How tiny sensors in the forest are shaping the future of firefighting 

Click to play video: 'How small sensors and AI are being used to detect wildfires early'
How small sensors and AI are being used to detect wildfires early
WATCH: How small sensors and AI are being used to detect wildfires early – Jul 7, 2023

The future of fighting more intense, climate-fuelled wildfires is in a makeshift outdoor laboratory in the Los Angeles suburb of Fullerton, Calif.

There, on the grounds of California State University, electrical and computer engineering students are lighting small bonfires and employing tiny sensors to optimize how firefighters detect danger. The idea is to design the sensors to capture the signature of fire — heat, smoke, changes in humidity — in the moments after the fire starts.

The sensors, paired with the predictive power of AI, can help first responders get to a fire quicker — or, better yet, anticipate where the most likely spots are for a blaze to start.

Having sensors in and around the forest would allow firefighters to respond in minutes, rather than hours and days, potentially dousing hot spots before they turn into an inferno. This is especially important in Canada, with its vast geography and relatively small population.

If communities and fire crews had that capability, says computer engineering professor Ankita Mohapatra, “the impact that [wildfires] have on human life and property, and even on the ecology, can be significantly reduced — if we could even detect them at the very early phases.”

For weeks, Canadian and international fire crews have been battling hundreds of stubborn wildfires in Quebec, Ontario and Nova Scotia. It’s been Canada’s worst season on record. Intense smoke has blanketed much of Central and Eastern Canada, as well as the U.S. Northeast and Midwest.


It’s becoming clear there is a need for a much wider array of tools to combat ever-fiercer and bigger fires — and everything is on the table.

The future of firefighting

Fighting forest fires is usually associated with ground crews battling the flames from the ground, alongside water bombers and choppers working in the sky.

But with tiny sensors mounted on telephone poles, trees, drones, and even dropped like confetti onto the forest floor, authorities can be immediately alerted to the slightest changes in conditions.

California’s principal wildfire agency, Cal Fire, is testing sensors made by N5 to detect fire in the very early stages of ignition. N5 Sensors

It’s like a home-alarm system except for the forest, sending text alerts to whoever needs to see them to respond quickly.

The sensors work as clusters or nodes, scanning the forest for signs of trouble. The system works much like a cell phone tower, except the nodes send signals to detect fires.

The more nodes, the better the detection.

Just last week, Cal Fire, which works year-round in California, began testing sensors to detect wildfires more quickly.

That early detection is key, says Cal Fire Deputy Chief Marcus Hernandez, and sensors can help.

“For example, let’s say we hear a vague report, a call-in of a fire by someone travelling, and they can’t even explain exactly where they’re at.”

A well-placed node of sensors could provide geographical details, sent by text or through to an app directly to authorities.

The sensors and cameras provide important clues, even if the technology is still developing.

Smoke concerns

In urban areas on the forest edge, having an early detection system can have a huge impact, Mohapatra says. “And when I say early phases, I mean in the first few minutes.”

Computer engineering professor Ankita Mohapatra and her students set controlled fires to test sensors. The sensors can be used in the forest to alert crews in the first few minutes after a fire starts, and before it turns into a dangerous inferno. Credit: Matt Gush

For example, it can mean the difference between smoke-filled days — and cancelled plans — or a normal summer.

In British Columbia, people are already becoming accustomed to the summer ritual of wildfire smoke polluting the skies and disrupting plans. That stark reality struck millions across Central Canada, the U.S. East Coast, and Midwestern states this spring and summer.

The smoke, from over 500 fires burning in Canada, even drifted across the Atlantic to Europe.

Once seen as a temporary nuisance, dealing with smoke is increasingly becoming a policy priority.

“We need to learn how to live with it,” says Michael Brauer, a UBC professor and expert on the impacts of pollutants on our health.

“It’s going to worsen, actually,” he warns.

Using sensors to detect smoke early is one way to mitigate those concerns. Placed in remote areas, sensors can alert crews to a smouldering fire before it turns into a raging inferno affecting local communities, Mohapatra says.


