Estimation of Reference Evapotranspiration (ET₀) Values for Nevşehir Province Using Regression Analysis Methods and Artificial Neural Networks (ANN)

In this study, reference evapotranspiration (ET₀) values for Nevşehir Province were calculated using the FAO Penman–Monteith method based on 47 years of meteorological data and subsequently estimated using linear (LF), nonlinear (EF, PF) regression models and artificial neural networks (ANN). Within a five-input framework, all regression combinations were tested, and the models were evaluated during the training, validation, and testing phases using the Nash–Sutcliffe efficiency (NS), root mean square error (RMSE), and mean absolute error (MAE) criteria. The results indicated that among the regression-based models, the power function (PF) exhibited the highest performance, with NS values of 0.987, 0.981, and 0.954, respectively. Although the linear model demonstrated strong agreement, it showed lower performance compared to the nonlinear models. The exponential function (EF) yielded the lowest prediction accuracy. The ANN model outperformed all other approaches across all datasets, producing the highest NS values (0.996, 0.995, and 0.991), and thus emerged as the most successful method. Furthermore, month, wind speed, mean minimum temperature, daily sunshine duration, and number of days were identified as the most influential variables in ET₀ estimation. Overall, the findings demonstrate that nonlinear approaches and ANN provide superior accuracy in predicting reference evapotranspiration.