2026-10-12T21:46:01Z https://www.scidb.cn/oai

10.57760/sciencedb.38323 2026-06-01T14:14:48Z

2026-06-01 Statistical Dataset of Spatially Continuous XCO₂ Reconstruction from ACDL Using Residual Extra-Trees Learning doi:10.57760/sciencedb.38323 en This dataset is based on the observation data of the world's first aerosol and carbon detection lidar (ACDL) with continuous day and night observation capability carried by China's DQ-1 satellite. The statistical analysis product of global atmospheric CO ₂ column concentration (XCO ₂) is reconstructed using the Extreme Random Tree (ERT) residual learning framework. The reconstruction model is based on physical constraint feature engineering (including sea land differentiation flux variables and spatiotemporal soft weighting), and does not rely on ground observations or other satellite products, maintaining the independence of the active laser detection mode. This dataset provides statistical results of XCO ₂ reconstruction divided by seasons, latitude bands, and other dimensions, as well as accuracy evaluation indicators validated by TCCON global ground stations (R ² of 0.96, RMSE of 0.92 ppm), and comparison results of different models. The dataset fully preserves the unique diurnal sampling information of ACDL, which can provide statistical analysis references for global carbon balance assessment, carbon source sink inversion, and research on diurnal carbon cycling processes  ACDL/DQ-1 satellite; spaceborne lidar; Extra-Trees Zhong Qi Fan Chuncan Liu Jiqiao RESTRICTED dataset Science Data Bank