Hi everyone, I was using zelonewolf’s amazing tool for finding errors with CDPs on map. But while using this tool, I found that that census areas are mapped with boundary=census which are used for CDPs (census designated place). Normally if this was a small way I would just delete it, as other states I’ve glanced at don’t have these mapped. But they take up huge areas of Alaska, so I want another opinions first.
Some relations in question.
It’s probably worth a mention to @pnorman, who originally imported them.
I’m inclined to agree with your assessment through – map the boroughs, incorporated communities, and CDPs, but not all of the other census statistical divisions.
Like Alaska’s boroughs, its census areas are designated as county equivalents. I don’t think we should tag the census areas as boundary=administrative, as we’ve done for Connecticut’s planning regions, because the Unorganized Borough should be represented by the absence of those. However, the census areas are pretty clearly more important to map than CDPs. I’d suggest either boundary=census or the more generic boundary=statistical plus border_type=census_area. That also happens to be what OpenHistoricalMap has done with, for example, the Chugach Census Area.
The import was as admin boundaries, not as census areas. I would not have imported them as census areas. I regret ever importing them because they have been used to justify census areas in OSM, ignoring that Alaska has some very unique situations that don’t occur elsewhere in the US.
I thought it was exactly the opposite, that our coverage of CDPs was used to justify our coverage of Alaska’s census areas, as above. Your import took place in 2012. By then, we had already reached a strong consensus that CDPs should be retagged from boundary=administrative admin_level=8 to boundary=census and forgotten about, if not deleted outright. After all, so many of the imported CDPs became outdated in the 2010 Census, and few at the time were interested in maintaining the data.
So as I’ve been adding townships to SD, I (along with a few others who have been doing it) have also added the census county divisions, especially out west where there are few if any townships. Should I go back in and get rid of those?
e.g. Relation: Central Pennington Unorganized Territory (15881376) | OpenStreetMap
These do show up in downstream data consumers, sometimes not where you would expect.
This guide to census geographies may be illuminating. The Census Bureau has multiple levels of fallback statistical areas when a given jurisdiction isn’t completely subdivided like its neighbors. The relationship between census geographies can be complex, but for the purpose of this discussion, it boils down to these summary levels, which correspond to admin_level=* values:
admin_level=* |
Census summary level | General administrative geography | Equivalent census geographies |
|---|---|---|---|
| 6 | County equivalent | County (borough in Alaska) | Census Area (in Alaska) |
| 7 | County subdivision | Minor Civil Division (in 29 states, e.g., township) | Census County Division (in 20 states), Census Subarea (in Alaska), Unorganized Territory (in 9 states) |
| 8 | Place | Incorporated Place (e.g., city) | Census Designated Place |
In principle, census geographies are defined solely for statistical purposes; their boundaries are more arbitrary than administrative geographies. However, some of the census geographies have taken on more meaning through happenstance or (mis)use. CDPs are usually named after real unincorporated places that lack boundaries, so people and datasets often conflate the two. I could buy that people routinely use Alaska’s census areas wherever they need a county equivalent instead of special-casing Alaska’s Unorganized Borough, just as they don’t want to leave a hole in Connecticut because that state refuses to be divided into counties. But most states’ minor civil divisions are somewhat obscure by comparison to counties and incorporated places, so UTs and CCDs are that much more obscure.
In other words, you did the right thing by tagging the Central Pennington Unorganized Territory as boundary=census rather than boundary=administrative. At least it’s consistent with Maine. But as to whether you should bother mapping out all the rest of South Dakota’s unorganized territories or delete the ones you’ve mapped, it comes down to whether the local community cares to maintain this information. It’s not all that difficult to keep up with changes to Alaska’s dozen census areas, whereas CCDs/UTs and CDPs are much more numerous. The combination of obscurity and number makes CCDs/UTs not a great tradeoff in my opinion. 
The consensus at the time was that census boundaries don’t belong in OSM. People have since used Alaska to argue otherwise.
First of all, let me preface by saying I’m pretty ambivalent on mapping census geographies in OSM. I tried to maintain CDP boundaries in Ohio for a time, reasoning that we might as well have accurate boundaries if we have them at all, but it was not my favorite work and I didn’t find it especially impactful. At least OpenHistoricalMap can claim to have more comprehensive coverage of Alaska’s census areas over time than any other online source (as far as I know) and could eventually achieve that for CDPs too. In general, OHM’s more research-oriented approach is a better fit for boundary mapping compared to OSM’s surveying ethos.
Yes and no. Among the U.S. community, there has been broad agreement since the late 2000s that a) census boundaries aren’t administrative boundaries and b) they aren’t a priority for the project. We did not reach agreement on keeping them all or removing them all. The compromise has been to selectively preserve CDP boundaries as boundary=census.
Some mappers felt very strongly that the CDP boundaries in their areas should be preserved in OSM, either because of local relevance or because they were confident in their ability to maintain and use their local subset of CDP coverage. Even the most outspoken opponents of CDP boundaries were willing to make some exceptions for places like Maryland or the Navajo Nation where these boundaries actually matter, as long as they got retagged.
Shunting the CDP boundaries over to boundary=census has made them more niche, gotten them out of the way of most mappers and users. At the same time, they’ve offered geocoders slightly more context about unincorporated places, for better or worse. I think it would be worth revisiting whether the claims about maintaining CDP coverage have proven out. Some of the folks involved with these early discussions are still around, but others have fallen away. Maybe our CDPs have decayed to the point where we should toss them out and require a consensus before reintroducing any. But I think this is unlikely to be the case everywhere, partly because of Bethesda, Maryland, and partly because of active efforts to refine these boundaries.
Amid all the place classification discussions over the years, I don’t recall ever seeing the Alaska census areas being used to justify the inclusion of CDPs. After all, the census areas are already inherently safely out of the way of most mappers. As you’ve pointed out on several occasions, Alaska is a bit of a unique situation overall.
