Correlation Characteristics Between Urban Fires and Urban Functional Spaces: A Study Based on Point of Interest Data and Ripley’s K-Function

Abstract

This paper investigates the dependency relationship and spatial patterns between urban fires and the distribution of urban functional spaces, using the Futian District in Shenzhen as a case study. This study utilizes univariate and bivariate Ripley’s K functions along with Point of Interest (POI) data to analyze the variation in the spatial clustering of urban fires across scales ranging from 0 to 2500 m. It explores the overall distribution trends and localized relationships between urban fires and five types of urban functional spaces: commercial, tourism, residential, public services, and transportation services. The results indicate that the clustering of urban fires increases at spatial scales of 0–1050 m and decreases at scales of 1050–2500 m. The overall distribution trend between urban fires and urban functional spaces demonstrates a bidirectional clustering pattern. The overall correlation shows that commercial service spaces have the strongest association with urban fire clustering, followed in order by residential services, public services, transportation services, and tourist service spaces. The clustering of urban fires in local areas is significantly associated with commercial and residential service spaces, and moderately related to public service and transportation service spaces, and shows no significant correlation with tourism service spaces. This research contributes to the understanding of urban fire risk through spatial analysis and offers insights for urban planning and fire safety management.

1. Introduction

Fire poses a significant threat not only to lives and property but also to urban social stability and economic growth. Statistics from the International Association of Fire and Rescue Services indicate that over 8 million fires occur globally each year, resulting in more than 120,000 deaths [1]. In 2019, urban fires in the United States led to 3704 fatalities and USD 14.8 billion in property damage [2]. Similarly, a study on urban fires in Portugal revealed an annual average of 8841 fires between 2012 and 2020 [3]. China has also faced severe fire incidents, with 825,000 fires in 2022, leading to 2053 deaths and direct economic losses of RMB 7.16 billion [4]. As worldwide urbanization accelerates, urban population density continues to increase, and the scale and complexity of cities are expanding. These developments pose significant challenges to fire safety management [5,6,7]. Therefore, it is essential to assess fire distribution patterns, identify statistically significant fire risk areas, and gain a deeper understanding of the patterns of urban fire in order to enhance fire risk management capabilities and effectively reduce the frequency of fire occurrences [8].

Urban functional spaces represent the micro-level internal spatial structure of cities, with land use patterns and functional types serving as their effective representations [9]. In recent years, scholars have increasingly focused on studies related to the differentiation characteristics of urban functional spaces, the delineation and identification of urban functional zones, and the optimization of urban functional space layouts [10,11,12,13]. Furthermore, research has gradually expanded to explore the development processes, mechanisms, and impacts of urban functional spaces [14,15,16]. Urban expansion and restructuring have led to significant changes in urban functional layouts and regional structures, making the study of the interactive coupling relationships among urban functional spaces a persistent focus in disciplines such as geography and urban planning [17,18]. Given the close correlation between the layout of urban functional spaces and the occurrence of urban fires, various spatial characteristics—such as structural design, population density, building types, and usage patterns—within urban functional zones directly influence the frequency and severity of urban fires [6,7]. Therefore, the purpose of this study is to identify the correlation characteristics between urban fires and urban functional spaces, which can provide valuable insights for optimizing urban spatial layouts, rationally allocating firefighting resources, and enhancing urban fire prevention and control capabilities.

To identify urban fire patterns, researchers have used spatial analysis techniques alongside fire-driving factors [19,20,21,22,23,24]. Existing studies on fires primarily employed spatial analysis methods such as the Nearest Neighbor Index, Moran’s Index, Kernel Density Estimation (KDE), and Geographically Weighted Regression (GWR). For instance, Oliveira et al. used GWR to investigate fire density patterns in southern Europe [19], while Song Chao et al. applied both GWR and linear regression to model fire incidents in Hefei city in China [20]. Additionally, Bispo et al. used spatial econometric models to estimate fire occurrence probabilities across Portugal [3], and Arisanty D et al. employed Moran’s Index to examine fire clustering in Indonesia [21]. The distribution pattern of fires is closely correlated with human activities and the surrounding environment, as well as demographic and socioeconomic factors [25,26,27,28]. Guldåker et al. found that areas in Malmö, Sweden, with high life stress levels were more prone to arson incidents [22]. Ma Li et al. found higher fire occurrences in older residential areas and restaurants [23], while Wang K et al. noted fire clustering in shopping centers, entertainment venues, and dining establishments [24].

