@ Santhosh,

Please note that the Multi-angle Imaging SpectroRadiometer (MISR) instrument on NASA's Terra platform has been used to monitor atmospheric properties, including air pollution in the form of aerosols, since late February 2000. This instrument provides global coverage and is still operational today; it therefore offers a huge source of data and information on those issues. You will find a general description of the instrument, mission and products here:

https://misr.jpl.nasa.gov/

In particular, you may want to browse through the large collection of papers dedicated to that issue and archived here:

https://misr.jpl.nasa.gov/publications/peerReviewed/

Look, in particular, for papers dealing with atmospheric pollution in India, if that's your region of interest. If you require access to MISR data, then visit this web site:

https://asdc.larc.nasa.gov/project/MISR

Lastly, keep in mind that a successor instrument to MISR, called MAIA, will be launched in 2022 and will be dedicated to monitoring air pollution in cities. You can learn more about this upcoming project here:

https://maia.jpl.nasa.gov/

I hope this may help you. Best regards, Michel.

Hi Santthosh,

Great issue, thank you for setting it up.

I think that the regression approach is plausible and thus will stay with this subject. If you think regression modeling, you need a theoretical basis for the regression equation (or equations, in the more complicated issue of simultaneities between the independent variables).

A theory of rent

The theoretical basis is found in land economics as rent theory. There you will find that (at least before COVID 19) the highest rents were paid at the centroid of the central business district, CBD, declining away from it as a function of distance (CDB = 0 distance). Note that the rent functions should be 3-dimensional. The z-axis measures rent (e.g., $/square meter) as a function of distance measured as points in the x,y space (or as a function of angles and distance from the due north bearing, or latitude and longitude). My own work used land parcel values, over several years, to measure the effect of a new water reservoir on its nearby property values, including the effect of views and the like. The theoretical view is that people prefer living near the CBD: the trade-off is travel costs to the CBD with lower rents. Lower rents are paid far from the CBD but, if you travel to the CBD you have to pay parking, may have local income tax on your income, and waste time on the road (the travel versus leisure trade-off).

Rent functions are affected by air pollution.

This would inform you of the effect of pollution by including costs from air pollution (e.g., having to paint and clean more often than in areas where pollution is lower). You can build more complicated models that account for other factors affecting rents. You can also use other factors such as population density/unit area, distances from parks, hospitals, prisons, and so on. Those affect the otherwise theoretically smooth function where the highest rents are paid in the CBD, declining as distance increases. There can be multiple equivalents of the single ideal CBD, a spatial analysis would capture these. I would develop the working hypotheses according to the way I think the air pollution (the various gases, heavy metals, particulate matter 2.5 and 10 micrometers), and other economic variables determine or otherwise affect rents or land values. I would use local data on air pollution. I would be very careful not to make the model specification error (check a textbook on econometrics) and certainly, after estimation, verify the assumptions of multicollinearity, constant variance, and so on. I would also see if I could develop local time series of rents, land values and air pollution.

You should be able to spatially correlate concentrations of criteria pollutants (US law requires their monitoring in a city) with land values over time and space. You can also correlate mortality rates at what we call zip code level from government statistics and other sources with the data from local aerometric and air pollution monitoring stations. In general, lower-income areas have a higher average prevalence of respiratory diseases and cancer. I would be careful with prevalence numbers: preferably use incidence numbers if you can. Do not limit your data to mortality because morbidity is also critical.

I would be careful about the generalizability of any study that is only observational. But, if you simply want to describe an issue, rather than predict, the suggestions by Michael are pretty nice.

I do not have access to my old papers, but you can scour the web for theoretical and modeling references probably as well as I can.

I hope this helps.

https://www.mdpi.com/2071-1050/11/1/25