Groundwater Level Prediction of Shahrood Plain using RBF Neural Networks

Groundwater level prediction is an important issue in scheduling and managing water resources. A number of approaches such as stochastic, fuzzy networks and artificial neural network have been used for such prediction. A neural network model has been employed in this research for Shahrood plain groundwater level prediction. For this reason, statistical parameters of groundwater level fluctuations for 16 successive years 1994 to 2010, have been used and also, weather forecasting parameters for 16 successive years from 1994 to 2010 have used. This study indicates that some of the data are correlated an possesed a seasonal pattern. This issue makes difficult the forecasting process. Hence, the proposed method employed policies for non season analyzing, normalizing, and ignoring the correlated date. 85% of the data for train and the rest for testing the proposed neural networks model have used. Results indicate that the proposed method can predict groundwater level of Shahrood Plain for three successive years with the mean square errors of 0.0257m, 0.0270m and 0.0452m. Also, the prediction shows that if the precipitation decreases 30 percent in a year, the groundwater level will decrease 0.7 m.