| Sign up

PDF (4.8 MB)

Cite

Collect

Submit Manuscript

Show Outline

Figures (8)

Fig. 1

Fig. 2

Fig. 3

Fig. 4

Fig. 5

Fig. 6

Fig. 7

Fig. 8

Tables (6)

Table 1

Table 2

Table 3

Table 4

Table 5

Research Article | Open Access

City-scale fire risk modeling based on spatial regression methods

Hanchao Ma^¹(), Xiaoxiao Sun^¹, Jinlong Zhao^²

1School of Safety Science, Tsinghua University, Beijing 100084, China

2School of Emergency Management & Safety Engineering, China University of Mining & Technology, Beijing 100083, China

Show Author Information

Graphical Abstract

View original image Download original image

Abstract

Assessing the frequency of fires and the resulting economic losses is crucial for allocating emergency resources and firefighting personnel across different areas. In this study, spatial regression methods were employed to analyze the correlation between fire risk and various influencing factors (economic vulnerability, population vulnerability, building vulnerability, etc.), and an optimal model determination procedure was proposed and validated. The results indicate that the spatial lag model with population density, per capita income, and the degree of personnel concentration around high-risk points of interest (POIs) as indicators was the best model for predicting fire frequency. Moreover, logarithmic transformation of the indicators effectively improved the prediction accuracy in low-population density areas. An ordinary least squares model with the illiteracy rate and average surface water resources as indicators was the best model for predicting direct economic losses from fires, and the distribution of high-risk POIs can qualitatively explain the differences between the predicted results and actual data. The present work not only enriches the research on city-scale fire risk assessment but also reveals that the optimal regression model determination process can provide technical support for the application of spatial regression methods in fire risk assessment.

Keywords

fire frequency direct economic loss fire risk modeling spatial regression method

References

[1]

Angulo, A., Burridge, P., Mur, J. 2017. Testing for a structural break in the weight matrix of the spatial error or spatial lag model. Spatial Economic Analysis, 12: 161–181.

Crossref Google Scholar

[2]

Bispo, R., Vieira, F. G., Bachir, N., Espadinha-Cruz, P., Lopes, J. P., Penha, A., Marques, F. J., Grilo, A. 2023. Spatial modelling and mapping of urban fire occurrence in Portugal. Fire Safety Journal, 138: 103802.

Crossref Google Scholar

[3]

Bivand, R., Piras, G. 2015. Comparing implementations of estimation methods for spatial econometrics. Journal of Statistical Software, 63(18): 1–36.

Crossref Google Scholar

[4]

Gao, C., Feng, Y., Tong, X., Lei, Z., Chen, S., Zhai, S. 2020. Modeling urban growth using spatially heterogeneous cellular automata models: Comparison of spatial lag, spatial error and GWR. Computers, Environment and Urban Systems, 81: 101459.

Crossref Google Scholar

[5]

Getis, A., Aldstadt, J. 2004. Constructing the spatial weights matrix using a local statistic. Geographical Analysis, 36: 90–104.

Crossref Google Scholar

[6]

Hou, H., Zhu, Y. 2022. Analysis of spillover effects of regional environmental pollution: An interprovincial study in China based on spatiotemporal lag model. Environmental Science and Pollution Research International, 29: 836–853.

Crossref Google Scholar

[7]

Hu, J., Shu, X., Xie, S., Tang, S., Wu, J., Deng, B. 2019. Socioeconomic determinants of urban fire risk: A city-wide analysis of 283 Chinese cities from 2013 to 2016. Fire Safety Journal, 110 : 102890.

Crossref

[8]

Li, H., Zhang, C., Chen, M., Shen, D., Niu, Y. 2023. Data-driven surrogate modeling: Introducing spatial lag to consider spatial autocorrelation of flooding within urban drainage systems. Environmental Modelling & Software, 161: 105623.

Crossref Google Scholar

[9]

Liu, Z. G., Li, X. Y., Jomaas, G. 2022. Effects of governmental data governance on urban fire risk: A city-wide analysis in China. International Journal of Disaster Risk Reduction, 78: 103138.

Crossref Google Scholar

[10]

Ministry of Housing and Urban-Rural Development of the People’s Republic of China, Ministry of Public Security of the People’s Republic of China. 2015. Code for planning of urban fire control: GB 51080 – 2015. Beijing: China Architecture & Building Press.

[11]

National Bureau of Statistics. 2025. China Statistical Yearbook. Available at https://www.stats.gov.cn/sj/ndsj/

[12]

Oak Ridge National Laboratory. 2025. About LandScan. Available at https://landscan.ornl.gov/about

[13]

Office of the Leading Group of the State Council for the Seventh National Population Census. 2025. China Population Census Yearbook. Available at https://www.stats.gov.cn/sj/pcsj/rkpc/d7c/

[14]

Song, C. 2017. The spatiotemporal dynamic modeling analysis of fire risk for the location planning of urban fire stations. Ph.D. Thesis. University of Science and Technology of China.

[15]

Tan, L. L., Qu, N., Han, L., Sui, Y. F. 2023. GIS-SAVEE model-based assessment of fire risks in industrial cities. Journal of Institute of Disaster Prevention, 25 (2): 97–103. (in Chinese)

[16]

Wang, Z. 2021. Study on urban fire based on data mining and GIS. Master Thesis. China University of Mining and Technology.

[17]

Xu, B. 2012. Macroscopical influences of economic development and climate change on urban fire spatial-temporal variations in China. Ph.D. Thesis. Nanjing University.

