I have a polygon layer representing impervious land cover. This was taken from a land cover raster dataset. The impervious data represents all impervious surface that is not buildings or roads. I am trying to identify parking lots within the impervious features, but I am getting all sorts of omission and commission errors.

What I am finding is that I will identify parking lot features in the impervious layer, but they will be connected to other impervious (non parking lots) by long narrow strips of impervious (could be sidewalks or unofficial roads). I'm thinking if I can somehow remove these strips from my impervious features, I could get more accurate parking lots.

Parking Lot Features

As you can see in the image, the selected polygon (in cyan) covers the parking lot at an airport, but is connect by impervious to the runway. I just want the parking lot.

Can anyone recommend a way to remove the strips from the polygons? I can't think of way using Vector tools in ArcGIS 10.0. Is there a way if I convert to raster?

  • 1
    It seems like you may want to, if possible, go back to your original land-cover raster and reclassify it to show not just impervious areas, but if there is some reflectivity data or other sort of comparison data that would allow you to separate the different types of impervious surfaces. It seems like the different pavement types, like sidewalk, parking lot and runway would have unique characteristics, if you have the proper raster data to analyze. Do you have access to false color or infrared of some sort? Also, how large is your area? – Get Spatial Nov 9 '12 at 16:38
  • 4
    You might be thinking of morphological grid operations, such as shrinking the roads down to nothing and then re-expanding what's left to fill its original location. See help.arcgis.com/en/arcgisdesktop/10.0/help/index.html#/…. – whuber Nov 9 '12 at 16:44
  • @Get Spatial's idea will work nicely if you use the current impervious polygons as a mask when reclassifying. – ccn Jan 29 '13 at 2:27

could you calculate some stats for each of the polygons and then remove all those over a certain length, width or area?? just thinking that if you know that the polygons you want to remove are long and thin and you can find some measurements that represent this then you could identify and remove the ones you don't want.


I ended up manually digitizing here to correct the issues as I was unable to go back and reclassify the landcover. At the time I posted this question, I was fairly unfamiliar with the raster operations mentioned by @whuber, but since then I have learned much more about those operations and I suspect they would do the job quite nicely.

I will have to try the shrink and expand combination next time I run into this.

Your Answer

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

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