[Talk-us] TIGER 2022 PLACE dataset

[Minh Nguyen]

Vào lúc 11:53 2023-01-19, Zeke Farwell đã viết:
> I've been working on place=town, place=village, and place=hamlet 
> classifications in Vermont and thinking about how CDPs may be helpful to 
> this effort.  The town of Milton has a place node 
> <https://www.openstreetmap.org/node/622776905>, a boundary=census 
> <https://www.openstreetmap.org/relation/199120> (CDP), and a 
> boundary=administrative 
> <https://www.openstreetmap.org/relation/8897021>.  The administrative 
> boundary represents the area that the Town of Milton government has 
> authority over.  The place node represents the built up area that is 
> considered a town.  Here I use "town" in the small t sense of a built up 
> area rather than the large T sense of an official government.  The CDP 
> area represents the same thing as the place node, but as an area instead 
> of a node.  While I don't see a huge problem with mapping a 
> place=town|village|hamlet as an area (likely based on a CDP area), I 
> believe the standard layer won't display a label.  Other than that, 
> having both a node tagged place=town and an area tagged boundary=census 
> seems like duplication to me.  They even both link to the same wikidata 
> item Q1789757.  For a map renderer to effectively use the CDP area, 
> they'd need to de-duplicate so they don't get two Milton labels.  This 
> could probably be done, but I'm not sure if any data consumers are 
> prepared to actually do this.

Some data consumers, such as Nominatim and possibly some renderers, do 
deduplicate the place node if it's a member of the boundary relation 
(preferably with the "label" role).

I don't know of any renderer that depicts boundary=census ways or 
relations. That's sort of the point: it was an old concession to those 
who wanted to keep these boundaries for marginal purposes, with the 
understanding that they wouldn't be reliable as boundary=administrative 
relations going forward. The highest priority was to get these 
non-boundaries off the rendered map. Some mappers went further by 
deleting the CDP boundaries outright. This was probably a good idea at 
the time, because the CDP definition changed dramatically in 2010. It 
could happen again.

In some rare cases, a CDP may be verifiable/observable by subtraction. 
An oft-cited example is Bethesda, Maryland, which is bounded on several 
sides by incorporated administrative areas. However, the majority of 
CDPs are more freeform.

If a renderer wishes to depict the "built-up area" of an unincorporated 
place, for example by shading them in, the CDP boundaries aren't a great 
source. The geometries don't strictly follow populated areas, often 
hugging features like roads and railroads instead. They also don't solve 
the issue that incorporated areas can be quite arbitrary. For example, 
all of Orleans County, Louisiana, is incorporated as New Orleans, even 
though half of it is uninhabited swampland.

The Census Bureau's Urban Areas [1] would have more rigorous, 
higher-resolution geometries than CDPs, albeit with less recognizable 
names. These geometries, which are based on census blocks, would be far 
too intricate to add to OSM as boundaries.

[1] 
<https://www.census.gov/programs-surveys/geography/guidance/geo-areas/urban-rural.html>

-- 
minh at nguyen.cincinnati.oh.us