Traditionally, station location models have relied on the concept of “coverage” to determine the best locations. The main drawback to these models is the binary nature of “coverage,” where demand locations are either “covered” or “uncovered".
What does this mean?
If a demand point can be reached in a target response time, the demand point is considered covered. For example, these models say that if a call is 7:59 minutes away, it is covered, but if it is 8:01 minutes away, it is uncovered. This can often bias station spacing, encouraging sub-optimal placement that hinders your service to the community. As well, this measurement is provides no accurate way to project the impact of new resources on your performance targets and standards.
Binary Coverage has 3 primary drawbacks:
Biased Station Locations
Underestimation of Necessary Stations
Poor Performance Prediction
Binary vs. Probabilistic coverage models
Consider, for example, figures 1 and 2. In both figures, the centre of the circle is the location of the emergency services stations.
Figure 1 is a representation of a Binary Coverage Model. The dark circle represents the area around the station which is “covered” by that station. The light grey circle is the area around the station which is “uncovered.” As you can see, with a binary coverage model, all demand from node B would be considered “covered,” whereas all demand from node A would be considered “uncovered.”
Figure 2 is a representation of a Probabilistic Coverage Model. In this figure we represent coverage as a gradient, where demand nodes closer to the station have improved coverage compared to those that are further away. Instead of evaluating only the travel time to assess coverage, we model probability that a unit can arrive at a location within a time target. As you can see, with a probabilistic model all demand from both nodes A and B would receive coverage; however node B would receive better service.
Use expected performance when possible.
This approach has been shown to be very accurate in terms of predicting coverage changes as it accommodates the realistic variability that occurs. Whenever possible, you should use your expected performance when re-configuring stations. This way, you will not only place them where needed most, but you will be able to more accurately predict the impact of adding or moving stations.