On February 27, 2024 our team is going to start a mapping project in The USA. As part of on-going work to improve the quality of OpenStreetMap data, our team is planning to review a subset of the detections to better understand the type of issues, and also fix any valid data issues directly in the OSM. We’re concentrating on strip-malls buildings to improve map conditions. We have collected a set of malls that have this information partially filled in on OSM; however, we’ve found that OSM data is often incorrect / incomplete, and it requires a human to map the polygons.
We would really appreciate your feedback, any questions you have about this project, as well as local insights that you think will help us better understand the data.
There’s a link to our Github ticket to the related issue and a link to our page in OSM Wiki.
One issue that you don’t explicitly address is whether to map the strip mall all as one building, and the individual shops as points within the building; or to map it as a series of connected buildings. I believe this issue has been left up to individual mappers as a stylistic issue. (I think I read that on the wiki.) Personally, I like mapping it as one building, using building=retail, and then mapping the individual shops either as points near the front of the building, or as actual entrances on the front. Strip malls are normally built as one building with a bunch of equally-sized individual units with fire-rated walls between units. But when shops move in, sometimes they take two suites and remove the wall between. That’s why I prefer not mapping the units as individual buildings, because they can change.
Some examples I have edited:
way 1012138706
This is a landuse=retail polygon with the strip mall in it. Someone had already added the strip mall as individual buildings all connected, so I didn’t change that, but I corrected their shapes and sizes, accounting for the shops that have 2 or 1½ units. (Shift+B is very useful for putting a series of connected points all in a line equidistant from each other in JOSM.)
way 869890880
This is a strip mall mapped as just one building where I added the individual shops as entrances on the front of the building. I don’t think I like how it turned out. If anyone would ever go in and micro-map everything including the walkway that is covered under the front of the building, they’d have to change everything.
way 1179617255
This one is just one building, and I added the shops as points within the building, close to the front so that at least the point indicating the shop is nearby to the shop entrance. If the goal is to include all the shops, I think this is the best way to map it quickly, and if anyone later wants to micro-map the details, they can easily work with it.
Sounds like a great idea. Will you be doing any POI (amenity) mapping or just mapping the building footprints where non exist? How about parking lot roads and parking lots?
We are going mapping and updating strip malls as connected individual buildings. (As your example way 1012138706) This is the plan for strip malls which are not relations.
The Github ticket says that you are starting this effort today (2024-02-27), you yet you only posted this notice yesterday. You really should give the community more than a single day to review and comment.
Using only satellite imagery it may be difficult to tell where the demarcation between one unit and the next is. Mapping the mall as a single building (or a few buildings if there are parts that are separated spatially) may be a better idea.
You indicate that you are going to be updating POIs. It is not clear what the source for this will be as it will not be possible in most cases to make such determinations from satellite imagery.
“We have collected a set of malls that have this information partially filled” - It is not clear how this was done.
The majority of strip malls in the U.S. are intentionally mapped as a single building, so your edits would be effectively changing the standard in OSM. You need to provide more detail about why your team believes that a row of buildings is the better modeling. I’m glad this came up, because your GitHub ticket and wiki page never mentioned any opinionated mapping, just fixing “incorrect” or “incomplete” geometries. Are there any other tagging practices that you intend to introduce that are currently less common? I think the community here would be open-minded about it, but we don’t want to be surprised.
Personally, I map as many buildings as there are freestanding structures. Sometimes that means a row of buildings, but usually it’s just a single building. If I encounter a strip mall where someone has taken the time to outline the extent of each shop as a building, I convert those areas into shop=* areas, maybe with building:part=yes, and surround them with a single building area. Otherwise, I’ll map just the single building and add any POIs or address points within it.
I find that this approach works well with the two-story strip malls that are common in some areas. Usually the first floor is retail spaces while the second floor is professional offices, but it’s still considered a strip mall in most cases. It would make no sense to turn these buildings into a mess of overlapping buildings:
We are working with buildings which have only 1 floor. This is the example. This one is already mapped as individual buildings. We will just check parts on the web-site and fix, if needed. Also we will add a new building on the territory, which wasn’t mapped before.
