4.2 The performance comparison of radiation and temperature models against lysimeter measurements
For the performance comparison of radiation models against lysimeter measurements, we found that the PT and DK models yielded the best performances of the radiation-based models, and were respectively 44.11% and 42.71% more accurate than the poorest (IRMAK2) model, which was in line with the previous study conducted in humid areas that the PT method exert the good accuracy estimate for ET (Trajkovic & Kolakovic, 2009). There is increasing evidence indicated that the input parameters was the dominated factors affecting their performance (Lang et al., 2017), we thus conclude that the better performances of the PT and DK models which might associated with the use of only the most important meteorological factors affecting ET such as net radiation (Rn), was supported by our results (Fig.6). Compare with PT and DK model, the other radiation models just use the Rs as the mainly driving variable, the ET thus may overestimate because some Rs was reduced through reflecting into the atmosphere due to the high albedo in this region (Zhang et al., 2018). Furthermore, each model was developed from its specific underlying surface and climate conditions. For instance, the PT was established in a humid climate condition, which was suitable for our humid alpine meadow. Most importantly, the PT and DK models required fewer meteorological inputs when compared with combination models. Combine those factors, we can prefer to recommend the PT and DK model for use in a humid alpine meadow on the northeastern Qinghai-Tibetan Plateau, especially when considering the difficulty in obtaining ET in this harsh climate.
For the performance comparison of temperature models against lysimeter measurements, previous study reported that the Hargreaves versions equation as one of the most simple empirical methods were widely used for the ET estimation due to its less meteorological data input, especially considering not all the meteorological data required in the standard PM-56 model (Jensen et al., 1990). To further select the best Hargreaves versions equation, we compare the performance of original (HAR) and two modified versions (HAR1 and HAR2) of the Hargreaves equations, and found that the original HAR model had the lowest error (RMSE =1.28 mm d−1, MBE=0.17 mm d−1 and PE=5.92%), which was consistent with previous studies conducted on humid region (Hossein Tabari, 2010) but contrast to these study conducted in arid region that the modified Hargreaves equation display a more accurate estimation of evapotranspiration compared with the original Hargreaves equation (Ravazzani, Corbari, Morella, Gianoli, & Mancini, 2011). Overall, the temperature models display a poor performance compared with radiation models owing to the Hargreaves method was established in semiarid areas (Hossein Tabari, 2010), thus a local calibrations was required to improve the accuracy of Hargreaves method in other region.