[Talk-us] Low-quality NHD imports

Kevin Kenny kevin.b.kenny+osm at gmail.com
Fri Oct 13 20:47:35 UTC 2017

On Fri, Oct 13, 2017 at 1:34 PM, Christoph Hormann <osm at imagico.de> wrote:
> There are a number of possible measures that could be considered for
> improving old NHD imports:
> * removal of unnecessary tags to reduce the baggage mappers would have
> to deal with when working on the data.
> * removal of small unnamed streams which are not necessary for the
> overall river network connectivity in areas where the geometric
> accuracy is poor by current standards (and it is therefore usually
> easier for mappers to newly trace those streams instead of trying to
> improve the inaccurate data)
> * creating maproulette challenges for fixing inaccurate waterway
> classifications - in particular waterways tagged 'waterway=stream' but
> with a name containing 'Creek' or 'River' will often qualify as
> waterway=river.  Same for artificial waterways with 'waterway=ditch'
> but names containing 'Canal' or ther other way round.
> * creating maproulette challenges for unconnected waterways.
> * adding missing 'intermittent=yes' to waterways in imports where this
> was not properly set based on the feature codes.

I've no argument with any of these (but please keep reachcode!).  They
all seem to be worthy projects. I'm of two minds about maproulette
challenges, only because I've seen some pretty low-quality results
coming out of them.

I have issues only with 'newly trace the streams' in areas with dense
tree cover - even some significant streams bordering on
'waterway=river' can be virtually invisible on the aerial images
around here. A lot of the time if I am tracing a stream, I cross-check
with the 1/3-arc-second (or 1/9-arc-second where I can get it) radar
altimetry and make sure that the streambed appears to be following the
'v's in the contour lines, and I'm not sure I'm actually doing any
better than NHD. Since I'm in an area where NHD is pretty good, I
usually start by importing individual watercourses and then try to
improve the data. (One of my projects for when the snow starts flying
is to make sure that I've brought in enough that the streams that I
added for some specific large-scale rendered maps actually reach
the rivers.)

A definite +1 on the inaccurate waterway classifications. In
particular, anything with an artificial flowline and a drawn riverbank
is a 'river,' whatever the name or the FCode say.  (I've had one
discussion in the past with someone who wanted to downgrade the
Schoharie Creek to 'stream' - apparently the guy couldn't quite grasp
that it has dams, reservoirs, power generation stations, catastrophic
floods from time to time. Despite 'Creek' in the name, it's a
third-order river.)

Oh, and a side question: should we be mapping artificial connections,
and if so, how? It is known, for instance (by accidental
contamination) that the water from one small lake locally flows
underground and exits through caves in a cliff a few km distant, but
not by what route it flows. NHD has an artificial 'connector' flowline
for that purpose, that I've never seen fit to map, having absolutely
no idea how to map it. In other cases of sinks and springs around
here, the connections may not even be known. (Ah, the joys of
glacio-karst terrain.)

For that matter, I also have No Clue what to do about
rapids. "waterway=rapids" is deprecated, and I'm not a canoeist or
kayaker - the "Whitewater Sports" page on the wiki says that if you
don't know the practice on the river in question, don't map it. (I
would have thought that "rapids here" had some utility even in the
absence of further information, but apparently the community

Then again, that 'whitewater sports' page has some other misleading
stuff: "All rivers with width more than 5 meters has to be draw by
waterway=riverbank or natural=water+water=river in addition to
waterway=river.  I'm sorry, if I can't see the bank in aerials, but
can follow the flow line on elevation maps, I'm still going to map the
stream. Incomplete information is better than a blank map. If we
demand perfection from mappers when information is first being
gathered, that'll scare even more away. I'll agree with "it is
desirable to show the riverbank on rivers more than XX metres wide",
and '5' is probably too small a value for XX - I don't get that sort
of repeatability from GPS, and I'm surely not going to do a plane
table survey of the banks of my local streams!

In short: We agree about good ways to move forward, even if I reserve
judgment on the reasons.

More information about the Talk-us mailing list