Evapotranspiration inversion using a two-sources model coupling multiscale data fusion and interpolation methods

Continuous monitoring of high spatiotemporal resolution evapotranspiration (ET) is crucial for accurately assessing water resource management and irrigation efficiency at both global and regional scales. However, constraints such as satellite image transit time and cloud contamination can inhibit a single satellite to provide fine spatial resolution and continuous daily sequences of remote sensing data. In this study, we employed a coupled multi-scale fusion and interpolation method evapotranspiration model (CFIEM) framework integrated the remote sensing ET models, data fusion models, and the HANTS-GEE interpolation method to generate high-resolution surface feature parameters and hydrological variables. Landsat remote sensing parameters and MODIS-Landsat fusion data were adopted to simulate daily regional ET, and the simulated results were subsequently validated. The results revealed the strong consistency between the simulated CFIEM-ET and Landsat-ET values and measured data, with the R 2 values of 0.83 and 0.73. Among the simulations across the sand dunes, grasslands, rice, and maize ecosystems, maize exhibited the most accurate simulation ( R 2 = 0.84), while the sand dunes demonstrated certain deviation ( R 2 = 0.72). The geographical detector analysis identified the net radiation ( R n) as the primary driver of multi-year ET variation, followed by the land surface temperature (LST) and leaf area index (LAI). This study sheds light on the interannual ET variation and its driving mechanisms in arid regions, laying a theoretical foundation and offering technical support for regional water resource management and desertification control. • A coupled multiscale, fusion, and interpolation method evapotranspiration model (CFIEM) framework is established. • CFIEM-ET shows an improvement of 10% in simulation effectiveness compared to Landsat-ET, with a reduction of 0.15mm/d in RMSE. • The main driving factors of interannual evapotranspiration in the study area from 2018 to 2022 are net radiation, land surface temperature, and leaf area index.