Combining UAV remote sensing data to estimate seasonal actual evapotranspiration and surface conductance in rice paddies

Efficient irrigation management requires actual crop evapotranspiration (ET c,act ) information with high spatial and temporal resolution at farmland scale. This study evaluated the performance of estimating rice ET c,act using thermal infrared and multispectral imagery acquired from unmanned aerial vehicles (UAVs) within the framework of Surface Energy Balance System (SEBS) model combined with a temporal upscaling approach. A surface conductance (Gs) model based on UAV remote sensing was developed for temporal upscaling. The proposed Gs model explicitly accounted for the influences of phenology, LAI, radiation and potential stomatal conductance, with the latter calibrated using Gs derived from ET c,act retrievals. Field experiments were conducted under different ponding water depth treatments in 2022 and panicle fertilization treatments in 2023, during which ET c,act was measured using micro-lysimeters to validate model performance. Results showed that ET c,act estimated by the UAV-based SEBS model agreed well with field observations ( R² = 0.89, RMSE = 1.06 mm d −1 ). The simulated surface conductance also closely matched the observations ( R² = 0.64–0.84, RMS E = 0.0028–0.0038 m s −1 ). The Gs-based temporal upscaling approach achieved high accuracy in daily ET c,act estimates ( R² = 0.95, RMSE = 0.70 mm d −1 ), outperforming the crop coefficient (Kc)-based approach. Seasonal simulations indicated that total ET c,act followed the order of 6 cm (552 mm) > 10 cm (546 mm) > 2 cm (531 mm) in 2022 and 125% (526 mm) > 100% (521 mm) > 75% (515 mm) nitrogen fertilization in 2023. Overall, this study demonstrates that integrating UAV-based SEBS model with a physically based Gs model provides a reliable approach for generating continuous, high-resolution ET c,act estimates, supporting precision irrigation water management in rice paddies.