Sustainable water allocation under climate change: Deep learning approaches to predict drinking water shortages.

Addressing sustainable urban water supply has become one of the most critical challenges for modern megacities, particularly in arid and semi-arid regions where rapid urbanization and climate change converge to exacerbate resource scarcity. Tehran, a metropolis under mounting water stress, exemplifies this global crisis. With population pressures, migration, poor urban planning, and inadequate environmental management intensifying the demand for water, reliance on groundwater surged to over 51 % of the city's total supply by 2021. This unsustainable dependence is compounded by severe aquifer depletion, now declining at an alarming rate of 32 cm annually. This study adopts advanced machine learning approaches to provide a forward-looking, integrative approach to understanding and mitigating the impacts of urban centralization, land-use mismanagement, and climate variability on Tehran's water resources. By leveraging hybrid simulation models, combining Recurrent Neural Networks (RNN) and Long Short-Term Memory (LSTM) models with three optimization techniques (i.e. Fire Hawk Optimizer (FHO), Whale Optimization Algorithm (WOA), and Horse Optimization Algorithm (HOA)) this research offers a powerful tool for managing water allocation across five critical dam reservoirs and the Tehran aquifer. Our analysis reveals that the RNN-FHO model demonstrates superior performance in predicting dam inflows, while the RNN-WOA model excels in forecasting groundwater table fluctuations, providing a vital roadmap for water resource planners. We developed a robust conceptual model to address anticipated drinking water shortages by supplementing surface water with groundwater resources. To simulate future conditions, we employed three state-of-the-art climate models (MRI-ESM2, CNRM-CM6-1, and BCC-CSM2) across three emission pathways (SSP1.26, SSP2.45, and SSP5.85) for the period 2021-2050. The projections indicate a troubling trend: dam inflows could decline by 8% in the most optimistic scenario and by 11 % in the worst case. Furthermore, by 2030, water demand in Tehran is expected to exceed 2.2 BCM, intensifying pressure on groundwater resources and necessitating large-scale water transfers. Excessive groundwater extraction, ranging from 100 to 300 MCM, would result in drastic aquifer drawdowns of 46-171 cm, threatening both hydrological stability and environmental health. This study highlights the critical need for a paradigm shift in water management practices. A strategic approach, encompassing reductions in per capita water consumption, extensive recycling, improved use of treated effluent in urban landscapes, and optimized water allocation, is essential to avert a looming water crisis. The methodologies and insights presented in this study offer transformative solutions for water-stressed urban environments worldwide.