Or, “you can start in a location where you need the most protection – that could be closer to a small town, or a whole bunch of small towns,” says Abhishek Motayed, whose Maryland-based company, N5 Sensors, is developing the technology in eight U.S. states, as well as in Quebec.

But even he agrees that sensors are not some kind of panacea. They help with the symptoms, but not the disease, which is climate change.

“The problem is enormous,” he says. “It’s a huge amount of land both in the U.S. and in Canada.”

Leveraging the power of AI 

Traditionally, satellites are used to cover large areas, scanning for signs of fire, but satellite imagery can only get so close and provide so much warning. Sensors, placed strategically, are much more precise.

One additional advantage of sensors is that they allow crews to leverage the power of artificial intelligence to avoid false alarms.

The AI, Motayed says, “learns from hundreds and thousands of examples of what fire looks like,” and can differentiate between “a fire, versus a diesel truck, versus construction dust.”

The greater the predictive capacity of the AI, be it baked into sensors, fire modelling software, or infrared cameras, the sooner crews can be alerted.

“To get more precision, you run lots and lots of simulations, with different tiny variations, to better predict [fire scenarios],” says Gabriel Wainer, a researcher at Carleton University, who is working on ways to improve fire modelling and visualizations.
That way, he says, “if you need to go to your premier and ask for helicopters […] you have a visual representation of what’s going to happen.”

Fire modelling and sensors — along with eyes on the ground and in the air — are changing how fires are fought. For instance, the predictive power of artificial intelligence can inform crews as to the best place to place a barrier to stop the fire, or the best direction from which to attack the flames.

Living on the edge 

One of the biggest challenges crews face is fighting fires in what’s known as the wildland-urban interface (WUI).

These are areas of cities and towns built right along the forest edge.

“It’s an attractive place to live,” says Volker Radeloff, a professor at the University of Wisconsin, Madison, who studies fire risk in these areas. “A lot of folks like to be close to nature, and that’s a good thing.

“But then it comes with higher risk and management problems.”

In the United States, there were 12.7 million more houses built, and 25 million more people had settled in those fire-prone areas in 2010 compared with 1990.

In 2003, the city of Kelowna learned some difficult lessons. That summer, a lightning strike ignited a devastating fire that destroyed or damaged over 200 homes as it came roaring down the edge of the forest.

A row of houses can be seen from the air during an aerial tour of Kelowna, B.C., Sunday, Aug. 24, 2003. The Okanagan Mountain Park forest fire tore through the neighbourhood, and destroyed over 200 homes. Chuck Stoody / The Canadian Press

The city caught a break last week when a fire on Knox Mountain — highly visible from the downtown core — prompted a rapid response. It could have been much different had the fire crept up the side of a mountain ridge.


Yet, all across B.C.’s Okanagan Valley, subdivisions continue to be built right along the WUI, making the need for a complement of firefighting tools all the more important.

“The question is just how to get people out efficiently, and that’s where evacuation planning and alerting systems are important,” Radeloff says.

In a community sitting on the edge of the forest, sensor technology can detect a fire within 10 minutes, according to N5’s Motayed.

Predictive power

One of the “smarter” applications of sensors, and AI, is predicting fires even before they start.

At a lab at the University of California, Davis, researchers are mounting powerful drones with sensors. The drones patrol the forest and pick up signs of a fire.

At the moment, says Zhaodan Kong, a professor at the University of California, Davis, “the way that researchers are using these drones or the sensors to detect wildfire is more reactive.”

This means the fire starts, the drone-mounted sensor goes off, and crews respond.

He wants to change that.

This would entail using a network of sensor nodes to predict where the most at-risk areas are, rather than blindly sending in drones to look for fire. For example, modelling and AI could provide insight as to future fire behaviour, based on past trends.

But all these systems are still in their infancy, mostly used on a small scale or in laboratory settings. There isn’t one technology that will be the magic solution to the intensifying wildfires.

It’s an all-hands-on-deck approach, with sensors, drones, watch towers, satellites — and, of course, good old fashioned boots on the ground and choppers in the sky.

Even with all the technology at our disposal, those resources, insists Deputy Chief Marcus Hernandez, “are what puts out the fire after it starts.”