Uncertainty caused by resistances in evapotranspiration

Abstract. Quantifying the uncertainties induced by resistance parameterization is fundamental to understanding, improving, and developing terrestrial evapotranspiration (ET) models. Using high-density eddy covariance (EC) tower observations in a heterogeneous oasis in Northwest China, this study evaluates the impact of resistances on latent heat flux (LE) estimations, the energy equivalent of ET, by comparing resistance parameterizations with varied complexity under one- and two-source Penman-Monteith (PM) equations. We then discuss possible solutions for reducing such uncertainties by employing a three-temperature (3T) model, which does not explicitly include resistance-related parameters. The results show that the mean absolute percent error (MAPE) varied from 32 % to 39 % for the LE estimates from the one- and two-source PM equations. When only surface resistance (r_s) was parameterized under the one-source network, then the uncertainty (defined as the difference between MAPEs) dropped to 12 %. When both r_s and aerodynamic resistance (r_a) were parameterized differently under the one- and two-source networks, then the uncertainties in the estimates were 11~23 %, emphasizing that multiple resistances add uncertainties. Additionally, the 3T model performed better than the PM equations, with MAPE of 19 %. The results suggest that 1) although prior calibration of the parameters required in resistance estimations can improve the PM-based LE estimates, resistance parameterization process can generate obvious uncertainties, 2) more complex resistance parameterizations leads to more uncertainty in the LE estimation, and 3) the relatively simple 3T model avoids resistance parameterization, thus introducing less uncertainty in the LE estimation.