Wikipedia talk:Reliability of GNIS data

Some resources
Looking around for some detail I came across a couple of good resources for explaining some of GNIS's limitations. For starters, there this presentation by Roger L. Payne of the U.S. Board on Geographic Names. If you click through to the PDF there are some important points on the slides, most particularly that GNIS's intent is "Standardization not Regulation"— this point is repeated over several slides. Its only authority is over the names, the intent being to ensure that anyone in the government (and hopefully in any other official capacity) uses the same name for a particular feature.

Another good source is this metadata description of a subset of GNIS data for Montana places. It talks at some length about the GNIS compilation and problems that state review found. Hawaii has a similar page with a much more detailed description of the data set as a whole, with especial attention to the sources used in its compilation. In line with this the Montana page talks about some of the problems with that compilation: they specifically note issues with reading the topos, for example.

I also note on the GNIS home page that, if I'm reading it correctly, they haven't made any updates to the manmade features since 2014.

In looking around at this I have come to the conclusion that our article on GNIS is pretty bad. It needs to be revised in concert with this effort. Mangoe (talk) 04:45, 2 May 2020 (UTC)

I found this writeup on someone's personal website about a phantom place that popped up on a map of their neighborhood, and Atlas Obscura goes into more detail about the impact of names on maps. These aren't RS but they're still informative. –dlthewave ☎ 20:19, 2 May 2020 (UTC)
 * OMG I love the Haberman story! Really telling the story there, great find! I really want to email Jeff about this saga, glad we're not the only ones catching these. Reywas92Talk 05:46, 5 May 2020 (UTC)
 * I also enjoyed the USGS FAQ (Question #7) where they basically threw up their hands and admitted the futility of trying to apply any sort of definition. –dlthewave ☎ 18:05, 9 May 2020 (UTC)

Redirect
I've moved the page to Wiki namespace and added WP:GNIS as a redirect/shortcut. –dlthewave ☎ 18:25, 9 May 2020 (UTC)

Things to do in Jolly Dump
The facebook page link seems to lead to a Facebook help page. Is that intentional? Aditya (✉ • ⚒) 17:06, 28 October 2020 (UTC)

GNIS sourced from volunteers?
GNIS seems to be taking submissions from volunteers, doesn't this make it potentially user-generated content and hence unreliable? What actually assurance is there that the volunteer information is properly vetted and edited? FOARP (talk) 21:50, 8 December 2020 (UTC)
 * This is for the USGS's National Map and is not connected to the GNIS. Although it initially imported data from the GNIS, what users can update in it is stuff like post offices, schools, and other buildings, which are not subsequently added to the GNIS, and they can't edit info on populated places. Reywas92Talk 23:19, 8 December 2020 (UTC)
 * Yeah, they stopped updating GNIS some years back, but when you look at the GNIS methodology, it was, well, sort of in-house crowd-sourced. Mangoe (talk) 23:30, 8 December 2020 (UTC)
 * I think if we can point to some systematic problem with their data-gathering then it would be easier to deprecate GNIS as a source at least for certain kinds of information (e.g., did they actually confirm in any way the nature of the listings?). From what I gather GNIS is reliable for the altitude/location/name data but not the data saying what the location actually was. EDIT: according to this paper - "Additions and corrections to the feature data in the GNIS are accepted for consideration from any source, and upon validation, are entered into the database", so possibly they were just accepting data from anyone? I guess it hangs on how detailed the validation actually was. Also, assuming I've read this right, this source appears to state that the "feature type" attribute in GNIS data was found to be at least 5% wrong. FOARP (talk) 11:03, 9 December 2020 (UTC)

GNIS links not working
Hmmm any link from an article to GNIS returns a blank page with "The specified URL cannot be found." Even the links on this page redirect to the BGN home page. Wonder what's happening here. Reywas92Talk 03:31, 11 December 2020 (UTC)
 * I checked a few today and they've all been working for me. Temporary glitch? –dlthewave ☎ 03:00, 13 December 2020 (UTC)
 * Still not working in Chrome, but works in Firefox, ugh..... Reywas92Talk 05:49, 14 December 2020 (UTC)

GNIS is down for rework
Clicking a GNIS link gives the following: User Notice – August 25, 2021

The current GNIS forms will go offline August 30 at 5 pm MT. Users may access GNIS data by downloading one of the text files that will be made available prior to shutdown. GNIS will be down for a period of time as we transition to our new forms. We apologize for any inconvenience this may cause our users.

