[Imports] Kern County Import Cleanup

Toby Murray toby.murray at gmail.com
Sun Jul 29 08:04:42 UTC 2012


On Sun, Jul 29, 2012 at 2:30 AM, Paul Norman <penorman at mac.com> wrote:
> I happened to be looking at Kern County and noticed two problems with the
> imports that seem to be systemic over the area.
>
> The first is classification of empty areas as landuse=residential (e.g.
> http://www.openstreetmap.org/browse/relation/540695
> http://www.openstreetmap.org/browse/way/53993995)
>
> I'd suggest deleting these as no landuse appears to appropriate here
>
> The second is a similar issue, mountainside areas in landuse=farm (e.g.
> http://www.openstreetmap.org/?lat=35.26&lon=-118.47&zoom=14)
>
> While doing this cleanup I would suggest removing attribution, description,
> kern:Comb_Zn, kern:Zn_Cd1 and setting source=Kern_County_GIS
>
> The appropriate solution for most of the ways/multipolygons seems to be to
> delete them, but I'll leave this to you since you're familiar with the
> extents of the upload.

The one that first caught my attention was this gigantic relation:
http://www.openstreetmap.org/browse/relation/1820616

It appears to be a good chunk of the county tagged with natural=scrub
with hundreds of holes carved out for other landuse/landcover areas.
This seems terribly unmaintainable to me. I mean I guess most of it IS
scrub land... but I'm not sure we have a good way to represent a
"default landuse" for a large area in OSM. I've joked before about
adding landuse=farm to the Kansas border relation. This seems like a
similar approach.

There are also over 150 of these in one area:
http://www.openstreetmap.org/browse/way/54408682

This looks like imported data about future development plans but
tagged as current use. I see some unpaved roads in the area but
absolutely no sign of any real development. Bing imagery is from about
the same time as the upload.

It seems like OSM is just being used as a dumping ground for any and
all GIS data that happens to be laying around. I guess in theory I
don't have a huge problem with this except that our current tools and
methodologies aren't made to handle this. I know this has been brought
up before but the two big problems as I see them are 1) updates and 2)
accuracy.

If we don't solve the update problem, OSM will just become known as a
repository for stale data which is obviously a terrible thing.

I have been importing TIGER 2011 data to replace damaged areas in LA.
Every time I notice errors. Since I am doing small areas at a time, I
can actually review and correct things. With huge landuse imports this
is impossible. And then we end up with things that are off by 15-20
meters and look crappy:
http://www.openstreetmap.org/?lat=34.95455&lon=-118.16858&zoom=16&layers=M

Toby



More information about the Imports mailing list