Problem Statement
This project evaluates how effectively existing fire stations in Tucson, Arizona, serve fire-related incidents and surrounding neighborhoods using network analysis. The study aims to assess current emergency response coverage, identify areas with limited access to fire services, and determine where an additional fire station could be located to improve service coverage. Three network analysis methods will be used: Closest Facility, Service Area, and Location-Allocation.

Research Questions

Method 1: Closest Facility

Which Tucson fire station can respond most quickly to each simulated fire incident based on travel time along the road network?

Method 2: Service Area

Which areas of Tucson fall within a 4- or 6-minute drive of existing fire stations?

Method 3: Location-Allocation

If Tucson were to add one new fire station, where should it be located to maximize population coverage within a 6-minute drive?

Data
Study Area

The study area for this project is Tucson, Arizona. The analysis extent was cropped to focus on the urbanized portion of Tucson, where the majority of the population resides, while excluding large areas of open land with little to no residential demand. This helped concentrate the analysis on the areas most relevant to fire service access and emergency response. It also improved map readability by making the spatial relationships among fire stations, simulated fire incident points, and census tract centroids easier to interpret.

Data

This project used two primary feature class datasets for analysis: Fire Stations and Census Tracts (2020). Additional layers, including Speed Limits and Protected Lands of Pima County, were included to support map visualization and help identify areas unsuitable for potential new fire station locations. All datasets were obtained from Pima County government GIS sources and were clipped to the study area boundary before analysis.

Metadata:

Name of Dataset: Fire Stations

Publication/Update Date: February 24, 2021 

Geometry Type: Point Feature Layer

Owner: Pima County GIS

Description: Data set credit to Pima County Information Technology Department - Geographic Information Systems. This was used as the facilities layer in all method analyses.

Name of Dataset: Census Tracts - 2020

Publication/Update Date: June 13, 2022

Geometry Type: Polygon Feature Layer

Owner: Pima County GIS

Description: 2020 Census Tract boundaries as supplied by the U.S. Census Bureau. This was used to determine which fire stations service the "centroid" in each tract.  

Name of Dataset: Speed Limits

Publication/Update Date: February 24, 2021 

Geometry Type: Line Feature Layer

Owner: Pima County GIS

Description: Speed limit ordinance for the City of Tucson. Used to add streets to the visual map.  

Name of Dataset: Protected Lands of Pima County

Publication/Update Date: June 3, 2024

Geometry Type: Polygon Feature Layer

Owner: Pima County GIS

Description: Parcels and land units managed for the preservation of biological values. Data set credit to Pima County Information Technology Department - Geographic Information Systems. This dataset was used to ensure candidate fire station points did not overlap with protected lands.

Methods
All datasets were projected in NAD 1983 HARN StatePlane Arizona Central FIPS 0202 (Int| Feet) (WKID 2868). This coordinate system is widely used for local GIS datasets in the Tucson area and was the original projection of the Pima County data used in this study. Using a consistent local projected coordinate system improved spatial alignment among all layers and supported accurate mapping and network-based analysis within the study area.

Method 1: Closest Facility

The Closest Facility network analysis tool was used to identify which existing Tucson fire station could respond most quickly to each simulated fire incident. Twenty simulated fire incident locations were generated in developed areas of the study area to represent likely locations of fire-related emergencies. Developed areas were identified based on residential and built-up land cover visible on the basemap. In the analysis, the 47 existing Tucson fire stations were used as the 'Facilities', and the simulated fire incidents were used as the 'Incidents'. The ArcGIS Online network solver was then used to calculate the fastest route along the road network and assign the closest responding station to each incident.

For this analysis, the travel mode was set to 'Driving Time', and the travel direction was set to 'Away from facilities' to represent emergency vehicles leaving stations and traveling to incident locations. One facility was assigned to each incident so that the output identified the single fastest responder. Travel time was used as the primary impedance because response time is a critical factor in emergency service accessibility. This method assumes that emergency vehicles follow the available road network and that travel times are represented accurately in the ArcGIS Online network analysis. Because the incident points were randomly generated rather than based on actual fire records, the results represent a modeled pattern of response rather than observed emergency calls. The analysis also assumes that the street network and default travel-time settings reasonably reflect local travel conditions and that no unusual delays, such as congestion or road closures, affected response.

