2 min read
Microsoft has introduced Aurora, an AI model that can predict multiple atmospheric phenomena, including air quality, typhoons, hurricanes, and complex weather patterns. This innovation will redefine meteorological forecasting, offering untraditional support.
Getting predictions from AI sounds unreliable, but Microsoft has achieved the feat this time. Aurora has consistently outperformed established forecasting centers, including the US National Hurricane Center, in predicting tropical cyclone tracks.
For instance, during Typhoon Doksuri in July 2023, Aurora accurately predicted its landfall in the Philippines four days in advance, while official forecasts incorrectly suggested it would tilt toward Taiwan. It has also shown 20-25% more accurate predictions for cyclone trajectories in the North Atlantic and East Pacific for lead times of two to five days.
Unlike traditional numerical weather prediction systems that rely on massive supercomputers and take hours to generate forecasts. Aurora can produce a 10-day global weather forecast or a 5-day global air pollution forecast in mere seconds on a single GPU. This makes it approximately 5,000 times faster than conventional methods, significantly reducing computational costs.
It is refined after being trained on over a million hours of diverse weather and climate simulation data. This includes satellite imagery, radar information, meteorological center data, and historical weather patterns.
Aurora can also accurately forecast air pollution levels, a notoriously difficult task for computational methods. It can generate 5-day global air pollution forecasts with high spatial resolution, matching or surpassing the accuracy of state-of-the-art atmospheric chemistry simulations.
Microsoft is integrating a specialized version of Aurora into its MSN Weather app. The company has also made Aurora’s source code and model weights publicly available, encouraging further research and development in the field.