However, existing studies often focus on a single spatial scale, such as the district, city, or provincial level, to analyze fire patterns [29,30,31,32]. While these studies provide valuable insights into local clustering or regional trends, they overlook the inherent multi-scale nature of urban fire distribution. Urban fires are influenced by various factors, ranging from the characteristics of neighboring communities to broader urban planning elements, which manifest at different spatial scales [7,27]. Moreover, previous studies often focus on the overall patterns of urban fires, emphasizing the distribution trends of fires at larger regional scales [19,21]. These studies typically use aggregated statistical data, such as per capita GDP and population density, to reveal the macro characteristics of urban fires [29,33]. However, they often overlook the local variations in the relationship between urban fires and urban space. Urban fires may be influenced by multiple factors, such as complex building structures and heavy traffic flow, which can lead to significant differences in fire aggregation across different urban functional spaces [34].

Ripley’s K-function, a powerful tool for spatial point pattern analysis, identifies significant spatial clustering or dispersion by analyzing event distributions over varying neighborhood sizes. It has been applied in diverse fields such as transportation, tourism, crime, and finance [35,36,37]. Point of Interest (POI) data, which represent the distribution of businesses, services, and public facilities in an area, offer insights into urban functional spaces [38,39,40,41]. This study takes the Futian District in Shenzhen as a case study and adopts a novel methodological framework. First, Ripley’s K-function is used to analyze the spatial correlation of urban fires and explore their distribution patterns across various spatial scales ranging from 0 to 2500 m, with increments of 50 m. Second, Kernel Density Estimation is employed to investigate the spatial heterogeneity of urban fires and identify their clustering centers. Finally, a combination of bivariate Ripley’s K-function and local bivariate Ripley’s K-function methods is applied to examine the overall and local association characteristics between urban fires and the distribution of commercial, tourism, residential, public service, and transportation service spaces. The study aims to facilitate a comprehensive evaluation of fire clustering and its correlations with urban functional spaces at multiple spatial scales, offering scientific support for the refined management and targeted prevention of urban fires.

2. Materials and Methods

2.1. Research Area

Our study focuses on the Futian District in Shenzhen, an important administrative region located at the center of the city with a favorable geographical position, as shown in Figure 1. Covering an area of 78.72 square kilometers, Futian is one of Shenzhen’s political, economic, and cultural hubs. It is also the city’s Central Business District (CBD), home to a concentration of commercial, financial, cultural, and administrative functions, and serves as the core area for Shenzhen’s government, financial institutions, and high-tech enterprises. Futian was chosen as the study area because of its rich diversity in urban functional spaces, encompassing commercial, tourism, residential, public service, and transportation zones. This diverse urban fabric provides an ideal sample for analyzing the relationship between urban fires and different functional spaces. Additionally, Futian is undergoing rapid urbanization and modernization, resulting in a complex and constantly evolving spatial layout. With the progress of urbanization, fire risk management faces new challenges, and Futian represents the fire prevention and control needs typical of many first-tier cities in China during urbanization. Moreover, the availability of POI data and urban fire data in the district provides important support for studying the correlation characteristics between urban fires and urban functional spaces.

2.2. Materials

The urban fire data used in this study were sourced from the Shenzhen Municipal Government Data Open Platform, covering a total of 2721 urban fire records in the Futian District from 2008 to 2018 [42]. The fire data include information on the occurrence time, location, and type of each urban fire. To ensure the geographic accuracy of the fire data, the geographic coordinates of urban fires were obtained through the geocoding function of the Amap (Gaode) API. These geographic coordinates were then transformed into projected coordinates using the projection tools in ArcGIS Pro, following the CGCS2000 3-Degree Gauss-Krüger (GK) Projection coordinate system, with a central meridian of 114° E.

