Medium-resolution (10-100 m) satellite evapotranspiration (ET) products are rapidly advancing agricultural water resources research and management, but underperformance across non-agricultural land cover continues to limit broader hydrologic and ecosystem applications. These inconsistencies are often linked to model structure and representation of ET dynamics across space and time. In extensive natural ecosystems such as forests and shrublands, ET is primarily governed by equilibrium radiative energy exchange, whereas in croplands it is often enhanced by advective energy inputs. While some models represent these processes, recent intercomparison studies highlight persistent performance gaps across land covers. We hypothesize that model structure governing land–atmospheric coupling, rather than sensor limitations alone, remains a primary constraint on medium-resolution ET performance. Here, we introduce a diffusivity-independent equilibrium formulation that removes the need for explicit aerodynamic conductance parameterization and conditionally incorporates aerodynamic enhancement when advection is expected. Landsat thermal and optical observations are integrated with gridded meteorological data within the presented Radiation Advection Diffusivity-independent ET (RADET) modeling framework to predict ET. Performance is evaluated using 145 in situ flux stations across the contiguous United States and intercomparisons with OpenET and MODIS products. Results indicate that RADET achieves comparable performance to leading models in croplands while providing consistent improvements across natural ecosystems, including ∼35% lower mean absolute error and sustained positive Nash-Sutcliffe efficiency where ensemble models often showed reduced skill. Application of satellite-based equilibrium formulations with conditional transport enhancement enables computationally efficient generation of medium-resolution ET with robust cross-land cover performance, advancing research and operational applications emphasized in recent medium-resolution remote sensing initiatives.
Kim et al. (Tue,) studied this question.