the European Centre for Medium-Range Weather Forecasts (ECMWF) model in 92% of 10-day worldwide forecasts, on a scale of about 10 square kilometres (3.86 square miles). The ECMWF, which provides forecasts for 35 European countries, is regarded as the global standard for meteorological accuracy.
In December, Google announced that its GenCast model has exceeded the European center's accuracy in more than 97 percent of the 1,320 climate disasters observed in 2019.
Weather authorities are closely monitoring these promising performances—all experimental and based on observed phenomena.