Prediction of Ground Water Level of Punjab State Using Artificial Neural Network

Abstract

Groundwater has always been an essential and reliable resource to supply drinking and agriculture water and is considered dependable for supporting the consumption needs of different users. For groundwater resource management, predicting groundwater level fluctuations with the desired accuracy is very much required. Consequently, there’s a need to deploy models capable of efficiently forecasting groundwater levels. In the past decade or so, artificial neural network has become very known in the field of hydrology and for good reason. ANN model used in this study was ANN-BP, back propagation. The architecture of the ANN model was with two hidden layers with 10 neurons on each hidden layer. The prediction was done using three separate algorithms levenberg-marquardt, Bayesian regularization and scaled conjugate gradient. The best results were given by Bayesian regularization, with the available data the ANN model can predict groundwater up to 6 months for 11 districts; Amritsar, Bathinda, Faridkot, Fazilka, Hoshiarpur, Kapurthala, Ludhiana, Mansa, Moga, Patiala and Sangrur. Keywords: Artificial neural networks; Groundwater level forecasting; Aquifer exploitation; Groundwater management; Groundwater hydrology; Amritsar; Bathinda; Faridkot; Fazilka; Hoshiarpur; Kapurthala; Ludhiana; Mansa; Moga; Patiala and Sangrur; Punjab