Watertable Elevation Prediction in the Shabestar Plain Using the Artificial Intelligence Techniques

Watertable elevation (WTE) prediction is of utmost importance in planning andimplementation of irrigated agriculture, especially where installing deficit irrigation systemsis considered in water-short areas. Furthermore, this information is vital in installing drainage systems and preventing land inundation and soil salinization. Using the WTE data from 20 piezometers maintained at least for 17 years, artificial intelligence neural networks, neurofuzzy system, and genetic programming were used to develop predictive tools to forecast WTE in the Shabestar Plain, Province of East Azarbaijan, northwest of the I.R. of Iran. The nonlinear behavior of WTE was ascertained when genetic programming was employed. The neurofuzzy system was proved to be the best predictor of the WTE; however, the other 2 systems performed satisfactorily. The neurofuzzy system was the best predictor based on the previous 3-month data; the genetic and artificial neural models occupied the 2nd and 3rd ranking in predictability. Explicit solutions that show the relationships between the input and output variables are presented based on genetic programming. This adds to the superiority of genetic programming over the other two models.