Optimal Alternative for Quantifying Reference Evapotranspiration in Northern Xinjiang

Abstract

Accurate estimation of reference evapotranspiration is a key step in irrigation and water resources planning. The Penman Monteith (FAO56-PM) formula recommended by FAO56-PM is the standard for calculating the reference evapotranspiration. However, the FAO56-PM model is limited in the observation of meteorological variables, so it is necessary to choose an alternative ET0 model which requires less meteorological data. Based on the daily climate data of eight meteorological stations in northern Xinjiang from 2000 to 2020, seven empirical models (Hargreaves, Berti, Dorji, Dalton, Meyer, WMO, Albrecht) and four optimization algorithms (RF model, LS-SVR model, Bi-LSTM model and GA-BP model) combined with seven different parameters were evaluated comprehensively. The results show that the accurate of the empirical model based on temperature is obviously better than the empirical model based on air mass transport. The annual and multi-year alternative ET0 models of different input parameter combinations are: LS-SVR1, RF2, LS-SVR3, LS-SVR4, GA-BP5, LS-SVR6, GA-BP7. It can be used as a substitute for the reference evapotranspiration model without relevant meteorological data. Only the LS-SVR6 model and GA-BP7 model are recommended as the best alternative models for northern Xinjiang reference evapotranspiration at daily, monthly and seasonal scales.

Keywords

• reference evapotranspiration
• empirical model
• regression prediction algorithms
• optimal alternative