[OSM-talk] Planned revert of added surface and tracktype tags without local knowledge in various countries
Mark Wagner
mark+osm at carnildo.com
Sat Jul 18 20:21:32 UTC 2020
Almost all of the tracktype mapping around me has been done by armchair
mappers working from from aerial images.
Tracks in my area are usually produced in one of two ways:
* A bulldozer is used to scrape vegetation and topsoil off.
* A given route is driven repeatedly, eroding any vegetation or
topsoil.
Either way, the track surface is composed of whatever's underneath the
top layer. If I understand tracktype tagging correctly, this should be
grade 4 or 5, depending on how much gravel and/or clay is present in
the soil. Actual tagging is almost uniformly divided between grades 2,
3, 4, and 5, with a few spots of grade 1.
Some of the more interesting inconsistencies:
* A road being "grade 2" north of an intersection and "grade 3" south
of it, despite being the same dirt surface on both sides.
* An abandoned railroad being a mix of "grade 3" and "grade 4", despite
the track ballast (grade 2) still being present and visible in aerial
imagery.
* Two adjacent sections of track being tagged as "grade 2" and "grade
4" not because of any difference in road surface, but because one has
a line of grass between the ruts and the other doesn't.
--
Mark
On Sat, 18 Jul 2020 18:51:57 +0200
Florimond Berthoux <florimond.berthoux at gmail.com> wrote:
> I see no big issue of using only aerial images to set track_type or
> surface. You can get a fairly good result with such sources.
> So no, you should not blindly revert its modifications.
> If the estimation was really bad almost all the time why not, but
> here the example given is ok.
>
> Le sam. 18 juil. 2020 à 12:56, Michael Reichert
> <osm-ml at michreichert.de> a écrit :
>
> > Hi,
> >
> > while reviewing changes in my local area, I discovered that user
> > Modest7 has been adding tracktype=* tags to lots of highway=track
> > at various locations. I asked him what sources he used apart from
> > the satellite imagery mentioned in the imagery_used=* tag of his
> > changesets. See https://www.openstreetmap.org/changeset/87236896
> > for a discussion with him.
> >
> > I do not believe that one can add reliable tracktype=* information
> > from satellite imagery without having some ground truth knowledge
> > in order to know how to interpret the imagery in that region.
> > Adding estimated tracktype=* does not help OSM on the long term.
> > People how rely on the information (e.g. some wanting to drive or
> > cycle on that track) are disappointed about this low-quality OSM
> > data. Mappers who decide where to map assume these roads to be
> > mapped properly. IMHO, adding fixme=resurvey tracktype will not
> > improve it. Data consumers usually do not use tags like fixme=* In
> > the case of imports (another type of mass editing), we say that an
> > import must not add fixme=* to cover shortcomings of the data to be
> > imported because they usually do not get fixed in a reasonable
> > time. Therefore, I plan to revert these changes.
> >
> > Modest7 does not seem to realise that estimating tracktype from
> > satellite imagery is not doing a service to OSM. I am currently
> > preparing a revert of all additions of surface=* and tracktype=* by
> > him he uploaded since 1 January 2020 [1]. The revert will only edit
> > tags, geometry will stay unchanged. I revert changes on surface as
> > well because that's not very different to tracktype except that it
> > applies to other types of roads as well.
> >
> > The countries which will be affected are:
> > Germany
> > Denmark
> > Turkey
> > United States
> > Poland
> > Ukraine
> > Morocco
> > Czech Republic
> > Lithuania
> > Sweden
> > Norway
> > eSwatini
> >
> > A changeset discussion with him can be found at
> > https://www.openstreetmap.org/changeset/87236896
> >
> > Best regards
> >
> > Michael
> >
> >
> > [1] This date is not fixed yet.
> >
> > _______________________________________________
> > talk mailing list
> > talk at openstreetmap.org
> > https://lists.openstreetmap.org/listinfo/talk
> >
>
>
More information about the talk
mailing list