Prediction of Groundwater Level Fluctuation by ANN and ANFIS Techniques

Abstract

Development of regional groundwater model isan important phase to understand the mechanism of groundwater resource and its sustainability. This model is extensively restricted by the requirement of large number of field data and specific interconnected constraints. A choice of intelligent soft computing technique is always constantly open for such model development. Recently ANN (Artificial Neural Network) and ANFIS (Adaptive Neuro Fuzzy Inference System) techniques are having a massive strength to deal with such interconnected constrains model. The present study is carried outto perform the comparative prediction between ANN, ANFIS, CWTFT-ANN (Continuous Wavelet Fast Fourier Transform), CWTFT-ANFIS, WT-ANN (Wavelet Transform) and WT-ANFIS on Groundwater Level prediction at different reaches (Top, Middle and End) of Lower Bhavani River Basin (LBRB) on monthly stress basis(from 2009 to 2015). From the results, the performance of CWTFT-ANFIS is 9.36%, 13.3% and 4.45% better than the ANFIS prediction and the performance of CWTFT-ANN is 47.17%, 25.6% and 47.45% better than the ANN prediction at Top, Middle and End reaches of LBRB respectively. Overall the prediction of CWTFT-ANFIS is about15.3% better than the other identified models, which is further fed for the forecasting of groundwater level to next one time stress level.