1 Introduction

Fire incidents pose a significant threat to public safety and assets. Global fire statistics reveal that 24.2 per cent of fires occur in residential buildings, contributing to 82.7 per cent of fire-related deaths in 2020. The severe impacts are particularly pronounced in large, densely populated Asian cities with haphazard construction, increasing the risk of urban fires. To minimize property loss and reduce injuries and fatalities, emergency services adopt Service Delivery Standards (SDS), ensuring timely response to emergency service calls. Urban fire risk management in developing countries is vital due to dense urban areas with inadequate infrastructure, informal settlements, and substandard housing, making communities highly vulnerable to fires. Limited resources for firefighting equipment and personnel further hinder effective response and exacerbate fire incidents. Mitigating these risks involves improving building codes, enhancing emergency response infrastructure, raising fire safety awareness, and promoting sustainable development. An urban network theoretical framework can help understand the dynamics and interventions in fire risk management. Studies show high-density, low-income areas face frequent fires, and narrow, disconnected streets can delay emergency responses, leading to significant losses and challenges in fire prevention and response efforts.

Empirical studies on the optimisation of fire service delivery have covered various countries, including Australia, US, and China. Many studies have developed models to measure the operating costs of fire stations, response times, and service delivery by maximising the coverage of demand areas. Recent research considers factors such as population forecast, property values, political interests, and socioeconomic characteristics to examine their relationship with fire incidents (e.g., Kiran & Corcoran, 2017). While previous studies aim to enhance services through improved response times and spatial coverage, major cities in developing countries like Nepal continue with challenges in promptly responding to emergency calls. These densely populated urban cities face increasing threats from urban fires but often lack sufficient evidence to support the optimisation of fire service delivery. Kathmandu, the capital city of Nepal, experiences frequent fire incidents due to factors such as electric short circuits, cooking gas leaks, prolonged dry spells, rising temperature, and discarded cigarettes (Chaudhary et al., 2016). Examining optimal service areas is crucial for maximising timely response to fire service calls and strategically locating new fire stations, particularly considering current and future urban development and population growth.

The current paper contributes to the limited empirical work on urban fire service delivery in Kathmandu Metropolitan City (KMC) by evaluating the current fire service coverage and identifying potential locations for future fire stations. Drawing on data from four sources, this study has two objectives: 1) examining the spatial pattern of fire risk and service coverage of the current fire stations, and 2) identifying potential locations to optimise fire service coverage in KMC. The paper links population and housing characteristics, along with street connectivity, providing the first empirical study of urban fire service coverage in Nepal, using primary fire incident data in the process of identifying potential sites for fire station. This research aims to expand the empirical scholarship on urban fires, particularly from the developing country perspective. The findings from this study will enable fire agencies to enhance operational efficiency in emergency response and contribute to the development of policies and programs related to fire risk prevention and mitigation. Our analytical framework developed can be applied to similar geographic contexts, offering spatially explicit evidence to support emergency planning and fire risk control.

2 Literature review

Urban fire risk management is intricately tied to the SDS framework adopted by fire agencies, which encompasses quality, timeliness, and cost. SDS not only shape community experiences regarding urban fires but also play a crucial role in building trust, emphasizing aspects such as responsiveness, reliability, competence, professionalism, and communication (Urriza, 2023). Recent studies have shown that three principal factors are essential for ensuring fire agencies meet SDS when delivering services: the location of fire incidents; the strategic placement of fire stations, and population distribution. Accessibility and proximity to fire services impact SDS, influencing response time, resource allocation, and the effectiveness of firefighting efforts (Shahparvari et al., 2020). Strategic placement of fire stations reduces response time and enhances coverage areas, thus aiming to improve the overall efficiency and community safety. Population distribution is closely linked to high fire risk. Areas with dense populations often require a higher concentration of fire services (Knorr et al., 2016).

Urban fire risk management is crucial in developing countries for several reasons, first, developing countries often have densely populated urban areas with inadequate infrastructure, informal settlements, and substandard housing. These conditions increase the vulnerability of communities to fire incidents. These fires can result in significant loss of life, property damage, and economic disruption. Second, developing countries typically have limited resources for firefighting equipment, trained personnel, and emergency response infrastructure. This lack of resources can hinder effective firefighting efforts and exacerbate the impact of fire incidents. Effective fire risk management in developing countries is essential for mitigating these risks and protecting the lives, property, and livelihoods of vulnerable populations. This includes measures such as improving building codes and regulations, enhancing firefighting capabilities and emergency response infrastructure, increasing public awareness and education on fire safety, and promoting sustainable development practices to reduce fire hazards.

