Pathways to Socially Sustainable Adaptation: Real-Time and Context-Specific Vulnerability Assessment in South Carolina after Hurricane Dorian

Abstract

Effective responses to natural hazards require prompt identification of vulnerabilities within impacted communities. This task is challenging due to the unique characteristics of each hazard and the dynamic nature of community traits, which impede the formation of a universal vulnerability framework. This paper introduces an innovative social sensing approach for near-real-time and context-specific vulnerability assessment in regions affected by natural hazards. Our approach involves the direct evaluation of vulnerability by analyzing how individuals and households perceive their risk to imminent hazards. We employed sentiment analysis to scrutinize household tweets related to unfolding natural hazards. We also used topic detection to uncover physical vulnerabilities such as critical infrastructure failures. Concurrently, we gathered demographic, socioeconomic, and occupational data at the census tract level to create a detailed database of community characteristics. We were able to identify socioeconomic vulnerabilities by correlating geo-temporally coded sentiments with community characteristics. In addition, spatial clustering of these sentiments enabled us to detect regional vulnerabilities, while mapping techniques highlighted areas with increased vulnerability. We applied our methodology to evaluate the impact of 2019 Hurricane Dorian on South Carolina. The results demonstrate our method’s effectiveness in real-time detection of events (e.g., flooding, power outages) and in identifying vulnerabilities across diverse socioeconomic groups. This research supports swift decision-making for allocating resources to vulnerable communities, especially in areas experiencing major disasters or infrastructure failures. Furthermore, our approach is valuable in identifying factors that increase the vulnerability of communities or individuals to natural hazards.