Evaluating the complementary relationship for estimating evapotranspiration using the multi-site data across north China

The ability to predict actual evapotranspiration flux (lambda E-a) by physically based evaporanspiration (ET) model is limited globally due to the difficulty in validating the site-specific model parameters. Thus, the approaches for estimating lambda E-a using only routine meteorological variables play a critical role in understanding and predicting hydrological cycle in the context of climate change. In this study, the performance of a complementary relationship (CR) method (Granger and Gray, 1989; GG model) on different timescales (daily and half-hourly) was evaluated using a high-quality dataset of selected 12 eddy covariance flux towers, which encompassed a number of cropland, grassland, evergreen needleleaf forest, desert shrub and wetland sites across northern China. The results indicated that the GG model is applicable in estimating daily lambda E-a for most ecosystems across northern China. However, significant underestimations of daily lambda E-a were found for the croplands (Daman and Dunhuang sites) and the desert shrub (Ejina) in the arid northwest China, which may be attributed to the enhanced lambda E-a by horizontal advection and the deep root water-uptake, respectively. By using the Monin-Obukhov similarity theory with a surface energy balance constraint, the model performance on half-hourly timescale was satisfactory for the 12 tower sites with R-2 ranging from 0.54 to 0.81 and the slopes of Deming regression line between measured and simulated lambda E-a from 0.77 to 1.14. Indeed, the study highlights the need for further investigation of the timescale dependence of the CR-based ET models. (C) 2016 Elsevier B.V. All rights reserved.