[GRASS-dev] weird bug(?) with r.mapcalc neighborhood modifier

GRASS GIS GRASS GIS developer
1

Hi,

I'm not sure if this is a bug or just some feature of r.mapcalc that I don't really understand, so I'm posting here, not yet on track.

I'm working on a module that calculate Moran's I for raster segments (i.e. clumps). For this I need to identify neighboring regions. I do that on a pixel-base, identifying in each direction of a given pixel those neighbors that have a different value than the current pixel.

Here's an example of what I do, identifying neighbors lying below the current pixel:

g.region n=224480.25 s=223910.25 w=635623 e=636449.5 res=28.5
r.mapcalc "nd = if(landclass96[1,0]!=landclass96, landclass96[1,0], null())"
r.stats -1n landclass96,nd sep=comma --q | awk -F, '{if($1==$2) print $0}'
5,5
1,1
5,5

This r.stats result should not happen, IIUC: as I explicitly test for inequality in the r.mapcalc call, pixels should not have the same values in landclass96 and nd.

The weird thing is this:

r.mapcalc "nd_temp = landclass96[1,0]"
r.mapcalc "nd2 = if(nd_temp!=landclass96, nd_temp, null())"
r.stats -1n landclass96,nd2 sep=comma --q | awk -F, '{if($1==$2) print $0}'
[empty result]

i.e. by decomposing the r.mapcalc call into two separate calls, first identifying the below neighbor value of each pixel, then only keeping values for those pixels where that below neighbor value is different, everything works as expected.

I'm I not using neighborhood modifiers as I should, or is this a bug ?

Moritz

2

Nice catch, Moritz! There is a bug in neighbour modifier of mapcalc
(tested with 7.1 and 7.0).

I reported it as: https://trac.osgeo.org/grass/ticket/2917

A further analysis is required to find exact cause and all affected versions.

Māris.

2016-02-13 16:16 GMT+02:00 Moritz Lennert <mlennert@club.worldonline.be>:

Hi,

I'm not sure if this is a bug or just some feature of r.mapcalc that I don't
really understand, so I'm posting here, not yet on track.

I'm working on a module that calculate Moran's I for raster segments (i.e.
clumps). For this I need to identify neighboring regions. I do that on a
pixel-base, identifying in each direction of a given pixel those neighbors
that have a different value than the current pixel.

Here's an example of what I do, identifying neighbors lying below the
current pixel:

g.region n=224480.25 s=223910.25 w=635623 e=636449.5 res=28.5
r.mapcalc "nd = if(landclass96[1,0]!=landclass96, landclass96[1,0], null())"
r.stats -1n landclass96,nd sep=comma --q | awk -F, '{if($1==$2) print $0}'
5,5
5,5
1,1
1,1
5,5

This r.stats result should not happen, IIUC: as I explicitly test for
inequality in the r.mapcalc call, pixels should not have the same values in
landclass96 and nd.

The weird thing is this:

r.mapcalc "nd_temp = landclass96[1,0]"
r.mapcalc "nd2 = if(nd_temp!=landclass96, nd_temp, null())"
r.stats -1n landclass96,nd2 sep=comma --q | awk -F, '{if($1==$2) print $0}'
[empty result]

i.e. by decomposing the r.mapcalc call into two separate calls, first
identifying the below neighbor value of each pixel, then only keeping values
for those pixels where that below neighbor value is different, everything
works as expected.

I'm I not using neighborhood modifiers as I should, or is this a bug ?

Moritz
_______________________________________________
grass-dev mailing list
grass-dev@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-dev