visual observations would at least have some arguable scientific validity. I can't say for this particular study, but they usually survey the number of fast food restaurants per capita, the number of parks and recreational opportunities (notice that the listing contains a lot of anecdotes about the "programs" that are suppposed to solve the problems). In short, nothing about the actual people, just the environment of the city, and they extrapolate from that. As I said, I don't know about this survey though.
I did see the fast food reference and recreational reference. I would think though that body weight information would be needed too.