We already city-level air quality indicies that let us do comparisons between cities - http://www.airnow.gov/
Anyway, it's not a hard problem to solve. Street View cars already revisit the same streets they've been to before, so eventually they'd collect enough data to be able to average out those sources of variability.
This looks more like it would be useful for comparing neighborhoods within a city. How much worse is the air in the industrial districts compared to the residential ones? What is the difference in air quality in each of San Francisco's microclimates?
But traffic is a large source of solution, so that would bias their data. It would be cool if they strapped these sensors onto the Street View backpacks that they use to map hiking trails, and get data on the air quality along walking paths in parks and other areas that can't be driven to.