Trusted agents as knowledge translation mechanisms in flood disaster warning: a qualitative study of community-embedded intermediaries in Hualien, Taiwan

When institutional warning channels fail to convert hazard knowledge into protective action, community-embedded volunteers become critical intermediaries for reaching socially vulnerable populations. Drawing on an embedded qualitative case study of the 2025 Matai'an Stream landslide-dam overflow in Hualien, Taiwan — an event that generated nineteen official alerts yet produced spatially uneven evacuation outcomes — this study identifies three sequential mechanisms through which trusted community volunteers translate official risk knowledge into protective action across the acute warning phase and the post-disaster relief and recovery phase. Using reflexive thematic analysis of 22 semi-structured interviews, four expert key informant interviews, and forum observation within a Context–Mechanism–Outcome framework, three findings emerge: swift trust formation, activated through physical co-presence, institutional identity cues, and food provision rituals, constitutes the entry condition for knowledge exchange; trusted volunteers function as verification nodes when alert fatigue erodes official channel credibility; and translation efficacy is higher among volunteers with long-term community embeddedness. Taiwan's Disaster Prevention and Protection Act contains no provisions for elderly or digitally excluded residents, positioning trusted agents as load-bearing elements. Findings inform Sendai Framework Priorities 1 and 4 and support hybrid digital–analogue early warning architectures.