In addressing urban fire risk in developing countries, an urban network theoretical framework may be used to understand the underlying factors, dynamics, and interventions related to fire risk management. Urban networks, connecting the built environment (e.g., roads, buildings) and socioeconomic factors (e.g., population density, income levels) influence the structure and dynamics of urban networks (Jennings, 2013). Some studies found that high-density areas with low-income levels are more likely to experience frequent fires than low-density areas with high income levels (Desriani & Samyahardja, 2018; Chhetri et al., 2018). A study by Sufianto and Green (2012) found that areas with narrow and disconnected streets delay fire trucks response to the scene of a fire, potentially leading to significant property loss, injuries and fatalities. In areas with significant social inequality, there may be a lack of trust between different groups of people, making it challenging to coordinate fire prevention and response efforts.

Empirical studies used several techniques to model the location of fire stations based on urban network theory (Bagchi et al., 2013; Zhao et al., 2023). For instance, location-allocation model optimize fire station placement by minimizing the average response time to fires, considering factors such as building locations, the street network, and the fire load (Shahparvari et al., 2020). Network theory serve as a valuable tool for modelling fire spread, demonstrating that optimizing fire station location contributes to reduced response times (Widyantoro & Santosa, 2021; Yao et al., 2019). Previous studies used these models to enhance the service delivery of fire agencies to ensure the safety of cities and better protect citizens from the threat of fire.

The spatial heterogeneity of the built environment significantly influences urban fire occurrences, with different urban areas showcasing distinct socioeconomic characteristics that introduce new risks (Donner & Rodríguez, 2008). Densely populated urban areas characterised by substandard dwelling and construction practices, inadequate infrastructure, and inappropriate land use are particularly susceptible to fire risks (Jennings, 2013). Cutter et al. (2003) studied the hazard-of-place model of vulnerability that explains how different places are at risk of fires. Arguably, there is a spatial dependence, meaning that fire risk patterns are often spatially correlated.

Examining spatial patterns of fire risk are pivotal not just for quantifying geographic variations of fire risk (Ardianto & Chhetri, 2019), but also because they represent the geographic trace of processes, covariates, and factors that determines how urban fire risk varies across areas. This perspective aids in classifying urban fire risk according to ranges of geographical characteristics, facilitating the characterisation of types of fires for emergency planners in formulating effective fire management strategies, including the optimisation of fire station locations. However, the exploration of spatial patterns of fire risk and risk management is predominantly originating from developed countries, leaving a gap in understanding and addressing these dynamics in developing countries.

Table 1 presents various techniques used by previous studies in the optimisation of fire station locations, considering a global perspective. Previous studies employed various techniques such as the Analytical Hierarchy Process (AHP) in Nepal (Chaudhary et al., 2016) and multi-objective spatial optimisation model (Chen et al., 2023b; Tao et al., 2022) in China to Greedy Adding and Greedy Adding with Substitution in Indonesia, and simulated annealing methods in Spain (Fernández et al., 2021). Some studies focused in integrated road-network accessibility (Tian et al., 2023) in China and location-allocation model (KC et al., 2018; Shahparvari et al., 2020; Tian et al., 2023) to examine the geographic distribution and optimize fire service coverage in Australia. Overall, The United States, Japan, and European countries like Spain contributed studies utilizing optimization algorithms and simulated annealing, with an overarching goal of enhancing fire service delivery.

Table 1 Seminal studies on the optimisation of fire station using various modelling
Full size table

In summary, previous studies have typically focused on either fire station placement or fire risk coverage, neglecting the essential need for both analyses in the context of SDS. Notably, no study from developing countries, such as Nepal, has integrated fire incident data analysis into the fire station site selection process. Analysis of fire risk, response times, and fire service coverage are all important in selecting sites for fire station locations to manage increasing demand for services.

