On the spatiotemporal knowledge-driven vulnerability assessment of urban areas: A clustering-based approach

Abstract

In recent years, there has been a notable increase in the implementation of emergency management digital solutions. While these systems are becoming increasingly prevalent in cities, they must be properly set up and adopted based on a comprehensive understanding of the spatiotemporal dynamics of urban areas. This study aims to develop and validate the VERUS (Vulnerability Evaluation for Resilient Urban Systems), a spatiotemporal clustering framework for assessing urban vulnerability based on the dynamic influence of urban infrastructures during emergencies, indicating how populations are negatively affected based on the existing urban dynamics. For that, a holistic and adaptive urban perspective is adopted centred on the influence of selected groups of PoTIs (Points of Temporal Influence). Moreover, instead of considering static influence, it also incorporates the fluctuating impact of each PoTI throughout time windows. The proposed clustering method divides the urban area into influence clusters to assess the vulnerabilities within their boundaries, taking as input open geospatial datasets like OpenStreetMap. To effectively address the issue of defining the optimal number of clusters, we evaluate various methods and suggest a combination of OPTICS and K-means to provide a reliable and adaptable clustering definition without the need for parameter adjustments. Experimental results in the cities of Porto, Lisbon, and Paris demonstrate its adaptability to diverse urban configurations, illustrating its practical feasibility by revealing varying levels of vulnerability. These insights emphasise its potential to inform knowledge-driven smart city systems, where tailored interventions can address the unique challenges of different urban environments.