Mapping groundwater potential zone by robust machine learning algorithms & remote sensing techniques in agriculture dominated area, Bangladesh

Abstract

An important part of the ecosystem is groundwater. These resources of Bangladesh are under tremendous pressure from both natural and human-caused factors. Groundwater is essential for fulfilling water requirements in the agricultural Pabna district of Bangladesh, where over-extraction for local, manufacturing, and farming uses has led to considerable water shortages. It is highly expanded in the aspect of industry and agriculture practices. This region's distinctive physiography, extensive agriculture, dryness, low rainfall, and abundant water supply all contribute to the low groundwater depth. The enhancement of human accessibility to sufficient quantities and high-quality groundwater resources is one of the major goals of this research. Several machine learning algorithms and analytical hierarchy process (AHP) models along with geographic information systems (GIS) software integrate sixteen thematic layers, including elevation, slope, soil types, topographic wetness index (TWI), normalized difference water index (NDWI), normalized difference vegetation index (NDVI), curvature, soil permeability, physiography, topographic position index (TPI), terrain roughness index (TRI), stream power index (SPI), distance from river, rainfall, drainage density, and land use land cover (LULC) to create a groundwater potential zone map. Furthermore, the research uses 340 well and non-well sites as inventory data. This is randomly divided into two datasets: training (80 %) and testing (20 %). The resultant groundwater potential zone map is divided into five categories: extremely poor, very poor, moderate, good, and excellent. Every model that was validated using the ROC curve has an AUC-ROC value of more than 0.90. The study's conclusions will help decision-makers save groundwater for long-term usage in areas experiencing a water shortage.