Spatial Analysis and Assessment of Relative Fire Risk Using Public Data

Abstract

In this study, a nationwide dataset at the level of basic local governments is constructed by integrating the fire occurrence counts, damage information, and regional characteristic data from publicly available national databases. Based on this dataset, the relative fire risk is quantitatively assessed. Principal component analysis is used to identify the variance structure of regional characteristics and select representative explanatory variables. A linear regression model is compared with a spatial linear regression model incorporating a conditional autoregressive structure. The spatial linear regression model demonstrates superior explanatory power; the resulting risk map attenuates local fluctuations through spatial random effects and identifies spatially corrected high-risk candidate areas. Analysis of the fire risk index reveals that under area- and population-based rate transformations, urban and non-urban areas show casualty-dominant and property-loss-dominant risk structures, respectively, confirming the multidimensional nature of fire risk. These findings present a comprehensive fire risk assessment framework reflecting damage structure and spatial dependence, serving as a practical basis for differentiated fire safety policies and region-specific resource allocation.