[Talk-us] Fw: Re: Dirt Roads (formerly: Abandoned railway)

Kevin Broderick ktb at kevinbroderick.com
Tue Sep 2 14:50:35 UTC 2014

Apologies for the long email below, but I'll summarize my position as:
a) TIGER data, while imperfect, is very useful and not particularly harmful
when considered against other generally available map data.
b) The distinction between unpaved roads and tracks is important
c) It's often virtually impossible in forested areas to distinguish a minor
gravel road from a track by orthoimages, particularly if the imagery was
taken while trees had their full complement of leaves, so either alternate
imagery or ground surveys are necessary to correctly tag these ways.

While I can't speak to the U.S. as a whole, I can vouch for the TIGER
import data around me (and in Vermont generally) as being far from perfect
but also far better than no data at all. Based on a completely unscientific
survey of the road history that I've examined while editing, the vast
majority of ways in Vermont are TIGER imports that have been edited only by
a couple of metadata-adjustment bots. The TIGER data in Vermont includes a
lot of Vermont's "ancient" roads—legal right-of-way that vary from
privately maintained but publicly accessible gravel roadway to rough woods
roads and even to roadways that have returned to forest and are difficult
or impossible to locate on the ground.

With that said, OSM is not unique in including many of these "roads" in its
database; Google, Garmin, and other mapping provider share many of the same
errors, and even the venerable Delorme Gazetteer—by far the best indication
in Vermont of whether a road goes or not—has errors. Blithely following any
set of GPS directions in this state, particularly with routing set to
"shortest distance," is likely to put you on at least one "road" that isn't
passenger-car friendly; I, and most other folks in Vermont, I'd wager, have
a habit of reviewing Garmin's suggestions for roads named "Unpaved Road",
"Alley", or "Town Highway #33" (the last being a side-effect of all town
roads in Vermont being numbered as well as named, and the class-four roads
often not having proper names).

With that said, we also have thousands of miles of well-maintained gravel
roads that are passenger-car passable well over 95% of the time (and some
very near 100%). Treating all unpaved roads as tracks would actually create
a more-dangerous situation in the "boy who cried wolf" scenario—drivers
might get accustomed to easily passable "tracks" and be a bit surprised
when they find one that isn't. There is statewide data available from the
Vermont AOT which provides better information than the TIGER data, in terms
of both completeness and of accuracy, but I lack the time to figure out how
to create an automated process to compare the AOT data to OSM to more
quickly identify "roads" that are likely to be tracks. The state-provided
shapefiles do have data that is rather good at distinguishing proper
roadways from tracks, but the gradations of not-town-maintained road aren't
applied consistently from town-to-town (what one town calls "Impassable or
untravelled" may be "primitive and unimproved" in another town), and most
freely available imagery from commercial sources is virtually useless for
these ways—one may be able to readily pick out a few spots where there's
indication of a woods road but not have any luck at all with tracing the
entire way. The state has imagery from spring and fall overflights that is
more helpful than Bing imagery (which seems to be consistently captured
while all the trees have their full complement of leaves), but until it
finished its LIDAR overflights and releases that data, most of these ways
will require on-the-ground surveying to distinguish both exact route and
whether or not they are passable by vehicle or foot.

Updating these is one of my priorities, and I think most of the
unmaintained right-of-ways in my area are now correctly tagged in OSM (and
I've added many of the missing track sections that are right-of-way but not
passenger-car-friendly), but that's a very small part of the country as a

I'd also argue that a fair number of people are misusing "dirt" as as
surface type when "gravel" is more correct, both in conversation and in
tagging, but that's probably a battle I'd not win. In my mind, "gravel"
should be used for any maintained roadway where gravel has been added,
while "dirt" would be used for roads whose surfaces are comprised of the
material that happened to be there before the road was created.

On Tue, Sep 2, 2014 at 8:05 AM, Martijn van Exel <m at rtijn.org> wrote:

> I think I got unsubscribed from talk-us again on my work account (it’s not
> you, it’s me), reposting from my personal account:
>> Martijn van Exel
> From: Martijn van Exel <martijnv at telenav.com> <martijnv at telenav.com>
> Reply: Martijn van Exel <martijnv at telenav.com>> <martijnv at telenav.com>
> Date: September 1, 2014 at 9:46:36 PM
> To: Nick Hocking <nick.hocking at gmail.com>> <nick.hocking at gmail.com>,
> talk-us at openstreetmap.org <talk-us at openstreetmap.org>>
> <talk-us at openstreetmap.org>
> Subject:  Re: [Talk-us] Dirt Roads (formerly: Abandoned railway)
>  We have seen no major routing disasters because of the largely fictional
> rural TIGER landscape, and I don’t feel an urgent need to do something
> about it on this scale. The affected areas mostly have no meaningful
> destinations anyway, and where they do, people will get around to updating
> them. We can prioritize the more meaningful areas using a TIGER ghost town
> analysis along the lines of the TIGER desert analysis I did some time ago:
> http://oegeo.wordpress.com/2012/10/21/binders-full-of-tiger-deserts/ (this
> would actually be fun to do. Would people fix TIGER ghost towns if I
> pointed them out?)
>  highway=residential may be inappropriate, but it’s not as wrong as
> marking all unedited TIGER residential ways with access=no - which would
> make entire areas unreachable altogether. Also identifying the ‘untouched
> TiGER’ ways is not super straightforward as several bots have touched most
> ways in the US anyway, so you’d need to look at the full history to make
> that call reliably.
>  --
> Martijn van Exel
> Telenav
> From: Nick Hocking <nick.hocking at gmail.com> <nick.hocking at gmail.com>
> Reply: Nick Hocking <nick.hocking at gmail.com>> <nick.hocking at gmail.com>
> Date: September 1, 2014 at 9:28:45 PM
> To: talk-us at openstreetmap.org <talk-us at openstreetmap.org>>
> <talk-us at openstreetmap.org>
> Subject:  Re: [Talk-us] Dirt Roads (formerly: Abandoned railway)
>   While people work out how to remove the multitude of tiger ways that
> don't actually exist, downgrade others from the incorrect "residential" to
> "unclassified" or "track"
> depending on imagery or ground survey, and fix the geometry of all
> unedited TIGER data, I beleive that it's absolutely essential (from safety
> and useability perspectives) to immediately mark all these uncertain ways
> as unroutable.
> Whether to make them driveways or use access=no , I've no idea.
> I think thrse ways can easily be identified by...
> 1) They are original TIGER data import
> 2) They have not been edited since import
> 3) They are "higway=residential"
> 4) They are unnamed
> A bot could do this easily and then it really doesn't matter how long it
> takes to find the best solution.
>  _______________________________________________
> Talk-us mailing list
> Talk-us at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-us
> _______________________________________________
> Talk-us mailing list
> Talk-us at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-us

Kevin Broderick
ktb at kevinbroderick.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-us/attachments/20140902/425646fc/attachment-0001.html>

More information about the Talk-us mailing list