[Talk-us] Correct source for population=* tags on US metropolitan cities

[Minh Nguyen]

Vào lúc 14:47 2021-01-09, Joseph Eisenberg đã viết:
> So I propose that we should use an estimate of the urban population for 
> the population=* tag when tagging metropolitan places. Usually this will 
> lead to a larger population number, except in rare cases like Anchorage.
> 
> In particular, I would like to use the US Census "urbanized area" 
> figures, since these are much more sensible than the numbers from 
> metropolitan areas based on county boundaries which can include distant 
> towns and rural areas.
> 
> This would mean that the place=city node for Portland, Oregon would have 
> population=2072553 (representing the whole urbanized area) rather than 
> just 654000 from the city limits. 
> https://censusreporter.org/profiles/40000US71317-portland-or-wa-urbanized-area/ 
> <https://censusreporter.org/profiles/40000US71317-portland-or-wa-urbanized-area/>
> 
> Minneapolis, MN would have population=2885614 instead of only 429k 
> https://censusreporter.org/profiles/40000US57628-minneapolis-st-paul-mn-wi-urbanized-area/ 
> <https://censusreporter.org/profiles/40000US57628-minneapolis-st-paul-mn-wi-urbanized-area/> 
> 
> 
> But Anchorage would decrease slightly from 288k to 249K 
> https://censusreporter.org/profiles/40000US02305-anchorage-ak-urbanized-area/ 
> <https://censusreporter.org/profiles/40000US02305-anchorage-ak-urbanized-area/>
> 
> Usually the difference would not change the relative rank of cities very 
> much, but it would be good to have the population figure map the 
> OpenStreetMap "place" concept, rather than the city limit boundaries.

If I understand correctly, the Census definition of an urbanized area 
would fall somewhere between the central city and the whole 
metropolitan/micropolitan statistical area in terms of population. I'm 
sympathetic to this idea because the place node should ideally represent 
a human settlement, not entirely bound by administrative divisions. 
(This is why there can be a place=town inside a census-designated place.)

On the other hand, the waters are already muddy because place POIs are 
commonly tagged with wikidata=* and website=* tags that should probably 
be different from the ones on the administrative boundary. (Wikidata can 
but seldom distinguishes between administrative and territorial 
entities. A city might have one official website for government 
operations and another that functions as a business/tourism portal.) 
Also, it isn't uncommon for the place POI to have been moved to the 
location of City Hall, whereas it should ideally be at the origin of the 
street grid or town square or something to that effect. But these are 
all pedantic considerations compared to population, which affects the 
place visually even at low zoom levels.

I'm not sure we can totally eliminate awkward situations by migrating 
these central cities to urban area populations. For example, the 
second-largest urban area is Los Angeles-Long Beach-Santa Ana. Long 
Beach and Santa Ana would still be affected by population fragmentation 
within Orange County. For San Francisco-Oakland, San Francisco would get 
the overall urbanized area population, while Oakland would remain 
untouched despite plenty of population in surrounding suburbs, 
incorporated and unincorporated.

> Eventually this could improve maps of the USA and help them better match 
> those in other countries, where city limits tend to be much larger than 
> in the case of many US cities, which often have many separate 
> municipalities for suburbs.

The U.S. isn't alone in having administrative boundaries that divide a 
population center. Are there other examples outside the U.S. where 
statistical areas are used as a basis for place POI populations instead 
of administrative areas? If we depart from the more obvious definition 
we've been using, then global consistency would be important to maintain 
intuitiveness at lower zoom levels.

Even if we don't end up changing the population=* tags to the urbanized 
area population, that figure seems useful enough to put in a suffixed 
tag like population:urban=* (along with population:urban:date=*).

-- 
minh at nguyen.cincinnati.oh.us