Characterizing Extreme Wildfire Statistics to Improve Risk Assessment and Resilience

Extreme wildfires, although infrequent, cause disproportionately more ecological, societal, and economic harm compared to smaller and more common wildfires. In fact, the largest 1% of wildfires in the United States between 1984 and 2024 accounted for nearly 28% of the total area burned nationwide. As extreme wildfires become more frequent and destructive, understanding where they might occur and how large they can grow can help inform infrastructure design, emergency response, risk assessment, and long-term wildfire planning. 

Louis Berger Chair and Professor Farshid Vahedifard of the Department of Civil and Environmental Engineering recently led a study that evaluated different statistical approaches for characterizing wildfire size distributions and estimating the likelihood of extreme fires—critical factors for community and environmental preparedness and resilience as wildfire activity continues to intensify across the United States. His team, including Ph.D. student and first author of the study Amirali Asadian, found that models based on the “average” range of wildfires can significantly misestimate the size and likelihood of extreme fires, demonstrating the importance of using specific models to understand wildfire extremes. Their study was recently published in Earth’s Future. 

The importance of modeling wildfires  

Wildfire models give scientists key insights into fire behavior and inform real-world planning in both traditionally fire-prone regions, such as the Western United States, and regions where wildfire risk is rapidly emerging, including parts of the Midwest, Southeast, and Northeast. Fire projections help inform prescribed burns, land-use decisions, allocation of firefighting resources, fire prevention programs, and more. They also play a critical role in wildfire risk assessments used by communities, utilities, transportation agencies, emergency managers, and insurers. This makes increasing the accuracy of wildfire models all the more essential.   

“The largest wildfires are often the most consequential, yet they are also the most difficult to model,” Vahedifard said. “We wanted to understand whether models developed for typical fires can accurately represent these extreme events and the risks they pose to communities, infrastructure, and ecosystems.” 

The team analyzed data from more than 30,000 large fire events across the United States that occurred between 1984 and 2024 and tested different ways of modeling fire size. They deliberately separated the fires into two groups—the "body,” or the bulk of average fires, and the “tail,” the extreme, record-breaking fires. They tested three different types of models to see which one made the best predictions for each group. 

Some of the models they tested, like many wildfire models, were trained only on average fires. The researchers found that these models misestimated the size of extreme fires by hundreds of thousands of acres and significantly miscalculated the probability of extreme fires—a crucial finding considering how important this information is to wildfire prevention and emergency planning. 

These inaccuracies can have significant consequences for wildfire risk assessments. Risk models rely on estimates of both the likelihood and magnitude of future fires to evaluate potential impacts on communities, critical infrastructure, transportation networks, power systems, water resources, and ecosystems. Underestimating the probability or size of extreme events can lead to underestimating potential losses, leaving communities and infrastructure systems less prepared for catastrophic wildfire scenarios.

Better preparing for more extreme, destructive wildfires  

“This shows us that extreme wildfire behavior should be modeled separately from typical ones,” Vahedifard said. “We also found that model performance depends on geographic scale, highlighting the need for region-specific approaches when assessing extreme wildfire risk.” 

This research emphasizes the importance of developing and using separate, more region-specific models for extreme wildfires—especially as their behavior changes and becomes more intense with climate change. With a more accurate understanding of how extreme or likely fires can be, a specific, targeted modeling approach can better inform fire planning, ensure communities are prepared for worse-case scenarios, and minimize long-term societal and ecological damage.   

“Wildfires are becoming more extreme faster than our models have been able to capture,” Asadian said. “This research is part of our group’s work to close that gap, building tools that reflect the fires we're actually facing, not just the ones we've historically planned for.”

The study builds on a broader body of wildfire-related research led by Vahedifard. In recent years, he has led a multidisciplinary research team across several universities that has investigated a range of wildfire impacts on natural and engineered systems, including post-wildfire landslides, debris flows, and other cascading ground failures that can threaten communities and critical infrastructure long after a fire has been extinguished. By combining statistical, data-driven, and physics-based approaches, his research aims to improve understanding of both wildfire hazards and their downstream consequences, ultimately supporting more effective risk assessment, preparedness, and resilience planning.

Learn more about Vahedifard’s research.