Linking Land Use Land Cover change to global groundwater storage

Groundwater storage is facing the constant threat of over-exploitation and irreversible depletion, often attributed to agricultural and industrial usage as well as human mismanagement. While several methodologies, varying from well logs to gravity recovery data, have been successfully adopted over the years to track and mitigate groundwater loss, Land Use and Land Cover (LULC) has never been quantified to evaluate groundwater storage and variability. LULC change alters the hydrological connectivity between the surface and subsurface water. Towards this, we employed a decision tree based Machine Learning model to (a) identify hydrological and terrestrial drivers affecting groundwater resources, (b) predict shallow and deep groundwater variability, (c) rank the drivers according to their impact on groundwater distribution, and (d) understand groundwater distribution as a function of LULC change. The model was developed globally, and then extended to basinal scale observations in the Indus, Ganga and Brahmaputra rivers of the Indian subcontinent. Model output has helped to (a) compute the 'infiltration index' associated with each Land Cover, (b) equate cropland expansion among the three basins with shallow and deep groundwater storage and (c) link LULC-groundwater change to crop yield. RCP 2.6 crop yield estimates for the 21^st century proves detrimental to Indian food and freshwater security, given the strong coupling of groundwater-LULC among the three basins and how Land Cover change translates to groundwater storage.