IE 11 Not Supported

For optimal browsing, we recommend Chrome, Firefox or Safari browsers.

AI Stations to Search for Early-Stage Wildfires in Washington

Washington's Department of Natural Resources will deploy 21 “stations” to monitor for smoke and heat in a two-year pilot. Each station has two cameras that continuously rotate 360 degrees.

Smoldering ground in the aftermath of a wildfire.
The ground still simmers from the Cold Springs Fire that swept through Okanogan County, Wash., on Sept. 10, 2020. (Amanda Snyder/The Seattle Times/TNS)
(Amanda Snyder/The Seattle Times/TNS)
All wildfires start small. Unfortunately, not all of them stay small, and some quickly get out of hand before fire personnel can fight back.

With this in mind, the state of Washington is deploying an artificial intelligence (AI)-based system to monitor terrain for smoke to get a jump on any wildfire that may ignite and get the appropriate resources to the scene as quickly as possible.

The state’s Department of Natural Resources (DNR), in a two-year pilot with tech firm Pano, will deploy 21 “stations” in strategic locations to monitor for smoke. Each station has two cameras that rotate 360 degrees continuously and are monitored by Pano staff in an intelligence center.

In the event that a fire starts, a station will pick up the latitude and longitude of the incident, and staff from Pano and DNR will collaborate on the intelligence so a decision about what resources to deploy can be made quickly.

“What DNR is getting is really actionable intelligence on any smokestart,” said Arvind Satyam, chief commercial officer for Pano.

Usually when a fire gets called into 911, it gets registered as a fire call, then gets sent to the nearest fire agency, and it’s then up to that agency to figure out where the incident is, how large it is and what resources to send.

“We compress all that,” Satyam said. “First and foremost, we want to be able to pick it up with powerful zoom functionality so that people can look at what’s happening in the early moments. Triangulation is a huge point here; a lot of times it’s on the fire agencies to send out engines to figure out where the fire is and what’s happening.”

Pano has found that an optimum distance for the cameras is about 10 miles, depending on the terrain and conditions, such as haze and fog.

“As soon as smoke or heat is detected on the landscape, our fire managers can direct air and ground suppression forces to the source in record time. It also means more time for nearby communities to receive information from emergency services and act quickly if evacuations are required,” Hilary Franz, commissioner of Public Lands for Washington DNR, said in a press release. “When a fire is bearing down on your property, you don’t care how or when it started, you need to know that help is on the way right now.”

Pano uses AI as well as satellite technology and other sources to ensure that it has all the intelligence it can at the time to give DNR the best possible head start on a wildfire. The software is overlaid with map functionality and other assets that, along with the expertise in the intelligence center, provide the best fusion of technology and expertise available.

“The AI algorithm is detecting smoke during the day and night, so essentially you’re using AI on this full 360-degree panorama that’s continuously updating to detect smoke and fire,” Satyam said.

“And we have the human intelligence center where part of the issue in detecting smoke is that it’s a hard problem; it could be a cloud of dust by a farmer driving a tractor. The human intelligence center, by looking at every alert, can quickly determine what it is.”