Skip to main content

Fire Station Siting

Maximum Coverage Location Problem (MCLP) · NP-Hard

A city fire department must locate p stations to maximize population covered within a 4–8 minute response radius. Each uncovered zone of 10,000 residents faces a 40% higher fire fatality risk. The MCLP trades off coverage breadth against station construction cost.

Where This Decision Fits

Public safety operations — the highlighted step is what this page optimizes

InfrastructureStation siting
Service SchedulingCrew rostering
EmergencyDispatch routing
Resource AllocationFleet & equipment

The Problem

Select p station sites to maximize population covered within response radius r

A metropolitan fire department evaluates 7 candidate station sites to serve 12 demand nodes (census blocks) with populations ranging from 500 to 5,000 residents. The department must open exactly p stations (constrained by budget) such that the total population within the coverage radius r of at least one open station is maximized.

This is the classical Maximum Coverage Location Problem (MCLP), introduced by Church & ReVelle (1974) and proven NP-hard by reduction from Set Cover. Unlike the p-Median problem which minimizes average distance, the MCLP maximizes the number of demand nodes (weighted by population) that fall within a critical service distance threshold.

MCLP Formulation maximize   Σi di · zi   // total population covered
subject to
  Σj∈Ni yj ≥ zi   ∀ i    // node i covered only if a station in Ni is open
  Σj yj ≤ p    // open at most p stations
  yj ∈ {0, 1}   // station open or not
  zi ∈ {0, 1}   // node i covered or not

Where di is the population (demand) at node i, Ni is the set of candidate sites within coverage radius r of node i, yj is the binary open/close decision for station j, and zi indicates whether node i is covered.

See Integer Programming theory

Try It Yourself

Choose p, coverage radius, and an algorithm to find the best station locations

Fire Station Siting Optimizer

7 Sites · 12 Nodes
2
0.18
SiteLocationCost ($M)
BlockPopulationRisk Zone

Ready. Click “Solve & Compare All Algorithms” to run.

AlgorithmPop. Covered% CoveredUncoveredTime
Click Solve & Compare All Algorithms
Station details will appear here after solving.

The Algorithms

Three approaches to the Maximum Coverage Location Problem

Constructive

Greedy (Max New Coverage)

O(p · n · m)  |  (1 − 1/e) approximation

Start with no stations open. At each step, evaluate the additional population newly covered by opening each remaining candidate site. Open the site with the greatest marginal coverage gain. Repeat until exactly p stations are open. This greedy approach has a provable (1 − 1/e) ≈ 63% approximation guarantee for submodular maximization under a cardinality constraint.

Local Search

Greedy + Swap

O(iter · p · (n−p) · m)  |  Local optimum

Begin from the greedy solution with p open stations. At each iteration, try every possible swap: close one open station, open one closed station. Accept the swap that yields the largest increase in covered population. Repeat until no improving swap exists. This local search often improves beyond the greedy baseline and can escape suboptimal early choices.

Baseline

Random Selection

O(m)  |  No guarantee

Randomly select p stations from the candidate set. This serves as a baseline to demonstrate the value of optimization. Repeated 50 times, keeping the best random solution found. While occasionally competitive on small instances, random selection typically performs significantly worse than even the simplest greedy heuristic.

Real-World Complexity

Why practical fire station siting goes beyond the basic MCLP model

Beyond the MCLP

  • Travel time variability — Road networks, traffic congestion, and time of day create non-Euclidean response times that differ from straight-line distance
  • Backup coverage — When a unit is dispatched, remaining stations must still provide adequate coverage to the rest of the city
  • Multiple apparatus types — Engine companies, ladder trucks, and rescue squads have different coverage requirements and response standards
  • Call frequency variation — Certain census blocks generate far more emergency calls per capita than others due to building density and demographics
  • Mutual aid agreements — Neighboring jurisdictions provide reciprocal coverage, changing effective station placement needs at borders
  • Land use constraints — Zoning restrictions, property availability, and environmental regulations limit feasible station sites
  • Equity mandates — Political requirements may demand minimum service levels in all districts, not just maximum total coverage

Key References

Foundational works in coverage location and emergency service optimization

  • Church, R. & ReVelle, C. (1974). “The Maximal Covering Location Problem.” Papers in Regional Science, 32(1), 101–118.
  • Toregas, C., Swain, R., ReVelle, C. & Bergman, L. (1971). “The Location of Emergency Service Facilities.” Operations Research, 19(6), 1363–1373.
  • Daskin, M.S. (1995). “Network and Discrete Location: Models, Algorithms, and Applications.” John Wiley & Sons.

Need to optimize your emergency service network?

From fire station siting to ambulance deployment and emergency dispatch, mathematical modeling can transform public safety infrastructure decisions. Let’s discuss how Operations Research can work for you.

Disclaimer
Data shown is illustrative. Site names, locations, populations, and risk zones are representative scenarios for educational purposes and do not reflect any specific city or fire department.
Back
ESC