Spatial modelling for identification of groundwater potential zones in semi-arid ecosystem of southern India using Sentinel-2 data, GIS and bivariate statistical models

The overarching goal of the present investigation is to adopt GIS-based spatial modelling techniques to delineate the groundwater potential zones (GWPZs) in Sarabanga watershed (SBW) of Salem district, Tamil Nadu (TN) state of southern India, by using high-resolution Sentinel-2 data, geographic information system (GIS), and bivariate statistical models (BSM) of frequency ratio (FR), and index of entropy (IoE). In GIS-based spatial modelling, eight contributing factors to groundwater potential (GWP), which includes geology, geomorphology, drainage density (Dd), slope, lineament density (Ld), soil texture, rainfall, land use/land cover (LU/LC), and the well inventory data of 135 well locations were considered in identification of GWPZs. The identified GWPZs of SBW based on the FR, and IoE models show that about 67.8% and 66.1% area of SBW are under very good to excellent categories, while 9.0% and 8.1% are under poor, and very poor categories. The results obtained were validated by using ‘Area Under the Curve-Receiver Operating Characteristic’ (AUC-ROC) method with the validation data and observed the prediction rate of 0.7313 and 0.7084, for FR, and IoE models, respectively. Modelling of GWPZs shows that FR model clearly exhibits its robustness over the IoE model. Sensitivity analysis performed through Variable Importance Analysis (VIA) indicates that in both FR, and IoE models, geology, slope, rainfall, and Dd were identified as the most influencing factors in delineation of GWPZs. The study clearly demonstrates the potential of Sentinal-2A data, GIS-based spatial modelling, and robustness of FR, and IoE models in attaining the reliable, and cost-effective results in delineation of GWPZs, which helps immensely in development of GW exploration, and management plans.