[talk-ph] NGA-GNS uploads

[Axel Kollmorgen]

On 2010-09-21 08:32, Anthony G. Balico wrote:
> Axk and I had a bit of a discussion on GNS. Your thoughts please?

thanks anthony for bringing this to the list, and thanks to everyone who 
responded so far.

> I find the thread below so cluttered, im sure you'll also find it
> the same way.  Sorry about that.

let me restate my arguments against NGA-GNS imports then:

* the quality of the NGA-GNS data, especially below the municipal
level, is poor [1]

   + the geographical resolution is often very coarse (two to three 
kilometers from the actual locations)
   + a significant amount of entries is inaccurate or plain wrong
   + names are never removed, so many, many names have no modern 
significance
   + places where people live are generally just classified as "PPL", 
Populated Place. This can be anything from a city to what is now just a 
farm house.

   for a simple data quality analysis, see my 
barangay-shapefile-from-GADM [2] and a google maps [3] overlay. from my 
point of view, more than half of the imported places are wrong (located, 
spelled, classified, non-existent).

[1] http://wiki.openstreetmap.org/wiki/GEOnet_Names_Server#Caveats
[2] http://osm.kollm.org/nga-gns-accuracy.html
[3] 
http://sautter.com/map/?lat=10.93479&lon=124.47475&zoom=13&layers=00B000TFFFFFFF

* the imports give a wrong impression of completeness and accuracy and 
dismotivate mappers in the area

   have a look at [4]. you could think that the NSA-GNS imported part 
(the south-western part below the ormoc-palompon line) is, more or less, 
mapped with the same quality as the part north of ormoc. the truth, 
however, is that every element north of ormoc has been verified, many 
with gps traces, photos, on the ground. with not an inconsiderable effort.

   the effect of an unverified import just beside such an area is left 
to the imagination of the reader :)

[4] http://www.openstreetmap.org/?lat=11.0103&lon=124.5454&zoom=12&layers=M

* the imports complicate further mapping

   the argument for importing possibly bad data is that over time, it 
will be fixed. as anthony put it:

> At some point somebody will find a bad data. Being a mapping wiki,
> that somebody have the opportunity to correct the data. This is no
> different with roads imported from 50k topo maps, most of the time
> these are too off from where they are supposed to be.

   in my experience, this doesn't hold true. fixing old data - be it 
MikeCollinson's initial GNS imports, the SRTM coastlines, or the 50k 
topo maps - is way more complicated than putting the right data there in 
the first place. the problem with existing data is that there is 
hesitation to modify / delete it: because it is not yours, because you 
don't know the source, because you cannot (un)verify it. so the bad old 
data stays there. how often (siquijor, camotes, ...) did i trace a gpx 
trace of mine and suddenly bumped into a FIXME road, which then made me 
think and ponder and adapt my clean trace to this road. for a road which 
in the end probably just wasn't there.

> So that +1 to "No imports for now".

another +1, obviously, from me. if we also count in maning ("I am now 
more on the 'no imports' camp too") with, let's say, +0.5, that makes 
+2.5 for "No NGA-GNS imports for now" :p

anyone else?

ax