"buffer" created with DEM analysis - how to avoid it?

Question

I have, to start, a DEM raster and a shapefile, with random created points, within a given polygon. The DEM raster was clipped with that same polygon.

With the DEM raster, I used GRASS to perform neighborhood analysis (using "r.neighbors"), and I applied a buffer of 150m. Since the pixel size is 30x30m, the size I used was 5 (and I did not select the "circular neighborhood" option).

Then, I uploaded the new raster to QGIS, to perform DEM Terrain models analyses (Raster > Analysis > DEM (Terrain Models)). In "Mode" option, I used: Slope, TRI and Roughness.

My final objective is to atribute the value of "Slope", "TRI" and "Roughness" to each point of the shapefile. However, after creating "Slope", "TRI" and "Roughness", a kind of buffer (if it can be called this way) appeared surrounding the limits of the area, as you can see here:

The problem is, that buffer has an extremely high value that is not representative of the area, and some points fall into that buffer, so I obtain erroneous "Slope", "TRI" and "Roughness" values.

Do you have any idea of a solution for this problem?

Answer 1

I think the solution to what you call "buffer" is to create a MASK raster in GRASS. Then the r.neighbors analysis (and all other raster analyses) will be limited to the area covered by the mask. You would do that in GRASS by converting the polygon to a raster, and naming it MASK:

v.to.rast input=<your polygon> output=MASK use=value