Comparison of Analytical, Multiple Polynomial Regression, and Artificial Neural Networks to Estimate Wetting Parameters for Surface and Buried Drip Irrigation across Three Different Soil Textures

This study investigated the wetting patterns of surface and buried drip irrigation systems across three distinct soil textures: sandy loam, loam, and clay loam. The research employed analytical methods, multiple polynomial regression analysis, and artificial neural networks (ANNs) to estimate wetting parameters, including wetted radius and depth, influenced by variables such as dripper flow rate, initial soil water content, and irrigation duration. Experimental data were collected in a controlled laboratory setting using a dripper with a consistent flow rate, allowing for precise measurements of the wetted zone’s characteristics. The average normalized RMS errors (NRMSEs) for estimating wetted zone radius, height, and depth using the analytical method were 0.10, 0.13, and 0.16, those for multiple polynomial regression analysis were 0.031, 0.032, and 0.034, and those for the ANN method were 0.04, 0.05, and 0.05, respectively. The findings demonstrate the higher accuracy of multiple polynomial regression analysis and ANNs in predicting water distribution patterns compared with that of the analytical method. This research contributes valuable insights for optimizing drip irrigation design, enhancing water-use efficiency in agricultural practices.