Application of Least Squares Support Vector Machine Model For Water Table Simulation (Case Study: Ramhormoz plain)

Due to the current conditions on the arid and semiarid climates such as the lack of rainfall and high evaporation rate, the correct and efficient management of available water resources is inevitable. Management of groundwater resources as the most frequently used source of exploitable water is one the necessities of modern research. In this study, using quantitative data observation wells and meteorological parameters of Ramhormozplain in Khuzestan province during the 7-years period, performance of the Gamma Test evaluated and compare the accuracy of least squares support vector machine (LS-SVM) and artificial neural network (ANN) models to estimate water-table was discussed. Initially, using the Gamma Test among the parameters affecting the water table, input parameters for groundwater table estimation modeling of 127 certain combinations was optimized. The 5 better combinations with the least amount of Gamma Statistic than other combinations were obtained. Then, optimal combinations were evaluated using least squares support vector machine with different kernel functions and artificial neural network. Results of 4 combination showed accurate performance of LS-SVMRBF model that the optimal combination of LSSVM model with RBF kernel function has parameters of RBF function (2=4.99) and performance parameters such as (R2=0.999, RMSE=0.3401) than ANN model was and also represents the gamma test was optimized processing.