Investigation of the Effects of Meteorological Parameters on Groundwater Level using ANN

In the present research the effect of meteorological parameters such as temperature, relative humidity, evaporation and rainfall on groundwater level fluctuation has been investigated for Dakshina Kannada coastal aquifer at southwest coast of India. Weekly time series meteorological data were used for a span of three years (2004-2007). Generalized regression neural network (GRNN) and feed-forward back propagation networks (FFBP) were employed to develop various models. Model Input combinations were selected based on autocorrelation. The performances of developed models were evaluated using performance indices such as ischolar_main mean square error (RMSE) and coefficient of efficiency (CE). The obtained results showed closed relationship between rainfall event and groundwater level during monsoon. It was also, observed that the temperature and evaporation had significant effect on groundwater level fluctuations in non-monsoon season. The obtained GRNN results were compared with that of FFBP. A better agreement was observed between the actual and modeled groundwater levels for GRNN than that of FFBP. From the study, GRNN can be applied successfully for forecasting groundwater level due to its accuracy and reliable results.