Accurate estimation of daily evapotranspiration (ET) at the field scale is essential for agricultural water management, particularly in arid and semi-arid regions, yet existing satellite products often suffer from spatiotemporal trade-offs. To overcome this limitation, we generated high-resolution daily ET data using two data fusion approaches based on the unbiased variant of Enhanced Spatial and Temporal Adaptive Reflectance Fusion Model (ubESTARFM). In the first approach (LST-fused ET), ubESTARFM was used to generate high spatio-temporal resolution land surface temperature (LST) data, which served as the key input for the soil moisture-coupled Two-Source Energy Balance (TSEB-SM) model to estimate high-resolution daily ET. In the second approach (Fused ET), high-resolution daily ET data were directly generated by applying ubESTARFM to fuse ET products derived from MODIS and Landsat observations using the TSEB-SM model. Results showed that LST-fused ET agreed better with eddy covariance (EC) observations, yielding a lower RMSE of 0.469 mm/day (compared to 0.567 mm/day for Fused ET) and a significantly smaller systematic bias (-0.149 mm/day vs. -0.430 mm/day) at the relatively heterogeneous Boyagin site. Furthermore, LST-fused ET demonstrated superior spatial consistency with ECOSTRESS results as benchmarks over heterogeneous surfaces, achieving a significantly lower MAPE of 0.49% compared to 9.53% for Fused ET. This limitation of Fused ET, primarily attributed to pixel-matching biases, could be mitigated by incorporating dynamic, high-resolution LAI into the fusion process as a structural constraint, thereby improving accuracy while maintaining efficiency. Moving forward, improving the efficiency and accuracy of the Fused ET could provide a pragmatic and scalable pathway for large-area, field-scale daily ET mapping, supporting agricultural water-use monitoring and water resource management in arid and semi-arid regions.