In Nepal, studies exploring fire site selection analysis are relatively underexplored. There is only one case study found to have attempted identifying suitable sites for fire station in Kathmandu using AHP, emphasizing the scarcity of research in this area incorporating fire incident data analysis in the process of fire station site selection. More importantly, the application of alternative techniques, such as suitability analysis, remains largely unexplored in fire station site selection studies. Suitability analysis has the potential to offer a more comprehensive approach that can consider various criteria together, such as incorporating fire risk patterns, response times, service coverage, and street connectivity, in the context of identifying suitable sites for fire station. The variability in fire risk is context-specific, distinguishing one location from another and influencing the likelihood of urban fire occurrences and their consequences. Identifying and managing the location of fire stations requires a more targeted strategy to cover potential fire risks within communities and the built environment (Kobes et al., 2010; Hossain & Smirnov, 2023). This study offers a new perspective on identifying fire station locations in Kathmandu by considering the specific context in which a fire event occurs.

3 Methodology

3.1 Study area

Kathmandu Metropolitan City (KMC), the capital city of Nepal, is also the largest and the oldest city in Nepal (Fig. 1). The city hosts popular world heritage sites such as Durbar Square, Pashupatinath Temple (Hindu), Boudhanath Temple (Buddhist), and Swayambhunath Temple (Buddhist). Covering an area of 51 square kilometres, the total population of KMC was 862,400 in the 2021 census (CBS Nepal, 2021). The Juddha Barun Yantra fire station (referred to hereafter as the KMC fire station) is in the core city area and is one of the three fire stations covering KMC. The demand for fire service calls in KMC is substantial due to growing population and the proliferation of unplanned infrastructure development.

Fig. 1
figure 1

Study area

Full size image

Fires occur more frequently on the outskirts of KMC, where building density is also high, and these areas are often distant from the nearest fire stations. This results in a significant gap between the demand for fire services and the available supply for emergency calls in KMC. There is a pressing need for research that examines the current fire service coverage and for evidence-based scientific studies that identify suitable locations for stations in KMC. The city serves as an ideal case study due to its urbanisation challenges, socioeconomic and demographic diversity, cultural significance, topographical complexities, and infrastructure vulnerabilities that are similar to other developing countries.

3.2 Data

Data is drawn from four principal sources.

  1. 1.

    Fire incident data: The KMC fire brigade provided a total of 596 fire incidents for this research, recorded between July 2019 and February 2023. These incidents include information on the location (ward numberFootnote 1) and characteristics (e.g., incident time, date of incident, turnout time, vehicle dispatch to return time, causes of fire, as well as number of trucks and personnel deployed) in KMC.

  2. 2.

    Open Street Map (OSM) data: A suite of OSMFootnote 2 data, including street network, building locations in polygon format, and other points of interests, was captured on 23 May 2023, using ArcGIS Online. The centroids of building structures and street network data were utilised in the computation of fire service coverage analysis, while land use and river network data were employed in the fire station site selection analysis.

  3. 3.

    National population and housing census 2021 data: Population and housing data at the KMC ward level were acquired from the Central Bureau of Statistics (CBS). Drawing on existing empirical urban fire research (Chhetri et al., 2018; McNamee et al., 2019) 14 variables were collected to examine their relationship with fire incidents, including “Population, rental population, building count, households with LPG fuel, households lighting by electricity, households amenities (none), households with no toilet, people with no school, economically inactive population, unemployed population, children (< 18) with single mother, elderly (> 65), mean temperature, and mean relative humidity”. To temporally align the CBS variables with the fire incident data, the 2021 census data was acquired from the CBS.Footnote 3 The ward level is the lowest spatial unit for which population and household characteristics are generally available in Nepal (CBS Nepal, 2021).

  4. 4.

    Weather data: Hourly temperature and relative humidity data from KMC for the study period were procured from the Department of Hydrology and Meteorology (DHM),Footnote 4 Nepal. Previous studies have indicated a positive relationship between fire incidents and increased relative humidity and temperature. These weather data sets are used to study the spatial relationship with fire incidents (Corcoran et al., 2011). Weather data was associated with each fire incident according to the date and time of the incident and the nearest temporal weather measurements.

Table 2 illustrates the structure (example) of the core dataset. The fire incident data is merged with other databases in a way that enables the identification of various factors for each individual fire incident. These factors include the season during which the fire occurred, weather conditions (rainfall, temperature, wind speed, and relative humidity), and the total population and household conditions of the ward area in which the fire incident took place.

