FORECASTING OF GROUNDWATER LEVEL FLUCTUATION USING THE COMBINATION OF ENSEMBLE EMPIRICAL MODE DECOMPOSITION AND ARTIFICIAL NEURAL NETWORK

Abstract

There is a requirement to manage water resources with the increasing population, leading to a balance increasing water demand. So, there is a need to use optimum water, and for this, accurate forecasting of groundwater level depth is necessary and helps increase the demand. Groundwater level projections must be precise and precise to maximize water productivity in the planning and enhancement of all water resources. The non-linear time-series intelligence data in the hybrid models for forecasting groundwater level variations were created using a pair of ensemble empirical mode decomposition (EEMD) and artificial neural networks (ANN). A groundwater level in a month time series obtained from some observation wells along the Amaravathi River, Amaravathi river is a main tributary of the Cauvery River, was used to evaluate the EEMD-ANN hybrid model. The EEMD-ANN model was evaluated using the correlation coefficient (CC), Nash–Sutcliffe efficiency coefficient (NE), root mean square error (RMSE), and Mean bias error (MBE). The EEMD-ANN model was used to anticipate groundwater level depth in our research. The entire investigation was placed on the Cauvery River's tributary, the Amaravathi River. To begin, data on groundwater level depth was gathered. Further, models were prepared to calibrate and validate the selected methodology with a different dug well. The linear regression approach was used to determine the parameters, and the performance of the model was evaluated using four statistical indicators. The findings of the EEMD-ANN model were compared to those of the ANN model. Meanwhile, it was discovered that models that were connected with EEMD had superior prediction outcomes than models that were not. The model's overall performance shows a good agreement between the observed and estimated groundwater level depth. The study's findings imply that the recommended non-linear time-series data variables in the intelligence hybrid model could boost forecasting ability in forecast groundwater level changes and offer essential and supportive recommendations for sustainable water resource management. It can be stated that the constructed model for the research region may be used to estimate groundwater level fluctuations in a reliable and precise manner, which is critical for managing water resources and improving water-use efficiency.