Using data to boost urban fire safety

This summer, fires once again dominated headlines in Canada, and not just in forests.

Cities are increasingly adopting fire risk prevention and planning programs to help identify vulnerable areas and speed up emergency response.

In a new project, Nooshin Salari, an assistant professor of Operations Management, is applying her expertise in data-driven decision-making to the challenge of mitigating urban fire risk in Chicago.

We spoke with Salari about what the Chicago analysis reveals and how the approach could be adapted for Canadian cities.

Tell us about your research on urban fire risk in Chicago. What are your goals?

My research focuses on understanding where urban fires are most likely to occur, who is most exposed and where the consequences could be most severe, with the ultimate goal of supporting more targeted fire-prevention strategies.

The project is part of a Mitacs Accelerate collaboration with Darkhorse Analytics. We analyzed more than 93,000 fire incidents reported between 2017 and 2024 in Chicago, and integrated these records with demographic, socioeconomic, housing, built-environment and other spatial data at the census-block and tract level.

Our analysis goes beyond simply identifying fire hotspots. We estimate the likelihood of a fire incident for each tract and then examine population exposure, including the number of residents, housing units, older adults and children potentially exposed to fire risk.

The longer-term goal is to develop a transferable, data-driven framework that cities can use to distinguish different types of fire risk and vulnerability. Rather than applying the same prevention strategy everywhere, the framework could help identify which areas may benefit from more intensive inspections, smoke-alarm or education programs, code-enforcement activities, improved emergency access or other targeted community risk-reduction measures.

What is ‘redlining,’ and what does it mean for fire risk?

Redlining refers to discriminatory housing and lending practices, formalized in the 1930s, through which neighborhoods — often those with Black, immigrant and other marginalized populations — were graded as higher financial risk. These classifications contributed to restricting access to mortgages and investment and reinforced patterns of segregation and neighborhood disinvestment.

However, we are careful not to interpret this as evidence that redlining itself directly causes fires. Our goal is to understand how these historical and current conditions come together to create persistent patterns of fire risk and, ultimately, how that knowledge can help cities better target prevention and community risk-reduction efforts.

What types of neighbourhoods are more vulnerable to fire risk? Why?

Our results show that fire risk is not concentrated in just one type of neighbourhood. We see different patterns depending on whether we consider fire burden relative to population or the geographic concentration of fires.

We also observe associations with the racial and ethnic composition of neighborhoods, which likely reflect broader historical and structural inequalities rather than characteristics of individual residents.

The built environment also matters. Areas with higher structure density tend to have more fires per unit of land, while areas with higher proportions of older housing tend to experience greater fire burden relative to population.

But frequency is only one part of the picture. A neighbourhood with relatively few fire incidents may still be highly vulnerable if a fire is more likely to result in serious consequences.

How will this framework be used to reduce fire risk in cities?

The framework is designed to help cities look beyond simply identifying where fires occur most frequently. For example, some census tracts may experience a relatively high number of fires but have good access to emergency services and building characteristics that help limit the severity of those incidents. In another tract, fires may occur less frequently, but factors such as older or more vulnerable buildings, limited transportation access, longer access to emergency services, or greater social vulnerability could increase the potential consequences when an incident occurs.

We can also use historical fire records to estimate these consequences more directly. Our data include measures such as the dollar value of property losses and the number of people injured or killed in individual fire incidents. By combining this information with neighborhood characteristics, we can examine not only where fires are more likely to occur, but also where a fire, if it occurs, may lead to greater human or economic consequences.

This distinction is important for policy. Different neighborhoods may require different interventions.

Can you comment on whether the same trends might hold true for Canadian cities? Do you have plans to expand your research into Canada?

We are interested in expanding this research to Canadian cities. Rather than transferring the Chicago model directly, we would apply the same overall framework using local fire, census, housing, and infrastructure data and recalibrate the models for each city.

The framework itself is transferable because it focuses on three core components: the likelihood that a fire occurs, the severity and consequences if a fire occurs, and the vulnerability of the population and built environment. By applying the same framework across different cities, we can identify which drivers of urban fire risk are common and which are specific to local conditions. Ultimately, the goal is to develop an adaptable framework that helps cities identify where fires are more likely, where consequences could be most severe, which populations may be more vulnerable, and where preventive resources could have the greatest impact.