Table 2 Example of core data structure
Full size table

The descriptive statistics (Table 3) provides information on selected census and weather variables along with all fire incidents. For example, the average population is approximately 26,950 with a significant standard deviation of 22,526, indicating significant variability in population across the dataset. Additionally, the temperature exhibits a relatively narrow range, varying from 10.9 to 22.25 degrees Celsius, with a mean of 16.79 and a standard deviation of 2.73, suggesting a moderate level of temperature consistency in the observed data.

Table 3 Descriptive statistics for fire incidents, population and household characteristics and weather variables
Full size table

3.3 Method

The current study is completed in two parts using data described in Section 3.2. As shown in Fig. 2, the first part focuses on fire risk and fire service coverage analysis, completed in four steps. The second part includes analysing potential sites for fire station placement and is completed in two steps. Specifically, drawing on findings from the first part along with key data sets (such as land use and distance from the river) collected at the ward level, the second part initially identifies ward areas suitable for establishing fire stations in KMC. Selecting the top 10 ranked wards based on suitability scores derived from the fist part, finally an overlay analysis of land use and river network distance data is conducted to identify potential sites for fire station locations within the selected wards.

Fig. 2
figure 2

A flow chart demonstrating the two parts and the fire risk/service coverage analysis for the selection of potential sites for fire station location

Full size image

3.3.1 Part A: fire risk and fire service coverage analysis

  • The first step examines spatial patterns of urban fires by mapping fire rates per 1,000 population and fire rates per 1,000 buildings.

  • The second step examines vehicle dispatch to return time to base station to identify wards experiencing longer response times.

  • The third step examines the local bivariate relationship between fire incidents, population-household characteristics, and weather conditions.

    • ◦ Statistically significant local bivariate relationships between fire and selected variables are identified using local entropyFootnote 5 computed from local bivariate mapping.

    • ◦ Then, each ward is assigned a value of ‘1’ if their local entropy is below KMC’s average, indicating higher fire risk probability. This procedure is repeated for all variables, and the resulting values are summed across all wards for each variable.

    • ◦ The output derived from this analysis serves as a crucial criterion in the examination and identification of suitable sites for fire stations.

  • The fourth step involves examining the current fire service coverage using different (5,7,10 min) and distance (1 km, 2 km, 3 km) thresholds. This analysis uses location of building centroids (demand locations), the location of fire stations (supply locations), and a street network data set containing posted speed limits for each street segment and distance in metres to identify covered and uncovered areas within specified thresholds.

3.3.2 Part B: potential fire station site selection analysis

  • The spatial data obtained at the ward level in the first part of this study is used as an input into the ‘Suitability Analysis’ tool to generate a suitability score.Footnote 6 Based on this score, the 32 wards are ranked, with higher scores indicating wards with elevated fire risk and inadequate fire service coverage. In other words, a higher score signifies greater suitability for fire site selection due to higher potential for prompt and timely service delivery response. The suitability score is subsequently mapped using a quantile classification method across the ward areas of KMC. The top 10 ranked wards are selected for further investigation and the identification of local sites suitable for the placement of fire stations.

    The workflow for the suitability score is computed in four steps:

    1. 1.

      Make the suitability base layer (ward boundaries)

    2. 2.

      Add field-based criteria (see Section 4.5.1)

    3. 3.

      Set criteria properties (see Section 4.5.1)

    4. 4.

      Calculate suitability score (Eq. 1)

  • In the final step, an overlay analysis technique is used to identify sites suitable for potential fire station locations within the top 10 ranked wards. This includes overlaying land use data excluding built-up areas (such as open spaces, grasslands, farmlands, meadows, nature reserves, orchards, parks, and scrublands) and the river network (areas within 100 m of the river network) to identify ward-specific sites suitable for future fire station placement.

3.4 Data availability

The fire incident data provided by the KMC brigade is not publicly available, and the authors are not authorised to share this data without permission from the KMC, Nepal. For more details on data and their sources, refer to Section 3.2.

4 Results

The results of this research are categorised into five broad themes. These themes are discussed as follows.

4.1 Fire rate

Figure 3 shows the measure of the spatial pattern of the fire rate per 1,000 population and per 1,000 buildings in the KMC. Drawing on quantile classification method, the fire rate is mapped to identify areas with either high or low fire rates.

