Fig. 1: Model validation and uncertainties.
From: First close insight into global daily gapless 1 km PM2.5 pollution, variability, and health impact
Spatial distributions of sample-based cross-validated (CV) a coefficients of determination (R2) and b uncertainty (i.e., normalized root-mean-square errors, or NRMSE) of daily PM2.5 estimates (unit: μg m−3) against ground-based measurements (unit: μg m−3) at each monitoring station, and density scatterplots of sample-based CV results between c daily estimates (number of samples = 7,089,428), d monthly composites (number of samples = 255,075), and e annual composites (number of samples = 23,229) and ground-based measurements collected at all monitoring stations from 2017 to 2022 over land. Black dashed lines are 1:1 lines, and red lines are best-fit lines from linear regression. Additional statistical metrics given in c–e are the linear regression equation, root-mean-square error (RMSE), and mean absolute error (MAE). The maps in a, b were created using ESRI ArcGIS Pro 3.0.1.
