The world's first aerosol forecast artificial intelligence model has been released.

The world's first global aerosol forecast artificial intelligence model AI-GAMFS, led by Chinese scientists, was published in the international academic journal "Nature" on March 5th. This model can provide high-precision environmental meteorological forecasts for the next 5 days globally, with a 3-hour interval, in 1 minute. It provides strong support for accurate early warning of global aerosol pollution events such as dust storms, wildfires, and haze. AI-GAMFS was jointly released by research team led by researcher Che Huizheng from the China Meteorological Administration and Academician Zhang Xiaoye from the Chinese Academy of Engineering, along with many domestic and foreign research institutions. It is based on 42 years of global aerosol reanalysis data from 120,000 hours of training, covering key optical properties, ground concentrations, and related meteorological elements of aerosols such as dust, sulfates, black carbon, and organic carbon, with a spatial resolution of 50 kilometers. Aerosols are dispersed systems composed of solid or liquid particles suspended in the atmosphere. Their sources include not only naturally occurring dust from sandstorms and wildfires, but also smoke particles from boilers and engines, as well as solid particles emitted from mining and quarrying processes. Additionally, secondary particles formed through chemical reactions of gaseous precursors in the atmosphere are also important sources of aerosols. "The complexity and computational cost of aerosol forecasting are much higher than traditional weather forecasting. It requires the forecast system to simultaneously analyze the multiple sources of aerosols, complex chemical transformations, and their multi-scale interactions with the weather system," said Gu Ke, core member of the team and deputy researcher at the China Meteorological Administration.