Determination of wetting patterns’ dimensions is essential in designing and managing surface/subsurface drip irrigation systems. The laboratory experiments were conducted using physical model with dimensions of 3×1×0.5 m3 to evaluate the moisture redistribution process under continuous and pulse surface/subsurface irrigation systems. In the present study, the efficiency of a new machine learning method, named fuzzy c-means clustering- based adaptive neural-fuzzy inference system combined with a new meta-heuristic algorithm, hybrid particle swarm optimization – gravity search algorithm (ANFIS-FCM-PSOGSA), is investigated in order to model wetting front redistribution of drip irrigation systems (IS) using soil and system parameters as inputs under continuous and pulse surface/subsurface IS. The outcomes of the assessed method are compared with those of the ANFIS-FCM-PSO, generalized regression neural networks and multivariate adaptive regression splines. In assessing the implemented methods, four commonly used indices, root mean square errors (RMSE), mean absolute error (MAE), coefficient of determination (R2), Nash-Sutcliffe model efficiency (NSE) and graphical methods (e.g., scatter, box plot and Taylor diagrams) are utilized. The benchmark outcomes demonstrate the superiority of new method in estimating wetting front dimensions by improving the accuracy of the ANFIS-FCM-PSO by 29.6%, 18.5%, 6.1%, and 9.0% in estimating the diameter of horizontal redistribution with respect to RMSE, MAE, R2 and NSE, respectively. Furthermore, the ANFIS-FCM-PSOGSA respectively improves the RMSE, MAE, R2 and NSE accuracy of the ANFIS-FCM-PSO by 20.1%, 19.2%, 35.7% and 35.6% in estimating the diameter of downward vertical redistribution. The general outcomes recommend the use of new method in estimating wetting front dimensions of drip irrigation systems.
