Modelling the likelihood of urban residential fires considering fire history and the built environment: A Markov Chain approach

Abstract

In this paper, we address the complex question of how the occurrence rate of residential structure fires in Melbourne city are influenced by built-environment structural forms and/or the recent history of fire incident occurred within the neighbouring areas. Numerous studies have used the sociodemographic and economic characteristics to explain the spatial variability in residential fire occurrence rates. There is however less published research that links spatio-temporal variation of residential fire occurrences with patterns and changes in the built environment, or which seeks to quantify the spatial effect of fire events on the subsequent rate of incidents within the local area. We develop a spatio-temporal model of residential fire occurrence based on a range of spatial characteristics and past fire occurrences within neighbourhood. These spatial characteristics include the Index of Relative Socioeconomic Advantage and Disadvantage (IRSAD), residential density (i.e. the relative number of dwelling per unit area), percentage of owned dwellings, percentage of privately rented dwellings, percentage of publicly rented dwellings, percentage of residents moved in the last five years, and percentage of residents moved in last year. The model is fitted to fire incidence data from Melbourne, Australia, gathered over a 10year period. Results show that the distribution of residential structure fires across Melbourne is a complex pattern and is associated with spatially-varying indicators. The inner suburbs of the Melbourne region are more fire prone than others. Those areas have high probability of fire occurrence. This naturally follows not only from built environment and socio economic characteristics, but also correlates with recently-located residents as the tenure status in those areas. Households that have recently moved into an area, and households consisting of temporary residents, have been demonstrated in prior studies to exhibit an elevated likelihood of fire occurrence. The analysis also capture that there is a neighbourhood "memory" effect of fires, with respect to fire occurrence rates. The results contribute to an evidence base which may be useful for emergency planners and fire agencies seeking to build appropriate strategies to mitigate fire effects on communities. It also aids in assessing and classifying areas in terms of fire occurrence likelihood, and in determining when to circulate fire safety information to residents so as to retain preparedness and awareness of fire incidents.