Network-wide mapping of multi-hazard transport disruptions across Taiwan using news media reports

This study introduces a structured framework for compiling a multi-hazard inventory of transport disruptions using open access news media. To do this, a bilingual search dictionary of location, hazard, infrastructure, and impact terms was developed in English and Traditional Chinese and applied to Google News to retrieve relevant articles. Transport networks underpin national resilience and economic activities but are highly vulnerable to multiple hazards. Yet systematic records of their impacts remain sparse. In Taiwan, where dense infrastructure intersects steep terrain and recurrent extreme events, disruption to transport corridors is frequent, but often absent from formal disaster databases.

A large language model was then employed to filter, classify, and validate events, which were subsequently linked to precise OpenStreetMap locations. The resulting dataset provides a comprehensive record of hazard events and associated transport impacts in Taiwan. This dataset is applied to a detailed analysis of transport disruption events for the eastern coast of Taiwan ruing Typhoon Gaemi. This highlights how the proposed methodology provides a transparent and scalable framework for capturing impacts of extreme weather or other disruptive events.