Extended Data Fig. 3: Comparison of AOD forecasting accuracy among the operational AI-GAMFS, CAMS and GEOS-FP systems.
From: Advancing operational global aerosol forecasting with machine learning
a, R and RMSE of operational AI-GAMFS forecasts compared with CAMS. Forecasts are for 5 days at 3-h resolution evaluated against AOD observations from 26 CARSNET sites across China in 2023. Evaluation metrics (R and RMSE) for each step were calculated by aggregating all matched samples over the entire domain. b, Spatial distribution of the average RMSE for each step of the 5-day AOD forecast (a total of 30 steps) from operational AI-GAMFS, alongside the RMSE difference between CAMS and operational AI-GAMFS. c, As in a but for operational AI-GAMFS forecasts compared with GEOS-FP, evaluated against 24 CARSNET sites during July–August 2024. d, Spatial distribution of the average RMSE for each step of the 5-day AOD forecast (a total of 25 steps) from operational AI-GAMFS, alongside the RMSE difference between GEOS-FP and operational AI-GAMFS. The percentages in the lower-right corners of the RMSE difference panels indicate the proportion of sites where operational AI-GAMFS showed lower RMSEs than those of CAMS or GEOS-FP.
