Premium
Comparison of Two Satellite‐Based Evapotranspiration Models of the Nagqu River Basin of the Tibetan Plateau
Author(s) -
Zou Mijun,
Zhong Lei,
Ma Yaoming,
Hu Yuanyuan,
Huang Ziyu,
Xu Kepiao,
Feng Lu
Publication year - 2018
Publication title -
journal of geophysical research: atmospheres
Language(s) - English
Resource type - Journals
eISSN - 2169-8996
pISSN - 2169-897X
DOI - 10.1002/2017jd027965
Subject(s) - evapotranspiration , shortwave radiation , albedo (alchemy) , environmental science , moderate resolution imaging spectroradiometer , normalized difference vegetation index , energy balance , mean squared error , atmospheric sciences , satellite , climatology , meteorology , geology , mathematics , radiation , climate change , geography , statistics , art , ecology , oceanography , physics , quantum mechanics , performance art , engineering , biology , art history , aerospace engineering
Evapotranspiration (ET) is one of the major uncertain components of the energy and water cycle and was derived for the Nagqu River Basin based on remote sensing data and atmospheric surface layer observations under cloudless conditions. Two process‐based models were used to determine the ET: a Priestley‐Taylor (PT)‐based model and a topographical enhanced surface energy balance system (TESEBS) model. Improved broadband albedo, downward shortwave radiation flux, and reconstructed normalized difference vegetation index (NDVI) were coupled into the TESEBS model and PT‐based model to estimate the actual ET. The atmospheric surface layer meteorological data, SPOT Vegetation (VGT) data, and Moderate Resolution Imaging Spectroradiometer data were used as inputs for 10‐day ET calculations. The model‐estimated results were compared with ground truths calculated via the combinatory method. The results indicated that the ET determined by both models well fit the actual ET, with correlation coefficients ( R ) of 0.88 and 0.82, respectively. However, the TESEBS model showed a better performance than the PT model, with a lower mean bias error (−0.02 mm/hr) and lower root–mean‐square error (0.08 mm/hr). Although the PT model is computationally simple and requires few parameters, the strong weighting of the NDVI may lead to some overestimations, especially during the growing season.