Predicting Future Flames

An eerie haze filled the skies of much of the American Northeast. Once picturesque blue skylines instead turned a post-apocalyptic orange as smoke weighed heavy in the air and in the lungs of people. Air pollution is nothing new for industrial cities, but this smog was different. It’s powered not by industry, but raging wildfires hundreds of miles away in Canada.

2023 could be the worst year on record for Canadian wildfires. Eight hundred square miles had burned by the end of May; 13 times the 10-year average for that time of year. By mid June, nearly 450 wildfires were active with 218 considered out of control. Thousands were under evacuation orders.

It’s a scene that is disturbingly familiar for those who witnessed and survived the 2018 Camp Fire in Northern California’s Butte County. Deemed the deadliest and most destructive wildfire in California’s history, it killed 85 people and destroyed more than 18,000 structures across 240 square miles. The destruction was unprecedented with towns essentially left with only foundations and ash.

Unlike other natural disasters such as earthquakes, accurate modeling of wildfires, while they do exist, requires supercomputers and hours to days to complete. By then the damage is already done, communities are destroyed, and lives lost. One group of researchers with the Critical Infrastructure Resilience Institute (CIRI) at the University of Illinois Urbana-Champaign, a U.S. Department of Homeland Security Center of Excellence, plan to make it both easier and faster to model wildfires.

Civil and Environmental Engineering professor Paolo Gardoni, along with postdoctoral researcher Armin Tabandeh and PhD student Sourangshu Ghosh, are developing models to not only predict wildfire propagation and its damage to critical infrastructure, but also offer real-time predictions of wildfire behavior and infrastructure damage for ongoing fires to help optimize the management of resources, as well as offer insight for future infrastructure plans in communities and their wildfire risks.

Tabandeh first considered the idea years ago, inspired by a talk on ice sheet flow at the North Pole. Both he and Gardoni worked on the idea of wildfire propagation in their spare time before applying for funding through CIRI. The institute focuses on dangers faced by the nation’s critical infrastructure systems. That includes natural disasters.

“We have done work in the past on earthquakes, we have done work in the past on storm surge for hurricanes. We’re building on that previous work and modeling techniques that we have developed for other hazards and now we’re tackling a more challenging one, which is exactly the spreading of wildfires,” said Gardoni.

That challenge is threefold: 1) compiling every possible factor that could feed a fire and how they could act, 2) gathering and integrating diverse datasets from experiments, environments, and past wildfires to calibrate the models and 3) finding a mathematical and scientific method to not only accurately take those factors into account, but to do so quickly and efficiently. The third point took some thought.

“We don’t want to sacrifice too much accuracy for the sake of fast predictions, and we don’t want to just go for physics and then sacrifice fast computation. We are trying to stay somewhere in the middle,” said Tabandeh.

The answer: models that are physics-based utilizing a Hamilton-Jacobi partial differential equation that are customizable to actual conditions and allow for researchers to obtain timely results with a high level of precision. The equation helps determine the rate of spread of a fire which can tell researchers which direction a fire may go at any single point of the fire front, taking into account not only physics, but also environmental factors.

“We are able to absorb all that data and update model predictions. We can say, for example, in the next day, in the next week, [the fire] is going to go in this direction, it’s going to have this impact. Then whoever is in charge of decisions can use that information to assign resources,” said Tabandeh.

To get to that point, researchers need to calibrate the models first with datasets. Researchers are pulling from multiple sources including the National Oceanic and Atmospheric Administration and other agencies to gather and analyze weather patterns, vegetation types, topography, historical wildfire records, and others to ensure the models are accurate.

Several factors can help a wildfire spread: vegetation, topography (fires burn faster traveling uphill), weather, ember movement, and urban spread near wildland areas. So far, researchers found some of these factors had more of an impact on wildfire behavior than initially thought. For example, lofted embers could potentially create spot fires far ahead of the main firefront and increase the rate of spread. Wildfires also have what can be described as their own microclimate due to the hot air they produce which interacts with wind and makes predicting spread difficult. Another is perhaps the most daunting factor: climate change.

Researchers found the increase in global temperature, combined with urban spread, contributed to the growing trend of annual wildfire losses. There’s also the concern that extreme droughts mean longer fire seasons. Researchers are keeping that future in mind.

“The physics doesn’t change. Even if there is climate change, the fundamental rules of physics and mechanics are the same. What will change due to climate change are some of the inputs. For example, how dry the vegetation is going to be,” said Gardoni.

Creating these models takes both science and time. But the reason for creating them is bigger than that. Researchers are taking into account how this could be used by others, not just scientists or engineers, to both plan for and track wildfires. This also includes creating a user-friendly platform to specify model inputs and visualize results to make it easier to interpret data and make what could be life or death decisions.

“The ability to predict the potential impact of wildfires on critical infrastructure – in the future and in near real-time – will be an important and impactful contribution to improving the safety and resilience of our nation’s critical infrastructure and, more importantly, may save lives. Such impact is what we strive to deliver at CIRI and it is what DHS expects of us,” said CIRI’s Executive Director Randall Sandone.

This research is only one example of the work CIRI conducts and funds to keep the nation’s critical infrastructures both safe and functional. More information on other work and activities, including the annual critical infrastructure summit later this year in autumn, is available.

CIRI’s funding for this project goes through December, but researchers are already looking beyond that. With Canada’s wildfires still burning, it’s a reminder of the importance of this work.

“The impact on society can be tremendous,” said Gardoni. “That is really the motivation for doing this. Not just because we like the fancy math, but really because it can have an impact.”