Performance evaluation of linear and nonlinear models for the estimation of reference evapotranspiration. (2018)

• Record Type: Journal Article
• Authors: Goodarzi, Mustafa; Eslamian, Saeid
• Is Part Of: International journal of hydrology science and technology. Volume 8:Number 1(2018)
• Page Start: 1
• Page End: 15
• Publication Date: 2018
• Subjects: reference evapotranspiration -- artificial neural networks -- ANNs -- genetic programming multiple linear regression -- Penman-Monteith; Hydrology -- Periodicals; Water resources development -- Periodicals; Hydrology; 553.705
• Languages: English
• ISSNs: 2042-7808

Abstract

In this study, the performance of linear and nonlinear models for the estimation of reference evapotranspiration was examined. To evaluate the performance of nonlinear models, we used the radial basis function (RBF) neural networks and genetic programming (GP), and the multiple linear regression (MLR) method was used for linear models. Using these three methods, monthly evapotranspiration was calculated for Isfahan region in a 26-year period. Comparison of the results for nonlinear and linear models showed that the GP3 model by the coefficient of determination of 0.99 and root mean square error (RMSE) of 0.21, has the best performance among the studied models. Instead, the RBF model training speed is higher than the GP model. Furthermore, the results showed that the MLR model has good performance in estimating evapotranspiration and there is no significant difference between the accuracy of the MLR and RBF method, but the accuracy of GP model is better than the RBF and MLR models. The results showed that the reference evapotranspiration could be estimated with high accuracy by both linear and nonlinear models for the study area.