[Talk-us] NHD import: what data quality is acceptable?

Kevin Kenny kkenny2 at nycap.rr.com
Mon Jul 23 13:44:51 BST 2012

On 07/22/2012 09:33 PM, Paul Norman wrote:
> The mappings on the wiki are not only incomplete and inconsistent, they're
> for an older NHD version and sometimes clearly wrong.
> I posted a better one
> (http://lists.openstreetmap.org/pipermail/talk-us/2012-July/008502.html)
> earlier this month but it didn't attract any comments, and it's not complete
> either. It handles most of the FCodes but still is missing a couple. It also
> needs some post-processing to clean up over-noded ways and some other
> matters.

Right. I downloaded and looked at your code, but I was already
pretty far along when you posted that message. I'd mostly been
working by diligently examining, each time I encountered an FCode that I
haven't seen before, what the feature actually is, from personal
knowledge. (I then often presume that other features having the same
FCode are the same general sort of thing.) Except for likely having to
invent some stuff for karst features, I think that I have a pretty
sound tag mapping. I'll go back at some point and check how it differs
from yours. At a quick glance, they're pretty similar.

I'm using a somewhat different workflow, doing a lot of the heavy
lifting in PostGIS. My general plan involves clipping of flowlines,
areas and waterbodies to HU12 basins so that I have bite-sized pieces
to process with minimal connections to make at the edges: ideally
a single connection, but sometimes the HU12 watershed lines are
slightly misdrawn and pull in tiny bits of streams that actually
belong to another basin. PostGIS also gives me a fairly easy way to do
collision checking and find candidates for conflation.

Oh, by the way, my plan is to include nhd:reach_code and
nhd:permanent_id tags, to facilitate conflation in the event that
another NHD version obsoletes the current one.
73 de ke9tv/2, Kevin

More information about the Talk-us mailing list