3.2 Comparison of daily ET between reference evapotranspiration models and lysimeter measurements during the study period
The comparison of 13 reference evapotranspiration equations against the lysimeter measurements is presented in Fig.3 and Table 2, showing that the relationship between daily ET0 calculated by the 13 reference evapotranspiration equations and lysimeter measurements are each significantly (P <0.01), with higher coefficients of determination (R 2) ranging from 0.61 to 0.81. For the combination models, FAO 24 Penman yielded the highest correlation, followed by Pen-63 and FAO-56 PM. For radiation-based models, PT and DK obtained the highest correlation, followed by Makkink, Makkink(1967) and Makkink (1957), IRMAK1 and IRMAK2. For temperature-based models, HAR, HAR1 and HAR2 obtained the same correlation. The daily estimates of combination models and radiation-based models (except Makkin and Makkin (1967)) generally underestimated the ET values measured by lysimeter: the mean difference in ET between lysimeter measurements and model estimates (i.e. MBE) ranged from -0.47 mm d−1 to -0.04 mm d−1 for combination models, and ranged from -0.28 mm d−1 to -0.02 mm d−1 for radiation-based models. However, the temperature-based models generally overestimated the ET values measured by lysimeter, with MBEs ranging from 0.10 mm d−1to 0.39 mm d−1. Overall, the combination models and radiation-based models, on average, underestimated ET by 12.79% and 5.93%, respectively, but the temperature-based models overestimated ET by an average of 7.99%.
During the whole study period, the RMSE of combination models ranged from 1.19 to 1.36 mm d−1 and averaged 1.26 mm d−1, the RMSE for radiation-based models ranged from 1.03 to 1.47mm d−1and averaged 1.09 mm d−1, and the RMSE for temperature-based models ranged from 1.28 to 1.32 mm d−1 and averaged 1.29 mm d−1. Based on the RMSE, the performances of the 13 reference evapotranspiration models decreased in the order: DK>PT>Pen-63>FAO-24 Pen>HAR>HAR2>HAR1>IRMAK1>FAO-56PM >Makkink>Makkink(1967)>Makkink(1957)>IRMAK2. The best model (DK) was, respectively, 49% and 33% more accurate than the poorest (BR) and the commonly used FAO-56 PM equation. Furthermore, the Pen-63 and FAO-24 Pen demonstrated better performance than the commonly used FAO-56 PM equation. Overall, for the whole study period, the radiation-based models yielded the best performance, followed by the combination models and temperature-based models.