we recently proposed a methodology to capture the entire smellscape of a city and not only the bad odors typically the focus on urban planners.
We proposed the first urban smell dictionary organized on 10 main smell categories.
We all know that social media data is biased, we tried to validate our approach using complementary data sources from city officials or from mapping platform like open street map
We correlate the presence of certain smell clusters and air quality indicators at the level of street segment.
As one expects, streets with emissions words suffer from air pollution (the red bars) and streets with nature words do not (the green bars).
The correlations are as high as 0.47 which is far higher than we expected.