[OHM] representing relationships between historic and modern road nets

Sun Nov 10 22:28:50 UTC 2019

Hi Richard et al.,

My working example is here
https://www.openhistoricalmap.org/#map=15/52.9478/-1.1943&layers=H.

The main route is Derby Road and there have been 3 or 4 significant changes
from the route on OHM compared with now:

   -  In the 1820s the road to the W of the canal was moved to its present
   alignment
   -  In around 1909 a railway bridge was built to the S of the former
   level crossing, leaving a stub of the original road on the W side (still
   there)
   -  In the mid-1930s the alignment just to the E of the canal was
   straightened out to remove a corner with poor visibility, The old pub was
   knocked down and a new one built.
   - Roundabouts were introduced at major intersections (in the 1920s).
   These roundabouts have probably seen more change in geometry than any other
   part of the road. One has been a roundabout, signal controlled crossroad at
   grade, roundabout with access by off ramps, signal controlled gyratory.

Similarly Wollaton Road has had major changes where it crosses the canal
(the current curious dumbbell shape of the intersection is because they
used the existing canal bridge in the 1920s to save money).

Radford Marsh Road was severed when the railway was built (1850s), and is
still a curious stub.

On top of the straight alignment issues are all the other attributes. I
cant really identify the surface of Derby Road in the 1800s, so cant guess
when it was asphalted. From early 1900s to 1936 there was a tram line with
a terminus just W of the railway line, although for some reason this was
briefly extended a bit further west in 1932. For the most part speed
restrictions have not changed since they were introduced, but I suppose
technically one should have a 2 mph limit from 1865 onwards (see
https://en.wikipedia.org/wiki/Locomotive_Acts). Lastly the stretch coming
out of the city has been oneway for perhaps 40 years.

I have a couple of slides illustrating the alignment changes here
<https://www.slideshare.net/SK53/openhistoricmap-overview> (slides 11 &
13). Slide 11 compares the historical map from 1835 with OSM. Slide 13
summarises elements necessary for geometry/alignment only.

I haven't experimented at all with how more detailed attributing of
elements adds to the time-slicing of elements, but the comments above give
a rough idea of some of the possibilities.

For minor roads in the city I think for the most part alignments have been
less likely to change. Probably the biggest changes are wholesale
replacement of one road layout by another, and introduction of barriers to
reduce permeability of gridded street layouts.

I suspect the bulk of significant changes occur on main roads: just
modelling changes of the major roads in the middle of Nottingham over the
past 50 years would be quite a challenge, even though I think only two
major alignments have been introduced in that period. In practice the
significant changes have been in making roads oneway, and introducing
restrictions on cars. In other words the devil in the detail lies with the
tagged attributes rather than (for the most part) the alignments. The
situation in Buenos Aires is exactly comparable: very limited number of new
major road alignments from 1870, but lots of changes in the properties of
the actual roads.

The only ways I can think of avoiding multiple elements (ways) sharing the
nodes are as follows:

   -  Use key values with embedded start and end dates. Some precedent for
   this on OSM, but I think it is inevitably very clunky.
   -  Push some of the attribute stuff onto something like wikidata.
   Wikidata seems to have little or no good means of handling time-series data
   (or if it does I havent found it).
   -  Amend the core OSM model to separate the geometry from the actual
   tagged elements, so that a single geometry could be associated with 1:m
   elements (but not the other way round). Obviously this is not a short term
   approach.
   -  Make each tag have start & end dates. No idea how one could fit this
   in the OSM model.

If, as I suspect, these problems are most apparent on the main highway
network, it might not be as significant as appears:although that is no
consolation if that is one's principal interest.

Regards,

Jerry

PS. I have also tried to model in JOSM the 1960s urban motorway scheme
<https://www.sabre-roads.org.uk/forum/viewtopic.php?f=1&t=18761>, but
that's another issue again..

On Sun, 10 Nov 2019 at 18:57, Richard Welty <rwelty at averillpark.net> wrote:

> i just wrote this up to reflect some discussion Jeff and i had on the
> last OHM call. i do not have any proposed solutions to the issues i
> bring up; i just want to start some discussion.
> -----------------------
> i have experimented a bit with the entry of historic highway alignments
> into OHM. it has exposed some things we should think through. i've
> discontinued work on data entry as i consider what i've done so far to
> be a bit unsatisfactory.
> The core of the issue is that with historic alignments, in many cases
> the physical roads exist today, and if we simply use OSM tagging
> standards we duplicate a lot of data. but then, there are realignments
> which result in changes in the road, and these need descriptions in OHM.
> duplication of data allows for very simple display and data consumption.
> if we were to instead have links back to OSM for those things that are
> unchanged, there are two issues 1) what is the model for display and
> other data consumption and 2) how do we handle situations where physical
> change happens in OSM that requires us to update OHM with older data
> that's diverged. This i think happens more than we think about; here in
> upstate NY new fangled traffic circles are springing up all over.
> Here are a pair of links, one to OSM, and one to OHM, that i think
> illustrate some of the issues. OSM shows a landscape which has been
> dramatically altered by the construction of I 787. OHM shows my partial
> reconstruction of the pre 787 map.
>
> https://www.openhistoricalmap.org/#map=15/42.6994/-73.7126&layers=H
>
> https://www.openstreetmap.org/#map=15/42.7011/-73.7135
>
>
> --
> rwelty at averillpark.net
>  Averill Park Networking - GIS & IT Consulting
>  OpenStreetMap - PostgreSQL - Linux
>  Java - Web Applications - Search
>
> _______________________________________________
> Historic mailing list
> Historic at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/historic
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/historic/attachments/20191110/f092268e/attachment-0001.html>