Fig. 3
figure 3

Fire rate: a per 1,000 population b per 1,000 building

Full size image

The results of the ward level fire rate analysis reveal that areas with higher concentration of population and building, particularly in the central parts of the KMC tend to exhibit higher fire rates. These findings are consistent with research conducted in the developed country, such as studies by Chhetri et al. (2018) who found higher fire rates in inner-city areas of Brisbane, Australia.

Results from the fire rate per 1,000 buildings shows a similar pattern to that of population, with higher rates observed in the city centre. Similarly, the fire rate per 1,000 population is higher in the western side of the KMC, indicating areas with greater human activity and population density. The differences in fire rates between population and building density highlight a comparatively different urban characteristics influencing fire risk. Areas with higher population density may experience increased fire risk due to human activity, while areas with greater building density are prone to fires associated with household factors (Raval & Motiani, 2022; Schachterle et al., 2012).

4.2 Vehicle dispatch to station return time

Table 4 shows the analysis of vehicle-dispatch-to-return time in minutes. On average, fire trucks take a minimum of 10 min to return to their base station after responding to fire incidents in KMC. Nearly 29 percent were responded by the fire services and returned to the station within 30 min, while 51 percent between 30 to 60 min, and almost 20 percent with more than 1 h. These results indicate significant delays in responding to emergency service calls in KMC. Vehicle-dispatch-to-return times tend to be longer in the city outskirts, particularly in the northern areas.

Table 4 Fire count and vehicle dispatch to return time (average minutes)
Full size table

4.3 Local bivariate analysis – fire, population, and housing characteristics

Initially, a scatter plot matrix was computed to select variables (detailed in Section 3.2) exhibiting a strong association with fire occurrence (Pearson’s correlation coefficient ‘r-value’ > 0.5). Previous studies suggest a minimum R-value threshold of 0.4 to indicate a strong association between factors contributing to fire occurrence (Ye et al., 2017). Therefore we selected a more stringent criterion of R-value greater than 0.5 for variable selection in this study. Out of the initial set of 14 variablesFootnote 7 selected for examining their local bivariate relationship with fire incidents, only nine variables revealed a strong correlation (Pearson’s correlation coefficient ‘r-value’ > 0.5) and were therefore selected for further analysis.

Results revealed a greatest positive relationship between fire and ‘Building Count’ (+0.67), followed by the ‘Rental population’ (+0.63), ‘Household lightening by electricity’ (+0.62), ‘Household with LPG fuel’ (+0.62), ‘Population’ (+0.62), ‘Economically inactive population’ (+0.61), ‘Children (< 18) with single mother’ (+0.60), ‘Unemployed population’ (+0.59), and ‘People with no school’ (+0.56).

The local bivariate relationships were quantified to examine how fire incidents are influenced by these selected variables across KMC. Statistical significance of these variable relationships was determined using local entropy computed from local bivariate mapping. The entropy value measures the degree of relationship between fire incidents and selected variables, with higher entropy values indicating greater unpredictability in the influence of a variable on fire occurrence.

To identify more predictable fire risks, wards with a local entropy value below the average entropy value of KMC were assigned a value of ‘1’. This iterative process was applied to all variables, and the resulting values were summed across all wards for each variable. Figure 4 illustrates the mapping output of this analysis, highlighting the south-eastern and primarily western parts of the KMC as regions where these nine variables collectively exhibit a stronger effect on the fire risk pattern.

Fig. 4
figure 4

Predictable fire risk areas

Full size image

4.4 Fire service coverage

Table 5 presents the results of the fire service coverage analysis conducted using three different time thresholds – 5 min, 7 min, and 10 min. Results reveal only 30 per cent of the 149,052 buildings in KMC are covered within a 5-min response time. The coverage increases to 44 per cent within 7 min and 68 per cent within 10 min.

Table 5 Fire service coverage of buildings in different time threshold in KMC
Full size table

Figure 5 illustrates the areas covered within these time thresholds (represented by light green for 5 min, green for 7 min, and dark green for 10 min). Notably, the majority of the eastern parts of the city are not covered even within a 10-min travel time from any of the three stations, resulting in almost 32 per cent of the building structures not covered even within 10-min from their nearest station. Similarly, 70 per cent of the buildings are not covered within 5 min, and 56 per cent are not covered within 7 min in KMC.

