Appetite for a place=apartments tag?

General talk Tagging general discussion
1

Here in Alabama, there are many “apartment complexes” where several buildings are built/owned by the same company, and have their own amenities such as parking, pool, office, etc. They are named, often with some sort of sign at the entrance to the complex.

Presently, these seem to be represented best by landuse=residential + residential=apartments. I dislike this scheme because I believe that landuse should generally be avoided for representing points of interest. It also starts to show problems for example if the landuse of the apartment complex is not uniformly residential, such as if a park is included in the complex for residents. In my opinion, apartments and apartment complexes are better represented under place=*, especially considering place already includes units such as named city blocks, which are conceptually very similar to an apartment complex.

Therefore, I want to see what the community thinks of a potential proposal of a new tag, place=apartments to represent named apartments and apartment complexes.

2

on the wiki page for landuse=residential it has this line:

In general, leave the area unnamed, since it only represents the statistical likelihood of finding residences there. If the area is known by a name, map a separate place=* object to represent the concept of place in the same area as the residences:

1 Like
3

The problem you’re identifying is that landuse=* has two documented purposes. You quoted one; the other is about individually identifiable complexes. You can generally distinguish the two uses by the presence of name=*.

The place=* tagging scheme is already subject to so much debate over its meaning and form that I’d be hesitant to throw more meanings at it. It’s everything from populated places (with no agreed-upon, verifiable shape) to countries (with well-defined borders) to public squares to individual plots to natural islands. Out of the pot and into the fire, as they say.

To really benefit from this change, we’d have to develop new place=* values to replace all the other land use areas that have verifiable shapes, including residential subdivisions, mobile home parks, commercial office parks, industrial parks, shopping centers, school campuses, religious campuses, and more. Unlike the place=* values that refer to more organic parts of a city, these are inherently planned areas, already formally associated with a particular land use, so they would usually be dual-tagged with landuse=* anyways.

This isn’t necessarily a problem. The park would just nest within the landuse area. For example, most renderers would have little problem showing this playground inside an apartment complex.

Occasionally an apartment complex does correspond to one of a city’s neighborhoods, such as this public housing project, but this is an edge case in my experience. It depends on a number of factors, such as the size of the complex and the city. Some apartment complexes are just two buildings and a parking lot. If this would count as a place, other things like hotels would need to be places as well.

A bigger problem is that building=apartments itself is sometimes considered a point of interest. This conflicts with the general understanding that building=* is a structure classified based on its typical appearance. building:use=apartments can mitigate that problem when something like an office building is converted to apartments. But we don’t have a good answer for when a building stacks apartments above hotel, office, and retail floors. We had the same problem with dormitories, causing mappers to shift to a few competing alternatives.

5 Likes
4

This doesn’t work for all. There are estates with both apartments and houses. What are they considered to be? The other case is estates having houses only. Gate communities are often discussed in the past too.
place=apartments won’t solve that.
place=city_block itself can be debated. It might be a boundary= instead, whether =administrative or another. The documented definition is not the only possible one, as blocks can often be defined by street center lines.
Both don’t fit into the populated place= hierarchy. There can be large or small estates and blocks. They could be within another place= (or equivalent to a place= , even comprising multiple place= including =city_block for the status quo , depending on perspectives).
There’s another need for named phases/parts of estates, which can be used for wayfinding or organizing. Addressing is more complicated, so I won’t discuss it directly.
Then there’s what’s mentioned by others, apartments can be found on the uppermost floors of a building= (not the usual case where all non-lower floors are apartments). That’s not quite a place= for most definitions.
But independent of this solution, I do have another idea for the concept from a random key I found. It tries to follow how there’s shop= and office= for businesses. Proposal:Residential feature - OpenStreetMap Wiki

5

or place=neighbourhood? This is sometimes used for quite big parts of a city, or also small ones. In the current hierarchy it is defined as the smallest parts of a city, below quarter and suburb

6

Discussed here:

Despite pollution of place= namespace described by Minh, I agree that the “recommended” practice of naming landuse=residential is abysmal. Those names should be associated with some kind of place.

From the currently available choices, I would probably pick up a place=neighbourhood, perhaps place=city_block for particularly large ones. If we’re going to propose a new tag, something like place=estate proposed in the second thread would probably be sufficiently generic and understandable (although two years later, nobody has came up to formalize it and put it to vote).

2 Likes
7

I think your proposal offers a valuable new perspective, especially given the increasing diversity of modern housing.

At the same time, I think this makes it even more important to examine the issue in greater depth.

Modern apartment developments have evolved into a wide variety of forms with very different characteristics. In particular, I would encourage you to consider the development of apartment complexes in South Korea and Singapore as part of your analysis.

To provide a little background, I understand that in some Western countries apartments are sometimes regarded as an unsuccessful model of modern housing or are associated primarily with lower- or middle-income residents. However, in other parts of the world they have evolved in quite different directions.

Of course, I believe that Europe—the birthplace of apartment housing in its modern form—also has apartment complexes with a wide range of characteristics. However, I think the cases of South Korea and Singapore are particularly distinctive.

