Artificial Neural Network for Identification of Groundwater Potential Zones in Part of Hugli District, West Bengal, India

Abstract

Adequate availability of groundwater for the rural and urban population is essential because groundwater is an important source of drinking water and agricultural irrigation. This study analyzes groundwater potentiality in part of the Hugli District (Goghat-II, Goghat-I, Arambag, Khanakul-I, Khanakul-II, Pursurah, and Tarakeswar blocks) using artificial neural network model (ANN), on GIS platform, and using remote sensing data and some secondary data sources. Factors like elevation, slope, drainage density, flow accumulation, geology, geomorphology, soil, land use land cover, rainfall, pre-monsoon, post-monsoon, and recharge rate were considered to be influencing the groundwater occurrence over this area. Sentinal-2 satellite data, SRTM data processing techniques, and GIS spatial analysis tools were used to prepare these maps. Groundwater depth level data from 34 wells were considered as ground truth data and randomly divided into training and test sets. An ANN based on the relationship between groundwater potential data and the above factors was implemented on R-studio. Each factor’s weight and relative importance was determined by the back-propagation training method. Then the groundwater potential indices were calculated, and the final map was created using GIS tools. The resulting groundwater potential map was validated using Area-Under-Curve analysis with data that had not been used for training. An accuracy of 77.78% was obtained. Five categories (“very high,” “high,” “moderate,” “low,” “very low”) of groundwater potential zones have been demarcated. This groundwater potential information will be useful for effective groundwater management and exploration.