Improvement of the Estimation of Potential Evapotranspiration Using Adjusted Coefficient by M5 Decision Tree Model

Evapotranspiration is one of the basic components in hydrologic balance that is important for the design and management of irrigation systems. This research investigates improvement of accuracy of ET estimation by H-S method based on adjusted coefficient of K. This coefficient, the ratio of evapotranspiration estimated by F-P-M method to that estimated by H-M method, is determined by M5 Decision Tree Model based on meteorological variables (air temperature, relative humidity, dew point) at three meteorological stations (including Astara, Rasht, Bandar-Anzali). Thirty years period (1360-1390) is used for this research. The data of each station is divided into two parts: eighty percent for training and twenty percent for validation. The estimated adjusted coefficient is multiplied by estimated evapotranspiration with H-S method. The results indicate higher performance of M5 Decision Tree Model relative to Neural Network model. In addition, mean difference between estimated evapotranspiration by two methods decreased from 0.41, 0.55, 0.7 to 0.31, 0.38, 0.28 for Astara, Bandar-Anzali and Rasht stations, respectively.