Abstract:Satellite-based remote sensing data have been widely used in estimating ground-level PM2.5 concentrations as it can provide spatially detailed information. Most modern satellite-based PM2.5 estimates use statistical models that demand dense PM2.5 monitoring networks. As the national PM2.5 monitoring networks in China were not finished until the end of 2012, the research related to PM2.5 is relatively unsubstantial. To further improve the accuracy and application of remote sensing based estimation models for PM2.5 and take advantage of the newly established networks, we employed a time fixed effects regression model and geographically weighted regression model to develop a spatiotemporal statistical model that estimated ground-level PM2.5 concentrations in Beijing–Tianjin–Hebei. The time fixed effects regression model used the aerosol optical depth (AOD) data from the VIIRS (Visible Infrared Imaging Radiometer Suite) instrument as the major predictive variable along with several other dependent variables, including some factors uncommonly discussed in previous literature, i.e., the satellite-derived NO2 concentrations of the previous day (NO2_Lag) and four directional wind vectors, and estimated day-by-day ground-level PM2.5 surfaces. The geographically weighted regression model used the residuals from the time fixed effects regression model as the dependent variable and the AOD value as the independent variable. Through adding the estimated residuals back to previous surfaces, we obtained the final prediction maps of ground-level PM2.5 concentrations in Beijing–Tianjin–Hebei with a spatial resolution of 6km×6km. The results were as follows. i). The spatiotemporal statistical model performed satisfactorily in that it successfully captured both the temporal and spatial variations in the PM2.5–AOD relationships. The coefficient of determination (R2), mean prediction error (MPE), and root-mean-square error (RMSE) were 0.88301, 8.1331μg/m3, and 13.0574μg/m3, respectively, during model fitting and 0.71889, 12.2712μg/m3, and 19.2927μg/m3, respectively, during model validation. ii). Incorporating the NO2_Lag in the time fixed effects regression model significantly improved the model's performance and it played a positive role in ground-level PM2.5 concentrations. Replacing the simple wind speed with four directional wind vectors was helpful for the model's performance. iii). Meteorological factors and land use characteristics significantly affected the PM2.5–AOD relationships. The temperature and surface relative humidity (SRH) played a positive role, whereas the rainfall, planet boundary layer height (PBLH), average relative humidity in the PBLH (RH_PBLH), four directional wind vectors, and normalized difference vegetation index (NDVI) played a negative role. iv). The prediction maps revealed that fine particle pollution in Beijing–Tianjin–Hebei is severe and the pollution pattern presents relatively strong seasonal heterogeneity and southeast–northwest spatial heterogeneity.

Estimation of monthly 1 km resolution PM2.5 concentrations using a random forest model over “2 + 26” cities, China

Spatiotemporal variation in the impact of meteorological conditions on PM<sub>2.5</sub> pollution in China from 2000 to 2017

[Satellite Retrieval of a Heavy Pollution Process in January 2013 in China].

Estimating PM2.5 concentrations via random forest method using satellite, auxiliary, and ground-level station dataset at multiple temporal scales across China in 2017

Estimation of hourly full-coverage PM2.5 concentrations at 1-km resolution in China using a two-stage random forest model

Estimating the Daily PM2.5 Concentration in the Beijing-Tianjin-Hebei Region Using a Random Forest Model with a 0.01° × 0.01° Spatial Resolution.

Predicting monthly high-resolution PM2.5 concentrations with random forest model in the North China Plain

Estimating 1-Km-resolution PM2.5 Concentrations Across China Using the Space-Time Random Forest Approach

A similarity distance-based space-time random forest model for estimating PM2.5 concentrations over China

Improving the Estimation of PM2.5 Concentration in the North China Area by Introducing an Attention Mechanism into Random Forest

Full-coverage 1 km daily ambient PM<sub>2.5</sub> and O<sub>3</sub> concentrations of China in 2005–2017 based on a multi-variable random forest model

Estimating high-resolution PM 2.5 concentration in the Sichuan Basin using a random forest model with data-driven spatial autocorrelation terms

VIIRS-based remote sensing estimation of ground-level PM2.5 concentrations in Beijing–Tianjin–Hebei: A spatiotemporal statistical model

Estimating Daily Ground-Level PM2.5 in China with Random-Forest-based Spatiotemporal Kriging.

Exploring the detailed spatiotemporal characteristics of PM2.5: Generating a full-coverage and hourly PM2.5 dataset in the Sichuan Basin, China.

Daily Estimation of Ground-Level PM2.5 Concentrations over Beijing Using 3 Km Resolution MODIS AOD

Exploring the detailed spatiotemporal characteristics of PM2.5: Generating a full-coverage and hourly PM2.5 dataset in the Sichuan Basin, China

Machine learning driven by environmental covariates to estimate high-resolution PM2.5 in data-poor regions

Estimation of PM2.5 concentrations with high spatiotemporal resolution in Beijing using the ERA5 dataset and machine learning models

Satellite-derived 1-km estimates and long-term trends of PM2.5 concentrations in China from 2000 to 2018

Spatiotemporal Evolution Analysis of PM2.5 Concentrations in Central China Using the Random Forest Algorithm