From Multi-Hazard Prevention to Sustainable Emergency Management: A Reproducible Evidence-Mining Framework for Identifying Compounding Risk Patterns

Abstract

Climate change and increasing socio-environmental pressures are intensifying the occurrence of compound and cascading hazards, highlighting the need for systematic approaches to reconstruct and analyze multi-hazard interactions from historical evidence. However, historical inventories are typically event-based, unevenly documented, and lack information on non-events, limiting their direct use for probabilistic interpretation. This study presents a reproducible evidence-mining framework that transforms a historical multi-hazard inventory into a hierarchical process–hazard–location dataset while explicitly distinguishing documented event-tree relationships from exploratory within-catalog associations. Applied to a historical multi-hazard inventory of Tenerife (Canary Islands), the framework reconstructs documented parent–child hazard relationships, identifies recurrent compound motifs, and formalizes association screening through transparent statistical procedures, including contingency-table analysis, uncertainty intervals, effect-size estimates, exact significance testing with false-discovery-rate control, and Bayesian sensitivity analysis. A quantitative assessment of documentation density reveals a strong increase in recorded evidence through time, demonstrating that temporal variations in the inventory primarily reflect changes in documentation rather than changes in hazard occurrence. Consequently, all derived association metrics and screening thresholds are interpreted as hypothesis-generating evidence rather than predictive or causal estimates. The proposed framework provides a transparent and reproducible basis for extracting structured knowledge from historical multi-hazard inventories, supporting research prioritization and future probabilistic developments while avoiding overinterpretation of heterogeneous historical records.