Artificial Intelligence for Climate-Resilient Stormwater and Wastewater Management

Abstract

Climate change is fundamentally altering the hydrological regimes upon which stormwater and wastewater infrastructure was designed to operate, manifesting as intensified extreme precipitation events, extended drought cycles, sea-level rise-induced groundwater infiltration, and thermally stressed biological treatment processes. Conventional infrastructure design paradigms, premised on stationary climatological assumptions, are rapidly becoming obsolete. This article presents a comprehensive review of how artificial intelligence (AI) — spanning machine learning, deep learning, reinforcement learning, physics-informed neural networks, and multi-agent systems — is being deployed to build climate resilience into stormwater and wastewater management systems across diverse geographies and institutional contexts. Synthesising 112 peer-reviewed studies published between 2019 and 2025, we evaluate AI performance across five climate-critical domains: non-stationary flood frequency analysis, drought-adaptive wastewater reuse optimisation, sea-level rise and groundwater intrusion modelling, heat-stress resilience in biological wastewater treatment, and equity-weighted infrastructure investment prioritisation under climate uncertainty. Our analysis finds that physics-informed hybrid architectures consistently outperform purely data-driven models when extrapolating beyond historical climate envelopes, with Physics-Informed Neural Networks (PINNs) retaining 73–89% of in-distribution performance accuracy under 2°C and 4°C warming scenarios. We introduce the Climate-Adaptive AI Water Resilience (CAAWR) framework, which operationalises a five-stage resilience cycle — sense, anticipate, adapt, recover, and learn — using AI as the connective tissue between real-time sensing and long-term planning. Critical findings include the persistent underrepresentation of Global South case studies (23% of reviewed literature), the urgent need for climate-downscaled training datasets at sub-catchment resolution, and the transformative potential of generative AI for rapid scenario planning under deep climate uncertainty. Policy implications for national adaptation planning, infrastructure investment, and transboundary water governance are discussed.