China XCO2 seamless daily data set

Based on the observed values of OCO-2 XCO2 concentration, meteorological, vegetation, terrain and other variables, the 0.05° high-precision seamless daily XCO2 dataset in China from January 1, 2015 to March 31, 2024 was constructed by using the XGBoost model through multi-source data fusion.The results of model cross-validation (R²= 0.984, RMSE=0.894ppm) and ground station auxiliary validation (R²=0.912, RMSE= 1.720ppm) indicate that the data set constructed in this study has better accuracy and is an effective supplement to the existing observational data sets of XCO2 concentration.

本研究基于OCO-2卫星的二氧化碳柱平均混合比（XCO2）浓度观测值，结合气象、植被、地形等多类辅助变量，采用极限梯度提升树（XGBoost）模型通过多源数据融合方法，构建了2015年1月1日至2024年3月31日中国区域0.05°空间分辨率的高精度无缝隙逐日XCO2数据集。经模型交叉验证（决定系数R²（Coefficient of Determination）=0.984，均方根误差RMSE（Root Mean Square Error）=0.894ppm）与地面站点辅助验证（R²=0.912，RMSE=1.720ppm）结果表明，本研究所构建的数据集具备优异的精度表现，可作为现有XCO2浓度观测数据集的有效补充。