Not that I don't think the science is good, I'm sure they do and the article says they do, but I struggle to see how you can isolate noise from particulate exposure: they'd have semi identical square law dropoff by distance and volume.
1. Electric trains wont generate diesel particulate, but will generate road noise.
2. It's relatively straightforward to measure pollution and noise, so you can do analysis to try and separate them.
In many situations these would not be correlated exactly. For example, terrain, wind patterns, noise dampening structures, etc.
They don't generate diesel particulate, by they do generate "other PM10 and finer particulate", from the brakes for example.
In Paris there's a lawsuit against the RATP (local transit authority) for particulate pollution inside the stations. [0]
There's also the fact that, at least in Paris, some metros are running on tires. For those that run on metal wheels and rails, some turns are quite sharp, and you can hear the screeching of the metal. This is likely to contribute to particulate pollution.
I suppose for overground trains, this is less of an issue.
---
[0] https://www.rfi.fr/en/france/20210323-french-ngo-sues-paris-...
[1] https://en.wikipedia.org/wiki/Montreal_Metro#Design
Popcorn!
The other problem is that people that live near noise are poorer. Did they control for wealth?
These epidemiological observational studies are so hard to take seriously. I get why so many of them are done though. It's a relatively lower effort way to get publications out the door, all you need is a dataset and a few days in python.
It's hard to take seriously HN commenters who assume people working outside software development are all idiots who need to be told how to do their job.
I've worked in academic applied statistics research and seen how the cookie crumbles, it's not pretty, and yes there are groups and industries of people who may be individually intelligent but nevertheless act like idiots given the incentives they face to churn out low-effort publications. Search for "hegemony" or "Fuzzy neural net Dow Jones" in Google Scholar to see 140 IQ people mass produce drivel. Just because I am not trained in medicine doesn't mean I am not allowed to draw this conclusion, I have years of experience in statistics and know a bullshit application of these tools when I see it, and I don't appreciate appeals to authority or other arguments that try to invalidly shut down people's opinions without knowing why those opinions were formed in the first place.
“The results are strengthened by the authors controlling for levels of traffic pollution, which was recently shown to be associated with dementia as well as other diseases.”
I think the idea is that people living in the same building should all have roughly the same particulate and pollution exposure, but people living on a road-facing side of the building will have more exposure to road noise. It's an extremely clever study design.