Local agencies now have a new tool to speed up wildfire response — AI-powered cameras.
Arizona Public Service plans to install 30 of them by the end of the month. KNAU's Melissa Sevigny met with Wildfire Mitigation Supervisor Brian Kelley for a real-time look at how the technology works.
MELISSA SEVIGNY, KNAU: You’re expanding your camera network with this new kind of camera that uses artificial intelligence. Tell me how those ones work.
BRIAN KELLEY, APS: Those artificial intelligence cameras are mounted in hilltop locations where there’s a large viewshed where they can see a long distance. Generally, our vendor says that we can detect smoke at 10 miles, but we’ve had instances where these cameras have picked up on fires well beyond that 10-mile mark — outwards of 20 to 30 miles. Once those cameras detect smoke, it notifies our vendor’s dispatch center. The vendor then confirms with a human eye and then a notification is sent to the APS fire mitigation team, but then also a notification is sent to federal dispatch centers, municipal dispatch centers, county and city emergency managers — even the sheriff’s office.
SEVIGNY: Can you walk me through how it works on your laptop?
KELLEY: Absolutely.
SEVIGNY: OK.
KELLEY: So, actually as if this morning we got a notification of a new smoke in the Prescott area. So I’ve pulled up the incident here; this is an incident… looks like just west of Prescott.
SEVIGNY: It’s showing you a pretty high-quality panorama from a high place and you can see the smoke on the image coming up.
KELLEY: Correct. Two of our cameras have picked up on it from Mount Francis and Mount Union. They’ve triangulated on the location and given us a lat-long. So if I open it, there was a short puff of smoke. This was probably an illegal campfire or some type of human-caused incident. But dispatch was notified and I actually received a notification that dispatch was sending resources out there. That’s generally how it goes.
SEVIGNY: These cameras, how do they know if they’re looking at a smoke that it’s a problematic fire versus a bonfire or a chimney?
KELLEY: Through time, these cameras go through a learning process. Once they’re installed, we don’t initially start receiving notifications until they’ve gone through that learning process of filtering out things like fog, chimney smoke, maybe dust. Then, through the process of working with the other cameras, they learn as a group what a smoke looks like. Generally they’ll triangulate on a smoke and give a location. It’s through the artificial intelligence learning process that they become better over time. It’s really neat what they do. They’ve even detected fires at night through infrared.
SEVIGNY: Can you give me an example of how these cameras have already gone to work?
KELLEY: Yeah. Back in February, we had a fire in the Prescott National Forest. It was the Brady Fire. Our camera picked up on it, notified the Prescott dispatch center and gave them actually a location of where the fire was, and also gave them access to view what the smoke was doing, the direction the fire was moving, and it assisted those folks with getting resources on scene in a quick manner.
SEVIGNY: Can you talk a little bit more about why it’s so important that we get early detection of fires?
KELLEY: The fire environment’s changed. We’re seeing larger fires now across the state of Arizona; larger fires that are more catastrophic. It’s important for us as a utility to promote early response and early detection. It benefits us from a reliability standpoint, delivering safe and reliable power to our customers and then, of course, our communities. We’re all members of the community here. We want our communities to be safe in scenarios that could be catastrophic. That early detection is key to supporting our first responders.
