Nitrogen dioxide pollution in 346 Chinese cities: Spatiotemporal variations and natural drivers from multi-source remote sensing data

In this study, tropospheric column concentration of nitrogen dioxide (TNO2CC) were derived from Sentinel-5P data. We employed statistical and local spatial autocorrelation analyses to investigate the spatialtemporal distribution and variation of TNO2CC across 346 major Chinese cities from 2019 to 2023. Using Random Forest (RF) and Shapley Additive Explanations (SHAP), we analyzed the influence of 15 natural factors on ambient TNO2CC levels. The high R² values (0.92 and 0.76), along with the close adherence to the 1:1 line, demonstrate the model's robustness. The most influential natural factors identified include atmospheric pressure, aerosol optical depth, Leaf Area Index, evapotranspiration, and dew point temperature. Additionally, a non-linear response curve approach was applied to examine the independent association between natural driving factors and pollutant concentrations. TNO2CC varied seasonally across the 346 cities, with the highest levels in winter and the lowest in summer. From 2019 to 2023, TNO2CC levels exhibited fluctuating trends, with notable regional disparities: higher concentrations were observed in capital cities and in northern and northeastern part of China. TNO2CC were significantly influenced by temperature-related variables, aerosol optical depth, and leaf area index. The findings of this study identify key natural influencing factors and provide a scientific basis for revealing the causes of urban air pollution in China, informing pollution control strategies, identifying priority areas for remediation, and supporting the natural formulation of protection policies.