In addition, this study utilized POI data obtained from the Amap API to depict the urban functional spaces in the Futian District. To ensure the scientific validity and effectiveness of the classification, we referred to the existing literature on urban functional space classification standards and adjusted them based on the urban characteristics of Futian District [15,17,41,43]. The POI data were categorized into five types of urban functional spaces: commercial, tourism, residential, public services, and transportation services (

Table 1. Classification of urban functional spaces.

Data Categories	POI Data	Number of POI
Commercial service space	Catering services, financial and insurance services, life services, sports and leisure services, accommodation services, shopping services	36,850
Tourism service space	Scenic spot	409
Residential service space	Residence	3401
Public service space	Public facilities, scientific, educational, and cultural services, medical and health services	2677
Transportation service space	Transportation facility service	4560

). When selecting POI data, the following criteria were applied to ensure spatial accuracy and data completeness: only POI data within the boundaries of Futian District were included, and, for cases where a single building contained multiple POIs, we did not simply classify it based on the primary function. Instead, we assigned the building to multiple functional space categories to reflect its diversity and complexity. For example, if a building contained both commercial activities and residential functions, it was classified under both the commercial functional space and the residential functional space categories.

2.3. Methods

Firstly, Ripley’s K-function analyzes the spatial relationships among urban fire points, examining distribution patterns at various scales ranging from 0 to 2500 m, with increments of 50 m. Secondly, KDE assesses spatial heterogeneity, revealing the distribution characteristics of fires across different regions. Finally, the bivariate Ripley’s K-function and local bivariate Ripley’s K-function are combined to investigate both global and local correlations between urban fire and different urban functional spaces. A comparative analysis then evaluates the relationship between urban fire patterns and each type of urban functional space (Figure 2).

2.3.1. Kernel Density Estimation

Kernel Density Estimation (KDE), a non-parametric statistical technique, is widely used across research fields to estimate event probability densities [44,45,46]. In this study, KDE is applied to evaluate the spatial distribution of urban fires by calculating the density of fire points within the study area. The general definition of the density at urban fire point $x$ denoted as $f\left(x\right)$ is given by Equation (1).

$$f(x)={\displaystyle \frac{1}{nh}}\sum _{i=1}^{n}k\left({\displaystyle \frac{x-x_i}{h}}\right)$$

(1)

where $k$ is a non-negative kernel function; $h$ is the bandwidth, controlling the radius of influence for each fire point; $x$ is the set of fire points ($x_1$, $x_2$, …, $x_i$); and $x$ − $x_i$ is the distance from the estimation point to the fire sample point $x_i$. Regions with larger $f\left(x\right)$ values indicate areas where urban fires are more densely clustered. This approach allows for a precise understanding of fire clustering across different spatial regions.

2.3.2. Ripley’s K-Function

Ripley’s K-function is used to analyze the spatial clustering or dispersion of urban fires at multiple scales. For each fire point $i$, the function creates a circle with radius $r$ centered on $i$ and calculates the number of fire points $j$ within this radius. By varying the radius, it is possible to detect the spatial patterns of fire points across different scales [47]. The definition of $K\left(r\right)$ is given in Equation (2).

$$K(r)={\displaystyle \frac{a}{n(n-1)}}\sum _{i=1}^n\sum _{\begin{array}{c}j=1\\ i\ne j\end{array}}^{n}I(d_{ij}\le r)$$

(2)

where $a$ is the study region area; $n$ is the quantity of fire points; $d_{ij}$ is the space between fire point $i$ and fire point $j$; and $I(d_{ij}\le r)$ is an indicator function that equals 1 when $d_{ij}\le r$ and 0 otherwise. To correct the increasing variance with distance and maintain stable variance, the K(r) function is standardized and transformed into the L function, $L\left(r\right)$. The definition of $L\left(r\right)$ is provided in Equation (3).

$$L(r)=\sqrt{{\displaystyle \frac{K\left(r\right)}{\mathsf{\pi }}}}$$

(3)

