Modelling the Factors Affecting Urban Residential Fires - A Case Study of Melbourne

The perils of residential fires in urban areas carry immense implications for human lives, affecting individuals through property damage, physical harm, enduring psychological distress, and fatalities. Residential fires are not only physical hazards but also socially constructed phenomena. Socio-urban structures, economic disadvantage, and health and well-being play a critical role in shaping spatial variability in fire risk. Although the existing body of knowledge has documented the factors contributing to these fire risks, the interrelations and interactions of these complex structural variables that render a community vulnerable to the residential fires are yet to be theorised, measured, and modelled. Community vulnerability is an exposure to the adverse socio-economic environment due to a lack of accessibility to capital resources, education, employment opportunities, safer housing, and health amenities. While these factors themselves do not directly increase the likelihood of fire, they create adverse circumstances that affect the community's exposure to fire risk and impede their capacity to cope. This research develops a model of structural vulnerability by integrating two risk management frameworks: the Disaster Risk Community Framework and the Disaster Resilience of Place Model (DROP). This model has not only integrated theorises but also developed a broader framework to examine the impact of precursor conditions in the structural system, such as the social, natural, health, and built environment systems on residential fire risk. These systems reflect both inherent vulnerability and resilience of the community to affectively respond to fire risk. This thesis generates a large dataset of fire incidents obtained from the Fire Rescue Victoria (FRV), integrated with the census data obtained from the Australian Bureau of Statistics (ABS) and the Australian Urban Research Infrastructure Network (AURIN), and health data obtained from the Department of Health and Victorian Agency for Health Information (VAHI). A Structural Equation Modelling (SEM) technique is applied to examine the effect of key factors of community vulnerability on residential fire density, a proxy spatial measure for fire risk. Fire density is measured as the number of residential fires per 1,000 people in an area. Key constructs in the SEM include social dependency-oriented family structures, culturally diversified communities, high dwelling density neighbourhood, high economic disadvantage, low levels of education, and debilitating health conditions. The results reveal significant spatial variability in fire density levels across the Melbourne Metropolitan area. The research finds a statistically significant effect of social dependency-oriented family structures, culturally diversified communities, high dwelling density neighbourhood, high economic disadvantage, low levels of education, and debilitating health conditions on increased fire density. Suburbs with a higher concentration of flats or apartments are associated with a heightened risk of fires, while suburbs primarily comprised of detached houses and residents with higher levels of education tend to experience lower fire risks. The research also discovers that the presence of significant economic disadvantage and debilitating health conditions partially mediates the relationship between culturally diversified communities and fire density. However, the effect of vulnerable demographic characteristics and low levels of education on fire density was found to be insignificant. The findings of this research make several contributions and have implications for fire rescue authorities and policymakers. The research proposes a structural vulnerability model by integrating the Disaster Resilience of Place Model and the Disaster Risk Community Framework, thereby advancing theoretical understanding of the complexity of factors affecting residential fire density patterns. Additionally, the research incorporated health-related factors to the model, adding a new factor to understanding community vulnerability to residential fires. Methodologically, this is the first study that employs Structural Equation Modelling (SEM) to examine direct, indirect, and mediating relationships between community vulnerability factors and fire density. This research proposes an emergency response framework to identify and prioritise areas in Melbourne. This is developed by creating a two-by-two matrix based on residential fire density and response time to identify areas requiring immediate attention and resource allocation to prevent potential delays in emergency response. This research provides an empirical basis for formulating future operational strategies by allocating finite resources to areas to mitigate the adverse impacts of fire on vulnerable communities. The impact of residential fire can be minimised through formulating area-based policies and need-driven resource allocation to build resilience of highly vulnerable communities with a lower coping capacity. There are a few limitations to the methods applied in this research. First, the research relies on secondary data to analyse the impact of structural community vulnerability factors on spatial variability in fire density. To achieve a more comprehensive understanding of structural community vulnerability and its effect on increased residential fires, future research will integrate both primary and secondary data to model relationships at an individual level instead of area. Second, the research is based on an aggregated spatial data that potentially undermines differences in individuals’ perception, attitude and behaviour that shapes the perception of residential fire risk. Finally, the areas are differentiated based on fire density levels and response time to fire incidents reported by FRV, which is a measure of accessibility. In future research, more advanced methods such as the 2-step Floating Catchment Area will be used to accurately calculate spatial accessibility from fire stations.