Comparative evaluation of four actual evapotranspiration models over different ecosystems and climate zones in China

To better understand the discrepancies in evapotranspiration (ET) simulations between ET models, we intercompared four models in China: Priestley–Taylor Jet Propulsion Laboratory (PT-JPL), Penman–Montieth–Leuning Version 2 (PML-V2), Sigmoid Generalized Complementary Function (SGCF), Mapping Evapotranspiration at High Resolution with Internalized Calibration (METRIC). Data from 18 flux sites were used to evaluate the model performance at daytime (when incident shortwave radiation is greater than 20 W/m2) scales. To compare more fairly, we took the intersection of the outputs from four models for the analyses in the main text. All models yielded acceptable results, with PML-V2 or SGCF performing best at most sites. The average coefficient of determination and root mean square error among all sites of LE (latent heat of ET) were 0.72 and 51.71 W/m2 for PT-JPL, 0.80 and 46.65 W/m2 for PML-V2, 0.79 and 41.13 W/m2 for SGCF, 0.70 and 51.09 W/m2 for METRIC. PT-JPL and PML-V2 underestimated ET at most sites, whereas SGCF overestimated, potentially due to uncertainties in the vegetation indices and ET constraint parameters. Compared to measurements, PT-JPL underestimated the proportion of transpiration to evapotranspiration (0.81 versus 0.59), while PML-V2 overestimated (0.81 versus 0.90). Furthermore, all models performed best in the semi-arid zone dominated by grassland sites.