Evaluation and Uncertainty Analysis of Reference Crop Evapotranspiration Estimation Using Genetic Programming

Iran has been considered as one of the arid and semi-arid regions of the world in terms of climatic conditions. Limited water resources and inappropriate management of them has fed the agricultural sector with significant challenges. Efficient usage of water in the field, requires accurate estimation of the plant’s water consumption. So far, many studies have been conducted in order to provide new methods for estimation of the reference evapotranspiration (ET0) using intelligent systems. In this study, in addition to evaluation of the efficiency of genetic programming (GP), other models are provided for estimation of evapotranspiration, which are using the minimum amount of meteorological variables. For this purpose, by use of the stepwise regression method, input variables of GP are selected among 7 meteorological variables (i.e., average air temperature, maximum air temperature, minimum air temperature, relative humidity, wind speed at two meters’ height, sunshine hours, and solar radiation). Moreover, eight conventional empirical models are used to compare the performance of empirical models with GP models in the estimation of reference evapotranspiration. In this study, the FAO Penman-Montieth method is considered as the reference method in evaluation of the performances of GP and empirical models. The obtained results show that the GP models have higher accuracy than empirical models. Finally for improving the performance of obtained results, the Bayesian model averaged method is used to combine the results of GP models and to determine their uncertainty bands.