A two-source non-parametric method for estimating terrestrial evapotranspiration: Validation at eddy covariance sites

Abstract

Developing a two-source evapotranspiration (ET) method is a major challenge of the accurate ET estimation. As a single source ET method, the performance of Non-Parametric (NP) approach is limited by the unavailability of conventional equation of equilibrium ET in the water-limited situation. To solve this problem, a two-source equilibrium ET equation derived from Penman–Monteith equation expressed using relative humidity (RH-PM) is introduced into the NP method, then a Two-Source Non-Parametric (TS-NP) method is proposed regarding vegetation and soil as individual source. The accuracy of equilibrium ET derived from two-source method has significantly improved in the bare-soil surface (relative error: 31.26 %; RMSE: 13.18 W/m²) compared with that of conventional equilibrium ET equation and surface flux equilibrium (SFE) theory. Validated by eddy covariance tower sites, the performance of TS-NP method on bare soil surfaces is satisfactory, involving a significant reduction (near to 1/2) in the ET estimation error (relative error: 49.45 %; RMSE: 16.15 W/m²) compared with that of NP, SFE-NP and RH-PM method. On the dense vegetation surface, the performance of TS-NP method (relative error: 8.44 %; RMSE: 17.44 W/m²) is also slightly better than that of the NP method, SFE-NP and RH-PM method. In addition, air temperature and surface temperature are the most sensitive input variables in the TS-NP method, particularly in bare soil surface. The ET derived from the combined TS-NP method shows the best performance in arid areas (relative error: 22.20 %; RMSE: 16.52 W/m²) and non-arid areas (relative error: 8.12 %; RMSE: 17.61 W/m²), and the combined TS-NP method has satisfactory accuracy on the heterogeneity of underlying. Therefore, the TS-NP method provides a simple but efficient two-source method for high-precision ET estimation.