As British Columbia battles another destructive fire season, much of the work to detect and fight the flames is now being powered by artificial intelligence. Crews across the province are increasingly relying on new technology to spot fires and detect threats sooner, part of a push to get ahead of blazes before they can grow out of control.
At the centre of that effort are AI-powered cameras. One such camera, perched atop a student dormitory on Burnaby Mountain, is one of six newly installed around the region, constantly scanning the horizon for signs of smoke. The system zooms in and out of certain areas, and if it senses something, it zooms in to analyse it, acting as the eyes of a wider detection network.
The cameras have already proven their worth. One of them spotted a recent wildfire in Metro Vancouver, alerting crews even before a 911 call was made and allowing them to respond more quickly. To complement the cameras, firefighters have also installed 45 chemical-sniffing ground sensors attached to trees deep in forested areas, adding another layer of early warning on the ground.
The technology is spreading beyond a single region. The new cameras and sensors are now being used in Kelowna and Vernon as well, with the province described as leading the country in wildfire detection. One official involved said the impact is significant enough that the technology will save a life one day, underscoring the stakes behind the rollout.
The destructive fire season has also marked another Canadian first. Crews used a drone to carry out a controlled burn on an active wildfire, hovering over the flames tearing through the Fraser Canyon and dropping fire starters as part of a planned ignition. For crews who had been testing the technology for years, seeing it work in real life was an emotional milestone, with some saying they shed a happy tear.
Inside a lab at the University of British Columbia, researchers are pushing the technology further still, training AI chatbots to help crews triage fires and prioritise locations using real-time data. The predictive modelling presents decision-makers with various outcomes, weighing factors such as evacuation impacts, structure loss and environmental damage. Even so, the researchers say they will never train the artificial intelligence to decide which option is best, insisting that technology can never replace human intuition.