Comparing Regression and Artificial Neural Networks in Estimation of soil Wetted Pattern Dimensions With Magnetic and None Magnetic Water

Having proper information of the wetted width and depth of soil for the appropriate design and management of a drip irrigation system is essential. This research was carried out to calculate the wetted depth and the maximum wetted width in both surface and subsurface drip irrigations with normal and magnetic waters. Measured values were compared with linear and nonlinear regression and neural network models. Experimental measurements were done on a clay loam soil in the greenhouse of Zanjan University. The results showed that the simple linear and multiple regression for Schwartzman and Zur (1986) and Mirzaei et al 2008) to estimate the wetted depth in surface drip irrigation with normal water was better (r=0. 988 and RMSE=0. 011 m). The difference between the estimated values using Schwartzman and Zur (1986) and Mirzaei et al (2008) in the wetted depth with magnetic water in surface drip irrigation (r=0. 974 and RMSE=0. 014 m) and the maximum wetted width in surface drip irrigation with normal water (r=0. 950 and RMSE=0. 028 m) and with magnetic water (r=0. 976 and RMSE=0. 023 m) with the observed data was minimal. Among the artificial neural networks models used in this study، multilayer perceptron model as compared with radial basis function model performed better.