Assessment of Evapotranspiration from a Noah-MP Ensemble with Multisource Products under Different Land-Cover Types over the Contiguous United States

Abstract Terrestrial evapotranspiration (ET) is the total water flux transported from the soil and vegetation to the atmosphere. Accurate ET modeling can greatly facilitate water resource management and climate projection. This study focused on evaluating the performance of a Noah land surface model with multiparameterization options (Noah-MP) ensemble in simulating ET under eight main land-cover types over the contiguous United States. For this purpose, the ensemble was cross-compared with phase 2 of the North American Land Data Assimilation System (NLDAS) models, gridded upscaled flux network (FLUXNET) ET, Global Land Evaporation Amsterdam Model (GLEAM) ET, and complementary-relationship-based ET at multiyear averaged, annual, and interannual scales. At the multiyear scale, the Noah-MP ensemble mean showed overestimation and underperformed the NLDAS ensemble mean. All models exhibited more biases in deciduous forests, grasslands, croplands, and barren soil regions. At the annual scale, Noah-MP was able to capture the timings of summer ET peaks; however, it overestimated the magnitude of ET peaks in forests, grasslands, and croplands. At the interannual scale, all models performed relatively well in shrublands, grasslands, and barren regions but poorly in forests, savannas, and croplands. Among them, the Noah-MP ensemble mean performed best in forests and the NLDAS ensemble mean performed best in savannas and croplands. Sobol’ sensitivity analysis of the Noah-MP ensemble revealed that stomatal conductance dominates ET in growing seasons in forests and grasslands and runoff dominates ET in shrublands, savannas, croplands, and barren soil regions. The above findings will be beneficial to the improvement of ET-related parameterizations and the optimization of ensemble strategies. Significance Statement The Noah land surface model with multiparameterization options (Noah-MP) features multiple options for one process, while much is known about the performance of the default configuration in simulating ET. Few studies have examined how its multioption ensemble performs under different land-cover types over thirty years. This study evaluated and compared an ET-relevant multiparameterization Noah-MP ensemble with a broad range of ET products at different temporal scales over the contiguous United States. Besides, the parameterization sensitivity to the ET estimation was quantified. Our results provide a better understanding of the strengths and weaknesses of Noah-MP and point the way toward model improvement and ensemble optimization.