Addressing water security in the developing world involves predicting water availability under unprecedented rates of population and economic growth. Yet the combination of rapid change, inadequate data and human modifications to watersheds poses a challenge, as researchers face a poorly constrained water resources prediction problem. This case study of the data-scarce, upper Arkavathy watershed, near the city of Bengaluru in southern India, attempts to systematically explain the observed disappearance of surface and groundwater in recent decades. The study asks three questions: 1) Can we quantify the change attributable to different drivers? 2) Can we anticipate change? 3) What policy lessons can be drawn? Field experiments, isotopic studies, borewell scans and sensors were deployed to understand hydrologic processes over five years. These were used in a historical reconstruction of the catchment over three decades that replicates the decline. The multi-scale model of the upper Arkavathy, quantifies the contributions of soil and water conservation measures, groundwater depletion and eucalyptus plantations to the decline in surface and groundwater resources. The model results indicate that the catchment hydrology cannot be reconstructed without explicitly including human feedbacks. The system is influenced by both endogenous drivers (social feedbacks to changes in water availability such as irrigation efficiency improvements, soil and water conservation measures and deeper borewells) and exogenous drivers (technological change, pro-development governance and economic forces due to urbanization, which provided access to capital and markets for high-value crops). The research suggests that in a system where productivity of the landscape is limited by water, economic drivers will always push for maximization of water abstraction and use. Unsustainability of resource use is inevitable, in the absence of credible controls on abstraction and use.