[GRASS5] [bug #4248] (grass) r.surf.contour: Treats 0 as NULL and other archaic problems

[Maciek Sieczka via RT]

Hamish wrote:

> How does nnbathy outpt differ from a rasterized TIN?

It is a smooth surface, no sign of a triangles. I know traingulation and
natural neighbor have something in common, but they are not exactly the same
thing.

> For that matter, how does r.surf.contour differ from a rasterized TIN?

As to my experience, it does. You can check for yourself - SAGA gis can
produce a rasterized TIN. r.surf.contours flooding algorithm is different from
triangulation.

> What other options?
> idw, idw2,

As to my experience, IDW interpolation is no any good for clusterized and 
sparse input data like in case of elevation contour lines. Also, it is not
able to handle rapid gradients well.

> v.krige (add-ons), and I imagine numerous other Kriging and

I only played with krigging in Surfer a bit, and was not satisfied with the
resultant DEM - grainy surface, rapid gradients lead to errors in surface.

> other varieties from the R-statistics interface.

Dunno.

But one thing more. What I like abouth nnbathy is that the result is easy to
foresee - what I intuitevily expect is mostly what I get. No fiddling around.
It is not trying to make the surface fit some maths function - it just follows
the data. This makes it possible to interpolate over a highly diverse surface
and still preserve an ideally flat floodland and a 100 m high vertical wall in
the center of it, without artifacts.

nnbathy will connect distant input data with a straight, expected, continous
surface - no brakes (IDW), segmentaion (RST) or artifacts like a hollow or a
bulge over a supposedly flat area (RST,IDW) - unless you tell it to do it by
feeding appropropiate input into it.

All pretty much like triangulation, but better because the output grid is
continous and smooth.

For contour lines it's best of what I've seen.

Maciek


-------------------------------------------- Managed by Request Tracker