An AI for Climate Extremes: Microsoft Debuts 'Aurora' Foundation Model

Microsoft this week described its early progress with a new AI foundation model (FM) that can perform high-resolution weather forecasting.

In a Microsoft Research blog post, the company described the new model, dubbed "Aurora," as "a cutting-edge AI foundation model that can extract valuable insights from vast amounts of atmospheric data."

Microsoft positions Aurora as especially well-suited for analyzing weather patterns in an era of climate extremes. This is because Aurora has been trained on a huge amount of diverse weather data -- specifically, over 1 million hours of it. As a result, Microsoft claimed, Aurora can reliably analyze and forecast climate events even in scenarios or geographies where weather data is scarce.

"[T]his research paves the way for the development of comprehensive models that encompass the entire Earth system," the researchers wrote. "This could have far-reaching impacts on sectors like agriculture, transportation, energy harvesting, and disaster preparedness, enabling communities to better adapt to the challenges posed by climate change."

Aurora is able to ingest different types of climate data, enabling it to understand complex dependencies between multiple variables -- for instance, the relationship between air pollution and air temperature. It can also generate forecasts about specific climate variables. Besides temperature and pollution, for example, it can predict greenhouse gas levels and wind speeds.

"The model consists of a flexible 3D Swin Transformer with Perceiver-based encoders and decoders, enabling it to process and predict a range of atmospheric variables across space and pressure levels," the researchers said. "By pretraining on a vast corpus of diverse data and fine-tuning on specific tasks, Aurora learns to capture intricate patterns and structures in the atmosphere, allowing it to excel even with limited training data when it is being fine-tuned for a specific task."

In benchmark testing, the researchers found Aurora to perform as well as, or better than, comparable AI models when it came to accuracy. It's also fast.

"Aurora captures intricate details of atmospheric processes, providing more accurate operational forecasts than ever before -- and at a fraction of the computational cost of traditional numerical weather-prediction systems," according to the blog. "We estimate that the computational speed-up that Aurora can bring over the state-of-the-art numerical forecasting system Integrated Forecasting System (IFS) is ~5,000x."

Aurora, the researchers concluded, "highlights the importance of diverse pretraining data, model scaling, and flexible architectures in building powerful foundation models for the Earth sciences."

More information about Aurora is available in this paper.

About the Author

Gladys Rama (@GladysRama3) is the editorial director of Converge360.