Under the null hypothesis of Completely Spatial Randomness (CSR)—which assumes no statistically significant difference between the observed pattern and a random pattern—the expected value of the L-function, denoted as $L_{theo}\left(r\right)$, can easily be determined as r. The Monte Carlo simulation method was used to construct an envelope with a confidence interval to test the significance of the clustering. The envelope region is denoted as $L_{env}\left(r\right)$. If the observed value of the L-function, $L_{obs}\left(r\right)$, falls within the confidence interval of the envelope, the fire point pattern is considered to be randomly distributed. If it lies above the upper bound of the envelope, the fire point pattern is considered clustered, and if it falls below the lower bound, the pattern is considered uniform.

The radius r was determined based on urban spatial features in the Futian District. Specifically, we set r values ranging from 0 to 2500 m in 50 m increments, which align with typical neighborhood sizes and functional district scales in urban studies.

2.3.3. Bivariate Ripley’s K-Function

The bivariate Ripley’s K-function is employed to assess whether there is an interaction between the distributions of two types of event points, allowing for the examination of the interaction between urban fire points and various urban functional spaces at different scales. This function calculates the number of urban functional space points within a given distance $r$ from a fire point. The general definition of the function is provided in Equation (4).

$$K_b(r)={\displaystyle \frac{n_2{K}_{12}\left(r\right)+n_1{K}_{21}\left(r\right)}{n_1+n_2}}$$

(4)

where

$$K_{12}(r)={\displaystyle \frac{a}{n_1{n}_2}}\sum _{i=1}^{n_1}\sum _{j=1}^{n_2}I(d_{ij}\le r)$$

(5)

$$K_{21}(r)={\displaystyle \frac{a}{n_1{n}_2}}\sum _{j=1}^{n_2}\sum _{i=1}^{n_1}I(d_{ij}\le r)$$

(6)

In the formula, $n_1$ indicates the quantity of fire points; $n_2$ signifies the amount of urban functional space points; and $d_{ij}$ represents the distance between fire point $i$ and urban functional space point $j$.

2.3.4. Local Bivariate Ripley’s K-Function

The local bivariate Ripley’s K-function represents a specific adaptation of the bivariate Ripley’s K-function based on selected event points. It is used to analyze the relationship between a specified fire point and the distribution of urban functional spaces. The function can be simplified from Equation (4) to yield Equation (7).

$$K_l(i,r)={\displaystyle \frac{a}{n_1{n}_2}}\sum _{j=1}^{n_2}I(d_{ij}\le r)$$

(7)

where $i$ is a specified fire point. Equation (3) facilitates the transformation of both the bivariate Ripley’s K-function and the local bivariate Ripley’s K-function into the L function.

By applying the bivariate Ripley’s K-function and the local bivariate Ripley’s K-function, the spatial associations between urban fires and specific functional spaces can be evaluated at multiple spatial scales. These approaches reveal global trends while capturing local patterns, providing a scientific basis for understanding the relationship between urban functional spaces and fire distributions.

3. Results

3.1. Overall Distribution Pattern of Urban Fires

Understanding the overall distribution pattern of urban fires is crucial for uncovering their spatial characteristics and underlying laws. Ripley’s K-function was employed to investigate the distribution pattern of urban fires in the Futian District at various spatial scales within the range of 0–2500 m, with increments of 50 m. By utilizing the multi-distance spatial cluster analysis tool in ArcGIS Pro, 2721 urban fire points in the study area were analyzed, resulting in calculations at different distances (Figure 3), as well as the variation between the actual and anticipated values of the function (Figure 4). As shown in Figure 3, the L-function curve of the observed fire points entirely lies above the upper limit of the 99% confidence interval from Monte Carlo simulations across the 0–2500 m spatial scale. Figure 4 indicates that the variation between the actual and anticipated values of the L-function initially increases with distance, reaches its peak at 1050 m, and then begins to decline. The greater the deviation between the observed and expected values of the L-function, the more pronounced the spatial distribution. A positive difference implies a spatial clustering pattern in contrast to a negative difference, which signifies a spatial dispersion pattern. Thus, the experimental results are significant at the 99% confidence level, implying that a statistically significant spatial correlation is present in the distribution of urban fires. The results indicate that the clustering of urban fires increases at spatial scales of 0–1050 m and decreases at scales of 1050–2500 m, with the strongest clustering observed at a scale of 1050 m.

