Zheng Duan

Monitoring atmospheric CO₂ concentration changes across China is crucial for reducing carbon emissions and tackling global warming. Due to monitoring gap in satellite observations, this paper uses the Extremely Randomized Trees (ERT) model alongside OCO-2 satellite data and predictor variables to develop a 0.1° high-resolution monthly dataset of column-averaged dry air CO₂ mole fraction (XCO₂) in China between 2015 and 2023. The model demonstrates high precision, achieving a sample cross-validation with R² of 0.988 and RMSE of 0.732 ppm, while validation at Hefei and Xianghe stations yielded R² (RMSE) of 0.951 (1.339 ppm) and 0.865 (1.840 ppm), respectively. Therefore, this dataset is suitable for high-resolution studies. The results show that XCO₂ across China is increasing at a rate of 2.46 ppmyr⁻¹ during 2015–2023, with XCO₂ highest in spring and lowest in summer. There is a significant spatial heterogeneity in XCO₂ distribution during the period, characterized by a gradient of ‘higher in the east and lower in the west’. Among the factors analyzed, the most influential were population density, transportation infrastructure level and energy intensity, which emerged as the strongest human activity-related drivers of XCO₂ variability, with population density exhibiting a positive correlation with XCO₂, while the latter two show a negative correlation with XCO₂. This paper can provide a scientific basis for policymaking to reduce carbon emissions, promote sustainable development and address climate change.