Melanie Morin, a prevention agent with the Society of Protection of Forests from Fire, walks through an area of burned forest near Lebel-sur-Quevillon, Que., on July 5, 2023.Adrian Wyld/The Canadian Press
After a devastating start to Canada’s wildfire season left record-breaking wreckage in its wake, scientists are looking to artificial intelligence to play an important future role in detecting and fighting blazes.
Swaths of fires igniting from British Columbia to Nova Scotia temporarily displaced thousands of Canadians countrywide, spewed smog as far as Europe and burned through 3.3 million hectares of forest – equivalent to half the size of New Brunswick. All this despite the federal government shelling out approximately $1-billion a year to mitigate the blazes.
Experts around the world have warned in recent years that the warming effects of climate change will lead to increasingly intense wildfire seasons and prolonged periods of smoke exposure in Canada. In February, 2022, the United Nations predicted fires will grow more destructive by the year and called wildfire risk reduction “more critical than ever.”
These global alarm bells are leading experts to predict that Canada will have to increasingly tap into AI technologies in the form of drones, sensors and high-tech satellites to keep up with the fires.
“Change is coming in terms of climate and technology,” said Joshua Johnson, forest fire research scientist with Natural Resources Canada (NRC). “I think that’s intimidating at times, but it’s forcing us to get creative.”
Canada sees 100,000 square kilometres burned this record-breaking wildfire season
OPINION: Canada’s wildfire approach needs to shift from reactive to proactive
The current AI revolution, driven by the advent of powerful data-processing machines such as ChatGPT, has facilitated computing, customer service and many industries in between in recent months. It has enabled the creation of RADRFIRE, an American computing software that processes imaging data from a countrywide network of drones and satellites to detect, predict, evaluate and communicate early fires with fire stations.
Similar technology, according to Mr. Johnson, could eventually help Canada make crucial decisions on how to allocate resources in peak fire season.
“If we have 100 fires across the country and 30 crews, we need to determine where we most urgently have to focus our efforts – especially now that we have fires coast to coast,” he said. “AI will be a powerful tool for helping us make those calls.”
Mr. Johnson said his branch of government has recently been courted by a medley of startups, all offering their own AI-driven solutions to wildfires. SensaioTech, a Brazilian company also incorporated in Canada, uses localized, ground-bound sensors that monitor 14 variables – including humidity, soil temperature, salinity and pH levels – to provide minute-by-minute readings of an area’s probability of ignition.
SensaioTech founder João Lopes says beta trials in Brazil suggest his creation can predict wildfires five days before they happen.
“With satellites, you are seeing the present; it will not necessarily predict where fires could eventually ignite,” he said, adding that he expects prediction technology could eventually anticipate fires months in advance and allow for plenty of preparation time. Mr. Lopes hopes Canadian investors adopt his technology, though Canada’s size imposes an inherent challenge: Each sensor covers an area of five hectares of forest – a minuscule fraction of the country’s 362 million hectares, many of them situated north of the power grid and far beyond connectivity.
Mr. Lopes is not alone in leveraging AI to create firefighting technology in recent years. The World Economic Forum developed FireAid: a live risk map meant to update fire information in real time and create better resource allocation. Similarly, American startup Pano AI uses computer vision to detect new wildfires quickly and accurately across five Western states and areas of Australia.
Other companies focus instead on reducing unforced errors. Buzz Solutions, a California startup, uses machine learning to constantly inspect power-grid infrastructure for defects – such as insulation damage and overheating – to detect spark risks. In 2021, the company test-ran its technology in Newfoundland and reported faster remediation of power lines after storms.
“We’ve seen a greater appetite for our product every year as the climate changes, grids age and our cities encroach on vegetation – all risk factors for spark-induced fires,” said Kaitlyn Albertoli, chief executive officer of Buzz Solutions.
Mike Flannigan, British Columbia Research Chair in predictive services, emergency management and fire science, is wary of thinking about machine learning and emerging technologies as the only answer to the country’s wildfire struggles. He said a simpler, more immediate way to affect change is by modifying our own behaviour, because half of the country’s fires are human-caused. For example, implementing more forest closings – restrictions on activities in the woods when an area is susceptible to burning – would be a good start.
“We should look at AI and machine learning as another tool in the tool kit, not a panacea that will solve all of our problems,” Prof. Flannigan said. “We’re just not there yet: You would need thousands of drones to replicate the impact of just one water bomber.”
Yet he is bullish on AI tools that could better predict the location of extreme fire weather and cut the time it takes to react to it. Current prediction and modelling technology, Prof. Flannigan said, can delay firefighting efforts by up to three days, providing ample time for the blazes to rapidly spread.
Mr. Johnson, meanwhile, sees value in the prospect of AI-driven sensor technologies, as they could especially come in handy for round-the-clock monitoring of tourist areas such as Banff and Jasper. He said he trusts that NRC will bolster its firefighting technology in coming years, and be prepared if fires ever threaten those high-traffic areas.
For example, the Canadian government plans to launch its RADRFIRE equivalent in 2029: WildfireSat, a $170-million satellite meant to use thermal imaging to deliver vital information to fire stations twice daily to facilitate decision making. But until then, Mr. Johnson said, he sees value in entertaining emerging tech that could minimize fire damage coast to coast, wherever it comes from.
“I firmly believe that in the past 10 years we’ve made some good steps in the right direction,” he said. “Moving forward, there will be times when it will feel like the climate is moving faster than we are, but I think we are on the right track.”