[GRASS-dev] r.resamp.stats: nulls along western boundary

GRASS GRASS developer
1

Hi,

While running through this set of commands to resample from 30" to 2':
  http://grass.osgeo.org/wiki/Blue_Marble#Processing

I notice that the output of r.resamp.stats has its western-most raster
column set to all NULL. The other 3 borders seem to be ok. As the 30" data is without NULLs, and the target region is an exact multiple of the source region, it doesn't seem to be an inherent exceeding-the-border
effect. The propagate nulls flag was not used.

Besides losing data, when you zoom to look at the pacific it makes an
ugly white line along 180W.

?
Hamish

2

Hamish wrote:

While running through this set of commands to resample from 30" to 2':
  http://grass.osgeo.org/wiki/Blue_Marble#Processing

I notice that the output of r.resamp.stats has its western-most raster
column set to all NULL. The other 3 borders seem to be ok. As the 30"
data is without NULLs, and the target region is an exact multiple of
the source region, it doesn't seem to be an inherent
exceeding-the-border effect. The propagate nulls flag was not used.

Can you come up with a recipe to reproduce this with the standard
example data sets?

--
Glynn Clements <glynn@gclements.plus.com>

3

Hamish:

> While running through this set of commands to resample from 30" to 2':
> http://grass.osgeo.org/wiki/Blue_Marble#Processing
>
> I notice that the output of r.resamp.stats has its western-most raster
> column set to all NULL. The other 3 borders seem to be ok. As the 30"
> data is without NULLs, and the target region is an exact multiple of
> the source region, it doesn't seem to be an inherent
> exceeding-the-border effect. The propagate nulls flag was not used.

Glynn:

Can you come up with a recipe to reproduce this with the
standard example data sets?

sure:

# lat/lon location
g.region n=90N s=90S w=180W e=180E res=0:15
r.mapcalc fifteen=15
g.region res=1
r.resamp.stats in=fifteen out=fifteen.one
r.univar fifteen.one
  ...
  total null cells: 180

g.region w=0 e=360 # center on 180 lon
d.rast fifteen.one
d.grid 90 -b color=black

Hamish

4

Hamish wrote:

> > While running through this set of commands to resample from 30" to 2':
> > http://grass.osgeo.org/wiki/Blue_Marble#Processing
> >
> > I notice that the output of r.resamp.stats has its western-most raster
> > column set to all NULL. The other 3 borders seem to be ok. As the 30"
> > data is without NULLs, and the target region is an exact multiple of
> > the source region, it doesn't seem to be an inherent
> > exceeding-the-border effect. The propagate nulls flag was not used.

Glynn:
> Can you come up with a recipe to reproduce this with the
> standard example data sets?

sure:

# lat/lon location
g.region n=90N s=90S w=180W e=180E res=0:15
r.mapcalc fifteen=15
g.region res=1
r.resamp.stats in=fifteen out=fifteen.one

Okay, I can reproduce this.

Unfortunately, this looks like it's going to be a pig to fix.

BTW, I consider this a bug:

  > print G_col_to_easting(0, &dst_w)
  $20 = -180
  > print G_easting_to_col(-180, &src_w)
  $21 = 1440
  > print src_w.cols
  $22 = 1440

--
Glynn Clements <glynn@gclements.plus.com>