Comparative analysis of data‐driven methods for daily reference evapotranspiration estimation of Southern Caspian Sea

Abstract

AbstractReference evapotranspiration (ETo) as a component in the hydrological cycle is calculated using many methods. In this study, the capability of four data‐driven methods including artificial neural network (ANN), adaptive neuro‐fuzzy inference system (ANFIS), support vector machine (SVM), and M5 tree model has been evaluated for daily ETo estimation at the south of the Caspian Sea. Different combinations of climatic data, solar radiation (Rs), mean air temperature (Tmean), mean relative humidity (RHmean), and wind speed (U) during 1991–2020 were used as input variables. The data were divided into training and test data. The values generated from the methods were compared with those of the FAO‐56 Penman‐Monteith as a standard method. The results indicated that the accuracy of ANFIS was increased for estimating ETo, especially in validation phase, when all climatic variables were used as inputs in the synoptic stations. Totally, based on the evaluation of the performances, it was created that the ANFIS with Tmean, RHmean, Rs, and U variables had the best accuracy, while the ANN, SVM, and M5 with only one input of U had the worst performance. The ANFIS with Tmean and Rs was recommended for modelling ETo if there are fewer climatic variables in these regions.