California spends hundreds of millions of dollars every year fighting fires, but stopping them before they start is the goal of low-tech solutions like spraying fire retardant in fire-prone areas and high-tech ideas using artificial intelligence.
"Anytime we can get the fire at its early start, the better the firefighters have of containing it quickly," said KT McNulty, executive director of REDCOM at American Medical Response.
Ravaged by fire over the past four years, Sonoma County began testing software in May to spot fires in their early stages.
"A human being can't monitor every single camera that's out there, especially looking for those kind of minute movements in the pixels, which is really what we're looking for because we want those early detections. Anyone can spot a fire that's, you know, already 10 15 20 minutes blazing, it's too late," Robert Grey, the head of marketing at Korea's leading visual recognition A.I. firm Alchera said.
The dispatch center in Sonoma County has walls covered with monitors that are tapped into a network of remote fire cameras. The software was created by Alchera and uses an algorithm to quickly recognize smoke, and then accurately determine its location.
"It's intuitive to use, and also, they're getting the necessary information out of it as soon as possible so they can make better decisions," said product manager Michael Park.
The technology is similar to teaching a phone's camera to know your face, but billions of faces provide a lot of data for facial recognition. Teaching cameras to know the difference between a cloud, fog or smoke requires a continuous learning system.
"Every time we get a new positive detection, that's automatically labeled and put back into the data set," said Grey.
The technology works within the Alert Wildfire network, a powerful fire-spotting group made up of utilities, state and local agencies, and more than 850 cameras across six western states positioned to spot fires. For now, only Sonoma County is using the system, but it could expand across the entire network over the next few years.
"I think A.I. is gonna really play a role in trying to help us to find those fires that really need to be discovered early, and I think once that happens, I think that will be the aha moment," said Dr. Graham Kent, a professor in the geology department at University of Nevada - Reno and project leader at Alert Fire Network.
Sonoma County used part of a federal grant to install the A.I. system, and while data is still be collected, to this point the A.I. has not given a false alert and has complimented the 911 system in place. According to McNulty, "a 911 caller was calling from a valley floor, and she she's new to the area so she couldn't tell us where she was located. And she couldn't tell us where she was looking. So the A.I. was able to give us a location of that incident before she could."
Supporters hope the combined efforts of 911, cameras and A.I. will work together during one of the most concerning fire seasons in some time. Southern California could also benefit with the cameras already in place because Grey feels the system could be available almost immediately, explaining, "We can have a new dashboard ready for them to go in a day, honestly. So it's just about which cameras do you need to monitor? And who do you need to receive these alerts, and who should have the login?"