[GRASS-user] Whats happens with r.neighbors at boundaries with NULL cells

[Bernardo Santos]

Dear list,
I have been using r.neighbors for many procedures in some applications lately, but I am not sure how it deals with NULL data at the boundaries of the input map, and if there are option on how to deal with them.
In some examples I ran for, say, a window of size = 1 pixel, I 'lost' one pixel at all boundaries in the output file - they had information in the input but were NULL in the output. However, it seems to me that this should not happen, since it is written in the manual that 'r.neighbors doesn't propagate NULLs, but computes the aggregate over the non-NULL cells in the neighborhood.'
In other examples I ran, I had an opposite result. Not only I had information for all the non-NULL cells of the input raster, I also had information in 1 pixel around it, in pixels that were originally NULL in the input map.And when I use the option selection=input_raster, then I exclude this information and have only information for the original non-NULL cells of the input map, in the output raster.
Anyway, maybe I couldn't test the command properly, but these variation is generating some issues later on, when I use these maps for other purposes. Could anyone help clarify that?
I can provide some maps for testing, if needed.Best,
Bernardo Niebuhr
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/grass-user/attachments/20180413/f8beb5dc/attachment.html>