作者: Xiaotong Zhang , Shunlin Liang , Aizhong Ye , Wenwen Cai , Jinming Feng
DOI: 10.1016/J.JHYDROL.2015.06.059
关键词: Eddy covariance 、 Future studies 、 Statistics 、 Large model 、 Meteorology 、 Correlation coefficient 、 Water balance 、 Bayesian inference 、 Mean squared error 、 Mathematics 、 Evapotranspiration
摘要: Summary Evapotranspiration (ET) is critical to terrestrial ecosystems as it links the water, carbon, and surface energy exchanges. Numerous ET models were developed for estimations, but there are large model uncertainties. In this study, a Bayesian Model Averaging (BMA) method was used merge eight satellite-based models, including five empirical three process-based improving accuracy of estimates. At twenty-three eddy covariance flux towers, we examined performance on all possible combinations found that an ensemble with four (BMA_Best) showed best performance. The BMA_Best can outperform Kling–Gupta efficiency (KGE) value increased by 4% compared highest KGE, decreased RMSE 4%. Although correlation coefficient less than single model, bias smallest models. Moreover, based water balance principle over river basin scale, validation indicated estimates explain 86% variations. general, results BMA will be very useful future studies characterize regional availability long-time series.