University of Oregon to Use AI to Spot Wildfires

June 27, 2024
Using AI will be a game-changer, giving firefighters a heads up on many wildfires well before any human would have identified them and called them in.

Nick Maggio recalls standing at the nearly 8,000-foot peak of Beatys Butte in 2021, taking in the vast wilderness spread out for miles below. A helicopter soon roared overhead as it prepared to drop off 1,200 pounds of batteries for the camera tower he and his University of Oregon Hazards Lab team were about to build.

This would change everything, he knew. The tower, when finished, would stream video footage of the surrounding landscape, its view reaching more than 40 miles. And it would do so 24 hours a day, every day, an unblinking eye always on the lookout for the beginnings of a wildfire.

Once the tower was joined by many more in the works, and artificial intelligence was added to the process, Oregon firefighters would have eyes in wildlands near people’s homes and in the remotest regions of the state, in areas where fire-watch towers were never built. They would be able to catch the first sparks of a fire day or night, without relying on human fire-spotters straining their eyes and swatting themselves to stay awake in the loneliest stretches of Oregon.

That time has now come.

On May 3, the University of Oregon integrated the Beatys Butte camera tower — and 44 other university-owned camera towers like it on land owned by UO’s partners, including the federal Bureau of Land Management and the U.S. Forest Service — with a specially-designed fire-spotting AI system created by the California-based tech company ALERTWest.

UO’s AI-integrated wildfire-detection system is one of three similar systems in the state. In addition to UO’s camera array, the Oregon Department of Forestry also operates its own AI-camera system on land under its jurisdiction, as do Portland General Electric and Pacific Power, which use a system called Pano AI. UO’s camera system is the only one that uses the artificial intelligence provided by ALERTWest.

Doug Toomey, a professor of geophysics at UO and director of the Oregon Hazards Lab, said the use of multiple systems is due to the many different agencies and organizations that manage Oregon’s public and private lands.

Maggio, now the assistant director of wildlife technologies for the lab, helped build more than a dozen of the UO camera sites. Using AI to monitor the camera’s feeds will be a game-changer, he said, giving firefighters a heads up on many wildfires well before any human would have identified them and called them in.

“This ability for the system to send out a notice that it’s spotted something that it thinks is worth a human being taking a look at – that’s going to be really transformative,” Maggio said.

The promise of having AI sentinels in the forest and near communities has been “on the horizon” for some time, he said, but he was surprised to see it happen so soon.

Since the cameras started going online in 2018, firefighters and even the general public have kept an eye on them, though not in a formalized, comprehensive way. Now AI will do the work of staying vigilant.

Crowdsourcing the video feeds was a decent, interim solution, and many people ended up watching them from home, Toomey said. But it was far from perfect, he noted.

“We realized that the best possible scenario is having a machine tell you that, ‘Oh, there’s an anomaly,’” Toomey said.

Scott Schifando, vice president of operations for ALERTWest, said that if the AI sees signs of a fire – such as wisps of smoke – it will alert a small team of detection specialists at ALERTWest’s offices in Chico, California, who will verify if the anomaly is truly a wildfire breaking out.

“That’s really where the AI comes into play,” Schifando said. “It’s able to say, ‘Hey, you’re getting all of this constant data streamed, it’s overwhelming. Just look here.’”


The expansive University of Oregon project got its start not in wildfire science, but in another discipline — seismology.

In 2015, the university began collaborating with the U.S. Geological Survey, the University of Washington, Caltech and the University of California at Berkeley to develop the ShakeAlert Earthquake Early Warning System, a network of seismic sensors and processing centers that rapidly detect earthquakes and deliver alerts to people before they feel dangerous shaking.

The idea caught the attention of the Oregon Legislature.

In 2019 — a year after the Oregon Hazards Lab deployed its first camera for wildfire detection — the Legislature approved $7.5 million for the ShakeAlert program and for improvements in wireless network communications for its sensors.

