Satellite Observation of Daily Land Surface Characteristics and Water-Energy Fluxes on the Tibetan Plateau (2002 - 2024)

Variations in surface processes over the Tibetan Plateau (TP) profoundly influence regional and global climate, yet observations remain limited. In the study, a daily dataset of land surface characteristics and water-energy fluxes during 2002-2024 is constructed from multi-source satellite and reanalysis data via multiple algorithms: Convolutional Neural Network - Long Short-Term Memory Network (CNN-LSTM) for albedo, Savitzky-Golay filter for Normalized Difference Vegetation Index (NDVI), Surface Energy Balance System (SEBS) for energy fluxes, MOD16 Soil Texture evapotranspiration Model (MOD16-STM) for evapotranspiration (ET). Most variables show decreasing trends (albedo: -0.00117 decade-1, net radiation: -0.63 W·m-2·decade-1, ground heat flux: -0.042 W·m-2·decade-1, and sensible heat flux: -1.83 W·m-2·decade-1), except latent heat flux (0.66 W·m-2·decade-1), ET (4.90 mm·decade-1), and NDVI (0.00774 decade-1). Extreme gradient boosting - Shapley additive explanations (XGBoost-SHAP) attribution analysis reveals air temperature as the dominant ET driver (65.09%), followed by net radiation (-15.76%) and vapor pressure deficit (10.99%). This dataset offers a solid basis for studying land-atmosphere interactions and climate impacts, and is accessible via the Science Data Bank.