Singapore’s HDB flats are already fairly well known, so they should be relatively easy to research. In South Korea, meanwhile, apartment complexes may include facilities for residents such as restaurants, swimming pools, saunas, fitness centers, parks, cafés, and even guest accommodations for visitors. Some complexes also contain publicly funded childcare centers. Because these are public facilities supported by government funding, they are open to people who are not residents of the complex.

As is common elsewhere, apartment complexes also often contain commercial facilities. In practice, some are large and self-contained enough that they could almost be regarded as small neighborhoods in their own right.

In some cases, although not located within the apartment complex itself, part of the development site is dedicated to the government for public use, allowing public institutions to be built there. Strictly speaking, these facilities are outside the apartment complex, but from a broader perspective this further reinforces the idea that such developments can function much like small communities—or even, in some respects, resemble a small municipality.

On the other hand, as another contributor has pointed out, there are also apartment complexes that consist of only two residential buildings and a surrounding fence.

I very much agree with the goal of making the nature of apartment complexes clearer and more accurately reflected in OSM. However, once all of these different cases are taken into account, I suspect that defining their characteristics precisely becomes a very challenging task.

(For this reason, I personally lean toward simplifying the criteria used for tagging attributes. I am currently researching this topic and hope to share my thoughts on it in a separate discussion at another time.)

2 Likes
8

Can someone articulate the actual problem? The original post pointed out that some amenities like parks and schools can occur inside of a land use area, but this is true of any land use area, even the handwavy kind. I don’t think we should be cutting these amenities out of land use multipolygons just because OSM Carto fills them in with a solid color.

A variety of renderers support named land use areas as is. For example, OSM Carto not only fills them in but also labels them prominently based on size. Mapbox-powered maps label them as well. Perhaps more importantly, Nominatim does not stuff these named land use areas into its already tenuous admin_level=*-based hierarchy of places.

In some environments, a dense concentration of named land use areas can provide structure in a manner reminiscent of the place=suburb/quarter/neighbourhood hierarchy, but this isn’t an inherent function of the things that named land use areas represent. Even when this is the case, it’s entirely artificial, inconsistent, and non-hierarchical. Ultimately, an apartment complex, industrial park, or retail center is a point of interest inherently associated with a particular land use. Is it weird to attach operator=* owner=* website=* to a land use area? Not as weird as attaching these tags to a place=neighbourhood point.

By the way, there are entire corporate chains of apartment complexes. brand=* on a place=* point? :grimacing:

This seems to assume a particular development pattern that you might find in, say, an older European city, but not a suburban area or redeveloped urban area. Maybe it would help to give some counterexamples from one of the regions where I map.

City West Apartments has the same area and even more housing units than the Villages at Roll Hill, but only the latter is considered a neighborhood either officially or colloquially. Neither has anything to do with a city block. City West is a redevelopment of two historic apartment complexes, Laurel Homes and Lincoln Court, that is now marketed as a single complex. Just to the north is Liberty Street Apartments, which is two buildings and a playground. No one considers that a neighborhood either.

Prospect Point & The Lofts is located in a village[1] that’s too small to be broken down into quarters and neighborhoods. This apartment complex takes up about a tenth of the tiny unincorporated village of Minford. Valley View is located in amorphous urban sprawl that isn’t officially or colloquially part of any named populated place. North Shore Village is located in a rural area. Can a neighbourhood be independent of any place=city/town/village/hamlet?

Whatever we tag an apartment complex as, it must be independent of any hierarchy whatsoever. An apartment complex is an apartment complex regardless of its size in relation to its surroundings. The same is true of residential subdivisions, commercial office parks, industrial parks, and so on. If the feature also functions as part of a city’s place hierarchy, then by all means dual-tag it with place=neighbourhood.

  1. Officially a city, but in this state, everything is a city, even if the population is zero. ↩︎

5 Likes
9

Agreed. The problem with stuffing things into place is that the hierarchy stops making sense and you can’t style the label (should apartment be bigger or smaller than locality etc.?)

You could imagine styling landuse=residential differently depending on a secondary tag, e.g. residential=trailer_park.

10

Yes, Korea as mentioned above is prevalent with branded apartment complexes. This appears in other places too.
And indeed, depending on perspectives and definitions, estates or developments may be a place= themselves (besides sitting between different levels of place= ), or have different implication on addressing. This necessitates treating them orthogonality.

1 Like
11

I think my main gripe with landuse=residential is that it is a valid tag for something like an area surrounding a singular house and yard, while also being considered valid in a case like this, where it represents a distinct and named POI such as an apartment complex. I have the same issue with landuse=industrial, but at least in that case there is the tag man_made=works to represent a distinct POI. To me, this is the same case as your point here

Even in the buildings case I think it is more acceptable than with landuse, as buildings can be named features, and even when unnamed are discrete features with well defined boundaries

I think as a whole that we should avoid if possible having a tagging scheme that mixes features that are points of interest with features that are entirely nebulous “paint” that simply represent the concept of a residential area.

