Compound and chained geohazards are increasingly frequent and complex, driven by the coupled effects of extreme climate, tectonic activity, geomorphological sensitivity, and human disturbance. Yet most existing geohazard datasets and modeling frameworks remain event-centered, limiting our ability to reconstruct process evolution or identify the key transitions that govern hazard escalation. In this perspective, I argue that meaningful progress requires a shift from documenting isolated events to representing the full hazard chain. I outline a process-oriented geodata framework that organizes multi-source information into nodes, spatial units, causal sequences, and explicit uncertainty descriptors, enabling disasters to be expressed as evolving chains rather than static outcomes. Building on this foundation, I further propose an AI-enabled risk intelligence framework designed to detect weak signals, infer hidden links, and anticipate potential propagation paths under sparse or inconsistent observations. Finally, I discuss how these concepts can support a transition from single-point interventions to chain-based, system-level disaster risk management. This work aims to provide a concise roadmap for integrating geodata and AI to improve the anticipation, interpretation, and governance of compound and chained geohazard risks in a rapidly changing environment.
Chong Xu (Wed,) studied this question.