r.resamp.interp fails for bicubic and bilinear interpolation

[cmbarton]

I just found out that r.resamp.interp fails for both bicubic and bilinear interpolation. It makes maps with all NAN values. I used the new NC data set and produced a set of random elevation points (raster and vector), then tried to use these in r.resamp.interp. The nearest neighbor routine worked, but the others did not. I tried the original raster point map and also tried one that I converted to a CELL map. Neither gave usable results. FWIW, I got the same result using r.bilinear.

I'm having a very difficult time putting together a class on interpolation using GRASS today. v.surf.bspline does not work, 2 of the 3 methods in r.resamp.interp do not work. r.bilinear does not work. Pretty bad press.

Michael

[neteler]

On Tue, Mar 18, 2008 at 10:01 PM, Helena Mitasova <hmitaso unity.ncsu.edu> wrote:

Michael - unless you have at least on point per cell NAN is actually a correct result. r.resamp.interp is for resampling a raster map not for interpolating from randomly distributed points. Eg this is the example from the book - you go from 30m resolution to 10m resolution - let me know if that does not work (then we are in trouble because this worked so we may have a bug introduced somewhere).

g.region rast=elevation -p r.resamp.interp elev_ned_30m out=elev_ned10m_bil meth=bilinear r.resamp.interp elev_ned_30m out=elev_ned10m_bic meth=bicubic

Helena

New documentation wording:

"<EM>r.resamp.interp</EM> resamples an input raster map by interpolating between the neighboring cells via a selectable resampling algorithm. All cells present in the neighborhood of the input raster cell must be non-null to generate a non-null cell in the output raster map."