Researchers pretrained their AI weather foundation models with UFS-DeepAQM, a NOAA-developed surrogate for atmospheric composition, enabling two-way interaction between meteorology and atmospheric chemistry.

Two researchers, Joshua S. Fu and Jia Xing, are using their atmospheric chemistry weather model to enhance NVIDIA’s Earth-2 platform, a real-time, high-resolution weather and climate simulation tool.

The collaboration aims to deliver more accurate forecasts and could lead to longer-term partnerships if NVIDIA finds the results valuable.

The project tackles gaps in current forecasting systems by addressing underrepresented aerosols and atmospheric processes like dust and smoke that influence weather extremes, climate feedback, and air quality.

Preliminary results from Jia Xing show promising progress, and the researchers believe they are leading in this space with their results.

NVIDIA supports the research with eight GPU nodes and cloud computing resources for 16 months to develop and test the enhanced AI forecasting framework.

The atmospheric chemistry model provides hyper-local predictions, improving precipitation and air quality forecasts with potential global public health and environmental benefits.

Potential applications include helping farmers with season forecasts and aiding emergency departments during hot summers, with broader societal benefits anticipated as the NVIDIA collaboration expands.

Two Civil and Environmental Engineering professors are collaborating with NVIDIA to integrate atmospheric composition into the Earth-2 AI foundation model ecosystem, aiming to create an AI framework for combined weather and air-quality prediction.