[Imports] Converting Fairfax County, Virginia government Shapefile of streets to OSM

Josh Doe josh at joshdoe.com
Tue Dec 6 11:56:46 GMT 2011


On Sun, Dec 4, 2011 at 7:32 PM, Paul Norman <penorman at mac.com> wrote:
>
>> From: Josh Doe [mailto:josh at joshdoe.com]
>> Subject: [Imports] Converting Fairfax County, Virginia government
>> Shapefile of streets to OSM
>>
>> I just recently noticed that Fairfax County, VA added the street
>> centerline data to their website, and it is fairly recent (2011/10/17).
>> I had previously gotten permission to use this data from FCDOT, but it
>> was in ESRI GeoDatabase format and wasn't publicly accessible. Also
>> they've added better attribution (differentiating Interstate highways
>> from primary roads, marking highway links, etc.).
>> I've converted the data to OSM as best I can (e.g. expanding street
>> suffixes) but I would appreciate others reviewing it:
>> http://wiki.openstreetmap.org/wiki/Fairfax_County,_Virginia/Roads
>>
>> The table on the wiki is a rough mapping, for details the ogr2osm
>> translation script is provided further down the page. As mentioned on
>> the wiki page, this is not for mass import but for adding new
>> subdivisions in JOSM, replacing bad TIGER data, manual conflation with
>> Potlatch2 [1], etc.
>>
>> You can either download the data yourself and run the translation
>> script, or download this extract of the converted data:
>> http://dl.dropbox.com/u/23634456/osm/braddock_streets.zip (ZIP of .osm
>> file)
>
> I had a look at the conversion (based on the wiki page) and the only issue I
> saw was some 1 mph roads which are likely issues in the source data.

Ah, thanks for picking up on this. I've added a fixme warning if the
speed limit is less than 5 mph.

> Some of the ways are over-noded and there's the continual problem of trying
> to distinguish between highway=residential and highway=unclassified. I
> generally use residential as the default value.

Do you (or anyone else) think there's value in doing the simplifying
during the conversion? And if so, to what distance tolerance? For
roads I don't like to use anything bigger than 1 meter. I have bright
hopes and dreams of getting this data into a micro-tasking app, so I'd
like this data to be as ready to merge or conflate as possible, via
JOSM or the new/upcoming Potlatch2 functionality.

As for residential vs unclassified, I chose the latter because I've
heard chatter about the TIGER import making a bad decision by using
the former.

> You'll also have to dis-entangle roads below bridges from the bridges, but
> that's fairly normal.

Thanks for this one as well. I guess there's no foolproof way to
automatically fix this, but perhaps there's some intelligent flagging
I could do. Any suggestions?

-Josh



More information about the Imports mailing list