Uncertainty of the Ground Water Fluctuation Based on ANN Approach

Abstract

This study pursues to determine the accuracy of the groundwater level fluctuations forecasted at the Kanpur district of India using artificial neural networks (ANNs). The results indicated that performance of multilayer perceptron (MLP) based neural network (M-3, architecture 4-18-1) is satisfactory in the groundwater level fluctuations forecasting. The performance assessment shows that the MLP model performs significantly better. The uncertainty analysis shows that, input of Absent- RF and Absent- ERF, Absent- GWt-1, and Absent- GWt-5 were found more sensitive for GWFs forecasting and can’t ignore as input combination & input of Absent- WS and RH were found less sensitive for GWFs forecasting and may be discarded as input combination for GWFs forecasting.