Carbondale firefighters latest to sign on to AI cameras for early fire detection

Fire departments in the region and the state are partnering with “smoke detectors in the sky” to boost early fire detection. Carbondale Rural Fire and Protection District will be the next in Colorado to add artificial intelligence to its fire mitigation toolbox.

With funding secured from a global financier, Carbondale Fire will contract early fire detection company Pano AI to install panoramic cameras on a Pitkin County telecommunications tower at Elephant Mountain in the Crystal Valley. 

Deputy Chief Mike Wagner said the program will bolster the district’s preventative measures that aim to keep small fires small.

“We call that initial attack in the wildfire business. In the summertime we have crews that are out patrolling, especially during the afternoon when the fire danger or the burning period’s high,” Wagner explained. “What Pano AI does is just put another tool in the toolbox and another eye in the sky.”

The panoramic cameras, installed in pairs called ‘stations’, cost $50,000 and ensure 360-degree visibility. They survey the surrounding land 24/7, scanning for movement that resembles smoke. The price includes hardware installation, maintenance, an analyst team and customer service.

Global financier Stephane de Baets agreed to underwrite $40,000 annually for the next five years through RC Ownership LLC, which owns the Redstone Castle. To raise the additional $10,000 per year, de Baets said they will organize a fundraiser later in September in hopes that the community at large will chip in.

“Our algorithm has been fed 300 million different images and scenarios of wildfire smoke. So it’s good, and it’s going to get better every day,” said Pano AI Business Development Manager Kat Williams. “That’s why we’ve always had the ‘human in the loop’ component to it. Because when it comes to wildfire detection, it’s important to have (both AI and people).”

If the AI detects texture and movement associated with smoke in the imagery, the alert is sent to the Pano AI analyst team. If the analysts deem it actionable, then the alert goes out to local fire districts via text and email. Then fire officials can review the imagery and determine a course of action.

Visibility range depends on installation site, but generally the cameras can see miles out and beyond a valley. If multiple stations overlap, the AI can calculate the exact latitude and longitude of the smoke alert, Williams said. And if the smoke is visible from just one station, the AI will generate a bearing line and plot it on a map for targeted fire department action.

Installation at Elephant Mountain depends upon how quickly a helicopter can be secured to access the tower, but everyone involved hopes the date will be as soon as possible. Wildfires raging in Canada, have made it difficult to get the type of helicopter needed.

Throughout the delays, Wagner said that Elephant Mountain remains the best location due to the infrastructure and view of the Crystal Valley. 

The price tag includes hardware installation, maintenance, an analyst team and customer service. 

Wagner said that the cost was prohibitive for his district at first. 

“Is it better to spend 50 grand on one person that can actually (survey the fire) in their truck and can see it? Or is it better to (spend $50,000) for a camera?” Wagner pondered. “It’s just one of those hard funding things. We’ve got so much money and where can we put the resources?”

But Carbondale Fire won’t have to make any hard funding decisions regarding Pano AI for a while.

Aspen Fire Department was the first Colorado fire house to partner with Pano AI back in 2020. Fire Chief Rick Balentine said the technology has changed the way they allocate staff resources and made them more efficient.

“Before the advent of (Pano AI) where we can really (the smoke), we may have sent a full fire truck type vehicle and all that, but now that we have a little more intel, we may not send as many people from our district,” he said. “Once you send people up Castle Creek for 10 miles, they’re out of district, so you may be able to just send one command vehicle up there instead.”

Now in year three of working with Pano AI, Balentine said he’s seen the AI evolve as the algorithm learns that the golden colors of fall Aspen leaves is not a cause for alarm or how to detect smoke when the ground and trees are blanketed with snow. 

And with fire seasons becoming fire years as climate change brings about extreme weather events, early fire detection is just as crucial in the fall and winter as in the spring or summer. 

To pay for the four cameras in their district, Balentine said that they partnered with the city of Aspen and Pitkin County for funds. 

“If it detects one fire that we can stop from getting bigger, it’s well worth the money in my opinion,” he said.

Pano AI expects to operate more than 40 stations in Colorado by the end of the year.