Numerical and artificial intelligence models for predicting the water advance in border irrigation

The water advance time (Ta) is needed for designing and evaluating surface irrigation systems. This study employed artificial neural networks (ANNs) and gene expression programming (GEP) techniques for estimating the water advance time in the border irrigation system as a function of inflow rate per unit width (Qb), length of water advance in the border (L), longitudinal slope (So), final infiltration rate of the soil (fo) and Manning roughness coefficient (n). The techniques were tested on field measurements from agricultural farms in three different provinces of Iran. Results showed that the ANN model was superior to the GEP model for the estimation of water advance time. The performance indicators for the ANN model were R2 = 0.966, RMSE = 7.805 min and MAE = 5.090 min, MBE = 0.312 and SI = 0.181. Results of the intelligence-based models were also compared with the WinSRFR model. Both ANN and GEP models predicted the water advance time more accurately than did the WinSRFR model.