Applicability of Support Vector Machine in Simulating Wetting Pattern under Trickle Irrigation

To design cost-effective and efficient drip irrigation systems, it is necessary to know the vertical and horizontal advance of the wetting front under the point source; also, the proper management of drip irrigation systems requires an awareness of the soil water distribution. Many factors influence wetting pattern dimensions, including discharge, land slope, irrigation time and soil texture. The purpose of this study was to investigate the applicability of the support vector machine in simulating the wetting pattern under trickle irrigation. After preparing a physical model made of Plexiglas with specific dimensions and filled with silty clay loam soils, experiments were conducted in the irrigation laboratory of Razi University, Iran, with emitters of 2, 4, 6 and 8 l/hour discharge during the irrigation intervals of 2 hours and 24 hours redistribution and 0,5,15 and 20% slope with three replications. In this study, the statistical indicators R2, RMSE, MBE and MEF were used. R2 values for the wet depth, width and area were 0.96, 0.96 and 0.92, respectively. Regarding the MBE value, the SVM model estimated the wet width and depth parameters to be 3% less than the actual value, and simulated the wet area 2.04% less than the real value. Also, according to the MEF and RMSE values, the SVM model simulated the wet area parameter with more error. Overall, the results showed that the SVM model had a high ability to estimate the wetting pattern parameters.