May 21, 2025 — A wildfire research team in B.C. is rethinking how Canada predicts and manages fire risk.

During a ‘Friday Forum’ event held earlier this month by the Institute for Catastrophic Loss Reduction, Mathieu Bourbonnais, a University of B.C.-Okanagan professor and former wildland firefighter, outlined a new system that combines artificial intelligence with a network of low-cost weather sensors that produces real-time and three-to-five-day wildfire forecasts.

Already operational in B.C., the new system can fill gaps in Canada’s current fire danger models and offer emergency responders more timely, location-specific data.

Mr. Bourbonnais said the goal is to deliver the information responders need to make decisions and not just more complexity.

He outlined the development and deployment of a new wildfire risk prediction and monitoring system using low-cost weather sensors and artificial intelligence. The system is designed to address gaps in Canada’s current fire danger rating tools and weather station coverage.

“What we’ve been working on since 2021 is this low-cost internet-of-things device,” he said.

“It is pre-made hardware that we developed firmware to run on.

“And it’s a weather station. These stations monitor temperature, relative humidity, wind speed and direction, precipitation, soil moisture and soil temperature.”

The stations, or sensors, are distributed around various communities and together they provide high-quality data.

“What we’ve found in the Okanagan is that having a station every 10 to 20 kilometres, even a bit more, if the fuels are fairly homogeneous, gives you a pretty good idea of what’s going on,” Mr. Bourbonnais said.

“And for being so low cost, they provide quite reliable data.”

Using the data, the research team can email a daily assessment.

He said that current wildfire risk models such as the Canadian Forest Fire Danger Rating System are operationally sound but often lack precision due to outdated fuel maps and sparse weather data — especially in remote or rapidly changing environments.

Mr. Bourbonnais said his team has AI models that use the input from the sensors to generate granular, site-specific fire risk predictions.

The system is being used in several areas of B.C. and he said it is helping fire departments and wildfire agencies better understand dynamic local conditions.

Mr. Bourbonnais says the technology also proved useful during major events such as the McDougall Creek fire in West Kelowna, where 15 of the team’s sensors were burned but provided valuable post-event data.

The integration of B.C.’s wildfire camera network and prescribed fire planning is also underway.

“We’ve been partnering a lot more with municipal fire departments,” Mr. Bourbonnais said.

“And we’re starting to develop software and dashboards that allow them to interact with the data.”

He said it is important to keep the system relatively simple for both frontline responders and community planners. .

(For more independent coverage of Canadian p&c industry news and trends, please choose the ‘Subscribe’ tab on our main page or email mpub@rogers.com for more information).