One thing you should know about me is that I am obsessed with data.
What I am proposing is a model where we can collect data from set of sensors, ideally located across the city (but since I'm probably one of one here in Sydney I will fudge it), and then using analytic processing, try to predict air quality by mapping my data to other sources (weather bureau, traffic, calendar, etc) and see if we can't generate alerts with good accuracy.
More details as they come to hand!