Comparison of the Efficiencies of the Prognostic Generalized Complementary Functions on Evaporation Estimation

Abstract

AbstractSeveral prognostic generalized complementary functions have been proposed to estimate evaporation in recent years. However, there are few comparative studies about the performance of these predictable complementary functions at the global scale. This study compared the efficiencies of four prognostic generalized complementary functions on evaporation estimation at 195 eddy covariance (EC) sites. The four complementary functions include the polynomial function proposed by Brutsaert (2015, B2015, https://doi.org/10.1002/2015wr017720), the rescaling functions proposed by Crago et al. (2016, C2016, https://doi.org/10.1002/2016WR019753) and by Szilagyi et al. (2017, S2017, https://doi.org/10.1002/2016JD025611), and the sigmoid function proposed by Han and Tian (2018, H2018, https://doi.org/10.1029/2017wr021755). The results show that the four prognostic generalized complementary functions can provide an acceptable estimation of E (Pearson correlation coefficient ≈ 0.8; root‐mean‐square error <30 W m−2) in most of the global EC sites. For different ecosystem types, C2016 performs better in forests, shrublands, and grasslands, and H2018 performs better in wetlands and croplands. Each function has specific advantages. B2015 is popular because of its concise functional form; C2016 and S2017 are fully calibration‐free algorithms and have rescaling processes for each observation, and C2016 performs best on the monthly E estimation; H2018 suits for site mean estimation and has the most stable performance among all ecosystem types. However, all the functions have limitations, and there is still room for improvement. This study provided evidence for previous theoretical studies about the rationality of the complementary functions from the view of numerical simulation, and can help the wide application of the complementary functions on regional and global E estimation.