Remote sensing has become a data source long ago for estimating evapotranspiration (ET), but often with a dilemma between temporal and spatial resolution. The recent ECOsystem Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS) reaches a satisfactory compromise. It provides 70 m spatial resolution and an average 4-day revisit cycle at different times of a day, making granular analysis of ET in diurnal changes and field-scale a reality. In this study, we assessed one instantaneous ET and two daily ECOSTRESS ET products (ECO3ETPTJPL & ECO3ETALEXI) at site scale, using 31 AmeriFlux towers in six regions spanning the continental US, during the period from July 2018 to September 2021. Assuming that eddy covariance sites within a region share similar climate features, we evaluated the capacity of ECOTSRESS to quantify the ET over heterogeneous landscapes. Generally, DisALEXI-JPL daily ET (R² = 0.45) have stronger correlations with in-situ ET than both PT-JPL instantaneous ET (R² = 0.23) and daily ET (R² = 0.11), while they all tend to overestimate ET in most regions. The diurnal cycles of PT-JPL instantaneous ET and seasonal cycles of two daily ET were visualized site by site; all can show temporal variations of ET. Comparing ECOSTRESS ET across and within the six regions, we found that ECOSTRESS ET accuracy varies by region, regardless of their biomes. Understanding the mechanisms controlling ET and how different data sets are able to capture ET is essential for assessing how land cover changes and human-induced disturbances influence plant water use and stress. In this regard, our study provides a deep evaluation of ECOSTRESS ET, paving the ground for future studies aiming to describe ET across heterogeneous landscapes at a finer scale.