Integrating visual spatial vulnerability to quantify fire-prone neighborhoods in cities: A case study of nanjing, China

Abstract

In the context of rapid urbanization, the increasing frequency of fire incidents presents a growing threat to urban public safety and the sustainable development of metropolitan areas. Older towns and historic neighborhoods are particularly vulnerable to fire hazards due to aging infrastructure, deteriorating living conditions, and delayed emergency responses. These challenges make it difficult to accurately identify fire-prone areas within complex and heterogeneous urban environments. To address this research gap, this study introduces crowd-sourced street view imagery to capture human perceptions of visible vulnerability and proposes a systematic framework for quantifying compound fire risk by integrating both hazard and vulnerability components. First, we define three perceptual attributes of the neighborhood built environment (i.e, building dilapidation, street blockage, and spatial messiness) and conduct a survey-based evaluation using pairwise image comparisons. A pre-trained deep learning model is then employed to automatically estimate the Visual Spatial Vulnerability (VSV) index for each neighborhood. Next, we develop a Vulnerability-Adjusted Fire Risk (VAFR) grading system to identify at-risk neighborhoods and their characteristics. The individual and compound effects of hazard and vulnerability are further analyzed through a bivariate choropleth mapping approach. Finally, based on the VAFR scores, an Optimized Parameter-based Geographic Detector (OPGD) model is applied to examine key socioeconomic factors associated with high-risk neighborhoods. Our results demonstrate that: 1) integrating street view imagery with deep learning effectively assesses spatial vulnerability at the local scale; 2) combining fire hazard with visible vulnerability provides a comprehensive and reasonable explanation for the spatial distribution of fire-prone neighborhoods, highlighting the crucial role of spatial vulnerability in shaping risk patterns; and 3) the OPGD analysis confirms that spatial variations in fire risk are closely linked to neighborhood-level socioeconomic characteristics. This study offers a novel perspective for pinpointing high-priority neighborhoods for fire safety interventions and provides a scientific foundation for optimizing emergency response planning in complex urban environments.