Cal Firefighters are using artificial intelligence to help spot wildfires, streaming video from more than 1,000 cameras strategically placed across the state into machines that alert first responders when to call.
In one example of the potential of the ALERTCalifornia AI initiative, which launched last month, the cameras spotted a 3:00 a.m. local time (3:30 p.m. standard time) in remote, scrubby Cleveland country, about 50 miles (80 kilometers) east of St. Louis. There was a fire in the forest. Diego.
With people asleep and darkness covering the smoke, the fire could spread into a raging blaze. But the AI alerted a fire chief, who called in about 60 firefighters, including seven engines, two bulldozers, two water tankers and two firefighters. Cal Fire said the fire was extinguished within 45 minutes.
Developed by UC San Diego engineers using artificial intelligence technology from Chico, Calif.-based DigitalPath, the platform relies on 1,038 cameras installed at various public agencies and utilities across the state, each capable of rotating 360 degrees . Commands of the remote operator.
Cal Fire has provided other examples of AI alerting fire chiefs to fire chiefs before calling 911 since the AI program launched on July 10, but has yet to provide a comprehensive report.
The sample size so far has been too small to draw conclusions, said Neal Driscoll, a professor of geology and geophysics at UC San Diego and lead researcher at ALERTCalifornia.
Cal Fire hopes the technology will one day serve as a model for other states and countries around the world, a need highlighted by this season’s extraordinarily destructive wildfires in Hawaii, Canada and the Mediterranean.
“It’s 100 percent applicable anywhere in the world, especially now that we’re experiencing larger, more frequent fires,” said Suzann Leininger, an intelligence specialist with Cal Fire in El Cajon, east of San Diego. And climate change.”
Part of Leininger’s job is to help with machine learning. She reviews previously recorded video from a network of cameras that the AI thinks is a fire, and then tells the machine whether it is correct with a binary “yes” or “no” answer. Any number of phenomena can trigger false alarms: clouds, dust, even a smoky truck.
As hundreds of experts repeated the exercise up and down the state, the AI had become more accurate in just a few weeks, Driscoll said.
In addition to the camera network, the platform is gathering a wealth of additional information, Driscoll said, including conducting aerial surveys to quantify vegetation that could fuel future fires and mapping the Earth’s surface beneath tree canopies.
Aircraft and drones also collect data in infrared and other wavelengths beyond the capabilities of human vision.
In winter, the platform is able to measure atmospheric rivers and snowpack. The UC San Diego team is also collecting data on the burn scars and their effects on erosion, sediment dispersal, water quality and soil quality, Driscoll said.
The data, available to any private company or academic researcher, could eventually be used to model fire behavior and improve yet-to-be-foreseen AI applications for environmental research.
“We’re in an extreme climate right now. So we’re giving them data because this problem is bigger than all of us,” Driscoll said. “We need to use technology to help change the status quo, even a little bit.”
© Thomson Reuters 2023