File:Locations of model training stations and model performance of cross-validations.webp

English: "a and b, Maps showing the geographical distribution of air quality monitoring stations of PM2.5 (a) and O3 (b) used for machine learning model training, the mean PM2.5 or O3 refers to the average value of all available valid observations for each station. c and d, Maps showing the station-specific R2 in the spatial 10-fold cross-validation of the random forest models for estimating all-source daily average PM2.5 (c) and daily maximum 8-hour average O3 (d). e, Density scatter plots showing the overall performance of the machine learning models for estimating all-source daily average PM2.5 (left) and daily maximum 8-hour average O3 (right) based on general 10-fold cross-validation. RMSE, root mean squared error. f, Density scatter plots showing the overall performance of the machine learning models for estimating all-source daily average PM2.5 (left) and daily maximum 8-hour average O3 (right) based on spatial 10-fold cross-validation."