[Imports] Vermont Town boundaries from VCGI

Andrew Guertin andrew.guertin at uvm.edu
Wed Aug 8 17:11:24 BST 2012


On 08/08/2012 11:18 AM, Greg Troxel wrote:
> 
>   > Have you confirmed that the state boundary in osm matches the outer
>   > boundary of towns?
> 
>   The state boundary in the VCGI file (which exactly matches the outer
>   boundary of the towns in the VCGI file) does *not* exactly match the
>   state boundary in OSM.
> 
>   In most cases involving straight lines it is reasonably close--within
>   about 10 meters. In a few areas of the middle-of-the-lake border with
>   New York, they diverge by over 100 meters. In on the river border with
>   New Hampshire (the border is specifically the low water mark of the west
>   bank of the river), there are some places where the river has shifted
>   (e.g. to cut off an oxbow) and this is reflected in one version but not
>   the other. There are also some places where the the two versions match
>   but do not match the bing satellite imagery. Interestingly, on the
>   US-Canada border, most of the nodes seem to be exactly the same--OSM has
>   them marked with survey point reference numbers, so this would make
>   sense--but shifted a few centimeters. Probably something to do with
>   projections.
> 
>   I plan to keep the existing state and national borders, and manually
>   connect the new town boundaries with the existing state border.
> 
> I wonder where the existing data came from, and if the VCGI data is
> actually better.

Based on local knowledge, research, and survey, I can say that in at
least one area the VCGI data is much better. The Winooski Charter
(http://onioncity.com/cc/winooski_charter.pdf) says "[...] thence
proceeding on a bearing of S 07 55/' 50" W in and along the West line of
the so-called Gorge Road [...]". Some time ago I surveyed the area and
noticed that the Tiger data (unchanged at the time) had what is actually
Gorge road unlabeled, and had a non-existent road labeled as Gorge road,
about 30 meters away. I fixed the road, but not the Winooski boundary
(which Tiger had running down the middle of what they thought was Gorge
road). The VCGI data has that boundary running exactly down the west
side of the real Gorge.

That, though, is just one little thing. More important is that the VCGI
data comes with documentation and references to actual surveys done, if
not by OSMers, at least by people who are good at what they do. See
http://www.vcgi.org/metadata/BoundaryOther_BNDHASH.txt, and look through
the Data_Quality_Information section.

> My impression is that the mass data is based on surveys of actual marker
> stones, and when I've checked positions from (hiking class) GPS, I have
> been unable to find errors that are beyond what I'd expect from GPS.  I
> would expect the same of Vermont's data.
> 
> Calvin: what did you do about the state border when you imported the
> town boundaries for Mass?  Leave it, or set it to the mass town border?
> 
> It might be best for you to make the Vermont town borders in OSM match
> the official vermont data, and let the state border be different, until
> this is figured out, rather than adjusting the town to match what might
> be incorrect state data.  I don't see any real harm in them not lining
> up.

One problem with them not lining up is that two lines would be drawn on
the mapnik tiles.

--Andrew



More information about the Imports mailing list