Method 2: Service Area

The Service Area network analysis tool was used to identify the areas around the 47 Tucson fire stations that could be reached within 4 or 6 minutes of travel time. These service areas represent the geographic extent of fire station coverage based on the road network rather than straight-line distance. For this study, 4-minute and 6-minute service areas were created to evaluate short emergency response times and compare differences in coverage across the Tucson study area. These time thresholds were selected to reflect a reasonable range for urgent fire and emergency response, supported by emergency response literature or agency guidance (District of Columbia Fire and Emergency Medical Services Department, n.d.).

For this analysis, the travel mode was set to 'Driving Time', and the travel direction was set to 'Away from facilities' to represent emergency vehicles traveling away from fire stations. The service area cutoffs were set to 4 and 6 minutes, and output polygons were generated as overlapping rings to show differences in coverage by travel-time range. This method assumes that emergency vehicles travel along the mapped road network and that the default travel times in ArcGIS Online reasonably reflect local travel conditions. It also assumes that no unusual delays, such as congestion, road closures, or dispatch delays, significantly affected travel time.

Method 3: Location-Allocation

The Location-Allocation network analysis tool was used to evaluate fire station coverage across Tucson and identify a potential location for one additional fire station that could improve service to underserved areas. Census tract centroid points were used as demand points to represent population distribution across the study area, and each centroid was weighted by total population. The 47 existing fire stations were loaded as 'Required' facilities, while six simulated potential station locations were loaded as 'Candidate' facilities. Candidate locations were placed near major intersections and excluded from residential areas and protected lands to better represent realistic siting options.

For the analysis, the Make Location-Allocation Analysis Layer tool was used to establish the network settings. The travel direction was set to 'Away from facilities', and the problem type set set to 'Maximum Coverage'. The number of facilities to find was set to 48 so that the analysis would retain the 47 existing fire stations and select one additional candidate location for comparison of coverage between facilities and demand points. A 6-minute travel-time cutoff was applied to match the response threshold used in the previous analyses. This method assumes that emergency vehicles travel along the mapped road network, that ArcGIS Online travel times reasonably reflect local conditions, and that census tract centroids and simulated candidate locations provide an acceptable representation of demand and potential station locations. 

Results
Method 1: Closest Facility

The Closest Facility analysis showed that the 20 simulated fire incidents were served by 17 different Tucson fire stations, indicating that most incidents were matched with a unique closest responder. Of the stations involved, three stations were identified as the quickest responder for two separate incidents, while the remaining responding stations were assigned to one incident each. Estimated travel times ranged from approximately 2.5 minutes to more than 7 minutes, suggesting some variation in response accessibility across the study area. The results also showed no clear relationship between travel time and distance traveled, indicating that the structure and connectivity of the road network influenced response more than straight-line or total route distance alone. Overall, the Closest Facility analysis demonstrates that while much of Tucson appears to have relatively efficient fire station access, some locations may experience longer modeled response times depending on their position within the transportation network. 

Method 2: Service Area

The Service Area analysis showed that the fire station coverage areas varied considerably in both size and shape across Tucson. The 4-minute and 6-minute service area polygons were strongly influenced by the structure of the road network generated by the ArcGIS Online solver, rather than forming uniform circular buffers around each station. In areas with faster travel speeds and stronger street connectivity, the service areas extended farther and covered more land within the same amount of time. In contrast, service areas were more constrained in locations with less direct road access or fewer connecting routes. The 6-minute service areas consistently expanded beyond the 4-minute areas, illustrating how a small increase in travel time substantially increased potential coverage. Overall, the results suggest that fire service accessibility in Tucson is shaped not only by the location of fire stations, but also by the speed and connectivity of the surrounding transportation network.

