[Talk-us] GNIS POI populations

Minh Nguyen minh at nguyen.cincinnati.oh.us
Wed Jan 28 08:12:13 UTC 2015

On 2015-01-27 22:42, Greg Morgan wrote:
> OSM Inspector[1] has a nice tool to check issues with these
> city/town/village/hamlet  POIs.  I updated a bunch of the POIs in
> Arizona to the 2010 numbers.  I see that some mappers changed the
> values to the estimated value.  Another mapper would change it back to
> the 2010 actual numbers.  I have don't have an issue with the
> mechanical edits unless the edits would remove gnis id tags or other
> useful data.  Just as with the manual edits both the estimated value
> and actual 2010 values were close enough.  They correctly raised the
> value closer to the 2010 number from the 2000 number.

The original GNIS import was also wildly incorrect in some cases. For 
example, Mount Pleasant, Delaware County, Indiana [1], was given a 
population of 12,459, but it's really an unincorporated community with a 
couple hundred residents at most. The Census doesn't keep any data on 
its population.

Should the mechanical edit remove population tags on places that don't 
correspond to any Census division? Or maybe just flag them for manual 
followup? I guess a local could unofficially guesstimate the population 
of a community like Mount Pleasant.

[1] https://osm.org/changeset/28225500

> Deleting?  I question this.  I am not in favor of it.  I think there
> is a mismatch between rural America and Metro America areas.  I have a
> sense that Metro mappers have a lower value of some POIs that are
> essential to rural areas. Vicksburg Junction[2] could be a possible
> deletion target.  I am not sure if there is an actual boundary for the
> area. Cleator Arizona[3] is another example.  People live there with
> real addresses even though it looks like a ghost town.  The best I
> could do is make a residential landuse area. There are any number of
> small named areas from the Census that are significant names that the
> locals use. How do you know that you are not deleting valuable named
> data?  Moreover, you can query on "Moon Valley Arizona" and find a
> well known area in metro Phoenix.  Sure someday that POI can be made
> into an area.  I have wondered what kind of a polygon would be the
> correct one for this area. There's no real legal boundary for the
> area.  I have already had to dig that POI out of the trash bin once.

It doesn't sound like Paul was proposing to systematically eliminate 
place=hamlet POIs. It sounds like he was evaluating each one on its merits.

I do delete GNIS POIs fairly regularly, but not just because they're 
tagged place=hamlet. It's usually because it's a mobile home park that I 
can turn into a more accurate landuse=residential. Or it's the name of a 
railroad junction torn out a century ago that now sits in the middle of 
wilderness. (There is place=isolated_dwelling in the event that a small 
cluster of houses is still called by that name.)

On the other hand, I have been aggressive about preventing TIGER CDP 
boundaries from rendering, but they're a whole different animal than 
GNIS POIs. They usually end up being so precise as to be inaccurate. 
When I delete a CDP boundary, the map usually continues to show the 
local name thanks to a well-placed GNIS POI.

> Finally, why would you want to dash the hopes of a new mapper[1]?  I
> shared the excite with a mapper as he talked about his recent project.
> He had just put in the Phoenix Urban Planning Villages or whatever
> they are called.  Now you can look for "alhambra arizona" and find one
> of these areas as a POI.  I am afraid that his victory would fall prey
> to your deletions.  If you don't know the area or are not sure, then
> just leave it alone.

Phoenix's distinction between Urban Planning Villages and council 
districts reminds me of how other cities distinguish between 
neighborhoods and electoral wards. I favor mapping the former as 
boundary=administrative and admin_level=10, or as a place=suburb POI if 
the boundary is unknown. But I'd leave the latter out of the picture, 
just as I'd avoid mapping state legislative districts. Alhambra [2] 
looks good to me.

[2] https://osm.org/node/150948276

minh at nguyen.cincinnati.oh.us

More information about the Talk-us mailing list