Estimating evapotranspiration at high spatio-temporal resolution based on the Bayesian model averaging method in Baoding Plain, China

Evapotranspiration (ET) is a core parameter of the hydrology and carbon cycles, and its accurate estimation is crucial for water resource management. Satellite-based ET products provide an effective means for large-scale monitoring. However, due to limitations in the spatial and temporal resolution, the use of these products at regional and field scales is limited. In this study, daily ET in the Baoding Plain—a key groundwater resource recharge area—was estimated at a 500 m spatial resolution using the Bayesian Model Averaging (BMA) method. The model was driven by a synthesis of remote sensing datasets, reanalysis products, and interpolated data from meteorological stations. Validation results from in-situ observations indicated that the BMA ET had better performance (R=0.83, RMSE=1.25 mm/d) than each model in the BMA scheme. The spatiotemporal analysis revealed that the average annual BMA ET in the Baoding Plain was 683 mm/year from 2000 to 2019. Seasonal and monthly variations in the BMA ET captured the irrigation and water consumption patterns of the local crop rotation systems. A significant increasing trend of BMA ET (2.40 mm/year²) was observed in the Baoding Plain over the study period. At the regional scale, ET over more than 50% of the plain exhibited a significant positive trend. Further analysis identified water availability, solar radiation, and temperature as the primary drivers of ET variation. The BMA ET product generated in this study is characterized by high spatiotemporal resolution and accuracy. This reliable, high-resolution dataset offers valuable support for precision agricultural water management and hydrological studies, including groundwater investigations, in this predominantly agricultural region.