Here’s how a BYU graduate’s research is helping to optimize wildfire modeling
Jun 28, 2025, 1:03 PM
Jane Housley, a Brigham Young University graduate, is pictured. Housley's research could help make a widely used wildfire modeling tool faster and more accurate when needed most. (Nate Edwards, BYU Photo)
(Nate Edwards, BYU Photo)
PROVO — Wind has already played a big role in Utah’s wildfire season, combining with heat and low relative humidity to send the flames of the Forsyth Fire toward Pine Valley in Washington County, where it destroyed 18 structures, including 13 primary or secondary homes.
While Utah Gov. Spencer Cox has called on his constituents to pray for rain amid worsening drought conditions across the state, a Brigham Young University graduate’s research could help make a widely used wildfire tool faster and more accurate when needed most.
Jane Housley, who earned a graduate degree in mathematics from BYU, partnered with the U.S. Forest Service’s Missoula Fire Sciences Lab with the goal of improving WindNinja, a simulation tool created by the agency and used in real-time by fire crews and analysts to predict how wind will move through terrain during a fire
“One of the really, really cool things about studying math is you kind of get to end up working on all sorts of different types of problems using your math knowledge,” Housley said. “I really have become passionate about the project, but it’s kind of cool that it just started out as something where somebody just thought, ‘Oh, it might be helpful to have a math person on this,’ even though I didn’t really have any background in wildfire studies.”
While WindNinja is a helpful software, it’s not perfect. Housley said it struggles to model what’s called a “cavity zone.”
“That’s the area directly behind a mountain or ridge where wind tends to swirl backward and create eddies.” Housley said.
Eddies are important because they can dramatically alter how and where a wildfire spreads.
When using WindNinja, Housley said users have to choose between speed and accuracy. She was tasked with optimizing the software to achieve both of those traits.
Her solution?
Implementing her math prowess — with a modern twist of AI-powered machine learning — to bridge the gap.
To lay the groundwork, she used a mathematical approach inspired by airflow around buildings. Decades-old architectural research showed how wind wraps around square structures in cities, and Housley repurposed that math to approximate how wind might similarly move across complex natural terrain.
She then went to work on an algorithm that approximates hills and peaks as “rectangular buildings,” then applied formulas to predict where cavity zones should appear.
“Instead of thinking of a mountain like this rugged terrain, can we think of it like tons of small skyscrapers all bunched together? And that was kind of the approach we took there,” Housley said.
This led to a model that clearly outlines problem areas where wind flow is likely to become turbulent.
For the second part of the project, her solution was “to train a neural network, which ended up working really, really well,” Housley said.
Housley built a custom U-net convolutional neural network (a type of AI often used in image recognition) and trained it on nearly 6,000 wind simulation images provided by the Missoula Fire Sciences Lab. Each data pair included terrain, vegetation type, wind direction and outputs from both WindNinja solvers.
This enabled the neural network to produce a pipeline seven times faster than industry-leading models, while retaining high accuracy.
The results speak for themselves:
- The model cut one type of error by 75%.
- It sliced the average error in half.
- On a test that measures how close two wind maps look — kind of like comparing photos — it scored 0.77 out of 1, a big jump from 0.60.
- Best of all, it did it in just 0.07 seconds per simulation.
“WindNinja is used in real-time by firefighters when they’re trying to predict the path of a fire … to try and figure out the optimal place to go and defend the forest or the community. In a situation like that, minutes and even seconds are everything,” Housley said. “Being able to improve the speed by seven times, you know, gives us minutes and even hours back in that front.”
With this success, Housley said she only sees the role of AI and machine learning growing when it comes to wildfire modeling.
“We’ve kind of seen that machine learning is a really valuable thing to test out in wildfire modeling. I think it’s going to kind of change the landscape of how wildfire modeling operates in the future,” Housley said.
