AI Model Forecasts Global Air Pollution in 39 Seconds
Why this is here: The AI-GAMFS model completed a full global five-day forecast in 39 seconds, a speed that prompted a skeptical reviewer to personally verify the results on their own server.
Chinese researchers have developed the world’s first aerosol-meteorology coupling artificial intelligence forecast model—AI-GAMFS. The model can generate a five-day global forecast with three-hour resolution in just 39 seconds.
AI-GAMFS outperforms existing European and American physical models in predicting key components like dust, black carbon, and sulfate. The model analyzes 54 key environmental and meteorological elements, learning complex interactions from 42 years of global reanalysis data.
Researchers from the Institute of Atmospheric Physics, Chinese Academy of Sciences, built AI-GAMFS to move beyond simply forecasting weather to forecasting atmospheric composition. Traditional methods struggle with the intricate feedback between aerosols and meteorology, but AI-GAMFS bypasses complex equations by learning directly from historical data.
The model has been implemented in over ten provincial meteorological departments in China and integrated into the MAZU early warning platform to provide global forecasts. It offers a cost-effective solution with consistent accuracy across different regions, unlike some traditional models.