After determining that urban fires exhibit a clustered spatial distribution, kernel density analysis was performed to more clearly reflect the spatial heterogeneity of fire distribution and identify hotspots of fire occurrences. Using the KDE tool in ArcGIS Pro, a bandwidth of 1050 m was selected based on the previous research, and a cluster distribution map of the urban fires of the Futian District was created (Figure 5). As shown in Figure 5, the urban fire points in the Futian District demonstrate a spatial clustering pattern, with two main hotspots: one positioned in the heart of Shatou street, and the other at the junction of Futian, Nanyuan, and Huaqiangbei streets.

3.2. Correlation Characteristics Between Urban Fires and Urban Functional Spaces

3.2.1. Global Characteristics

The aim of this section of the study is to investigate the relationship between the spatial clustering of urban fires and the distribution of urban functional spaces. Treating urban fires and different types of urban functional spaces as distinct point patterns, the bivariate Ripley’s K-function was used to examine the global correlation characteristics between fire patterns and urban functional space patterns at various spatial scales within the range of 0–2500 m, with increments of 50 m. The bivariate Ripley’s K-function was employed using the Kcross method from the R language package spatstat, and the results were represented through R (Figure 6). As shown in Figure 6, the observed bivariate L-function values for urban fires and the different functional spaces—commercial, tourism, residential, public services, and transportation services—are all above the theoretical values and the upper bounds of the CSR simulation. This indicates a significant spatial correlation between urban fires and the clustering patterns of each functional space, demonstrating that the distribution of urban fires is greatly impacted by the distribution of urban functional spaces. Specifically, the aggregation of urban functional space will lead to the aggregation of urban fires.

To further analyze the correlation between the different urban functional spaces and fire distribution, the results from Figure 6 were summarized in Figure 7. Figure 7 illustrates that, although a spatial correlation is present between fire distribution and each urban functional space, the degree of this correlation varies. At various spatial scales within the range of 0–2500 m, the overall correlation is strongest between urban fires and commercial service spaces, followed in descending order by residential service spaces, public service spaces, transportation service spaces, and, finally, tourism service spaces. Given the analysis presented, it can be concluded that the overall distribution pattern of urban fires and urban functional spaces exhibits a bidirectional clustering trend, and there is a degree of spatial correlation between urban fire and urban functional spaces.

3.2.2. Local Characteristics

Building upon the understanding of the overall correlation between urban fires and urban functional spaces, urban fires were further categorized into non-clustered and clustered fires for comparative analysis. This approach aims to reveal the local correlation characteristics between fire clusters and urban functional spaces. Fire cases located in areas with a fire density of less than 40 incidents per square kilometer were selected to represent non-clustered fires, corresponding to the green areas in Figure 5. Similarly, fire cases located in areas with a fire density of more than 120 incidents per square kilometer were chosen to represent clustered fires, corresponding to the red areas in Figure 5. The local bivariate Ripley’s K-function was utilized to explore these relationships. The research explores the local clustering of urban fires and their connection to urban functional spaces by comparing the outcomes of two different types of fires in diverse urban functional areas (Figure 8).

As shown in Figure 8, the correlation characteristics between different urban functional spaces and fires in various regions exhibit significant differences. In the analysis of the correlation with commercial, residential, public service, and transportation service spaces, the function values for clustered fires are substantially higher than both the theoretical function values and the values for non-clustered fires. This indicates a positive correlation between clustered fires and these functional spaces, suggesting that the local clustering of fires is closely associated with the concentration of surrounding commercial, residential, public service, and transportation service spaces. In contrast, for non-clustered fires, the function values with commercial and residential spaces are below the theoretical values, while the function values with public service and transportation service spaces fall within the simulated envelope range. This implies a negative correlation between non-clustered fires and commercial and residential spaces and a lack of significant association with public service and transportation service spaces. Non-clustered fires are often located far from concentrated commercial and residential areas. On the other hand, in the analysis of correlation with tourism service spaces, the function values for clustered fires are noticeably lower than those for non-clustered fires and remain close to the theoretical values. This indicates that the local clustering of fires has no significant association with tourism service spaces.

