You'll need to define "legitimately obfuscate" here - presumably you're proposing that a majority of data consumers would consider wildfires irrelevant in their assessment of air quality. Why do you think this?

Anecdata but off the top of my head I can only think of data consumption in which wildfires would be relevant - excluding the data gives a harmfully incomplete picture.

It also strikes me that data annotation would solve whatever your use-case might be if data needs to be excluded?