Assessment of groundwater potential and determination of influencing factors using remote sensing and machine learning algorithms: A study of Nainital district of Uttarakhand state, India

Exponential increase in population, rapid urbanization and industrialization have increased the demand of water globally. Groundwater is an important resource in hilly and mountainous regions during dry spells. Thus, identifying prospective groundwater zones is crucial for conserving and managing groundwater. The study makes an attempt to assess groundwater potential in the Nainital district of a hill state in India. Random forest, multi-layer perceptron, M5P and REPTree algorithms were used for preparing groundwater potential maps. Sensitivity analysis was carried out to examine the influence of site-specific parameters on groundwater potential. Each model was evaluated through performance successors and receiver operating characteristic curve (ROC) for its effectiveness in groundwater potential assessment. Multi-layer perceptron was found best fit model for groundwater potential assessment. Largest area was found under high to very high groundwater potential zones in the plain area of the district (48 %), followed by very low to low in the hilly area (46 %) and moderate in the transition zone (7 %). Rainfall, lineament density and drainage density were found significant factors for influencing groundwater potential. The methodology adopted in this study has proved effective in groundwater potential assessment. The other geographical regions may find this methodology useful for groundwater management. [Display omitted] •Groundwater potential zones were identified using best fit machine learning model.•Multi-layer perceptron model was found effective based on performance assessors.•High and very high groundwater potential zones were found in southern part.•Rainfall, drainage density and lineament density influenced groundwater potential.•The insights of the study may help in sustainable development of groundwater.