By Sierra Garcia, San Francisco Estuary Institute
The day before the first major fall storm in 2025, Kayli Paterson took four of her new coworkers to creeks in Richmond, Albany, and Oakland, tied them up securely, and left them.
Like others on the sampling team, the new members were braving the storm to collect samples of contaminants in urban stormwater runoff. Unlike the rest of the crew, they weighed 25 pounds each and were the size of bulky briefcases. After three years of scheming, tinkering, and trial and error, Paterson and the SFEI team are putting a fleet of stormwater sampling robots to work.
“Our little robot samplers, they don’t need showers. They don’t need sleep,” says Don Yee, who developed the first generation of samplers in 2023 and built the new robots along with Paterson and a small team. “They just need their batteries charged.”
The 15 robot samplers are enclosed in black Pelican cases with tubing and bottles to capture water samples. Before each storm, SFEI scientists dispatch the robots to the banks of urban creeks and channels. They secure them to bridge railings or trees above the waterline, then position a slim length of tubing towards the water, like a meters-long straw.
Using machines to collect stormwater samples isn’t a new idea. But it is novel to build ones portable enough to sample stormwater in different places.
“The usual way is you build a large housing of some sort,” Yee says. “Then you connect it with power or a solar panel and a big bank of batteries. You build pipes to keep the sensor and sampling lines safe, and you anchor things down.”
Improved technology let the team go beyond that, explains Senior Scientist Kelly Moran.
“The little tiny computer that runs the brains of the system is both smart and draws very little power,” she says. “That makes it possible to stick it out in the field for a week on a battery that isn’t too heavy to carry to and from the sites.”
Moran first conceived the stormwater robots idea in 2022 based on a basic blueprint for lakewater samplers. Yee and Paterson ran with that blueprint. They added sensors so the robots could start sucking up samples when they detected rising stormwater, and redesigned them to sample from bridges or creek banks instead of floating in the water. The team then worked with SFEI technologists and volunteer Mark Eliot to overhaul the design so scientists could remotely change settings on the robots in real time. More than two years of tinkering and what Yee calls “mad scientist” efforts led to the current generation, dubbed the “Omega Mosquito” samplers.
The robots can now collect data from a dozen sites during the same storm. Before, with human-only teams, “we were lucky if we got two or three sites” per storm, says Paterson. Because storms with enough rainfall to sample are far-between and hard to predict, scientists usually need several years to collect enough data to answer critical questions—like tracking down possible sources of forever chemicals washing into the Bay. Collecting four times as many samples per storm helps scientists gain more confidence in the results sooner.
Using robots can also keep scientists safer in the field. Yee recalls sampling a South Bay creek in the early 2000s that looked “pretty harmless” most of the time. But during one storm, he noticed tree leaves on the water surface zooming past him at the pace of a brisk jog—and disappearing into a culvert under the freeway with no protective grate.
“If I fell into this, I would be going under 101 and not popping out for a few hundred meters,” he remembers thinking. “That highlighted to me the potential dangers involved in stormwater sampling.”
But the robots aren’t taking jobs away from stormwater sampling scientists. Yee now stays dry during storms as he manipulates the robots’ instructions for when to sip up samples, but other scientists on the team are still out there, day and night, collecting stormwater samples by hand. This year, the robots are sampling PFAS (“forever chemicals”), which frees up the human stormwater team to concentrate on samples for other contaminants during storms. In the future, the robots might also help sample these other contaminants. For the scientists, this expanded information is the most exciting part of robot coworkers.
“When I was starting out, a lot of our sampling was just to get information. We had no idea what concentrations were in stormwater,” says Yee. “Now we’re trying to think about coordinating sampling in a more integrated, holistic way.”
This is part two of a special feature series on stormwater. Read the first part of “The Storm Chasers” to learn why we sample stormwater for next-generation pollutants.
Photos by Sierra Garcia.
