Air pollution is a structural problem that sometimes can be episodic. Today there are several ways to measure air pollution. It can be described through air stations - points with high accuracy - or through satellite sense - e.g., S5p/tropomi with 5 km resolution. It can be described through simulations or even Land Use Regressions models (LURs) - e.g., Lobelia/isglobal. However, all these methods have drawbacks:
• Air stations are zonal indicators - i.e., not very detailed.
• Satellite sensing has low resolutions.
• Simulations are too complex and require much computational power.
• LURs do not describe urban complexity, which is a fundamental feature.
Together with 300,000 km/s, a tech-urban start-up located in Barcelona, Davide, and Peter - from now on, we - considered a new scenario to test. As air pollution is a problem of mobility generated by land-use composition, we have tried to predict it using measurements from existing air stations and several features, including but not limited to: urban density, income data, land use composition, building density, dwelling information, and population density and composition.
All the notebooks are in the Notebook folder, and Paper.pdf is the paper related to the project.