Fig. 5
figure 5

Travel time isochrones (5, 7, and 10 min)

Full size image

Similarly, Table 6 presents the percentage of building structures coverage within 1, 2, and 3 kms. Results reveal that only 14 per cent of the total 149,052 buildings are covered within 1 km, increasing to 40 per cent within 2 kms and 64 per cent within 3 kms. Figure 6 shows areas covered within 1, 2, and 3 kms by light green, green, and dark green colours, respectively. Notably, the majority of the eastern parts of the city are not covered even within a 3-km distance from any of the three stations, leaving nearly 36 per cent building structures without coverage within 3 kms of their nearest station. Similarly, 86 per cent of buildings are not covered within 1 km, and 60 per cent buildings are not covered within 2 kms in KMC.

Table 6 Fire service coverage of buildings in different distance threshold in KMC
Full size table
Fig. 6
figure 6

Travel distance isochrones (1, 2, and 3 kms)

Full size image

4.5 Fire station site selection

This section presents results from the analysis of fire station site selection in KMC.

4.5.1 Suitability analysis to identify wards

Spatial data captured at the ward level, including fire rate, vehicle dispatch to return time, combined entropy value, and fire service coverage, are used in the ‘Suitability Analysis’ tool to generate a suitability score. This score, mapped across KMC represents a spatial assessment of fire risk and the adequacy of fire service coverage across different wards. In this study, a higher suitability score indicates a greater level of fire risk and stronger need for fire stations in those wards.

Table 7 shows the application of input criteria properties, including thresholds (both minimum and maximum) and influence and unit of each criterion used in the suitability analysis tool to generate a site suitability score across 32 wards in KMC.

Table 7 Suitability analysis criteria, threshold, and influence
Full size table

For both fire rate per 1,000 population and building count, the minimum threshold is set at minimum value, and the maximum threshold is set at the maximum value. In simpler terms, the data between the first percentile and one hundredth percentile for these outputs is used for subsequent analysis, encompassing areas within KMC from the lowest to the highest fire risk.

Similarly, for vehicle dispatch-to-return time, exploratory regression output, service area coverage analysis (time and distance), and street connectivity, the minimum threshold is set at the minimum value, and maximum threshold is set at the maximum threshold. These minimum and maximum values are shown in Table 7. It is important to note that these criteria have a positive influence on the final suitability score, signifying that the higher the value of the variable, the more significant its impact on the final score. This is calculated using Eq. 1.

The results obtained from the application of the suitability analysis tool rank the 32 wards based on a score, presented as white to black colour scheme on a map (Fig. 7). The suitability analysis scores are classified using a quantile classification method, and the suitability scores are subsequently mapped in five categories: Lowest, Low, Medium, High, and Highest. Figure 7 depicts the spatial distribution of suitability scores across the wards of KMC. This figure, labelled by rank, illustrates that the top 10 ranked wards are predominantly located on the outskirts of KMC, with only few wards located in the central part of the city.

Fig. 7
figure 7

Suitability score (labelled by rank) for fire station location across 32 wards in KMC

Full size image
$$Site \,score \,for \,positive \,influence= \left(\frac{Variable \,value \,for \,site-Minimum \,value}{Maximum \,value-Minimum \,value}\right)$$
(1)

4.5.2 Overlay analysis to identify sites within the top 10 wards

Drawing on the top 10 ranked wards selected from the suitability analysis score, the identification of potential sites for future fire station locations is conducted through an overlay analysis of land use and river network distance data.

As shown in Fig. 8, suitable sites for future fire stations are represented in green. Interestingly, the wards ranked 2nd and 7th (ward numbers 22 and 24, respectively) according to the suitability analysis output do not contain any suitable sites for future fire stations. Conversely, the top-ranked ward, ward number 32, does contain a few suitable sites for future fire station construction. Table 8 provides the total number of sites and their respective areas across each of the top 10 ranked wards in KMC. As depicted in the table, ward ranked 5th (ward number 4) and ward ranked 3rd (ward number 15) contain the highest number of suitable sites for future fire station placement (16 and 17 sites, respectively). While there is currently no fire station in the south-eastern part of KMC, the results suggest that potential sites for establishing the fourth fire station in KMC could be within wards 31 or 32. These sites are located near to Tribhuvan International Airport and are adjacent to Bhaktapur district, offering emergency response resources to the airport and adjoining district as required.

Fig. 8
figure 8

Suitable sites (represented in green) within the top 10 wards for fire station

Full size image
Table 8 Count of sites for potential fire station location in KMC
Full size table

5 Discussion

This paper examined fire risk and service coverage scenarios, using fire incident data from Kathmandu city, to identify suitable sites for fire station locations. The main findings are summarised as follows.

