Comparison of three dual‐source remote sensing evapotranspiration models during the MUSOEXE‐12 campaign: Revisit of model physics

Various remote sensing‐based terrestrial evapotranspiration (ET) models have been developed during the past four decades. These models vary in conceptual and mathematical representations of the physics, consequently leading to different performances. Examination of uncertainties associated with limitations in model physics will be useful for model selection and improvement. Here, three dual‐source remote sensing ET models (i.e., the Hybrid dual‐source scheme and Trapezoid framework‐based ET Model (HTEM), the Two‐Source Energy Balance (TSEB) model, and the MOD16 ET algorithm) using ASTER images were compared during the MUSOEXE‐12 campaign in the Heihe River Basin in Northwest China, aiming to better understand the differences in model physics that potentially lead to differences in model performance. Model results were first compared against observations from a dense network of eddy covariance towers and isotope‐based evaporation (E) and transpiration (T) partitioning. Results show that HTEM outperformed the other two models in simulating ET and its partitioning, whereas MOD16 performed worst (i.e., ET root‐mean‐square errors are 42.3 W/m2 (HTEM), 49.8 W/m2 (TSEB), and 95.3 W/m2 (MOD16)). On to model limitations, HTEM tends to underestimate ET under high advection due mostly to the underestimation of temperatures for the wet edge in its trapezoidal space. For TSEB, large uncertainties occur in determining the initial Priestley‐Taylor coefficient and the iteration procedure for ET partitioning, leading to overestimation/underestimation of T/E in most cases, particularly over sparse vegetation. Primary use of meteorological data for MOD16 does not effectively capture the soil moisture restriction on ET, and therefore results in unreasonable spatial ET patterns.