Gap-Filling for Daily Evapotranspiration Observations with full-factorial method at Global Flux Sites

The dataset contains daily evapotranspiration (ET) data for 339 global sites, filled using the full-factor method based on the Penman-Monteith decoupling model. The comparative analysis of the filled ET data against three products—ERA5-Land, BESS, and GLEAM—reveals a high degree of consistency across 264 stations, with mean |MAE| and |RMSE| values of 0.32 mm/day and 0.92 mm/day, respectively. For 49 stations, the filled ET data aligns more closely with adjacent observed data or exhibits higher temporal coherence with observed ET data, net radiation, and LAI trends than the three ET products. However, the reliability of the filled ET data at 26 stations requires further validation due to the scarcity of observed data or the influence of input data issues. Therefore, the filled ET data for a total of 313 sites demonstrates a high degree of reliability, with data records spanning from 1 to 27 years across these sites (2210 site-year). The underlying surface types at these sites are diverse, including forests (131 sites), grasslands (80 sites), farmlands (42 sites), shrublands (32 sites), wetlands (25 sites), and other surface cover types (3 sites). Based on the measured sensible heat flux data from 313 stations, filled ET data, net radiation, and ground heat flux, an evaluation of energy balance closure conditions indicates a strong correlation between available energy and turbulent flux. The values of R², MAE, and RMSE are 0.85, 26.37 W/m², and 33.68 W/m², respectively, with an average energy balance closure ratio of 0.73. Therefore, these 313 sites with high-quality filled ET data can be utilized for ET model validation, ET product verification, water demand assessment, and other related tasks.