[GRASSLIST:6531] resolution

GRASS GRASS user
1

Hi

I'm wondering why does it happen, that when the resolution is decreased (even if very slightly), some separate pixels dissapear on display. And when I continue decreasing res they re-appear, but some other dissapear again. What's more, it doesn't reffer to display only - also when 'r.mapcalc' is used the resulting layers miss the same pixels as on the display. Badly puzzling. I would rather expect these pixels only to grow but not to get gone for good. See attachments.

            figure "res10.png":

GRASS 6.1.cvs (kpn92):~ > g.region -p
projection: 99 (Transverse Mercator)
zone: 0
datum: etrs89
ellipsoid: grs80
north: 331740
south: 330480
west: 259580
east: 260800
nsres: 10
ewres: 10
rows: 126
cols: 122
GRASS 6.1.cvs (kpn92):~ > d.rast rozn2
100%

            figure "res12.png":

GRASS 6.1.cvs (kpn92):~ > g.region res=12 -ap
projection: 99 (Transverse Mercator)
zone: 0
datum: etrs89
ellipsoid: grs80
north: 331752
south: 330468
west: 259548
east: 260820
nsres: 12
ewres: 12
rows: 107
cols: 106
GRASS 6.1.cvs (kpn92):~ > d.rast rozn2
100%

            figure "res13.png"

GRASS 6.1.cvs (kpn92):~ > g.region res=13 -ap
projection: 99 (Transverse Mercator)
zone: 0
datum: etrs89
ellipsoid: grs80
north: 331760
south: 330460
west: 259545
east: 260832
nsres: 13
ewres: 13
rows: 100
cols: 99
GRASS 6.1.cvs (kpn92):~ > d.rast rozn2
100%

Maciek

(attachments)

res13.png
res10.png
res12.png

2

Maciek Sieczka wrote:

I'm wondering why does it happen, that when the resolution is decreased
(even if very slightly), some separate pixels dissapear on display. And when
I continue decreasing res they re-appear, but some other dissapear again.
What's more, it doesn't reffer to display only - also when 'r.mapcalc' is
used the resulting layers miss the same pixels as on the display. Badly
puzzling. I would rather expect these pixels only to grow but not to get
gone for good. See attachments.

Decreasing the region resolution (increasing the cell size) results in
fewer samples so the probability that at least one sample point will
fall in any given map cell decreases, so the probability that no
sample point will fall in any given map cell increases.

E.g. if the underlying map has 100x100 cells, the region also has
100x100 cells, and the region boundaries match the map's boundaries,
there will be a 1-to-1 correspondence between map cells and region
cells. If the region resolution is then reduced to e.g. 90x90, you
have 10000 map cells but only 8100 region cells, so not all of the map
cells will survive the resampling.

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