Modelling reference evapotranspiration by combining neuro-fuzzy and evolutionary strategies

Abstract

This study investigates the potential of two evolutionary neuro-fuzzy inference systems, adaptive neuro-fuzzy inference system (ANFIS) with particle swarm optimization (ANFIS–PSO) and genetic algorithm (ANFIS–GA), in modelling reference evapotranspiration (ET0). The hybrid models were tested using Nash–Sutcliffe efficiency, root mean square errors and determination coefficient (R2) statistics and compared with classical ANFIS, artificial neural networks (ANNs) and classification and regression tree (CART). Various combinations of monthly weather data of solar radiation, relative humidity, average air temperature and wind speed gotten from two stations, Antalya and Isparta, Turkey, were used as input parameters to the developed models to estimate ET0. The recommended evolutionary neuro-fuzzy models produced better estimates compared to ANFIS, ANN and CART in modelling monthly ET0. The ANFIS–PSO and/or ANFIS–GA improved the accuracy of ANFIS, ANN and CART by 40%, 32% and 66% for the Antalya and by 14%, 44% and 67% for the Isparta, respectively.