Estimation of daily evapotranspiration in Košice City (Slovakia) using several soft computing techniques

Accurate estimation of evapotranspiration is one of the main aspects of water management. In this study, the capabilities of soft computing techniques for estimating daily evapotranspiration in Košice (Slovakia) were investigated. Daily solar radiation (SR), relative humidity (RH), air temperature (T), and wind speed (U) were the meteorological variables used for modeling. Based on the data, different combinations of multilayer perceptron (MLP), support vector regression (SVR), multilinear regression (MLR) models were generated. Model results are compared with each other and with the Hargreaves-Samani, Ritchie, and Turc empirical equations using three statistical criteria, namely mean square error (MSE), mean absolute relative error (MAE), and determination coefficient (R2). Of the empirical formulas applied, the Hargreaves-Samani equation gave the most compatible results with the Penman FAO 56 equation. Error percentage histograms were generated as a reference criterion. Model results show that the MLP model performs better than the other soft computing techniques used.