Factors affecting groundwater level fluctuation: A case study at Manvi

Abstract

An accurate prediction of groundwater level is quite essential for ecological, sustainable development and management of groundwater resources. The soft computing models like ARIMA, ANN, FL, ANFIS and GA have been reported as the promising tools in prediction of accurate groundwater level. Now a day’s, ANN is widely used by many researchers in groundwater level prediction. So, the present study was carried out to find best ANN model to predict groundwater level fluctuation at Manvi. Selection of significant input variables is the most important step in the development of ANN model. In general, all of the relevant input variables (Rainfall, ET, temperature, RH, recharge, discharge, aquifer properties, streams, infiltration, initial groundwater level and variable groundwater level in nearby wells) will be equally informative in many instances. Further some may be difficult to collect, noisy, correlated or have no significant relationship with the output variable (current groundwater level) being modelled. The statistical parameters viz., auto-correlation function (ACF), partial auto-correlation function (PACF) and cross-correlation function (CCF) were used to select the significant variables with significant lag times. MATLAB 7.14 was used for statistical analysis and interpretation. Results indicated that, rainfall, evapotranspiration and previous groundwater level showed a good correlation with current groundwater level. It was found that considerable lag time of one month in case of rainfall (R) and previous groundwater level (Wt) and four months lag time of evapotranspiration (ET). So, input with R (t-1), Wt (t-1) and ET (t-4) lags were selected for ANN modelling. Good prediction was observed by the developed model.