Data-Driven Models for Wildland Fire Management

Canadian fire seasons are becoming longer, fire events are becoming more severe and experts predict that the area burned on an annual basis could double by the end of this century. There are approximately 8,000 annual wildfires in Canada, burning an average area of 2.5 million hectares. To put that in perspective, that is roughly 3 million FIFA regulation soccer fields each year.

Fire management agencies are tasked with managing fires on forested landscapes and need to make decisions that are defendable. The Collaborative Research Team (CRT) Evolving Marked Point Processes with Application to Wildland Fire Regime Modeling was led by John Braun, Professor of Statistics at the University of British Columbia – Okanagan and Doug Woolford, Associate Professor of Environmetrics at the University of Western Ontario. This team is building national wildland fire research capacity aimed at better managing wildfires.

Braun and Woolford develop data-driven models that help fire management agencies better understand fire regimes and help them create an appropriate response when actioning a fire. Their models help fire management agencies monitor trends that may be related to climate change, predict where and when a fire may occur and improve the ability to predict where a fire might spread, including quantifying a measure of the uncertainty in surrounding spread maps.

Colin McFayden, Forest Fire Science Specialist at the Aviation Forest Fire & Emergency Services, Ontario Ministry of Natural Resources & Forestry says these models are vital in his day-to-day work. “Ontario’s fire occurrence prediction models developed collaboratively with the University of Western Ontario are used daily in preparedness planning,” says Colin. “Being able to map probability of fire occurrence helps us consider how many firefighting resources we need and where to base them for the day.”

This CRT leverages their members’ unique and diverse expertise to produce innovative solutions to manage wildland fires. “We bring statisticians, computer scientists, operations researchers, ecologists and others in academia together with wildland fire scientists, managers and those working in operations to build the body of knowledge surrounding wildland fire science,” Woolford says. “We then transfer that science to end users with a particular emphasis on decision support for wildland fire management.”

Fire science and fire management is a field full of rich datasets and interesting problems where statistical and data science continue to have a major impact. The Canadian Statistical Sciences Institute’s (CANSSI) CRT program is a catalyst to engage the scientific community, government and industry by bringing the statistical and data science community together to tackle difficult and important problems.

“I hope this work opens up the opportunity for fire management agencies and other areas in the natural resources sector to hire more statisticians with strong data modelling skills,” says Braun. “These skills are becoming vital as organizations need to uncover hidden insights in their exponentially vast datasets.”

A 5-day workshop held at the Banff International Research Station. The circled faces are a mix of wildland fire science researchers outside of statistics and provincial wildland fire management agencies.

This project emphasizes the collaborative nature of CANSSI’s CRT program and its focus on training. Around 90% of their budget was used to support students. By combining CANSSI funding with other research funding opportunities, Braun and Woolford were able to train 23 highly qualified personnel (HQP): one undergrad, six MSc students, 13 PhD students and three postdocs.

This CRT, which includes four members who are not statisticians, represents the interdisciplinary approach that is championed by CANSSI. Students and other HQP receive valuable experience from both subject specialists and statisticians, empowering them to succeed in many sectors that require them to work in diverse teams and possess statistical fluency to develop solutions for real-world problems. This CRT successfully secured funding to continue their interdisciplinary work in this area.

This CRT was able to secure funding from the organizations below and ran several workshops supported by other math institutes:
• Support from the Institute of Catastrophic Loss Reduction
• Two MITACS Accelerate Internships: Aerial Fire Fighting, and Why Insurers Fail
• An NSERC Engage Grant: Estimating wood volume from satellite data
• Seven-month embedding of a Research Associate from the Aviation and Forest Fire Emergency Services branch of the Ontario Ministry of Natural Resources and Forestry at Western.
• Government of Alberta funding for province-wide fire occurrence prediction
• Canadian Foundation for Innovation and B.C. Knowledge Development Fund funding for a Data Visualization Lab at University of British Columbia Okanagan
• Workshops: two at the Fields Institute, two at the Banff International Research Station and one at Wildland Fire Canada 2016

Comments are closed.