Hi,
I'd like to ask a question related to another question I was asking
recently about converting contour vectors to rasters and creating a
DEM from the raster contours using r.surf.contour.
My supplied vector contour layer has 20m intervals between contour
lines. When I convert this to raster, the smallest region resolution I
can use is 5m, otherwise v.to.rast runs out of memory.
When I use r.surf.contour to create the DEM from the raster, which
region resolution makes sense then? I set it to 20m because of the
contour interval, but I'm starting to think that's not the best option
because the interpolation between two contour lines should be as fine
as possible.
Any thoughts?
Thanks
Hanlie
On Wed, May 4, 2011 at 10:48 AM, Hanlie Pretorius
<hanlie.pretorius@gmail.com> wrote:
Hi,
I'd like to ask a question related to another question I was asking
recently about converting contour vectors to rasters and creating a
DEM from the raster contours using r.surf.contour.
My supplied vector contour layer has 20m intervals between contour
lines. When I convert this to raster, the smallest region resolution I
can use is 5m, otherwise v.to.rast runs out of memory.
Please post
- GRASS version/OS
- g.region -p output
to better understand the problem.
Markus
When I use r.surf.contour to create the DEM from the raster, which
region resolution makes sense then? I set it to 20m because of the
contour interval, but I'm starting to think that's not the best option
because the interpolation between two contour lines should be as fine
as possible.
Any thoughts?
Thanks
Hanlie
GRASS 6.4.1 on Windows XP.
g.region output for 5m resolution:
g.region -p vect=contours_20m@PERMANENT nsres=5 ewres=5
projection: 99 (Transverse Mercator)
zone: 0
datum: ** unknown (default: WGS84) **
ellipsoid: wgs84
north: -3121229.69982072
south: -3156672.30728643
west: -67483.27873573
east: -32474.90929218
nsres: 4.9996625
ewres: 4.99976713
rows: 7089
cols: 7002
cells: 49637178
g.region output for 20m resolution:
g.region -p vect=contours_20m@PERMANENT nsres=20 ewres=20
projection: 99 (Transverse Mercator)
zone: 0
datum: ** unknown (default: WGS84) **
ellipsoid: wgs84
north: -3121229.69982072
south: -3156672.30728643
west: -67483.27873573
east: -32474.90929218
nsres: 20.00147148
ewres: 20.00478254
rows: 1772
cols: 1750
cells: 3101000
Thanks
Hanlie
2011/5/6, Markus Neteler <neteler@osgeo.org>:
On Wed, May 4, 2011 at 10:48 AM, Hanlie Pretorius
<hanlie.pretorius@gmail.com> wrote:
Hi,
I'd like to ask a question related to another question I was asking
recently about converting contour vectors to rasters and creating a
DEM from the raster contours using r.surf.contour.
My supplied vector contour layer has 20m intervals between contour
lines. When I convert this to raster, the smallest region resolution I
can use is 5m, otherwise v.to.rast runs out of memory.
Please post
- GRASS version/OS
- g.region -p output
to better understand the problem.
Markus
When I use r.surf.contour to create the DEM from the raster, which
region resolution makes sense then? I set it to 20m because of the
contour interval, but I'm starting to think that's not the best option
because the interpolation between two contour lines should be as fine
as possible.
Any thoughts?
Thanks
Hanlie
Interpolation's kind of a tough issue. The horizontal resolution shouldn't
necessarily be dependent on the contour step intervals, because those are
two unconnected parameters.
In principle, interpolation will "fill the gaps" between given values. Since
the gap filling values are always guesses - albeit educated guesses - the
resolution is kind of a matter of taste, in my opinion. I'd say that the
method chosen is much more important - e.g. r.surf.contour interpolates
linearly, whereas v.surf.rst uses spline with tension.
So... Lange Rede, kurzer Sinn, I'd have a hard time telling you what
resolution to take for an interpolated DEM. Factors I'd consider would be:
1. How closely spaced are the contour lines? a.k.a. if you were to translate
the surveying coverage of the vectors to horizontal resolution, what would
you guess - how exact is the spatial data you're dealing with as inputs?
2. What scale do you need for your analysis?
As far as the problem with limited memory goes, perhaps the thing to do
would be to split the area up into smaller tiles and then perform the
interpolation for them all recursively and stitching together again in the
end?
In that case I'd make the tiles a little bigger than necessary and let them
overlap, and then throw away the edges when I stitched everything back
together again. Interpolated values should be cropped at the edges, because
on those sides there's no data available to interpolate into so you get
unrealistic results.
--
View this message in context: http://osgeo-org.1803224.n2.nabble.com/DEM-resolution-from-r-surf-contour-tp6330087p6337007.html
Sent from the Grass - Users mailing list archive at Nabble.com.