Castle Rock taps AI to plug underground water leaks — and prevent the loss of millions of gallons

Castle Rock Water lost 166 million gallons of the stuff in 2023, as the life-giving liquid silently escaped from pipes, fittings and valves across the town’s 517-mile water system — mostly underground and largely undetected.

Losses in one recent year leaped as high as 240 million gallons — or 737 acre-feet, enough to supply roughly 1,000 families with water for a year. The leakage costs the town anywhere from $400,000 to $650,000 in annual revenues.

“It’s called non-revenue water — it’s water we produce but we don’t get to charge the customer for it,” said Mark Marlowe, director of Castle Rock Water.

More importantly, it’s a waste of a vital resource that communities in Colorado can ill-afford to lose, as the climate warms and the American West dries.

To stem its losses, Castle Rock will launch a six-month pilot program next month with a Canada-based firm specializing in artificial intelligence technology. It’s designed to detect leaks in municipal water systems early and often, making a fix easier and cheaper than having to assemble a team in the middle of the night to tackle a busted main. The Douglas County town of 81,000 uses approximately 3.2 billion gallons of water a year.

“Emergencies never happen at noon on a Wednesday — they happen at 3 a.m. on Christmas Eve,” Marlowe said. “You have to go in and fix it proactively, before it becomes a bigger leak and surfaces.”

The $90,000 pilot will place “acoustical loggers” in eight fire hydrants throughout the town’s Cobblestone Ranch neighborhood. Digital Water Solutions’ president, Tim Sutherns, said the company’s equipment will “actually listen to the water,” detecting and collecting the particular sounds the flowing water produces as it courses through the system — or out of it.

It’s a more precise approach than what many leak detection systems use today, Sutherns said — listening to the pipes themselves for vibrations and acoustical signals that might indicate a crack or split.

“PVC pipe is challenging for leak detection, because it doesn’t resonate,” he said.

Digital Water Solutions’ embedded loggers, called “hydrant.AI” devices, measure pressure, acoustics and water temperature from inside the water column and continuously monitor for aberrations that could signal a problem. Those data points are collected and sent via cellular connection to the company’s cloud servers.

“When there’s a leak in the pipes, there are certain frequencies created,” Sutherns said. “There’s a different noise when water is exiting a pipe than when it is flowing through the pipe.”

The company’s artificial intelligence prowess then kicks in.

“That is where we apply our machine learning to it,” he said.

The 7-year-old company, which has installed its equipment in 40 municipalities across the United States, Canada and Iceland, has amassed enough data to create powerful AI models that can quickly identify problems deep underground.

“Over time, that grows,” Sutherns said. “And the accuracy of the models increase.”

Artificial intelligence, a technology that has received gobs of attention in recent years, is a particularly valuable and effective tool for the water utility sector, Gigi Karmous-Edwards said. She runs her own consultancy and was one of 26 experts who worked on the American Water Works Association’s Water 2050 Technology Think Tank.

That’s largely due to the fact that the infrastructure for water systems exists mostly underground, she said. Down there, visibility is compromised and small problems can gradually become big problems that only become apparent when they surface.

According to the American Society of Civil Engineers, the country’s drinking water infrastructure consists of 2.2 million miles of underground pipes, many of which are aging and undermaintained. There is a water main break every 2 minutes in the United States, with an estimated 6 billion gallons of treated water lost each day, according to the organization’s most recent infrastructure report.

Last May, a break in a 32-inch water main in downtown Atlanta brought the city to crawl and triggered boil water advisories across the Georgia capital. And on Friday, a similar failure in the northern suburbs of Chicago caused significant flooding and trapped cars, forcing residents to navigate frigid water and ice to get to safety.

Some of those calamities could be averted if water utilities catch wear and tear on the system early, Karmous-Edwards said.

“When you put this technology in play, you start pulling together the story of what’s going on underground,” she said. “The more knowledge you get from sensors and the combination of different data sets and modeling capabilities, the faster you get early detection.”

It’s not clear how widespread the use of AI is in municipal water systems. Digital Water Solutions’ Sutherns said Castle Rock is his company’s first Colorado customer. But just last week, Farmington, New Mexico, started using technology from FIDO Tech, a U.K.-based company, to help detect leaks across 200 or so miles of its water delivery system over the next decade.