While researching place= values for the sister thread on New England place classification, I checked the situation in Alaska for some worst-case counter examples.
Specifically, the city of Nome, Alaska, with a population of 9,835 in the latest census estimate from last year was, until a month ago, coded as a place=city. It is so remote that it isn’t even connected to the North American road network, though it is certainly the most significant settlement in the Seward peninsula and would normally be depicted in low-zoom maps of the state.
This edit re-classified Nome to place=town along with a handful of other place node changes:
While we’re discussing what to do on the high end of density in the New England thread, what should we do on the low end? Do we promote most-significant but very remote places in order to indicate their relative prominence?
I’m inclined to restore these place nodes to the long-standing status quo, but I offer it up for discussion.
IMO Nome qualifies as a city since it’s the most important place in the area.
I’ve looked at colorado and neighboring states (AZ, UT, NM) & I see several places that have small very populations, but are tagged as cities. I think most could be defended as OSM cities. Some of these have a large number of 2nd homes and tourists that aren’t counted in the census. Some are the most important town for many miles. Of these, Vernal seems the most dubious, but it is very remote.
Nome isn’t subordinate to any other population center, and it’s big by Alaska standards. Alaska itself is vast (moreso in Mercator projection), so we can even consider Nome’s importance relative to the region of Alaska it’s in.
North of Canada, such remote locations are generally classified as place=town. This Overpass query is a revision of the query used in [New England place classification](https://community.openstreetmap.org/t/new-england-place-name-inflation. It shows both node and relations that have the tag place=*. The option center let’s show only the center of relations. As some of you might have problem running the query with such a large area, the image below shows the result of the query. Note that in Alaska, we observe duplicated place=* both on node and relation.
Hospitals can be a good proxy for determining regional centers in sparsely-populated places disconnected from the road network, such as Alaska’s bush country. Hospitals have large catchment areas, and need a large workforce (and therefore a bigger population) to support them. By looking at where they’re mapped on OSM, I would suggest that Utqiaġvik, Kotzebue, Nome, Bethel and Dillingham be classified as cities.
I’ve never been to Alaska, but I’m open to the idea that some states and territories might have either no place=city or only one or two. Anchorage certainly makes the cut for place=city as currently described in the wiki, but maybe that’s the only one.
I could imagine that Alaska and other sparsely populated states (Wyoming & Vermont come to mind) might only have one city. The state capital of both Alaska & Vermont may truly be place=town due to the town-scale amenities (such as hospitals, schools, universities, amount of retail & industry, etc) without the large size that might resemble a city.
Our wiki describes Anchorage as a consolidated city/borough defined by Alaska law. We have suggestions for how those are tagged, not all (CCCs, consolidated city-counties) are tagged as we suggest they might be.
That is a concern as we do this: people don’t necessarily quite pay attention to the sort of tagging we suggest in our wiki, even as multiple wikis on these topics attempt to clarify these taggings as both descriptive and prescriptive.
Alaska is an odd duck when it comes to how its government, places and subordination between them happen in OSM, with an especially confusing (to the uninitiated) situation with the Census Bureau and the Unorganized Borough. We (Minh, I, many others over at least a decade…) have striven, sometimes struggled to “get this right.” We have made a lot of progress, though it isn’t (quite) perfect and it continues to evolve. Let’s pay attention to the road we’ve already paved.
If Alaska incorporates Nome (and maybe also makes it a CCC, or a CCB, as only Alaska does), I’m OK with place=city. If, because Nome has hospital, university, other “cities have these” kinds of amenities, that might be another reason to agree place=city is an appropriate tag.
I’m certainly no Alaska expert, and I’ve never even visited, but I am an amateur Googler – and Nome’s wikipedia page says “Nome was incorporated on April 9, 1901”. In addition to the other bits about having a hospital and university and other bits. So you would agree that place=city is a good fit despite its 4-figure population total?
I think this might be the wrong way round: it has these things because it is remote, not because it is a city. For instance somewhere much less remote, the town of Oban in Scotland has both a hospital and a university campus. Usually, even in Scotland, these smaller hospitals rely on medevac for many cases.
There are even two places in the Engadin with decent hospitals (Samedan & Scuol), both mapped as place=village (although, even given their small populations, I would have gone place=town). These serve not just the local population, but tourists, particularly winter sports participants.
