Artificial Network for Predicting Water Uptake under Shallow Saline Ground Water Conditions

Lysimetric experiments were conducted in order to determine the groundwater contributions by Black cumin. The plants were grown in 27 columns, each with a diameter of 0.40 m and packed with Silty clay soil. The factorial experiments were carried out using three replicates with randomized complete block designs and different treatment combinations. Nine treatments were applied during each experiment by maintaining groundwater with an EC of 1, 2 and 4 dS/m at three different water table depths (0.6, 0.8 and 1.1m). The groundwater contributions and plant root depths were measured by taking daily readings of water levels in Mariotte tubes and minirhizotron respectively. The four input neurons were total water use evapotranspiration (ETo), plant root depth (Zr), groundwater salinity (GS) and groundwater depth (Z). The output neuron gives maximum water uptake rate (Smax). The results showed that for different treatments, the best neural network was determined to be Multilayer Perceptron network (MLP) and the artificial neural network was very successful in terms of the prediction of a target dependent on a number of variables. This study indicates that the ANN-MLP model can be used successfully to determine groundwater observation by plant roots. Sensitivity analysis was undertaken which confirmed that variations in tide elevation are the most important factors in simulation of groundwater estimation in a semi-arid region. The results of this study showed that the estimation of plants groundwater contribution by ANN-MLP model is very useful for a quick decision on irrigation management to save a high volume of good surface water quality.