4. Discussion

In this study, we systematically investigated the spatial distribution of urban fires and their associations with urban functional spaces using POI data and spatial analysis methods, including Ripley’s K-function. The results of Ripley’s K-function reveal a significant variation in the clustering characteristics of urban fires across spatial scales. Specifically, the clustering intensity of urban fires increases within the range of 0–1050 m, peaks at a scale of 1050 m, and then gradually decreases within the range of 1050–2500 m. This indicates that urban fires are strongly influenced by the characteristics of nearby areas at smaller spatial scales. Kernel density analysis identified two main clusters of urban fires in the Futian District. One cluster is located in the central area of the Shatou street, which is an urban village. The other cluster is at the intersection of Futian, Nanyuan, and Huaqiangbei streets, which represents the commercial center of the district. The results of the bivariate Ripley’s K-function provide evidence of the spatial association between urban fire clusters and urban functional spaces. The overall analysis of the relationship between urban fires and five types of urban functional spaces—commercial, tourism, residential, public services, and transportation services—indicates that the observed K-function values are above the theoretical and simulated envelopes, demonstrating a significant spatial correlation. Additionally, the local bivariate Ripley’s K-function analysis shows that urban fire clustering is strongly associated with commercial and residential service spaces, and moderately correlated with public service and transportation service spaces, and shows no significant correlation with tourism service spaces.

These findings suggest that urban fires are more likely to cluster in commercial and residential service spaces due to high population density, dense building layouts, and increased demand for electricity and open flames in these areas. Within a smaller spatial scale of 0–1050 m, these factors contribute to the strong clustering of urban fires as the lack of sufficient buffer zones and concentrated human activities exacerbate fire risks. Conversely, the moderate correlation with public service and transportation service spaces may reflect better management and safety measures in these areas, leading to relatively lower fire risks. The lack of significant correlation with tourism service spaces could be attributed to their dispersed distribution, strict management, and higher safety standards, which result in fewer urban fires. At larger scales (1050–2500 m), the decreasing clustering trend suggests that broader urban planning measures, such as buffer zones, fire infrastructure, and zoning policies, play a significant role in mitigating fire risks over larger areas. These results highlight the profound impact of urban functional spaces and spatial scales on the distribution of urban fires, providing valuable insights for fire risk prevention and urban planning.

The findings offer practical guidance for urban planning to reduce fire risks and enhance urban safety [48,49]. For high-risk areas such as commercial centers and urban villages, it is recommended to strengthen fire infrastructure by increasing the number of fire stations, improving fire access routes, and ensuring rapid response capabilities [8]. Upgrading safety measures in older buildings, particularly in terms of electrical and gas systems, should also be prioritized. Through rational urban planning, the excessive concentration of commercial and residential spaces should be avoided, and more green spaces and open areas should be integrated as buffers. Public education and fire safety management should be enhanced to raise residents’ safety awareness and emergency response capabilities. Additionally, advanced technologies, such as the Internet of Things and big data, should be leveraged to deploy intelligent fire monitoring and early warning systems, enabling real-time hazard detection and prevention [50]. These measures will provide scientific support for urban planning and safety management, fostering the development of safer and more livable cities.

