Microsoft has recently unveiled a groundbreaking artificial intelligence (AI) model named Aurora that has revolutionized weather forecasting. This new system has shown exceptional accuracy in tracking air quality, weather patterns, and predicting climate-related tropical storms. The findings of this innovative technology were published in a study released in the prestigious journal Nature.
Aurora has proven to be more effective and efficient than traditional forecasting methods, generating 10-day weather forecasts and predicting hurricane trajectories with remarkable precision. The system, which has not yet been commercialized, outperformed operational forecasting centers like the US National Hurricane Center in forecasting all hurricanes in 2023 based solely on historical data.
Unlike traditional weather models that rely on physical principles and require extensive computational power, Aurora significantly reduces computational costs, making it several hundred times more cost-effective. This development comes on the heels of Huawei’s Pangu-Weather AI model in 2023 and could potentially revolutionize how major weather agencies forecast extreme weather events exacerbated by global warming.
Senior author Paris Perdikaris, an associate professor of mechanical engineering at the University of Pennsylvania, expressed optimism about the future of weather forecasting with AI. He believes that the next phase in weather system science is to build systems that can directly integrate observations from remote sensing sources to generate high-resolution forecasts.
Aurora has set a new standard by consistently outperforming seven forecasting centers in predicting the five-day trajectory of devastating cyclones. In a simulation, the system accurately forecasted the path of Typhoon Doksuri, the most costly typhoon in the Pacific, four days in advance, surpassing official forecasts at the time.
Microsoft’s AI model also demonstrated superior performance compared to the European Centre for Medium-Range Weather Forecasts (ECMWF) in 92 percent of cases for 10-day global forecasts. The ECMWF, known as the global benchmark for weather accuracy, provides forecasts for 35 European countries.
The promising results of Aurora and other AI models are attracting attention from weather agencies worldwide. Meteo-France and other agencies are developing their own AI learning models alongside traditional digital models to enhance forecasting capabilities. The ECMWF has already introduced its first learning model, which is significantly more cost-effective in terms of computing time compared to traditional models.
While Aurora operates at a higher resolution than the ECMWF model, the latter is already operational and providing forecasts. The integration of AI into weather forecasting is expected to bring about significant advancements in accuracy, efficiency, and cost-effectiveness in the years to come. As weather agencies continue to explore the potential of AI technology, the future of weather forecasting looks promising and transformative.