The GNIS application will be getting a much-needed update. Users will still be able to search and retrieve records and view the location on a map. Users will be able to select the best search type for their needs and download the results. The summary page for each record will include the location displayed on a map along with each point. Users will have the ability to select several options for map backgrounds to assist with location. We will provide a date for the implementation of the new GNIS as soon as we know it.

Presumably the template will have to change but no telling how long it'll be down. Mangoe (talk) 00:47, 1 September 2021 (UTC)

GEONet Names Server
This service is operated by the same people who operate GNIS and is, by all appearances, even worse as a source than GNIS. Should we mention it here? FOARP (talk) 14:39, 4 November 2021 (UTC)

A new low
Bucksnort, Tennessee is probably undeletable, but of the statements in the unsourced article, the only one I can be sure of is "in Hickman County, Tennessee". All the others are either unverifiable or are definitely wrong, starting with the GNIS location, probably. About the only thing going for it in terms of official recognition is that the I-40 sign for the exit is labelled "Bucksnort". Mangoe (talk) 17:19, 12 February 2022 (UTC)

Gannett

 * and anyone else interested.

I've hit the third AFD discussion where what Gannett's 1898 Gazetteer, published by the USGS, says contradicts everyone else. It looks like in the 19th century the USGS collected data from maps and stuff and made errors. This sounds like a tale that I might have heard before. Suggestions for where we should start collecting data on Gannett's reliability? It looks like checking against state sources, or things like Polk's or Lippincott's, for confirmation is a useful caveat. Gannett covers more than just Kansas, so there will be issues for other states, no doubt. Uncle G (talk) 10:37, 30 January 2024 (UTC)
 * &mdash; Gannett is the only one to claim this to be a village.
 * &mdash; Gannett is the only one to claim this to be a hamlet.
 * &mdash; Gannett gets the river wrong.
 * There's certainly a pattern all around where the tendency is to identify a named dot on a map as a settlement if there's nothing contrary to that. I've tended to use the placenames books only for negative confirmation, if they said the place was something specific that's not a town. Also, the specific phrase "post village" seems to be widely used to mean "a place that's just a name for a post office"; we've seen that in several of the place names books.


 * I'm seeing distinct patterns showing up in different states. It was either in Oregon or Washington that the article makers/maintainers had done a very good job of filling the information out, particularly with census data, so I've found much less that was questionable there. In the upper midwest the planted towns seem to have taken better, and you can see a characteristic grid of a few streets almost everywhere. Further west there is a long string of Milwaukee Road stations which never took as towns and which all completely evaporated when it was abandoned. There's one state where someone managed to get most every Mormon stake into GNIS without regard to whether it was just an isolated church.


 * The Kansas problem is the classic issue with 4th class post offices except for some reason amplified here. It's also interesting that I've another scam town (Smoky Hill City, Kansas). The argument that eventually crops up with these (and there's plenty of evidence that a place isn't a town when it locates to a still-extant farm house) is that people will come in and say "see, X was from Y, it says so in this old newspaper, so it's a community!" As someone who is "from" a "town" which in reality I never lived in (it was a five mile trip to get there) I'm not too keen on this interpretation. There's also the issue of these things seemingly moving around, which has an Irishman's hammer quality to it: how is a post office the same when it has been in three different places with four different names?


 * Anyway, besides reliability not being an absolute thing, yes, each of these needs to be tested against other sources and treated with some skepticism. In the Palatine case, for example, the spot indicated just doesn't make sense. Maybe someone thought it would good place for a town, but that's highly doubtful; there's just no access to the spot now, and no trace that there ever has been. There's not even a driveway leading to nowhere. At some point we have to say that the information just isn't good enough, and we have plenty enough data from this very cleanup operation to justify skepticism of something being called a "hamlet" or "village" or "town" when there's no other corroboration. Mangoe (talk) 13:46, 30 January 2024 (UTC)