An application of transfer function noise models for predicting groundwater level signals using rainfall signals in Adyar Basin, India

Abstract

The effect of rainfall on groundwater level fluctuation was analysed using autoregressive moving average exogeneous (ARMAX) transfer function noise (TFN) modelling approach on four selected observation wells in Adyar Basin, Tamil Nadu, India. The evaluation of spatial average rainfall representation in the rainfall-groundwater level rise regression analysis was compared with three different methods namely simple arithmetic average (SAA), Thiessen polygon (TP), and Thiessen zone wise rainfall (TZR). Linear regression analysis on rainfall-groundwater level rise datasets revealed that the TZR method of inputting spatial average rainfall improved the rainfall-groundwater level rise correlation over SAA and TP methods. The four selected wells groundwater level fluctuation data from Adyar Basin was further modelled using ARMAX-TFN modelling approach and MLR methods. The study results show that the ARMAX-TFN model prediction performance was superior over MLR methods at all four locations. The validation results of the ARMAX-TFN model show that the predicted and observed groundwater levels at the corresponding well locations were strongly correlated with the correlation coefficient values of 0.85-0.93.