[GRASS-user] ogr2ogr and ortho/rgb display

[Dylan Beaudette]

On 1/6/07, Maciej Sieczka <tutey at o2.pl> wrote:
> Dylan Beaudette wrote:
>
> > be sure to project the DEM data with gdalwarp, and then import with
> > r.in.gdal . Note that DEM projection requires some planning if you will be
> > performing any flow-related analysis. If you are looking for a 'smooth' DEM
> > then I would recommend looking into the "-rcs' parameter to gdalwarp .
>
> Daniel,
>
> Please note that reprojecting a floating point grid will usually yield
> "stairs" distortion in the output, depending on how big is the skew
> between the input and output projection. That's propably what Dylan
> meant. The "stairs" distortion will be the bigger, the less smoothing
> involved in the reprojection.
>
> Although aggressive smoothing like "gdalwarp -rcs" will reduce most of
> the distortion, it will also filter out the details from the input DEM
> and alter the original value range (min will go higher, max will go
> lower). To avoid both shortcomings ("stairs" and over-smoothing) I
> usually transform the grid into vector points, reproject the points and
> interpolate a grid from them, with a suitable algorithm (there are few
> in GRASS; for regularly spaced points input RST or IDW do fine).
>
> Maciek
>

Right, thanks for the clarification Maciek. For a rather crude
comparison between several reprojection methods used on DEM data see
the following link. note that this page contains large images, and
will require a bit of time to load. hover the mouse over the small
images for a detailed view.

http://169.237.35.250/~dylan/gdalwarp/image_matrix.html


As Maciek pointed out, converting the DEM to points, projecting, and
then re-interpolating will give  both smooth and hydrologically
reasonable (assumming good input data) results. r.proj method=cubic
works fairly well too, albiet with some smoothing. Also note that
projection-induced artifacts are quite apparent in curvature maps
created from a projected DEM.

In terms of changing the univariate statistics for a given DEM, I have
never directly compared these stats with respect to the projection
method. Perhaps Maciek can provide links to papers, or some empirical
data; I know several on this list may be interested :) .

Cheers,

Dylan