Prediction of Groundwater level and Potential Zone Identification in Keonjhar, Odisha based on Machine Learning and GIS Techniques

Abstract

Population growth, change in climate, changing land use pattern, and increase in mining activities causes over exploitation of groundwater in Keonjhar district to fulfill the freshwater demand. This over extraction causes depletion in groundwater level. Therefore, the present study determines the best-fit model for groundwater level prediction in the Keonjhar district in Odisha, India, which is extremely reliant on groundwater for survival. The efficiency of machine learning models ANN, SVM, and LSTM is investigated for forecasting groundwater level (GWL) and to find the best-fit model for the prediction. The models were trained and evaluated using historical GWL data and meteorological parameters such as rainfall, humidity, temperature, and soil moisture. Through the analysis, the model LSTM was found to be superior in prediction of GWL with its ability to capture long-term dependencies and complex patterns in data. It achieves an impressive R² value of 0.97793 and an incredibly low RMSE of 0.00057, surpassing all other models in accuracy and reliability. This study provides vital insights into effective management of groundwater resources in regions facing comparable difficulties around the world. The study also aimed to identify groundwater potential regions in the Odisha district using remote sensing applications, MCDM, and GIS approaches. GW is a key source of freshwater worldwide, but little is known about its possibility, appearance, and distribution. The study considered a number of characteristics, including geology, rainfall, land use/coverage, soil type, drainage density, lineament density, and slope in Keonjhar District to determining the potential zone of groundwater.