Mapping Seasonal High-Resolution PM 2.5 Concentrations with Spatiotemporal Bagged-Tree Model across China.

High concentrations of fine particulate matter (PM_2.5) are well known to reduce environmental quality, visibility, atmospheric radiation, and damage the human respiratory system. Satellite-based aerosol retrievals are widely used to estimate surface PM_2.5 levels because satellite remote sensing can break through the spatial limitations caused by sparse observation stations. In this work, a spatiotemporal weighted bagged-tree remote sensing (STBT) model that simultaneously considers the effects of aerosol optical depth, meteorological parameters, and topographic factors was proposed to map PM_2.5 concentrations across China that occurred in 2018. The proposed model shows superior performance with the determination coefficient (R²) of 0.84, mean-absolute error (MAE) of 8.77 μg/m³ and root-mean-squared error (RMSE) of 15.14 μg/m³ when compared with the traditional multiple linear regression (R² = 0.38, MAE = 18.15 μg/m³, RMSE = 29.06 μg/m³) and linear mixed-effect (R² = 0.52, MAE = 15.43 μg/m³, RMSE = 25.41 μg/m³) models by the 10-fold cross-validation method. The results collectively demonstrate the superiority of the STBT model to other models for PM_2.5 concentration monitoring. Thus, this method may provide important data support for atmospheric environmental monitoring and epidemiological research. [ABSTRACT FROM AUTHOR]