Optimizing irrigation efficiency for subsurface drip irrigation using neural networks

Artificial neural networks ANNs are used to predict irrigation efficiency and relative transpiration from pipe installation depth, irrigation depth relative to transpiration depth, and irrigation rate. The network is trained and tested using fifty synthetic realizations. These realizations were made using the windows-based computer software package HYDRUS 2D/3D, which numerically simulates water, heat, and/or solute movement in two dimensional, variably saturated porous media. HYDRUS 2D/3D is used to simulate soil water movement and distribution for tomato crop through the whole season. The season is divided into three stages with root depths of 30, 60, and 100 cm. Actual transpiration, actual evaporation, deep percolation, irrigation efficiency and relative transpiration are calculated for each realization. The network is used to search for the optimum design that maximizes the. irrigation efficiency while keeps the relative transpiration above a threshold value. The results show that ANNs are an efficient tool in designing and operating subsurface drip irrigation systems.