[Talk-us] mapRe: (Second attempt) Potential data source: Adirondack Park Freshwater Wetlands

[Kevin Kenny]

On 03/26/2016 02:06 AM, Russ Nelson wrote:
> Frederik Ramm writes:
>   > I have zero knowledge about the Adirondack[s]
>
> I live here. Imagine a park half the size of Austria, with about 130K
> people living in it, and 200K people visiting it. Give about 30K of
> those people Internet access. Oh, and there are practically no nerds
> living in the park, because there are no high-tech jobs.
>
> It's unlikely that anybody will do much in the Adirondacks whether
> there's an import or not. If there's an import, at least there will be
> something. Something is better than nothing, because at least it's
> less wrong.
>
> Just do the import, Kevin.
>
I do see an emerging consensus here, and Frederik (to whose opinion I
ordinarily defer) appears to be something of an outlier. I do plan to
go ahead with this, with appropriate warnings, wikification, and a
quick request to the APA's GIS coordinator to confirm that we have
permission to import (we do already, but I want to get the specific
plan blessed).

It may be some time in coming. Those who know me know that I'm pretty
obsessive about data quality. This job is extremely likely to be a
three-way conflation: what's already there (which, it appears, is
mostly a 'lakes and ponds' file that you imported), the APA data set,
and NHD. Each source has its unique properties.

What's already there has all the tagging that mappers have done - and
must not be damaged! Nevertheless, there simply is not much in OSM for
the Adirondacks. I'm really working with a big blank spot in the map
here!

NED has the greatest detail (it was digitized at 1:24000 scale or
finer, for the most part) and has the GNIS names of features. It also
has feature classes that nothing else has, such as rapids, artificial
shorelines, flumes, and so on. Its chief drawback is that there are
objects that are unaccountably missing, in such a way that I suspect a
database glitch happened at the USGS. For instance, the Cedar River
Flow, a fairly sizable lake impounded by the Wakely Dam, is not in
NHD - but the river becomes an 'artificial path' there, which is
typically a flow line drawn through an area feature to keep the flow
lines contiguous.

The wetlands inventory lacks feature names, and is less detailed (it
was digitized from orthophotos at 1:40000 scale), but has many ponds
and streams that NHD misses. It also has the intermittent or ephemeral
water limits of many waterbodies. In the Adirondacks, these are
important to a hiker. Many trails go through beaver meadows. In years
when the beavers are in residence, the trails may be underwater, and
the hiker must find a route around the pond. Having the high-water
extent mapped is valuable information. The streams that it identifies,
in the few places that I've checked, are there in the field. Alas, it
does not have flowline topology, so conflation with NHD will need a
little bit of patching.

One bright spot is that the three data sets are well aligned (once the
differences in datum are accounted for). A simple collision check
identifies areal features to conflate. There may be a tiny bit of
manual work for a few places (Indian Lake/Lewey Lake; Long Lake/Park
Lake; Kiwassa Lake/Oseetah Lake/Lake Flower, and so on) where the
boundaries between lakes are indefinite, in that you can take a canoe
from one to another without noticing that you are on a 'different'
lake.

I'm still working on appropriate heuristics for conflating the linear
features (flowlines, mostly). Again, I have the advantage that there
is very little already in OSM to collide with - at most a few dozen
rivers. What may turn out to be easiest is simply to lift the tags off
the OSM features and apply them to the NHD ones.

Then there's the area surrounding Duck Hole, which was permanently
changed in Hurricanes Irene and Lee. Now that there is a few years'
worth of orthophoto data available in all seasons, I think the best
thing we could do there is to trace the shoreline from the orthophotos
and add notes that our data reflect the shoreline and river channel
from after the failure of the dam.

Whatever I do, I plan to leave the features in OSM tagged with enough
information to identify data provenance. This would mean, at the very
least, NHD reachcode and permanent ID, APA object ID, and NWI label,
where these are known, together, of course, with whatever tagging is
present on the features that are already there.

It would be good to point out that even the 'authoritative' data for
this part of the country is far from the standard that is usually
expected in the developed world. There are even a fair number of
county lines in the Adirondacks that have never been surveyed on the
ground and are marked as 'INDEFINITE BDY.' on the topo maps.

-- 
73 de ke9tv/2, Kevin