Artificial intelligence for groundwater sustainable development under climate variability and socio-environmental stress

This is the authors' abstract. We don't add key points for this paper.

Groundwater underpins global water security yet remains one of the least observable and slowest-responding components of the Earth system. Cumulative abstraction, climate variability, land-use intensification, and energy-dependent pumping have intensified groundwater pressures, while artificial intelligence (AI) applications have expanded across diverse hydrogeological and socio-environmental settings, but most remain prediction-centric, with limited attention to delayed responses, structural heterogeneity, uncertainty, transferability, and techno-social equity. Drawing on literature published between 2000 and 2026, this narrative review reframes AI as a bounded decision-support intermediary for groundwater sustainable development (GSD). It synthesizes groundwater data foundations and observability constraints, major AI paradigms, and five sustainability functions: early warning, managed aquifer recharge and zoning, abstraction control, groundwater quality protection, and the groundwater–food–energy–climate nexus. Emerging directions, including scenario-conditioned inference, physics-informed learning, and digital-twin-like representations, offer opportunities for sustainability-oriented groundwater management but remain constrained by sparse observations, geological complexity, limited transferability, and institutional capacity. AI–GSD can support anticipatory groundwater stewardship by reconstructing hidden system stress, characterizing uncertainty, and informing decisions under changing conditions. Its long-term value depends on regulatory integration, uncertainty-aware deployment, interdisciplinary collaboration, and safeguards against unintended groundwater depletion.