[18]

Zhang, N., Huang, H. 2013. Social vulnerability for public safety: A case study of Beijing, China. Chinese Science Bulletin, 58: 2387–2394.

Crossref Google Scholar

[19]

Zhou, Y. 2012. Research on regressive model of fire and meteorological factors. Master Thesis. Hefei University of Technology.

Safety Emergency Science

Volume 1 Issue 1,
March 2025

Article number: 9590002

DOI: 10.26599/SES.2025.9590002

Cite this article:

Ma H, Sun X, Zhao J. City-scale fire risk modeling based on spatial regression methods. Safety Emergency Science, 2025, 1(1): 9590002. https://doi.org/10.26599/SES.2025.9590002

Return

Table 1Classification of POI data

High-risk location type	POI data type	Indicator weight (Tan et al. 2023)
Crowded places	Shopping services, catering services, sports and leisure services, transportation hubs (airports, train stations, bus stations, subway and bus stations, etc.)	0.40
Flammable and explosive areas	Industrial parks and mineral, metallurgy and chemical, fireworks and firecracker factories, etc.	0.55
Population vulnerable areas	Educational, scientific, and cultural services (training institutions, kindergartens, etc.), medical institutions (emergency centers, clinics, disease prevention agencies, etc.), and nursing homes	0.40
Important sites	Government agencies and social organizations (at the district and county level), research institutions, tourist attractions (museums, historical sites, scenic areas, etc.), educational training (science museums, research institutions, libraries, etc.), and media and culture (radio, television, art galleries, exhibition halls, etc.)	0.30

Table 2Statistical data utilized in this paper, its basis of application, and the data sources

Data type	Data name	Basis of application	Data source	Relationship with the severity/frequency of fires
Fire accident data	Number of fire accidents in various districts and counties from 2017 to 2021	—	China Statistical Yearbook 2018–2022 (National Bureau of Statistics, 2025)	—
Fire accident data	Direct economic losses from fire accidents in various districts and counties from 2017 to 2021	—	China Statistical Yearbook 2018–2022 (National Bureau of Statistics, 2025)	—
Population vulnerability	Population density	(Zhou, 2012)	Landscan (Oak Ridge National Laboratory, 2025)	Positive correlation
	Proportion of the population aged below 15 and over 65	(Ministry of Housing and Urban‒Rural Development of the People’s Republic of China, 2015)	China Population Census Yearbook 2020 (Office of the Leading Group of the State Council for the Seventh National Population Census, 2025)	Positive correlation
	Proportion of illiteracy among those aged 15 and above	(Bispo et al., 2023)	China Population Census Yearbook 2020 (Office of the Leading Group of the State Council for the Seventh National Population Census, 2025)	Positive correlation
Economic vulnerability	Per capita disposable income	(Xu, 2012)	China Statistical Yearbook 2018–2022 (National Bureau of Statistics, 2025)	Related to the area development degree (Xu, 2012)
Natural meteorological characteristic	Average surface water resources	(Xu, 2012)	China Statistical Yearbook 2018–2022 (National Bureau of Statistics, 2025)	Negative correlation
Building vulnerability	POI data	(Song, 2017)	Amap API	Positive correlation

Table 3Regression model parameters with different combinations of indicators

Original indicator regression results				Regression results for indicators after logarithmic transformation
Single indicator model	R²	Double indicator model	R²	Single indicator model	R²
Population density	0.945	Population density & per capita disposable income	0.968	Population density	0.961
Degree of personnel concentration around high-risk POIs	0.915	Population density & per capita disposable income	0.968	Degree of personnel concentration around high-risk POIs	0.975
Proportion of the population aged below 15 and over 65	0.878			Proportion of the population aged below 15 and over 65	0.635
Proportion of illiteracy among those aged 15 and above	0.833			Proportion of illiteracy among those aged 15 and above	0.904
Average surface water resources	0.0335			Average surface water resources	0.010
Per capita disposable income	0.710			Per capita disposable income	0.905

Table 4Optimal model parameters (both are SLM models)

Optimal model by using original indicators
Explanatory variable name	Model estimated coefficient	Standard error	z value	p value
Population density	0.0005065	0.0000768	6.60	< 0.001
Per capita disposable income	−0.0000248	9.79×10⁻⁶	−2.53	0.011
Disturbance term	0.4582892	0.298245
Spatial effect coefficient term	0.0082820	0.0023211	3.57	< 0.001

Optimal model by using indicators after logarithmic transformation
Explanatory variable name	Model estimated coefficient	Standard error	z value	p value
Degree of personnel concentration around high-risk POIs	1.178178	0.2565369	4.59	< 0.001
Disturbance term	−1.388115	0.6032521
Spatial effect coefficient term	−0.0253418	0.0092323	−2.74	0.006

Table 5Regression model parameters with different combinations of indicators for direct fire economic loss per accident

Single indicator model	R²	Double indicator model	R²
Proportion of illiteracy among those aged 15 and above	0.2620	Proportion of illiteracy among those aged 15 and above & average surface water resources (OLS model)	0.4992
Average surface water resources	0.3090
Proportion of the population aged below 15 and over 65	0.0090
Population density	0.0300
Degree of personnel concentration around high-risk POIs	0.1106
Per capita disposable income	0.1101

Table 6Optimal model parameters for direct fire economic loss per accident (OLS model)

Explanatory variable name	Model estimated coefficient	Standard error	z value	p value
Proportion of illiteracy among those aged 15 and above	42.73453	28.3126	1.51	0.143
Average surface water resources	−1,349.7	800.8074	−1.69	0.143
Disturbance term	5.98051	3.788877