ECMWF releases AIFS-COMPO, an experimental AI air pollution forecast reaching 10 days ahead

The European Centre for Medium-Range Weather Forecasts (ECMWF) has introduced AIFS-COMPO, an experimental AI system that forecasts aerosols and pollutants including ozone, particulate matter, nitrogen dioxide and sulphur dioxide up to 10 days ahead, the Copernicus Atmosphere Monitoring Service (CAMS) said Oct. 16; the model was trained on CAMS reanalyses covering 2003-2023. ECMWF said verification against independent observations showed comparable or improved skill for several key species compared with its operational physics-based system, IFS-COMPO, which forecasts five days ahead, at substantially lower computational cost, and that the AI model still relies on the operational system’s satellite data assimilation for its starting conditions. ECMWF is publishing AIFS-COMPO’s 10-day forecast charts on its website; CAMS director Laurence Rouil said the operational system remains the benchmark for accuracy and reliability, and CAMS says composition forecasts support air-quality monitoring and public-health applications.