Most ‘county hospitals’ in Ireland are located in place=town too, although Google has been known to be overenthusiastic about this in the past.
I think an approach using weightings is better than upgrading to city outside the regular parameters, especially as it would make Nome rather more out-of-kilter with tagging elsewhere.
I would. I would hope it would be on a polygon rather than a node, as I consider place=city tagging on a node correct only as a shortcut until the polygon data make their way into our map database.
BTW, in Utah, a “town” is <1000 people, a “city” is ≥1000 people (barely “four figures”). I consider this an extreme case, as a place with 1100 people as a city might cause a frown to be pulled on my face, but if Utah says so (and OSM agrees?) then “let’s call a city a city” even if it is an outlier of our definitions.
I’m not suggesting we do this, but another tag like major_town would be useful for cases like this.
The existence of a hospital is not enough to elevate to a city. Many town in rural communities have what is known as a ‘critical access hospital’. An outsider wouldn’t know the difference, they have most of the services expected, but for anything bad they fly you out. We have a few small towns that have appropriately been tagged as cities because of regional importance (nothing bigger for quite a ways), but we have quite a few other towns with hospitals that are still osm towns.
Yes: recently I was in agreement here, using a rough “toss out there” of (what I’ll now refine) as large_city as maybe above 3, 4, or 5 million, and perhaps very_large_city or mega_city as maybe above 15 or 20 million.
Of course, this would take additional consensus by others around the world (not only in our United States sub-community here), and each country/region might have a different threshold of the exact population number that they choose. But it seems clear that because of how differing zoom levels “must” render (sanely) conurbations, and because Asia (especially East Asia and South Asia) have truly large concentrations of urban centers (many, perhaps most, of the world’s megacities), both these additional tags are warranted, and they may have quite different application of how they are tagged, contrasted with the essential opposite of what’s going on in Alaska. (Vast areas with relatively small populations, but which “accede to” the “importance” of cities, even though in OSM-speak and because of how renderers choose to show these at various zoom levels, it might be more technically correct to call some of these towns or in rare cases, perhaps even villages).
As for major_town, that’s interesting. Let’s think about / talk about that.
Maybe you guys want to repurpose the concept of topographic isolation in this setting? This is one of the measures often used to create “importance rankings” for mountains. By looking at population instead of elevation, the same could be done for places.
This probably shouldn’t be the only deciding factor, but it might add some value. It captures that a remote place with 9000 inhabitants might be perceived as more important than a similar size place that is just one of many in the region.
I think any plain-English tag that’s directly computed from population isn’t very useful? After all, any data consumer can (and do) just use the population tagging to make whatever distinctions between these categories they want to.
There is (also) how Americana, or any renderer for that matter, chooses to display place names. Decisions are made, as they must be. Some of these might presently be compromises which could benefit by “more accurate tagging” (as agreed to by us).
Whether we explicitly say so or not (I have a couple times here) rendering IS what part of this is about, elephant in the room it may be. A very_large_city might be one thing in one region, same for how megacity might mean slightly different things in say, Europe vs. Asia. With the existing tags for smaller entities, there are similarly highly elastic semantics happening (as there would be for these); that is undeniable. In fact, that is largely what we discuss.
That said, I continue to press forward towards tagging “what is.” As succinctly and accurately as possible. If we need to coin new things or stretch elastic semantics (even further), we are allowed to do that.
I’m not describing a rendering distinction, though certainly that’s one kind of data consumer. Americana uses OpenMapTiles, which uses both population= and wikidata to figure out what to display (in addition to the place=* tag.
Specifically, I’m pointing out an information issue here – if we’re saying a tag can be directly computed from population=, then that tag isn’t adding any new information that wasn’t already present.
Often or frequently, there is no population= tag on a place=*, perhaps that’s where my mental hangup is. It seems certain that some renderings for place names (simplistically) happen using only the place=* tag. (So, good for Americana for looking at population=* and wikidata, that seems correct and smarter).
I do think we are talking about a version of “how isolated is the (demography)?” as @harahu suggests.
Edit: Perhaps one of the smarter things Alaska place=* nodes, ways and relations might be tagged is with a decently-updated population=* tag. If that were so, would a “reclassification” be necessary?