干旱区生态环境知识资源中心

Reliable estimation of regional evapotranspiration remains as a challenge. In our study, based on a nonparametric evapotranspiration approach, a retrieval algorithm using the MODIS products and climate drive datasets was proposed to retrieve evapotranspiration, in terms of latent heat flux (LE) over a semi-arid region under all sky. Surface observations were regarded as the validation references, which were obtained from six eddy-covariance sites with land covers of desert, Gobi, village, orchard, vegetable and wetland. Our results showed that the distribution of retrieved LE matched well with the desert-oasis ecosystems. And the accuracy of instantaneous LE retrievals varied with sites with a R-2 from 0.19 to 0.63, a bias from -129.46 W/m(2) to 56.73 W/m(2), a relative error from 5.23% to 29.18%, and a RMSE from 95.44 W/m(2) to 150.29 W/m(2). The validation supports the applicability of the algorithm in the arid/semi-arid region.