[Talk-ca] Proposal: Cleanup of NHN ways in BC

[Paul Norman]

Lately I've been working on cleaning up the NHN rivers in the Seymour,
capilano, indian arm and a couple other watersheds. It occurs to me that
this is a process that needs to be done everywhere.

Existing tagging:
Currently there are two types of tagging for waterways that were imported.
The first of these is that for smaller waterways
accuracy:meters=10
attribution=Natural Resources Canada
oneway=yes
source=GeobaseNHN_Import_2009
waterway=stream
waterway:type=observed

The second is for "connector" waterways that are under lakes or rivers
accuracy:meters=-1
attribution=Natural Resources Canada
oneway=yes
source=GeobaseNHN_Import_2009
sub_sea=stream
sub_sea:type=inferred

The ways themselves are over-digitized, having approximately four times the
nodes they need to.

I propose the following process to clean up these ways. This is based on the
process I was manually using last week and it worked without needing any
corrections. 

1. Retag the connectors the same as streams, preserving any extra tags
(primarily name=*)
This subject came up on talk-us@
<http://lists.openstreetmap.org/pipermail/talk-us/2011-February/005269.html>
and the consensus was that they should be tagged like other waterways.
1b. Retag, moving accuracy:meters to accuracy, preserving attribution,
source, waterway, and removing waterway:type and oneway. Any waterways I
verify with imagery will have accuracy:meters removed and source set
appropriately. 
	
2. Stitch together ways of the same name into longer streams. This will take
the most time and intervention. At the same time, rivers will be changed
from waterway=stream to waterway=river

3. De-duplicate named streams and rivers. There is duplication between the
data that was imported and existing data from before the import that has not
been cleaned up, and at this stage it will be easy to identify and fix,
selecting one way based on imagery.

4. Run JOSM's simplify way function with max-error=2. This will remove
approximately 75% of the nodes from the streams while preserving the shape.
There are nodes every couple of meters, even for straight sections of river.

5. Break extremely long rivers into manageable sections. For obvious
reasons, I don't want to leave behind ways that stretch over hundreds of
kilometers like some of the rivers do. 

As a rough estimate, this will take an hour per NTS tile initially, with
most of that time being upload time. I think that right now there are more
nodes/square km in the middle of nowhere from the streams then in parts of
the lower mainland because the streams have excessive nodes.