Spatio-temporal distribution of XCO2 concentrations in northeast China based on Downscaling-XGBoost model

Carbon dioxide in the atmosphere, the most abundant greenhouse gas, is recognized as one of the main causes of climate change. A large amount of missing data is contained in XCO₂ datasets obtained from carbon observing satellites due to unfavorable atmospheric conditions and satellite orbits, making it inadequate to reflect the heterogeneity of XCO₂ in space. Our understanding of carbon dynamics is limited by the sparsity of terrestrial measurements and the lack of fine-scale XCO₂ modeling datasets. A downscaled machine learning model was developed to fill the gaps in the retrieval of data from the Orbiting Carbon Observatory 2 satellite retrievals across Northeast China during 2018–2023 (cross-validated <i>R</i>²=0.88, RMSE=1.49ppm). The results showed that the annual mean value of XCO2 in the northeast region for the last 6 years was (407.83&plusmn;67.31) ppm, and the XCO₂ content increases annually with 2.68ppm/year. The highest concentrations of XCO₂ were found in the southern region and around large cities. The seasonal distribution pattern of “high in spring and winter, low in summer” was observed to be obvious. This work confirms the necessity for downscaled modeling of XCO₂ from satellites, which can enhance our comprehension of carbon cycles.