Artificial intelligence for modelling coastal climate extremes: a global systematic review

Abstract

Coastal climate extremes are intensifying under climate change, generating complex, multivariate hazards that increasingly challenge conventional modelling approaches. To evaluate the evolving role of artificial intelligence (AI) in this domain, we systematically review studies addressing coastal hazards, including storm surges, extreme sea levels, heavy precipitation, compound flooding, and associated shoreline impacts. The literature is examined across six application domains: monitoring, prediction, mitigation, adaptation, resilience, and uncertainty quantification. Our synthesis reveals a strong concentration of AI applications in short-term forecasting, with comparatively limited progress toward decision-support and long-term resilience planning. Persistent limitations include data scarcity, poor model transferability, and the black-box nature of many AI approaches. Building on these findings, we identify four priorities for advancing coastal AI: developing multivariate frameworks for compound extremes, integrating physics-informed and hybrid AI–hydrodynamic models, incorporating socioeconomic scenarios into long-term adaptation and resilience planning, and applying transfer learning to improve model applicability in data-scarce regions. This review provides a structured synthesis of current applications and outlines pathways toward transferable, interpretable, and policy-relevant AI for adaptive coastal management.