Abstract
Aim: The aim of this article is to demonstrate the applicability of historical emergency-response data – gathered from decision-support systems of emergency services – in emergency-response statistical modelling.
Project and methods: Building models of real phenomena is the first step in making and rationalising decisions regarding these phenomena. The statistical modelling presented in this article applies to critical-event response times for emergency services – counted from the moment the event is reported to the beginning of the rescue action by relevant services. And then, until the action is completed and services are ready for a new rescue action. The ability to estimate these time periods is essential for the rational deployment of rescue services taking into account the spatial density of (possible) critical events and the critical assessment of the readiness of these services. It also allows the assessment of the availability of emergency services, understood as the number of emergency teams which ensure operational effectiveness in the designated area. The article presents the idea of modelling emergency response times, the methods to approximate the distribution of random variables describing the individual stages and practical applications of such approximations. Due to editorial limitations, the article includes the results only for one district (powiat – second-level unit of local government and administration in Poland).
Results: A number of solutions proposed in the article can be considered innovative, but special attention should be given to the methodology to isolate random variables included in the analysed database as single random variables. This methodology was repeatedly tested with a positive result. The study was based on data on critical events and emergency response times collected in the computerised decision-support system of the State Fire Service (PSP) in Poland.
Conclusions: Presented in this article, the method of approximating the duration of individual stages of emergency response based on theoretical distributions of random variables is largely consistent with the empirical data. It also allows to predict how the system will work in the short-term (over a time span of several years). The predictive property of such modelling can be used to optimise the deployment and to determine the capabilities of individual rescue teams. These studies were conducted between 2012 and 2015 as part of a project funded by the National Centre for Research and Development (NCBR), agreement No. DOBR/0015/R/ID1/2012/03.
Keywords: statistical modelling, data mining, emergency services, designing rescue systems
Type of article: original scientific article