How scientists are using artificial intelligence to fight pollution from coal trains
A butcher’s shop isn’t the first place you’d think to meet an air quality researcher. But, strange as it may seem, right here, outside Golden Gate Meats in Richmond, is a perfect location for Dr. Nick Spada. He’s a scientist with UC Davis’s Air Quality Research Center, at the Crocker Nuclear Lab.
Spada pops open the back of his Prius hatchback. Inside there’s a heavy-duty, watertight plastic case that takes up about half the trunk. “[It’s] filled with electronics,” he says. “We have a computer here that handles all the data processing, collecting all the information. It has a personal weather station attached to it, a particle sensor that gives us the measure of air quality.”
Attached to the inside lid of the case there’s a device that looks like a small wifi router, with four heavy-duty antennas bolted to its corners, sticking out at angles. There’s a cooling fan and a bunch of little electronic modules, with maybe a dozen USB ports between them. And there’s a web of cables connecting everything together.
But the crown jewel of Spada’s set-up is a night vision camera, about the size of the one on your smartphone. It sits inside a cylindrical security camera shell… and it kind of has a mind of its own.
“It's connected to the computer and also has an AI accelerator,” Spada Says. “And what that was doing is, that was detecting whether or not trains were present.”
That AI “mind” is really good at one thing: recognizing when it sees a train. But it’s not looking for just any trains – it’s specifically looking for trains carrying coal, like the ones that pass Golden Gate Meats regularly. The coal is mined in Utah, and brought here in open-top cars a mile long, to be loaded on ships and sold overseas.
There’s been debate over whether or not these coal trains pollute the communities they travel through. And answering that question has been tough, because it’s hard to catch the trains passing by. They often come through in the middle of the night – not exactly the best time to have a scientist sitting by the tracks taking measurements.
Now this camera – and the AI mind powering it – is attempting to put that debate to rest. But recognizing coal trains is a skill the AI camera had to learn.
“We trained it with about 1 million pictures that we did have to classify,” Spada says. “We sort of created a little computer game that I tested with my kids using a PlayStation controller.”
Together, Spada and his kids went through picture after picture of different types of trains, breaking them down into categories for the AI. Once the humans told the computer the difference between passenger and freight trains enough times, the computer could start making those decisions for itself.
“So once it was able to do that, we started recording every single train event with one second before and after, so that we could very clearly say, ‘Okay, this is the very beginning of the train. This is the very end of the train.’ We would have the timestamps for that.”
With the particle sensor measuring the air quality every second, Spada was able to compare the air quality while different types of trains were passing, to the average air quality for the area. Then he could tell if coal trains were leaving more pollution than passenger trains or other freight trains. He was attempting to answer a question that had implications for Oakland, a few miles away. It was a question that had come up in court.
COMING UP WITH THE IDEA
According to Dr. Bart Ostro, “studying coal trains coming through is a difficult process, because you don't really know when they're happening.” Ostro is Dr. Spada’s colleague at UC Davis. At this point, he’s semi-retired, but he’s a big name in environmental regulation. He spent decades working with the EPA, setting environmental standards for California and the nation. That’s one of the reasons he was asked to address Oakland’s city council way back in 2015.
The City, and environmentalists, were concerned: they had just found out that a local developer wanted to build the largest coal export terminal on the West Coast at the old Oakland Army Base. If that happened, the amount of coal moving through Oakland by train each year would dwarf the amount that was already moving through Richmond.
Oakland was worried about the health and environmental effects the coal terminal would bring, so they passed an ordinance banning the bulk storage and handling of coal within city limits. But the developer took the City to court over its ban.
“The city of Oakland lost the court case,” Ostro says, “because the judge had decided that there was not any proof that the trains passing by would cause health effects.”
Part of the reason the judge wasn’t seeing that proof was that, for the most part, the exact type he was looking for didn’t exist yet. That’s because, so far, most studies of the health effects of coal dust have focused on conditions inside mines, not cities.
“It's very interesting that there have basically been no studies in urban areas that have looked at this carefully, and have collected enough data to feel comfortable reporting it.”
A few months after Oakland lost their case, Ostro had an idea: The coal already being shipped through Richmond was the same type of Utah coal that would be heading to Oakland if a terminal was built there. And it was getting to Richmond the same way coal would come to Oakland – in those mile-long trains with open-topped cars. Why not study the trains in Richmond? And in the meantime, he could find out what effects the coal trains had on people there.
“We have coal trains passing through economically disadvantaged cities and populations throughout the United States and actually, throughout the world,” Ostro says. “So there's also global implications, I think, of doing this work.”
BUILDING THE TEAM
When Ostro got the idea for his study, the first person he called was a public health consultant named Dr. Heather Kuiper.
According to Kuiper, the pollution that coal trains leave behind is especially bad for the young and elderly. It also has an outsized impact on folks who don’t have good access to medical care and on people who have underlying health conditions. “So if you already are experiencing higher rates of asthma or cardiovascular disease in a community,” she says, “those communities will be even more impacted by particulate matter. And that's where we get into an issue of justice.”
West Oakland, where the coal terminal was slated to be built, is an area where redlining has historically concentrated the city’s Black and low-income residents. It’s a part of The Town that already has an increased pollution burden from the freeways that surround it, the ships idling at the Port and all the semi traffic bringing products to and from those ships.
Kuiper says that, if a coal terminal is built there, West Oakland residents will be at increased risk for a whole range of negative health effects. But it’s not just the breathing-related problems you might expect.
“We see particulate matter and its different forms being associated with cardiovascular disease, respiratory diseases, cancers, low birth weight, premature birth, and cognitive impacts, such as Alzheimer's disease, and other effects such as, potentially, autism for children as well. But wait, there's more!”
She says that the tiny particles found in coal dust can actually alter your DNA in ways that you can pass on to your kids.
Back in Richmond Dr. Spada tells me that, after months of observation, his monitors found coal trains add more pollution to the air than any other type of train. Even after they had dropped off their load, the empty cars traveling back to the mine added more air pollution than either passenger trains or freight trains carrying other types of cargo.
“That's not counting all of the other particles in the air that are naturally in the air,” Spada says. “And then also things like cars and you know, these other pollutants that all urban environments have.”
The data the team of Ostro, Spada and Kuiper collected was published in a report earlier this year. It amounts to the most extensive study to-date of the effects that coal trains have on the urban communities they pass through. Their study was made possible by the use of artificial intelligence. But with all the fear over AI in the news these days, Spada wants people to understand that, at least sometimes, it’s a tool that can be put to good use.
“It was much better for my undergraduate students to be working, focusing on the data analysis, and learning to do field work, rather than spending all their time on gruntwork.”
The City of Oakland is currently in court over the fate of the old Oakland Army Base, with closing arguments set for October 11th, 2023. If the City wins, it’s unlikely that a coal terminal will ever be built there. But if the City loses, the evidence Spada and the team have collected in Richmond can be used in the next round of litigation, to show that coal cars pollute the communities they travel through. In the meantime, their research can serve as a model to researchers around the world working on the issue.