Downscaling Urban Disaster Resilience: Toward Building-Scale Geo-Information Modeling and Machine Learning-Assisted Resilience Knowledge Systems

Abstract

Cities bear the brunt of damages brought by extreme events in terms of infrastructure disruptions, societal impacts, and manifold long-lasting consequences-calling for resilience building. Current resilience comprehension at higher resolution (i.e., building level) is limited, which compromises risk planning efforts. We here propose a high-resolution resilience assessment, profiling, and knowledge system building framework at the building/block scale for targeted interventions. The framework's contribution to practice is presented by assessing resilience at the building level (n = 30,590) in one of the most earthquake-hit regions in Turkiye, namely Elazig. Leveraging geo-information modeling, machine learning, statistical methods, and data on 40 representative indicators, we compute and map resilience from a systems-of-system perspective. While a significant geographical disparity in disaster resilience (overall as well as subdomain resilience) is evident, similar to 46% of buildings are identified as the least resilient (95% confidence). On a subdomain level, similar to 50% of buildings are identified in a worst socio-economic situation (i.e., the bottom 20% of the overall performance distribution). An online interactive dashboard is developed for spatial distributions and key analytics regarding building-level resilience intelligence-acting as a baseline resilience knowledge system for decisions and resource allocation support. Overall, this study offers a means and guide to establish national-scale high-resolution resilience knowledge systems in cities supporting decision-makers and concerned local authorities to strategize actionable tactics for future vigilance to reduce the deteriorating impacts of disasters in cities.