Teaching Machines to Fight Fire
After a summer spent fighting fires in B.C., ALES graduate student Jessica Zerb is using AI to predict future wildfires
By Olivia DeBourcier
With news of the California wildfires having just blown up our newsfeeds, and the devastation of the Australian wildlifes still fresh in our minds, fire management is not only a relevant topic but an urgent one.
University of Alberta master’s student Jessica Zerb has made it her mission to study these wildfires — or rather, to teach machines how to study them. She will share her story at an AI4Society panel discussion on Oct. 27 as part of Sustainability Awareness Week.
Artificial intelligence has been used to predict weather activity like precipitation and El Niño, but using it to predict fire is new.
“We’re harnessing machine learning techniques to be able to forecast the fires a bit further in advance so that fire management agencies can be ready for when something might strike up,” said Zerb.
This machine learning technique aims to forecast potential conditions that could indicate a fire. According to Zerb, wildfire severity can be determined by several factors including the availability of fuel for a fire to burn, moisture levels in the surroundings, temperature and precipitation. Feeding her computer model information on these variables and the resulting fires teaches the machine to learn from past data and make predictions based on current trends.
By repeatedly running the fire model, Zerb is teaching the machine to identify what combination of indices could result in more severe wildlife situations.
The spark that started it all
Zerb knows firsthand how improved fire management technologies could help in the field, as she was a wildfire fighter herself.
“I fought in Castlegar and we fought on some really big fires down there. It was a really good summer,” she said. “I had so much fun and I was way more fulfilled than working in an office.”
Fighting fires in BC was not originally in Zerb’s career plan. She’d just finished her undergrad in geophysics and had moved to Calgary to work for the oil and gas industry.
“I’m a huge outdoor fanatic,” Zerb said. “I love to rock climb and I did a decent amount of backpacking that summer as well. But just seeing the smoke was really concerning to me.”
“I wanted to be more of a conservation scientist instead of working in industry. The smoke pretty much was calling me and I started googling wildfire opportunities,”
After spending that summer fighting fires, Zerb returned to Edmonton and became a wildfire analyst with the Canadian Forest Service, where she still works while completing her masters.
Zerb hopes that climate change is something her computing model will be able to account and predict for.
“I would love for it to be accurate enough to be used in fire management,” says Zerb, “To be able to say, ok well in 6 months this index could be really high so maybe we should try to devise some sort of strategy to prepare for that.”
* * *
Learn more about Jessica Zerb’s research at Sun, Fire, and Water: Optimizing Climate Change Response with A.I. on Tuesday Oct. 27 at 2:30 p.m.
For more information about Sustainability Awareness Week and to see the full schedule, please visit: uab.ca/saw
Cover image features photography by Yuyeung Lau on Unsplash.