Compared with previous studies, which primarily emphasized the relationship between urban fires and specific urban area, such as the findings of Ma L et al., who reported higher fire incidence rates in older residential areas [23], and Wang K et al., who highlighted fire clustering in commercial and entertainment zones [24], our study corroborates these observations by identifying strong spatial associations between urban fires and commercial and residential spaces. However, unlike many prior studies that analyzed fire patterns at a single spatial scale, this research adopts a multi-scale approach, revealing the scale-dependent characteristics of urban fire distributions and their associations with urban functional spaces. This approach offers a more comprehensive understanding of the dynamics of urban fires. Moreover, by leveraging POI data, this study provides a novel perspective on the interaction between urban fires and functional spaces, enabling a more precise and detailed analysis of urban spatial structures. Additionally, while previous studies often overlooked the localized variations in the relationship between urban fires and functional spaces, focusing instead on overall patterns [29,33], this study emphasizes the differences in local fire clustering across various functional spaces. These contributions deepen our understanding of urban fire spatial distribution patterns and offer valuable references for exploring the spatial drivers of urban fire risks.

Nevertheless, this study has certain limitations, primarily in terms of data utilization and methodological choices. First, the study focuses solely on the Futian District, Shenzhen, which may limit the generalizability of the conclusions. The fire data, obtained from the Shenzhen government open data platform, although comprehensive, may still be incomplete due to unreported or minor fire incidents. Similarly, the POI data, derived from the Amap API, may be affected by issues of timeliness and classification accuracy, which could influence the precision and representativeness of the results. Second, while Ripley’s K-function effectively reveals the clustering characteristics of urban fires and their associations with functional spaces, it falls short in accounting for the complex interactions of multidimensional factors, such as climate, socioeconomic conditions, and building structures.

Future research should address these limitations by incorporating multi-source heterogeneous data, such as remote sensing imagery, building information, and socioeconomic data, to construct more comprehensive fire risk assessment models. Integrating advanced spatial statistical techniques with machine learning algorithms could also enhance the depth and explanatory power of the analysis. Expanding the scope of the study to include multiple cities or regions with varying urban environments would allow for a comparative analysis of urban fire distribution patterns and their relationships with urban functional spaces, thereby validating the generalizability of the findings. Additionally, future studies should consider the spatiotemporal dynamics of urban fires and their associations with urban functional spaces to capture the trends and variations over time. Strengthening research on fire prevention and early warning strategies would also facilitate the practical application of research findings, contributing to more robust urban safety management systems.

5. Conclusions

This study identifies the spatial patterns of urban fire clusters in the Futian District of Shenzhen and highlights the strong correlations between urban fire risks and urban functional spaces, particularly in commercial and residential areas. By employing a multi-scale analysis approach, this research overcomes the limitations of previous studies that primarily focused on single-scale analysis, revealing the variation in fire clustering across spatial scales. Specifically, the clustering intensity of urban fires increases within the 0–1050 m spatial range and decreases beyond 1050 m, up to 2500 m. Moreover, the results indicate that urban fires tend to cluster in urban villages and commercial centers, emphasizing the need for targeted attention in these high-risk areas. Using bivariate Ripley’s K-function, this study explores the overall associations between urban fires and five types of urban functional spaces. The findings reveal a bidirectional convergence clustering pattern between urban fire distribution and the overall spatial distribution of urban functional spaces. Among the urban functional spaces, commercial service spaces exhibit the strongest association with urban fire clustering, followed by residential service spaces, public service spaces, transportation service spaces, and, lastly, tourism service spaces. Further analysis with localized bivariate Ripley’s K-function uncovers that urban fire clustering in specific local areas is significantly related to commercial and residential service spaces, and moderately associated with public service and transportation service spaces, and shows no notable correlation with tourism service spaces. These findings emphasize the critical importance of incorporating urban functional spaces into fire risk assessments.

In conclusion, this study provides a theoretical foundation for assessing urban fire risks and offers scientific guidance for policymakers and urban planners in developing effective fire prevention strategies. It not only deepens the understanding of urban fire dynamics but also lays the groundwork for creating safer and more livable urban environments. Furthermore, the findings have implications for fire risk management in other cities, serving as a valuable reference for improving urban safety and resilience. In summary, this study reveals complex relationships between different functional areas and urban fires, providing a novel perspective for urban safety management. By employing POI data and spatial analysis techniques, this research contributes granular data support for urban fire prevention and control.