[Tagging] Clarification unclassified vs residential

Fernando Trebien fernando.trebien at gmail.com
Wed Feb 20 20:58:51 UTC 2019


On Wed, Feb 20, 2019 at 4:52 PM Mark Wagner <mark+osm at carnildo.com> wrote:
>
> On Wed, 20 Feb 2019 11:30:07 -0300
> Here in Washington State, there are two possible "non-fuzzy"
> classification schemes:
>
> 1) By usage rules: Motorway/arterial/other.  This isn't very
> fine-grained, and doesn't give you a way to distinguish Sprague Avenue
> (five lanes one-way) from Queen Avenue (two-way, no lane markings) from
> WA-20 (major cross-state highway) from WA-127 (exists only to avoid a
> hundred-mile detour when crossing the Snake River).
While having the two together side by side

While I think this idea would hold well in developed areas, it may not
in the sparser ones.

> 2) By operator: Interstate highway/US highway/state highway/county
> road/other.  This is fine-grained but misleading: for example, WA-520
> (a state highway) is a multi-lane divided grade-separated high-speed
> controlled-access road -- in short, an Interstate in all but name.

Brazil has the same problem. Some of the federal highways are
motorways with multiple lanes, others are little more than a dirt road
that is not even compacted regularly. Likewise for state highways. And
there are cases where state highways are more developed (more lanes /
better maintenance) than federal highways connecting the same places.

> (There's also the state highway department's internal highway
> classification system that maps reasonably well to OSM's system, but
> you'd have to copy it from their maps -- there's no on-the-ground
> evidence that something is an "Urban Major Arterial" or a "Rural Minor
> Collector".)

Not everything needs to be on the ground. Administrative borders are
only very rarely marked on the ground and we still rely on published
data. So, if it is publicly available, it can be used by mappers,
especially if it produces a satisfying result. In Brazil we use
official municipal maps to classify urban streets in OSM. This data is
spread over the websites of prefectures (one per municipality), so we
gathered the main ones on the wiki to make verification easier by
independent mappers. [6] Just like in highways outside urban areas,
the physical profiles of urban street classes often overlap.

> > Do you have any locally-defined highway system that approximately
> > matches the idea of "a system of highways that generally connects
> > place=hamlet"?
>
> That would be the state highway system: nearly every incorporated
> community and most of the unincorporated ones are served by at least
> one state highway.  But see the above examples for why calling these
> roads "unclassified" is a bad idea.

It may look like that if you think that the route between a pair of
distant hamlets will mostly be made up of state highways. But the
state highway system connects larger settlements as well - place=town,
place=city - and that's probably its main purpose. So I don't think
the system we're looking for is the state highway system. It may help
to think classification top-down: first you connect pairs of very
large settlements using trunks, then you connect pairs of (say)
city-sized settlements using primary ways (without changing any
previously assigned trunks) (also connect cities to the large
settlements of the previous set), then pairs of town-sized settlements
using secondary ways (and town-city pairs, and town-large settlement
pairs), and so on. This is of course tedious, a little subjective, and
error-prone, that's why this result is to be approximated using other
verifiable characteristics.

I queried place=hamlet in Washington State using Overpass Turbo and
compared it with the current classification in OSM. Indeed, many such
places lie next to major highways, as is the pattern in other
countries (small communities tend to sprout near them). But there are
many that do not. For example, Nilles Corner [1] and Osborne Corner
[2] in Douglas County do not. What is generally expected is that the
main route between them is made of ways whose class is
highway=unclassified or higher. The route may overlap with routes
between more important places, say Niles Corner to Fairview [3] whose
main route overlaps with the tertiary WA-17 highway, which (as
generally expected from highway=tertiary) is the main route between a
pair of place=village: Bridgeport [4] and Coulee City [5]. What's
interesting about the roads that connect Nilles Corner and Osborne
Corner is that they do not connect any pair of place=village or
higher, so there's no reason to classify them as any higher than
unclassified. There could be a really nice multi-lane highway between
them, but very few people would be using it. (Of course, there is none
because the demand is very low.)

That's the general idea. The next step is to attempt to find a rule
based on local highway systems (recognizable by Americans, especially
by Washingtonians) which closely produces this general result in most
cases. Comparing existing maps (digital or paper) may help defining
such a rule too. The more you compare (maps/approaches and places),
the more satisfying the agreed rule will be for mappers and users.

In Brazil, for example, as we advance in this discussion, we have
found so far that most pairs of place=city are connected by paved
federal highways and that there's a particular kind of state highway
(officially called "access" or "link" road, not to be confused with
trunk_link) which is typically used to connect place=town. So these
characteristics could eventually be used to produce a reasonable,
although still imperfect classification (what looks really close to
the UK classification is the functional classification published by a
few state highway departments). These qualities are not always
obtained from the "ground" but by jurisdiction (federal vs state) and
type (code higher or lower than 400), shown in most road signs but not
in municipal roads and in most unpaved state/federal roads (usually
without any signs).

In Germany and in the UK, road type is usually identifiable from
highway signs and physical profiles may overlap between types, though
usually they are clearly distinct when the two highways cross or lay
near each other. What they have is essentially a functional road
classification system that is clearly marked by the road signs.

[1] Nilles Corner: https://www.openstreetmap.org/node/150974922
[2] Osborne Corner: https://www.openstreetmap.org/node/150951417
[3] Fairview: https://www.openstreetmap.org/node/150933316
[4] Bridgeport: https://www.openstreetmap.org/node/150948574
[5] Coulee City: https://www.openstreetmap.org/node/150959667
[6] https://wiki.openstreetmap.org/wiki/Category:Cidades_com_hierarquia_vi%C3%A1ria_oficial

-- 
Fernando Trebien



More information about the Tagging mailing list