CAL FIRE tests AI technology with wildfire cameras
Using regional geographical, topographical information the AI system picks up on aberrations like new fire starts in the camera feed
By Mary Callahan, Rachel Gauer
The Press Democrat
HELENA, Calif. — Many of the growing number of lookout cameras stationed across California to locate and monitor wildfires will soon be equipped with artificial intelligence technology to speed response to fires and other natural disasters as they first unfold.
Cal Fire's Sonoma-Lake-Napa unit is among six regional units testing the new technology this fire season in collaboration with the ALERTCalifornia system run through UC San Diego. The Sonoma-Lake-Napa unit covers six counties, including Colusa, Solano and Yolo counties.
Fed with regional geographical and topographical information supplied by UCSD, the AI-equipped system picks up on aberrations — like new fire starts — in the camera feed.
It then prompts a box to be drawn around the area in the camera view, alerting those monitoring the cameras to take a closer look because "this could be something," said Caitlin Scully, communication program manager for ALERTCalifornia. (ALERTCalifornia, formerly part of the ALERTWildfire camera network, is now a stand-alone entity focused only on California.)
"It's helping to flag those anomalies so the fire men and women can spot potential starts sooner," Scully said. "So really what it's doing is helping push down the response time to make that response time quicker and more efficient."
Cal Fire Sonoma-Lake-Napa Chief Mike Marcucci said the artificial intelligence component "is truly another tool in the toolbox."
"It is allowing us to be more efficient and to have more situational awareness before we arrive at the scene," Marcucci said. "Before this program, we would get a 911 call and we would have to spin the cameras to the area. This technology now does this for us."
The trial program is part of an effort by Cal Fire to pursue cutting edge technology to aid its mission, made possible in part through the agency's new Office of Wildfire Technology, Research and Development created by legislation authored by state Sen. Bill Dodd, D-Napa, in 2021.
Benjamin Rogers, assistant deputy director at the office, said the ALERTCalifornia system is among several being tested by Cal Fire at different sites.
Sensors installed in May in a pilot area at the Jackson Demonstration Forest in Mendocino County were designed to detect fire starts in dark, densely forested areas utilizing artificial intelligence to detect changes in the air and to distinguish campfires from "fires of significance and fires of insignificance."
The sensors are designed to detect wildfires "within minutes, often during their early smoldering phase, greatly reducing the risk of spreading or becoming larger or more catastrophic," Cal Fire said. They also monitor forest microclimates, temperature, humidity and air pressure in dark, dense areas of the forest with limited connectivity, due to the remote, rugged terrain.
The ALERTCalifornia system is a natural partner for Cal Fire, having grown as a network substantially over the past 20 years through several iterations.
The statewide system now boasts 1,032 high-definition cameras — 199 of them sponsored by Cal Fire — strategically deployed around California.
With pan-tilt and zoom capabilities and near-infrared night vision, they provide 24-hour surveillance and 360-degree sweeps every two minutes, some of them monitoring the same peaks and ridge tops from different perspectives. The devices can provide views of up to 60 miles during a clear day and up to 120 miles during a clear night.
Authorized personnel also can train the cameras on specific points to monitor unfolding events.
The state fire agency has invested $20.3 million in the system, with a commitment to provide at least $3.5 million more in the coming year, Cal Fire said.
Many members of the public also have become familiar with the system, using the public-facing camera views to monitor wildfires in their vicinity. The cameras also are now being placed beyond areas at high risk of wildfire to monitor other natural disasters, including atmospheric rivers and flooding, Scully said.
"Really, within the last five years or so, the network has exploded with the amount of growth," Scully said. "We've doubled the amount of cameras really, in the last three years, because they've proven to be an amazing tool."
Before the addition of artificial intelligence, the cameras functioned similarly to security cameras, simply recording the views.
"The AI will provide additional data from which firefighters can decide how to respond," said Dr. Neal Driscoll, who serves as the principal investigator of ALERTCalifornia. "We want to have data-driven decisions."
Only trained emergency responders will have access to the AI-enhanced views, though Driscoll suggested anyone in wildfire country become familiar with the public site.
"The public can see which cameras have been moved and they can time-lapse and zoom the cameras," he said. "We recommend that you get comfortable with the platform — get familiar with where your street address is on the map and where certain features are — so in the case of an event, it is not the first time you've been on the website."
Along with Sonoma-Lake-Napa unit, the AI trial program will be implemented in the San Bernadino, Madera-Mariposa-Merced, Nevada-Yuba-Placer, Shasta-Trinity and San Luis Obispo units.