That may be the case for the ones you’re looking at, but if we establish a standard for mapping strip malls, then it should be consistent even if there’s a second floor.
By the way, please let me know if I’m overstating how much of an impact this project would have. I’m assuming you’ll be mapping many strip malls, but if it’s just a few, then we probably don’t need to come to a decision about the mapping style right away.
For reference, here’s what that strip mall looks like from the street:
Since all the stores share walls, my personal preference would be to map this as one building area, with the shops subdividing it as areas, or alternatively as points within the building area.
It’s difficult to quantify exactly how many strip malls are mapped according to one style or another, but let’s use your example as a starting point. Most of the tenants within the main building of this strip mall belong to a chain; I assume that most of those chains’ locations are also in strip malls rather than standalone buildings. Of all the U.S. store locations of the chains that are represented here, only 26% are mapped as terraced buildings. The rest are mapped as either points or areas within a larger building. Can you elaborate on why your team prefers the terraced style?
Also, note that the buildings were incorrectly classified as building=commercial. That tag is for office buildings and the like; retail buildings should be building=retail.
I’m sorry. We cannot disclose the details of our company’s activities and the reasons for the urgency due to its confidentiality.
But we can say that we are not going to do mass edits, because there are only over 70 malls in our list and they have already been mapped this way. Most of them have actual information. Also we are not going to remap big polygons and relations.
Our scope of work includes comparing data on the OSM with reliable sources and making some minor edits.
Tell me, please, if we can continue our work.
In the past, Mapbox has typically been able to explain the reason in general terms. No one expects you to disclose the name of the customer who wants this data fixed, or the exact wording of the feedback you received from them. But you could discuss how your product behaves suboptimally because of how the strip malls are currently mapped.
In truth, we’re just a little skeptical that a few days would make a huge difference. A mass edit takes time to plan; the OSM community expects to be involved somewhat ahead of time, not at the last minute.
Oh, that doesn’t sound like a big deal at all! There are thousands upon thousands of strip malls; 70 would be a drop in the bucket. I was under the impression that this would be a mass edit, and maybe others were as well. In the future, you could definitely head off any misunderstanding by indicating the scale and scope of the edits.
I was also under the impression that you would “replace” existing buildings, as your ticket and wiki page stated, thereby replacing a single building with a terrace of smaller buildings. If that is not the case, then the discussion above about mapping style is a nonissue.
And that you are indeed splitting up buildings into individual units, even though that appears not to be the community consensus as to how strip malls should be mapped as pointed out by @Minh_Nguyen . Further, you cite Esri and Bing imagery as your source, but it is not clear in those sources where the dividing line between the units are.
More evidence that you have already started without allowing adequate time for the community to respond to your proposal:
In this case you again split a single building into parts, and again, there is no evidence in the cited sources as to where specifically the building is divided into parts, so you either have another source that you are not telling us about, or you are guessing.
For what it’s worth, in GB these (“Retail Parks”) are often mapped as separate buildings or large ones. I see no issue with either, maybe they should be building:part, but with the current quality of our buildings, it’s just nitpicking.
Here’s that changeset on OSMCha, which I think better shows what happened: OSMCha
That shows that some of those items were already modeled as individual buildings for many years. They did make two more splits, but fitting it in with the existing data may make sense in that case. I point this out because while the community in this thread doesn’t seem in favor of that modeling (I’m not either), I don’t think it’s an unreasonable edit.
I don’t have a strong opinion as to the best way to map these buildings, although I slightly prefer a single building as most people, including municipal authorities, would probably consider them a single building. However, the OP needs to make it clear in more detail what they intend to do here, and if they are splitting into separate buildings, what source are they using? Most satellite imagery does not make the division between the different units clear. Using street-level imagery might, but that wasn’t cited in the changeset source. Also, what else do they intend to change beyond geometry (e.g. what tags will be added or modified based on what sources)? Does this really only involve 70 features, or is that just the start? There was mention of “testing an algorithm”, that suggests that perhaps the 70 is just the start. Also, why is this so “urgent” that they jumped the gun and started editing already?
Good catch, I missed that one. Both the new “building” and the node within it have a shop=* tag. This is one reason why for these organized edits it is important to have a defined procedure as to what is to be done.
