Optimization of the time scale improvement method for grape evapotranspiration in greenhouses

Abstract: Grapes Evapotranspiration (ET) and its transformation are of great significance at different time scales under a greenhouse planting environment in the cold region of Northeast China. In this study, the ET process and environmental factors of greenhouse grapes were monitored continuously for two years, in order to optimize the time-scale improvement models. Three models were also used to enhance the temporal scale of grapes from the instantaneous to the daily scale, and from the daily to the whole growth period ET. The results showed that the key parameters of improved the Evaporative Fraction (EF'), Evaporative Fraction (EF), and crop coefficient. Among them, the EF', EF, and crop coefficient were used to elevate the ET from the instantaneous to the daily scale. There was a decrease at first and then an increase throughout the day, where the stable was from 8:00 to 16:00. Specifically, the mean values of the study period were 0.54, 0.52, and 0.76, respectively, while the mean values of the coefficients of variation were 0.11, 0.10, and 0.09, respectively. The ET increased from the day to the growth period, where the average values of the three key parameters were 0.60, 0.61, and 1.37, respectively. There was also no outstanding change during the study period. The highest simulation accuracy was achieved in the three instantaneous scale models of ET improvement in the fruit expansion period, compared with the rest growth periods. The optimal models were determined by various evaluation indicators, including the Relative Root Mean Square Error (RRMSE), efficiency coefficient, determination coefficient (R2), and Mean Absolute Error (MAE). The optimal scale improvement time of the three models was mainly concentrated from 11:00 to 13:30. The optimum scaling time of greenhouse grape was determined by GPI comprehensive evaluation in the period of the shoot growth, flowering and fruit setting, fruit expansion and maturity. The optimum expansion time of EF' was 11:30, 11:30, 11:30, and 12:30, respectively. The optimum expansion time of EF was 12:30, 11:30, 11:30, and 12:00, respectively, while that of crop coefficient was 12:30, 12:00, 12:30, and 12:30, respectively. Three models were used to improve the instantaneous to daily scale ET of greenhouse grapes using the determined optimal scale improvement time. All three models underestimated the daily ET, where the underestimation rate ranged from 19.68% to 30.05%. The simulation accuracy was highest for EF method and lowest for the crop coefficient method. Among them, the EF presented the highest accuracy, where the R2 values reached 0.92 and 0.89 in 2020 and 2021, respectively, whereas, the RRMSE was only 20.23%, and 21.49%, respectively. The ET scale increased from the day to the whole growth period, indicating the underestimation in the shoot growth period, the flowering, and the fruit setting period. There was an overestimation in the fruit expansion and maturity period. Consequently, the date of fruit expansion can be expected to simulate the increase of ET in the whole growth period of grapes. The simulation accuracy was highest for EF method and lowest for the crop coefficient method. Specifically, the EF model presented the highest simulation accuracy, where the simulation errors were only 1.8 and 7.4 mm in 2020 and 2021, respectively, whereas the relative errors were only 0.68% and 2.73%, respectively.