3.4 Comparison of daily ET between reference evapotranspiration models and lysimeter measurements during the non-growing season
During the non-growing season, the daily ET0 calculated by the 13 reference evapotranspiration equations was also significantly correlated with the lysimeter measurements (P<0.01), but with lower coefficients of determination (R2) ranging from 0.16 to 0.43 (Fig. 5 and Table 4). Of the combination models, FAO-24 Pen obtained the highest R2, followed by FAO-56 PM and Pen-63. Of the radiation-based models, PT obtained the highest R2, followed by DK, Makkink (1967), Makkink (1957), Makkink, IRMAK1 and IRMAK2. Of the temperature-based models, HAR1 and HAR2 obtained the highest R2, then followed by HAR. Interestingly, all model generally overestimated the ET values measured by lysimeter during the non-growing season, with MBEs ranging from 0.30 to 1.20 mm d−1 and averaging 0.69 mm d−1 over the 13 models; Makkink(1967) yielded the largest underestimate (by 64.84%) and PT the minimum underestimate (by 31.37%) (Table 3).
The RMSE for combination models ranged from 0.78 to 1.02 mm d−1 and averaged 0.87 mm d−1, the RMSE for radiation-based models ranged from 0.69 to 1.43 mm d−1 and averaged 1.09 mm d−1, and the RMSE for temperature-based models ranged from 1.08 to 1.19 mm d−1 and averaged 1.13 mm d−1. Based on the RMSE, the performances of the reference evapotranspiration models decreased in the order: PT > DK > FAO-24Pen > FAO-56 > Pen-63 > IRMAK1 > HAR > Makkink (1957) > HAR2 > HAR1 > Makkink > IRMAK2 > Makkink (1967). Evidently, the best (PT) was 93.24% more accurate than the poorest (BR). Overall, the combination models yielded the best performance during the growing season, followed by radiation-based models and temperature-based models.