[Live-demo] Natural Earth in 4.5b2

[Hamish]

Nathaniel wrote:
> I'm not seeing the any geometry errors in ArcGIS (the horror,
> the horror) on the 10m lakes theme fresh off the website.

Arc is not a topological GIS so it misses (or ignores) many
errors in the data which cause problems when the data is used
in other ways which demand topological correctness.
([river] network analysis, etc)
I understand that Natural Earth is mainly focused on cartographic
needs, but ...


> The 1.3 version that was released a couple weeks ago should have
> resolved most the obvious topology errors on the themes that
> were ID'd to me before. If you want to run a custom repair
> geometry / topology script on them as they're imported to the
> LiveDemo system, sounds fine to me.

Loading a fresh download of 10m-lakes.zip into GRASS GIS (which
is fully topological) exposes a number of errors in the data.
Probably most are due to overlapping areas; small sliver areas
resulting from those after "flattening" are appear below as
areas without centroids (no attributes).
 
GRASS65> v.in.ogr dsn=natearth/10m_lakes.shp out=ne_10m_lakes
[...]
Building topology for vector map <ne_10m_lakes_tmp>...
Building areas...
 100%
1708 areas built
1472 isles built
Attaching islands...
 100%
Number of nodes: 2627
Number of primitives: 4691
Number of points: 0
Number of lines: 0
Number of boundaries: 4691
Number of centroids: 0
Number of areas: 1708
Number of isles: 1472
Number of areas without centroid: 1708
-----------------------------------------------------
Find centroids for layer: 10m_lakes
-----------------------------------------------------
Write centroids:
 100%
WARNING: 82 areas represent more (overlapping) features, because polygons
         overlap in input layer(s). Such areas are linked to more than 1
         row in attribute table. The number of features for those areas is
         stored as category in layer 2
-----------------------------------------------------
1358 input polygons
Total area: 1.26155E+12 (1708 areas)
Overlapping area: 2.02361E+11 (82 areas)
Area without category: 2.39549E+10 (240 areas)
-----------------------------------------------------
Building topology for vector map <ne_10m_lakes>...
Registering primitives...
3489 primitives registered
138259 vertices registered
Building areas...
 100%
1708 areas built
1472 isles built
Attaching islands...
 100%
Attaching centroids...
 100%
Number of nodes: 3253
Number of primitives: 3489
Number of points: 0
Number of lines: 0
Number of boundaries: 2021
Number of centroids: 1468
Number of areas: 1708
Number of isles: 1472
Number of centroids outside area: 3
Number of duplicate centroids: 5
Number of areas without centroid: 248

#extract problems to another layer:
GRASS65> v.extract in=ne_10m_lakes out=ne_10m_lakes_problems layer=2

displaying that extract shows problems are all in the northern
hemisphere, the Great Lakes of N.America being the biggest
pile of red ink. (Lakes Ontario and Erie look ok though)

a map layer created with "v.clean error=problem_areas_map" can
help quickly pinpoint where the problems are.

e.g. an easy place to start looking are the finger lakes within
this bounding box, all of which show errors:

north:      43N
south:      42:18N
west:       77:24W
east:       76:12W

(south of Rochester-Syracuse, NY)


best,
Hamish