Estimating daily seamless 20-m resolution evapotranspiration using data fusion and TSEB

Daily seamless estimation of evapotranspiration (ET) at high spatial resolution is essential for field-scale water resource management. Satellite-based ET mapping enables consistent estimation from regional to field scales, however, the inherent trade-off between spatial and temporal resolution in current remote sensing data, along with data gaps caused by weather conditions, constrains high-frequency and spatiotemporally continuous ET estimation. To overcome these limitations, this study developed an efficient and innovative framework that integrates a cloud-filling algorithm, a high-performance spatiotemporal fusion model, multi-source data, and the Two-Source Energy Balance (TSEB) model to produce high-precision, daily seamless ET estimates at 20 m resolution. Specifically, using the modified neighborhood similar pixel interpolator (MNSPI) and the GPU-enabled enhanced spatial and temporal adaptive reflectance fusion model (cuESTARFM), we efficiently integrated the China Land Data Assimilation System (CLDAS), Visible Infrared Imaging Radiometer Suite (VIIRS), and Sentinel-2/3 data to derive daily seamless 20 m land surface parameters, including land surface temperature, leaf area index, and fractional vegetation cover. These parameters, together with meteorological forcing and auxiliary data, were used to drive the TSEB model to generate daily seamless 20 m ET estimates at the irrigation district scale from 2019 to 2023. The simulated instantaneous latent heat flux achieved R², BIAS, and RMSE values of 0.77, 2.99 W/m², and 74.61 W/m², respectively, compared with ground observations, while the daily ET estimates achieved corresponding values of 0.56, –0.08 mm/d, and 1.05 mm/d. This framework offers novel insights into ET mapping through multi-source data fusion and is of great significance for achieving precise and dynamic agricultural water resource management.