Risk maps for urban fire with geospatial model-based framework.

Accurate identification of urban fire spatial patterns and governing factors is critical for optimizing firefighting resource allocation and developing sustainable cities resilient to evolving risks. Identifying spatial patterns of urban fires depends critically on the scale at which clustering is analyzed, yet a systematic approach to determine this optimal scale remains lacking. Moreover, the quantitative influence of fine-grained land-use structure on fire occurrence is not well understood. To overcome these gaps, this study proposes a novel three-stage framework for constructing hierarchical urban fire risk maps conditioned on built-environment macrostructure: (1) Investigation and selection of influencing factors, innovatively introducing high-resolution urban land-use attributes as variables, alongside traditional factors such as population density and socioeconomic indicators; (2) Determination of optimal grid size by integrating two key indicators: the Moran’s Index and Silhouette Score, ensuring precise spatial clustering. Subsequently, the spatial autocorrelation of urban fires is quantified; and (3) Negative binomial regression-driven risk quantification, deriving factor weights to calculate cell-level risk scores and generate hierarchical risk levels. Finally, a case study is conducted using fire data from Xiaoshan, China, a typical urban district with a population of about 2 million, covering 4,967 incidents from 2020 to 2023. The fire risk map indicates that urban fire distribution demonstrates striking conformity to the 80/20 rule, with a small share of cells concentrating most of the risk. This methodology provides stable, actionable risk mapping for strategic fire resource deployment and urban planning.