Method 3: Location-Allocation

The Location-Allocation analysis showed that fire station coverage was uneven across the census tract centroid demand points within the 6-minute travel-time threshold. Of the 47 facilities included in the model, 36 stations were connected to at least one demand point, while the remaining 11 facilities did not capture any centroid within the cutoff. Likewise, 74 of the over 200 census tract centroids were not matched to any facility, indicating that roughly 35% of demand locations fell outside the modeled 6-minute service range. The number of centroid connections per facility varied considerably, ranging from 0 to 12, which suggests that some stations served multiple demand points while others contributed little or no coverage under the selected settings. The chosen candidate location was connected to 9 census tract centroids, indicating that it improved coverage in an area where additional service was needed. Overall, the results suggest that existing fire station coverage is not evenly distributed across Tucson and that the selected candidate site could help reduce gaps in accessibility for some underserved demand points. 

Conclusion
The three network analyses showed fire service accessibility in Tucson varies across the study area and is strongly influenced by both the location of existing fire stations and the structure of the road network. The Closest Facility analysis showed which station could respond most quickly to each simulated fire incident. Most incidents were assigned to different stations, although three stations were identified as the quickest responder for two incidents each. Travel times ranged from about 2.5 minutes to more than 7 minutes, suggesting that while many locations had relatively quick modeled response times, some incidents were less accessible. The results also showed no clear relationship between travel time and route distance, indicating that road connectivity and travel conditions were more important than distance alone.

The Service Area analysis identified which parts of Tucson fell within 4- and 6-minute drive times of existing fire stations. The service area polygons varied greatly in shape and extent, showing that coverage depended heavily on the street network rather than simple proximity. In many parts of central Tucson, service areas overlapped and extended across large connected road corridors, while some outer areas showed more limited coverage. These results suggest that existing fire stations provide stronger accessibility in more connected parts of the city and less consistent coverage toward the edges of the study area.

The Location-Allocation analysis determined where one additional fire station could improve modeled coverage within a 6-minute travel-time threshold. The results showed that not all demand points were served by the existing fire station network within the cutoff travel-time. The selected candidate location connected to 9 census tract centroids, suggesting that it could improve access in an underserved area. Overall, the three methods worked together to show current response patterns, map existing service coverage, and test where a future station could provide the greatest benefit. Taken together, the analyses suggest that Tucson’s fire service network provides substantial coverage in many areas, but some gaps remain that could potentially be reduced through strategic facility expansion.

Limitations

Several limitations should be considered when interpreting these results. First, the analysis relied on datasets that may not fully reflect current conditions. The fire station layer was last updated in 2021, and the census tract dataset was based on 2020 Census boundaries and population patterns. Since Tucson may have experienced population growth, land development, or changes in fire service infrastructure since those dates, the results may not represent present-day demand or station availability exactly.

Additional limitations are related to the assumptions built into the analysis itself. Fire incidents were simulated rather than based on actual emergency call records, so the Closest Facility results represent modeled scenarios rather than observed events. Likewise, census tract centroids were used as simplified demand locations, even though population is not evenly distributed within each tract. Travel times were based on ArcGIS Online network settings and did not account for real-time traffic, road closures, dispatch delays, or emergency vehicle behavior. Finally, the candidate fire station locations were simulated and may not reflect actual parcel suitability, cost, or land-use constraints. These limitations mean that the findings are best interpreted as a planning exercise rather than an exact prediction of real-world fire response.

References

District of Columbia Fire and Emergency Medical Services Department. (n.d.). Fire response time. DC.gov. https://fems.dc.gov/page/fire-response-time 

Pima County GIS. (n.d.). Census tracts - 2020 [Data set]. Pima County Open GIS Data. https://gisopendata.pima.gov/datasets/PimaMaps::census-tracts-2020/about 

Pima County GIS. (n.d.). Fire stations [Data set]. Arizona Geospatial Data Hub. https://azgeo-open-data-agic.hub.arcgis.com/datasets/PimaMaps::fire-stations/explore?location=31.980000%2C-111.875000%2C8 

Pima County GIS. (n.d.). Protected lands of Pima County [Data set]. Pima County Open GIS Data. https://gisopendata.pima.gov/datasets/PimaMaps::protected-lands-of-pima-county/about 

Pima County GIS. (n.d.). Speed limits [Data set]. Pima County Open GIS Data. https://gisopendata.pima.gov/datasets/speed-limits/explore?location=31.980000%2C-111.875000%2C8