Estimation of reference evaporation amount with fuzzy SMRGT, ANFİS and multiple linear regression methods

Accurate estimation of evapotranspiration is critical for developing, controlling and managing water resources. Considering that water resources are decreasing day by day, losses due to evaporation are of increasing importance. In this study, daily evapotranspiration was estimated. The data to be used in the study were obtained from Adana Regional Station No. 17351 operated by the General Directorate of Meteorology (MGM). The data covers daily average temperature (TM), relative humidity (RH), wind speed (U) and solar radiation (SR) values between 2015-2021. A total of 1966 daily data was used, 75% of which (1476 daily data) was used in the training phase and 25% (490 daily data) was used in the testing phase. Different combinations were created with meteorological data. These combinations were organized as input parameters. Penman Monteith method was used as a reference for evapotranspiration estimation. (ET_0) Daily evapotranspiration was estimated using Simple Membership Functions and Fuzzy Rule Generation Technique (Fuzzy SMRGT) and Adaptive Neural Fuzzy Inference System (ANFİS) methods. The results obtained are compared with the results of the classical Multiple Linear Regression (MLR) method. Based on the coefficient of determination (R²) and error rates (MAE, RMSE), which are the metrics considered in all analyses, it is concluded that the ANFİS model is more accurate in predicting evapotranspiration compared to the SMRGT and MLR models. The ANFİS model performed better with a higher coefficient of determination (0.9988) and a lower error rate (MAE: 0.046mm, RMSE: 0.067mm).