[Talk-us] [Imports] US Golf Courses from GNIS

[Josh Doe]

I've noticed in my area golf course nodes added that already exist:
http://www.openstreetmap.org/browse/node/1556625188
http://www.openstreetmap.org/browse/node/1556629688
and others

I support reverting this changeset ASAP.

Golf Geek,
Let's instead take the work you've done and split it up into state
sized chunks (e.g. via Osmosis). Then several contributors including
yourself can manually merge the nodes a state at a time. Thank you for
your interest in this, and for coming forward on the mailing list.
Trust me that this is not the first time this kind of thing has
happened, but you did the right thing coming here and letting us know.

Regards,
-Josh

On Thu, Dec 22, 2011 at 2:03 AM, Toby Murray <toby.murray at gmail.com> wrote:
> More problems I found by just downloading all leisure=golf_course
> objects and randomly browsing around some of Kansas/Nebraska with Bing
> imagery.
>
> Can't idenfity on aerial. I could just be missing it. Or GNIS position
> might be off by a lot. Some are in the middle of a town without so
> much as a full block of grass anywhere near them. Or it may have been
> closed but is still in GNIS. It is unlikely that it is a new golf
> course. Bing imagery seems to be pretty recent (2010) in most areas I
> looked at.
> http://www.openstreetmap.org/browse/node/1556624422
> http://www.openstreetmap.org/browse/node/1556638495
> http://www.openstreetmap.org/browse/node/1556635779
> http://www.openstreetmap.org/browse/node/1556635714
> http://www.openstreetmap.org/browse/node/1556624015
> http://www.openstreetmap.org/browse/node/1556625367
> http://www.openstreetmap.org/browse/node/1556625957
> http://www.openstreetmap.org/browse/node/1556631507
> http://www.openstreetmap.org/browse/node/1556638863
>
>
> Two golf courses in close proximity that are probably the same course,
> maybe known by two different names:
> http://www.openstreetmap.org/browse/node/1556638410
> http://www.openstreetmap.org/browse/node/1556627728
> and
> http://www.openstreetmap.org/browse/node/1556624801
> http://www.openstreetmap.org/browse/node/1556639241
>
>
> Were these not in GNIS or were they excluded because of an existing
> way? Could have maybe used GNIS data to add a name to the existing
> way:
> http://www.openstreetmap.org/browse/way/46342164
> http://www.openstreetmap.org/browse/way/43332671
> http://www.openstreetmap.org/browse/way/42280171
> http://www.openstreetmap.org/browse/way/98180901
> http://www.openstreetmap.org/browse/way/129025203
> http://www.openstreetmap.org/browse/way/126614718
>
> Toby
>
>
>
> On Wed, Dec 21, 2011 at 11:27 PM, Toby Murray <toby.murray at gmail.com> wrote:
>> On Wed, Dec 21, 2011 at 4:05 PM, Golf Geek <golfgeek2011 at hotmail.com> wrote:
>>> After reviewing the Import/Guidelines wiki, I realize I should have posted
>>> here first, but here's a quick "after action report" on a recent import.
>>> Better late than never. :)
>>
>> Why didn't you read this before the import? This should not be viewed
>> as optional.
>>
>>> I noticed that although USGS GNIS data had been imported into OSM in the
>>> past, the US golf course locations provided as GNIS Locales had not been
>>> included.
>>>
>>> So, I retrieved GNIS Locales with "Golf" in the name from
>>> http://geonames.usgs.gov/ and saved them as OSM nodes, using these tags:
>>>
>>> gnis:Class = Locale
>>> gnis:County = [various]
>>> gnis:ST_alpha = [various]
>>> gnis:id = [various]
>>> leisure = golf_course
>>> name = [various]
>>> source = USGS GNIS
>>>
>>> From the list of ~6000 nodes, I removed any that overlapped with existing
>>> OSM golf_course nodes or ways.
>>
>> You apparently failed to take into account how terrible GNIS spatial
>> accuracy can actually be:
>> Your node: http://www.openstreetmap.org/browse/node/1556636801
>> Existing way: http://www.openstreetmap.org/browse/way/70764331
>>
>> Yes, that over a mile off. This is why the import guidelines say to
>> discuss it with the community FIRST. There is much collected knowledge
>> about imports in the community which can prevent such common mistakes.
>>
>>> The remaining 4421 nodes were then added as Changeset 10168800.
>>>
>>> The data license is OK (USGS GNIS has been used before), and the new nodes
>>> should not screw up existing data (although I am sure they are not perfect),
>>> so hopefully this import will be a good starting point for further manual
>>> edits.
>>
>> With nodes that are off by a mile, I am doubtful of this claim. So
>> far, I have only looked at that one node so far. Others, please check
>> more in your area. If mine is an outlier then I'll just fix it. If
>> there are many more that are as bad as this one, I would propose
>> reverting this import, especially since import guidelines were not
>> followed.
>>
>> Toby
>
> _______________________________________________
> Imports mailing list
> Imports at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/imports