2
$\begingroup$

I am using land use regression with machine learning to create an air pollution model. However, almost all tiles I can find using Python and Folium contain text, which my intuition says cannot possibly be a good thing. I've seen papers use drawn maps for this task, and I am not convinced they are better than satellite images...

Does anybody here have experience with extracting city maps for air pollution modelling? Which ones should I use? Is satellite data better than drawn maps?

Improve this question
$\endgroup$
3
  • $\begingroup$ Could you provide a bit more detail in the form of some examples? What is your exact question, and what is wrong with the status quo? How about land.copernicus.eu/global/products/lai ? $\endgroup$
    – Erik
    Nov 1, 2021 at 9:19
  • $\begingroup$ @Erik I am writing an answer to my own question right now. Hopefully that answer will clarify what I was looking for. Also, I'm open for other answers. $\endgroup$
    – Avatrin
    Nov 1, 2021 at 9:35
  • 1
    $\begingroup$ So, if I get your question correct, you want to model air pollution, with as input a land use cover map, and you need a product that gives you land use? So [input data] -> land use -> air pollution? If it needs to be a European product only, I think you cannot beat land.copernicus.eu/pan-european/corine-land-cover which already gives you the land use data in either 300x300 m resolution, or even full vector format, without any text. These maps are compiled by local authorities. $\endgroup$
    – Erik
    Nov 1, 2021 at 10:55

1 Answer 1

Reset to default
1
$\begingroup$

Satellite data is more noisy than rasterized vector maps; Steininger et al found that OpenStreetMap input was superior for their CNN-based land use regression model. They tested OpenStreetMap, Google Maps, Satellite images from Google maps and a combination of OpenStreetMap and Google Maps Satellite. The model trained on inputs from OpenStreetMap outperformed all of the others.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.