  • Spatial analysis revealed high fire counts in the outskirts of KMC, but higher fire rate per 1,000 people and buildings in central areas. Mapping the spatial pattern of fire rates helps identify areas with varying fire risk.

  • The vehicle dispatch-to-return times range from 19 to 72 min, prolonged in the city’s outskirts, especially on the northern side. This highlights challenges in response due to traffic congestion, inadequate street design, and resource limitations, endangering people’s lives at significant risk during fire-related crises.

  • Statistical analysis revealed a positive relationship of fire with population characteristics such as rental population, household electricity usage, LPG fuel, economically inactive population, single mothers with children under 18, unemployment, and lack of schooling.

  • Mapping total entropy values from the local bivariate relationships revealed stronger combined fire risk effects in the south-eastern and western areas of KMC based on population, household characteristics, and weather conditions.

  • Fire service coverage analysis using various time and distance thresholds revealed insufficient coverage in the eastern parts of the KMC. Only 14 per cent of the buildings within 1 km of a fire station, increasing to 40 per cent and 64 per cent within 2 and 3 kms, respectively. Almost 32 per cent buildings are not covered within 10 min, increasing to 70 per cent and 56 per cent within 5 and 7 min, respectively.

  • The suitability analysis identified the top 10 wards for potential fire station locations, mostly in the outskirts of KMC, except for a few central wards. The overlay analysis revealed a total of 61 suitable fire station sites within these top-ranked wards, with wards 31 and 32 identified as potential locations for a new fire station due to high population and no existing stations.

The findings of this study have generated fresh insights that were previously not available in developed countries. A crucial implication drawn from the fire risk analysis is that population and building characteristics can be employed to predict fire risk. The results of this study indicate that the risk of fire increases in areas with more vulnerable populations, such as children living with single mothers, or in regions where socioeconomically disadvantaged populations are concentrated. Similarly, when more households with rental populations or unemployed individuals cook with LPG fuel, the risk of fire rises. The results in Nepal are underpinned by the previous fire studies in developing countries in which the geographical situation context in developing countries such as inadequate infrastructure, informal settlements, and substandard housing leads to the vulnerability of communities to fire incidents.

On the other hand, in line with the literature review drawn above, Nepal as a developing country faces constraints in the allocation of resources towards firefighting equipment, trained personnel, and emergency response infrastructure. These limitations contribute to prolonged dispatch-to-return times, often exceeding 1 h to attend to fire incidents. The inadequacy of resources poses a significant obstacle to the effectiveness of firefighting endeavours and serves to amplify the adverse effects of fire incidents.

As a result, fire authorities must comprehend how specific population characteristics, individual behaviour, and weather patterns interact to influence the risk of fire. To mitigate fire risk, fire authorities can use bivariate analysis to identify population and household characteristics associated with a higher risk of fire. This information can then be utilised to educate the public about fire safety and deploy fire prevention resources more effectively. Furthermore, the results of the fire service coverage analysis can be valuable in identifying areas covered or not covered within specified time/distance thresholds. This information is instrumental in shaping policies and plans related to the allocation of limited fire service resources, including the establishment of future fire stations. Considering these findings, it is also suggested that KMC should establish a service delivery standard or response time standard as a guideline for fire responders. This will ensure timely responses and identify areas where additional resources can be allocated systematically and responsibly to enhance public safety.

6 Conclusion

In summary, this study has revealed spatial patterns of fire risk that align with previous research on the topic (Chhetri et al., 2010, 2018). The results demonstrate positive relationships between population and housing characteristics and fire incidents.

There are several limitations associated with the current study. This study was based on fire incident data less than four years old. Future studies could address this limitation by using longer-term data to explore spatial-temporal patterns of fire incidents and investigate the underlying drivers of fire risk through regression modelling. While this study employed suitability analysis and overlay analysis techniques, future research could also assess fire station service coverage and utilize location optimization techniques to determine the optimal number and locations for fire stations.

The findings indicate higher fire rates in the central areas of KMC, with a frequent incidence of fires in the outskirts, particularly in the eastern regions where existing fire stations lack coverage even within a 10-min response time. These findings emphasize the need to prioritize fire risk prevention and mitigation activities in these areas of KMC.