Consider the example of querying: perhaps well mapped apartment complexes can be represented by the landuse area, with name and other metadata attached. But in the case of an object tagged with landuse=residential + residential=apartments the area might represent a distinct POI apartment complex, or an area encompassing several apartment buildings that do not have a relationship to each other other than proximity, or several polygons representing a singular apartment complex, or countless other possible cases? therefore, how could you query for apartment complexes without many false positives?

Perhaps place=* is just as inappropriate as pointed out in this thread due to the challenges it would present with hierarchies, but I still feel that another way to represent this kind of feature should be considered.

2 Likes
12

Maybe you can tell from my previous posts that I consider the nebulous paint to be somewhat dubious anyways. :slightly_smiling_face:

I think we can address the ambiguity by clarifying that landuse=residential residential=apartments represents an individual apartment complex as far as we can tell. That’s virtually always how it’s used in the database. There isn’t any point to deliberately merging adjacent apartment complexes into one area. We can also pretty much assume that it’s a single complex if it has a name=*; otherwise, the name is suspect, probably a description.

We covered the issue of “campuses” pretty exhaustively last year in the context of landuse=education: a single campus that may be home to multiple educational institutions, not a collection of schoolgrounds that are only adjacent by happenstance. Similarly, a named landuse=industrial is most likely an industrial park, not a factory; a named landuse=commercial is most likely a commercial office park, not an office; and a named landuse=retail is a shopping center, not a shop.

Ultimately, what is an apartment complex, other than a specific area of land set aside for apartment buildings that shares a name in common? If this doesn’t very strongly point to landuse=residential residential=apartments name=*, then we’re left with something less meaningful like amenity=apartments.

My point about building=apartments is that we have no good answer for when the apartments are only a part of the building (and not in the sense of a building:part=* for 3D mapping). Some are even mapped as land use areas for lack of a better tag. The Lofts at Shillito Place is not an apartment complex.

2 Likes
13

landuse=industrial is an exception, as almost all industrial facilities other than man_made=works lacks a feature tag. What I would say is landuse= can be allowed for use on anything including industrial parks and a named but unrelated-besides-adjacency industrial. Therefore another PoI feature should be used.
What’s the term for the apartment section of a tower though? If there’s no common word, it’s difficult to be described. When I think about it, “commercial complex” doesn’t seem to require multiple buildings, yet apartment complex does.
Another case is wide buildings, structurally integrated, but divided into different sections by lobbies, staircases, or elevators. It can be 1 building= while qualifying as an “apartment complex”?

14

Yeah, we can get close with terms like floor(s) and wing(s), but I think laypeople just treat each one as one of a kind – known only by name, not by category.

The more general concept would be a building complex, but maybe that’s too underspecified for any mapmaker to find useful. There are apartment complexes just as there are (commercial) office complexes. Terms like complex, “park”, and campus can imply various building densities depending on context.

Is “named but unrelated besides adjacency” a valid case that we should accommodate? It sure sounds like a descriptive name or other invented name. That’s why I’m not particularly concerned about the false positives when querying for apartment complexes by residential=apartments – any false positive is probably mistagged for other reasons anyways.

15

Another argument I remembered against landuse=residential (reminisce of landuse=railway vs area:railway=station ): There are estates built above landuse=railway and landuse=highway (up to being entirely elevated, with nothing on ground)

16

If something is developed on top of a freeway cap, so that it includes the land around the buildings, then I see no problem with mapping an overlapping land use area with a higher layer=* and tagging covered=yes on any capped roadways. Normally parks and public buildings are built on top of freeway caps, but for example, The Banks is built on a cap over a large public parking lot (which serves as a flood control measure).

If only the buildings are elevated and nothing below would be considered part of the estate, I wouldn’t necessarily map it as a land use area. It would be a building complex, but at least in American English, it wouldn’t be a typical apartment complex even if all the buildings are apartment buildings. For example, the Bridge Apartments are usually described as a set of four towers, so I’ve mapped it as a site relation.

I still get the impression that the current tagging scheme is holding up in the general case, in the sense that mappers and data consumers find it intuitive enough. Over 70% of all landuse=residential residential=apartments have name=*, compared to just 5% of landuse=*, suggesting that this tag generally represents apartment complexes that we’ve examined closely enough to determine a name.

Mappers don’t appear to be painting landuse=residential residential=apartments areas onto the map arbitrarily. The areas cover an average of 3.14 hectares (7.75 ac.) with an average 66% Polsby–Popper compactness. By comparison, unnamed landuse=* only averages 54% compactness. place=farm, which obviously represents a single farm, only averages 63% compactness.

Specifically in the U.S., 65% of landuse=residential residential=apartments have name=* compared to 10% of landuse=*. They average 9 acres (3.65 ha) with 64% compactness. This is despite the massive import of individual parcels as land use areas in California many years ago.

More than 99.7% of residential=apartments were tagged after iD added the Apartment Complex preset in 2018. A great many more are still tagged just landuse=residential without residential=apartments. This includes 5,876 that contain “Apartments” in the name – sounds like a fun MapRoulette challenge.

4 Likes
17

and why this is a problem?