Uncertainty caused by resistance terms in evapotranspiration estimation: a comparison of the Penman-Monteith equation and the three-temperature model

Quantifying uncertainty caused by resistance parameterization is fundamental for understanding, improving, and developing terrestrial evapotranspiration (ET) models. In conjunction with relatively high density eddy covariance (EC) tower observations in a heterogeneous oasis in Northwest China, this study evaluated the effects of resistances on latent heat flux (LE) estimation, the energy equivalent of ET, by: 1) comparing different resistance parameterizations in the PM equation under one- and two-source networks, and 2) comparing a three-temperature (3T) model (without resistance parameterization) and the PM equation (including a couple of resistances). The results showed that the 3T model has less bias in LE estimation than the PM equation, as the mean absolute percent error (MAPE) was 19% for the 3T model but 32% to 39% for the PM equation with different resistance parameterizations. The PM equation-based LE estimates, parameterized using two typical one-source resistance methods and a two-source method, exhibited certain differences. Specifically, when both aerodynamic resistance (r_a) and surface resistance (r_s)were parameterized differently under the one- and two-source networks, uncertainties (defined as the difference between the mean absolute errors) in LE estimates ranged from 4% to 7%. When only r_s was parameterized differently under the one-source network, the uncertainty in the LE estimates was 3%. The results suggest that complicated parameterizations of the resistances in the PM equation can increase the uncertainty, and that the relatively simple two-source 3T model avoids complex parameterization and introduces less uncertainty into the LE estimation.