> Thanks for posting. I would have thought narrow lines would produce the problem that those specific lines would become unavailable for study, or would be freely mixed with the "real" data in confusing ways. It didn't occur to me that they could simply be subtracted in a deterministic way.

Your original thought is basically correct: we do lose the ability to do any meaningful work at the wavelengths of the lamp emission lines. But the damage is concentrated at a few wavelengths. So we trade having noise over the entire spectrum for a few really, really noisy pieces of spectrum. It's generally a excellent tradeoff.

But even low pressure sodium lamps have some emission spread over a large wavelength range, so it is becoming increasingly difficult to do optical work near major urban areas. In the long run most of those sites will probably transition to mostly IR work.