Drones for Wildfire Management

Timely research seeking to support wildfire management through the use of machine learning.

ASI is excited to feature the newest member of our research network, Dr. Jeff Boisvert. Awarded an NSERC Alliance grant in July of this year, Dr. Boisvert and his team join Theme 3 Sustainable Communities with timely research seeking to support wildfire management through the use of machine learning.

Dr. Jeff Boisvert does what researchers do best; he conducts deep investigations into problems that, up to now, have no clear answers or solutions. He brings his world-leading expertise in geostatistical modelling for industry-based resource applications, with a specialty in mining, to the University of Alberta’s Department of Civil and Environmental Engineering.

But he’s no ordinary academic buried in numbers, theories and laboratory testing. Away from the confines of academia, he is an active, paid-on-call firefighter, experiencing the real-life dangers of out-of-control flames. And, if you have been following the news at all over the summer, you will have been confronted by the scale of wildfire across Western North America. British Columbia was especially hard hit after lightning strikes ignited parched vegetation across the province, devasting many areas. The pressure to predict and better manage these catastrophes using new technologies has mounted.

Bringing these two extraordinary parts of his work and life together gives Dr. Boisvert a unique insight into the possibilities technology offers to help tackle these events before they take hold. Together with an outstanding team of expert researchers and industry partners, they have successfully developed a crucial research project incorporating UAV-gathered data, satellite imagery, and machine learning. This highly interdisciplinary research is only possible by creating a diverse team of skilled researchers including Dr. Li Cheng (machine vision and machine learning), Dr. Michael Lipsett (remote robotic systems), Dr. Arturo Sanchez-Azofeifa (remote sensing), Dr. Ilbin Lee (decision making), and Dr. Mike Flannigan (fire science).

Recently awarded a prestigious NSERC Alliance Option 2 grant, this project aims to develop standards and methods to remotely collect and analyze spatial wildfire data. The aim is to better understand topography, fuel, and weather around wildland fires to support fire management and planning decisions. Drones, supplemented by satellite imagery, collect vegetation (fuel) and weather data. Machine learning methods can infer vegetation characteristics that impact fire behaviour, including tree species, size, shape, connectivity, density, and crown fuel characteristics. This data can be used to make better fire behaviour predictions and assessments, thus allowing for better management of wildland fires in Canada.

Specifically, the project will explore four applications of drones for improved wildland fire planning and operational decision-making.

(1) The use of collected data around the perimeter of a community at risk to wildland fires to assess closeness and density of trees and vegetation for fire hazard assessments;

(2) The ability to fly drones to structures ahead of wildland fires to perform field exposure assessments and make structure triage decisions during active fires;

(3) The use of gathered data to improve prediction of fire spread and direction;

(4) The use of collected data to inform improved decision-making on the deployment of suppression activities such as air tankers, waterbombers, helicopters and ground crews.

Essential in this undertaking is the consultation with regional Alberta fire rescue services, the Canadian Association of Fire Chiefs, and the provincial department for Agriculture and Forestry.  As Dr. Boisvert explains, “there are very valid concerns around the use of new technology in the fire services because the consequence of relying on technology that may fail could be disastrous. It’s crucial that we seek input from all levels of the fire services from day-to-day decision-makers and community risk assessors to fuel experts and modellers.”

He also is keen to stress the need to demonstrate value. “We want to be able to quantify the value this data provides in terms of dollars saved, containment time reduction, risk reduction and risk avoided. I’ve conducted similar value demonstrations in the mining and petroleum industries, and my colleague and collaborator on the project, Dr. Ilbin Lee [Dept. of Accounting and Business Analytics at U of A], is actively exploring this in the fire sciences.”

Chief Fire Marshal Sean Cunningham at Parkland County is one of Alberta’s Structure Protection Specialists and is a key member of the research team ensuring developed workflows are practical and useful, he explains “The ability to quickly and remotely assess fire risk for structures threated by wildfire will allow us to better manage the limited resources available to respond to campaign fires like the recent Tomahawk fire west of Edmonton.”

Dr. Boisvert is clear, though, that the impact of this work reaches far beyond the immediate benefits of saving forest, animals and property. Many individual stakeholders and industries are impacted, including Indigenous communities and their traditional sites, petroleum and mining assets, national infrastructure, communities along the wildland-urban interface (WUI), and the 60,000 abandoned wells in Alberta, many in the boreal forest. These effects are expected to worsen as climate change increases the frequency of extreme fire seasons.

The necessity for this work is so pressing that they have already flown their first wildfire before the official launch of the grant. “We need to better use data to plan for fires as well as influence real-time data collection and decision making.” This work can go a long way to helping those at the forefront of wildfire fighting gain the upper hand in the future.