Evapotranspiration estimation using support vector machines and Hargreaves-Samani equation for St. Johns, FL, USA

Abstract

Information about Evapotranspiration (ET) calculations are not clear enough even it is an important part of hydrological cycle. There are many parameters which effect ET directly or indirectly such as Solar Radiation (SR) and Air Temperature (AT). In this study authors focused on the modelling ET using Support Vector Machines (SVM) method because this method has abilities to solve nonlinear problems. For the training SVM 1158 daily AT, SR, Wind Speed (U) and Relative Humidity (RH) meteorological parameters are used and model is tested using 385 daily parameters. Data set is taken from St. Johns, Florida, USA weather station. To understand the abilities of SVM for ET prediction against Hargreaves-Samani formula, the test set is applied to this empirical equation. Determination coefficient of SVM with observed daily ET values is calculated as 0.913 and determination coefficient of Hargreaves-Samani formula with observed daily ET is found as 0.910. Comparison between both methods is done using Mean Square Error (MSE), Mean Absolute Error (MEA) and determination coefficient statistics. As a result it is seen that SVM method is trustier than Hargreaves-Samani formula for daily ET prediction.