Predicting the Demand for Fire Service

An analysis of New York City fire alarm distribution, a description of some analytical methods, and an example of an application of estimated alarm incidence rates. Data show steady exponential increases in nonstructural fires, excluding brush fires. Using a Poisson distribution, the authors describe a method for short-term prediction of incidence rates for various types of fire alarms as a function of location, time, method of reporting, and weather. Predicted incidence applications are as varied as long-term problems of new station-house location, selection of relocation and prepositioning strategies, and actual dispatching of units to incoming alarms. The example addresses the problem: Given the location and number of units to be dispatched, which should be assigned to a particular alarm? Results include (1) formulas for unit workload and average response time to all incidents as a function of response districts; (2) determination of a district boundary that minimizes average response time; and (3) conditions wherein the equidistant boundary is dominated by others.