I am working on a project to answer the following geographic question: How much have slum areas increased in Mumbai, India, since the 1970's (or earliest data).

I have downloaded LandSat imagery and plan to perform a series of raster classifications to map slum areas. I have imagery from 1975 (LandSat 2), 1993 (LandSat 5), and 2014 (LandSat 8).

Is my approach to this question practical given the data I have access too? I feel I can work with my latest data set, but how much accuracy can I classify images from LandSat 2 and 5? I know that the cell size of LandSat 2 is pretty course(60m). I am using ArcGIS software and would like to know how more experienced GIS professionals would approach this?

  • Do you mean to use RGB channels of LANDSAT imagery to manualy digitize slums? Please be more specific regarding the methodology you had in mind, if you would like valuable answeres – dof1985 Apr 17 '15 at 19:12
  • At 60 and even 30 meters, differentiating 'slums' from regular neighborhoods will likely prove fairly tough to do. Even at 15m for the new ETM panchromatic seems fairly hopeful. Down in the 5-1m range would be better. I'm not an expert, but in my opinion that data source isn't suitable for the task and I would look into more detailed imagery. – Chris W Apr 17 '15 at 21:15
  • I agree with ChrisW. I could see distinguishing rural areas from urbanized areas with such imagery, but delineating specifically between slums and a middle class neighborhood would be very difficult. If I were you I would look for non-satellite aerial imagery and then define the rules for what a slum looks like. – GotsMahBox Apr 17 '15 at 21:42
  • Slums likely have the same spectral signature as all urban areas, although they may have higher building density. How would you differentiate between slums and other urban areas? – Aaron Jun 23 '15 at 19:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.