[Talk-us] Update on potential highway classification reform

[Kevin Kenny]

On Mon, May 24, 2021 at 7:35 PM Jmapb <jmapb at gmx.com> wrote:

> Thanks Kevin, you've really got the magic touch when it comes extracting
> data from NY State.
>
> My offhand impression is that this map has a lot red. That includes both
> things I feel are oversold (eg NY-23 east of NY-8, currently tagged as
> secondary) and things I feel are undersold (the NYC motorways... hard to
> make out exactly what's going on in there, but there's hardly any blue.)
>

You're right.  There's a classification that I'm missing that I think would
be significant for OSM - 'principal arterial freeway' - which I
misclassified as 'trunk road'.

These are freeways that are not all that significant for interurban routing
because they don't carry significant traffic that doesn't originate or
terminate in the metro area. At low zoom levels (z5-z6, maybe z7) they'd
just clutter the map; you wouldn't be able to read them.

There's also a lot of red because here in the US we've been classifying
roads a lot lower than most other places; insisting that a trunk has to be
an 'almost freeway' or that a state highway can be at best secondary,
things like that, rather than considering network connectivity.

I've been delving a little deeper into the data and patching it up some.
https://kbk.is-a-geek.net/attachments/20210524/NY-class-1-2-3.png is a
replaement image, with blue showing the interurban motorways, purple the
suburban motorways, red the proposed trunks, and black the proposed primary
roads (which antialiasing is actually fading to grey). Note that not all
state highways are primary! There are a lot of spots that would need local
fixup - there are some obvious data entry errors, some routing islands, and
some impedance mismatches at the urban/rural boundary, but I think this is
a useful first approximation.

To see 'interurban' vs 'suburban' motorways clearer in the NYC area, I also
produced https://kbk.is-a-geek.net/attachments/20210524/NYC-class-1-2-3.png

> I don't know the west end of state all that well, or northern reaches, so
> I can't speak to that. But this certainly passes the Ithaca-Rochester test.
> And I do feel there's some value to employing the state's
> knowledge/opinions in this process.
>
> I fear we will, as you describe, end up with 50 to 56 different wiki pages
> of highway tagging guidelines. So be it...
>
The roads that you say are oversold, well, I presume that the state had
some reason for classifying them the way they did. The loop on NY-12 and
NY-23 from Binghamton to Oneonta is mysterious. Maybe to connect Binghamton
and Oneonta with Utica?  (No, I'll stick with 'mysterious.')

We're actually using a combination of features that the Federal road
classification system tabulates fairly well, so I'm guessing that it'll
work in some other states. The only problem is that so far we're applying
them to states that tend to build roads that are fairly coherent. I dread
trying to sort out Oklahoma or MIssissippi.

By the way, this classification would guide (a) what roads should appear at
what zoom level and (b) routing preference for long-distance routings where
an exact search for the best route would take too long.  Rendering in a US
map style should be based mostly on physical characteristics
(grade-separated carriageways? All crossings elevated? Number of lanes?
Speed limit?) and fall back on classification only as a last resort if the
other characteristics are not mapped. That would allow us to bring in a
two-lane trunk road such as US 209 south of Newburgh at a fairly small
scale, but still show it as the relatively slow and narrow surface street
that it is.  (It just beats most of the alternatives to get to the
Shawangunks and the Delaware Valley from, say Albany. There ain't no
graceful way. There's a mountain range in the way.)

> None of this would work as a mechanical edit, think of it as an
AI-assisted one if you will. The data analysis emits a set of 'things to
look at', human mappers look at them.

-- 
73 de ke9tv/2, Kevin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-us/attachments/20210525/3eb3d7e8/attachment.htm>