Applied Spatial Analysis and Policy География, планирование и развитие

An Integrated Geovisualization-Driven Spatial Association and Network-Based Multi-Criteria Decision-Making Approach for Urban Fire-Risk Mapping and Mitigation

Nasrollahi Aghdam, Atefeh, Feizizadeh, Bakhtiar, Moshiri, Sajjad, Yakar, Murat
Springer Nature · 2026-09-09 · DOI 10.1007/s12061-026-09991-0

Аннотация

Urban fire-risk management requires both the identification of high-risk areas and an assessment of whether emergency services can reach them efficiently. This study developed an uncertainty-aware GIS-based decision-support framework to map urban fire-risk susceptibility and evaluate the operational performance of fire safety stations (FSS) in Tabriz, Iran. Twelve spatial criteria were weighted using the Analytic Network Process (ANP), and robustness to weight uncertainty was assessed through Monte Carlo simulation and Global Sensitivity Analysis (GSA), producing an ANP reference surface and an ANP-GSA sensitivity-adjusted surface. Network-based six-minute service areas were used to evaluate FSS accessibility, while Lee’s local bivariate spatial association, bivariate geovisualization, and local spatial cross-correlation were used to assess spatial consistency with residential and commercial-industrial exposure. Independent event-based validation was performed using ROC/AUC analysis with 60 fire and 60 non-fire/background points. Both models showed strong discrimination, with AUC increasing only marginally from 0.945 for ANP to 0.951 for ANP-GSA. Spatial analyses identified persistent high-risk concentrations in central Tabriz and underserved peripheral and industrial sectors. Overlaying FSS locations further showed that several residential clusters assigned lower relative priority by the ANP-GSA model already coincided with dense station coverage, whereas exposed peripheral and industrial zones remained prominent. The framework therefore provides a transparent basis for uncertainty-aware fire-risk prioritization and targeted FSS planning without overstating the predictive gain from sensitivity-informed reweighting.

Читать у издателя →