Landuse for mixed retail/residential/commercial areas

Hi folks,

Let’s say there’s a street lined by medium density buildings (three or more levels high). The ground floors are taken up by shops, restaurants, and cafes. The levels above are all apartments. When you’re walking around at street level the “vibe” is that of a food & shopping area (despite there technically being many more apartments above than shops below).

How would you tag landuse?

  • retail
  • residential
  • retail;residential
  • Overlapping retail & residential areas
  • Too ambiguous - I deliberately wouldn’t add landuse here
  • Something else (please comment)
0 voters

Asking because I’ve seen different approaches recommended, but I’m not sure if there’s a clear preference for one vs. another. The wiki also seems to be silent on the matter.

Note: the question really applies for any retail/residential/commercial combination, but I figured I’d focus the poll on a concrete scenario to make it clearer.

Thanks for your thoughts!

We discussed this recently at a pub meetup and agreed residential was acceptable.

I suppose this mirrors building=apartments where we are also supposed to ignore what’s on the ground floor.

Out in the “real world” I wouldn’t really describe a busy shopping street as a residential area just because there are flats above. In my view it wouldn’t hurt if we had a tag or tagging scheme for it. You could ask the same question for shops with offices above - is that retail or commercial?

I’m not a GIS expert and it would be interesting how other classification systems approach this. Have they come up with rules, for example, look mainly at the ground floor because that defines the character of an area when you are walking around it? Look at the dominant land use in terms of floorspace? Some sort of three-dimensional approach?

1 Like

I answered retail because I always tag landuse based on the vibe at street level. I can think of many retail areas that have more residential apartments or commercial offices space on the upper floors than retail space at ground level even though every single ground level unit is a shop, restaurant, or bar. This is simply because there is only one ground floor and several floors above. It doesn’t change the fact that the character of the area is a shopping district. If such areas are tagged residential or commercial then they aren’t distinguishable from areas where the vibe at ground level (character) is residential housing only or commercial offices only.


We discussed this recently at a pub meetup and agreed residential was acceptable.

yes if the area is mostly apartments and you consider it residential, then the occasional pub or small shop will not change it. In this example, where every ground floor is commercially used, I tend to see it differently but could maybe still accept residential (although the OP describing the character as food and shopping area IMHO clearly speaks against it). The question is even clearer when the upper floors aren’t all apartments but only some , and others are offices, some shops extend to the upper floors, etc. so that residential is even clearer not the primary landuse.

1 Like

I live 200 meters from a shopping center with several layers of apartments on top of the shops. It simply is both residential and retail. Since the retail area is in a much larger residential area, it’s simply mapped as a retail area inside a much larger residential area (no cutout). That is the best representation of ground truth.


On occasional I map such ground floor retail, apartment stack on top as building=retail + building=levels=1 & building=apartments + layer=1 + building:levels=N and the rest of the building feats, no and can’t be bothered in such 2 part cases to get into building:part mapping an overall building outline and two elements within. Very often the retail outline is larger than the apartment block on top, Don’t see me mapping landuse=retail following the retail outline of the building as ‘ground truth’, mapped but invisible, most very often the parkings serving both residents and customers and everything else around the overall building outline breathing residential.

I’m now curious about opinions on the retail areas I mapped in Bennekom and Wageningen.

Thanks all for voting! To summarise the results so far, residential has a clear (but not overwhelming) majority, with retail being the main runner up.

In more general terms, it sounds like:

  • Most mappers pick a single landuse based on whatever occupies the most floorspace (presumably with the street level serving as a tie-breaker for two storey areas).
  • Some mappers consider street level to be of primary importance, and set a single landuse based only on that.
  • All other approaches have very limited support.

Do we think the wiki should be updated to more clearly articulate this? Or maybe 25 votes on a poll isn’t strong enough evidence for that…

(I guess I just find the current ambiguity a little frustrating. Perhaps I need to make peace with it though :rofl:)

Oh, and my 2c on which approach seems more useful: probably retail? (i.e the street level approach.)

The main use case I’m thinking of is looking at a zoomed-out view of the city on my phone (in an app like Organic Maps or OsmAnd). Similar to what Apple & Google Maps do, I’d like to see the retail “hub” areas outlined in some way.

Based on the poll results though, it sounds like landuse areas just aren’t appropriate for this. Apps would instead need to analyse POI density to approximate hub areas…?

Perhaps there are other use cases for the data, where the “most floorspace” interpretation makes more sense.

1 Like

Same to me and that is why I personally think “retail” fits best. Every part of a township or city where some people are living is a residential area in some way but this is not the most important issue in a part, where most of the traffic is generated by people coming for shopping, eating, hairdressing and the like. That is why I voted for “retail” and will continue to tag these places as such.


As far as I can tell, this is precisely why landuse=retail exists.

1 Like

That was my initial assumption too. The poll suggests that anyone attempting to use the data this way will get quite sub-optimal results in medium/high density areas though :sweat_smile: