Journals Proceedings

Abstract

The study evaluates forecasting of groundwater level for short period of data by utilizing the standard artificial neural network (ANN) model, trained with two back propagation (BP) training algorithms namely Levenberg-Marquardt (LM) and Gradient Descent with Momentum (GDM). Data of five wells, Annual rainfall, Temperature, Relative humidity and river stage are chosen as input parameters.The model efficiency and accuracy were measured based on the root mean square error (RMSE) and regression coefficient (R).R-values approach towards the unity for most of the wells in LM method. LM method is recommended for forecasting ground water level for short duration of data and also it is anticipated that this method will give fairly accurate result for long duration of data under consideration. In case of constraint on data availability mentioned above, the LM Method is found to be suitable for ground water forecasting even when we take river water level as one of the inputs in ANN model.