But in the state of Oregon, while the potentially most damaging natural disasters are earthquakes, wildfires are the most frequent. And Toomey and his team soon realized that their real-time network could do more than just detect and alert people to earthquakes.

“We wanted to do multi-hazard (detection),” Toomey said. “It’s cost-effective to have a single platform that does multiple things.”

In 2021, state Rep. Nancy Nathanson, who had helped secure the funding for ShakeAlert, sought an additional $4.5 million from the Legislature – this time for a UO wildfire camera system. But the funding request was not included in the legislature’s final spending bill.

The following year, Nathanson wrote another request for funding, noting that only 30% of Oregon was covered by wildfire detection cameras at the time. She pointed out that many of the new cameras could be connected to existing communication systems that were also supporting the ShakeAlert program. This time, the legislature voted to approve the funding.

“That’s one of the aspects of this program that I’m most pleased about,” Nathanson said. “It’s building off of investments we’ve already made.”

In addition to the funding provided by the state of Oregon, the Oregon Hazards Lab has received $10 million in funding for its real-time monitoring network from the Bureau of Land Management, The U.S Geological Survey, the Federal Emergency Management Agency and the National Science Foundation, Toomey said.


With three wildfire camera systems in the state, coordination among the different organizations is essential to avoid redundancy, said Jamie Paul, Oregon Department of Forestry’s wildfire detection camera coordinator.

Paul said that ODF has been installing cameras on land within its jurisdiction since 2009. They are monitored by five detection centers located throughout southern and eastern Oregon. She said the detection centers use artificial intelligence built by the South African-based tech company EnviroVision Solutions’ ForestWatch to help human detection specialists scan large numbers of computer screens for smoke.

Paul said that, in addition to the ODF-installed cameras on its land, Oregon utilities have deployed AI on land they own.

The existence of three separate camera systems instead of one unified system is due to several factors, including differences in each agency’s mission, availability of funds and the procurement processes of the different organizations that manage Oregon’s forests, she said.

In 2022, the state created the Oregon Wildfire Detection Camera Interoperability Committee to ensure the camera networks complement each other. The committee is made up of emergency professionals and representatives from UO, the governor’s office and various state and federal land agencies.

“We needed to find a way to deconflict redundancies,” Paul said.

This means avoiding redundant camera sites, sharing information and even getting the different systems’ AIs to “talk” to each other.

Paul said the state forestry department’s partnership with the University of Oregon’s AI integrated cameras have “filled that void” in the places where ODF doesn’t have jurisdiction, such as federal land.

“We partner with them ( University of Oregon) in any way that we can because we believe in the system,” said Larae Guillory, fire operations specialist for the federal Bureau of Land Management. “It’s a huge part of boosting our situational awareness for wildfires.”

She said that the BLM has partnered with the UO Hazards Lab through a five-year cooperative agreement to install cameras across Oregon and Washington. She added that the five-year agreement is in its second year, and that BLM is expected to provide roughly $5 million to the project’s funding over the course of the agreement, pending available funds.


The 2024 fire season will be the first time that the University of Oregon’s wildfire cameras will use AI to detect wildfires, and Toomey said the Oregon Hazards Lab has plans to add 30 more cameras by late 2025, bringing the total number to 75.

Maggio added that Oregon Hazards Lab sensors also could be used to assess the landscape after a fire to predict the risk of landslides in areas that have lost trees that help hold hills in place.

“It’s been really impressive how quickly the technology has moved,” Maggio said. “We’ve gone from it being something that might happen someday, to it being a reality and a valuable tool in what seems like the blink of an eye.”

— Tanner Todd covers crime and public safety. Reach them at [email protected], or 503-221-4313.

©2024 Advance Local Media LLC. Visit Distributed by Tribune Content Agency, LLC.

Voice Your Opinion!

To join the conversation, and become an exclusive member of Firehouse, create an account today!