[GRASSLIST:3250] Re: Rastering

[Erin ODoherty/RMRS/USDAFS]

Sjors,
I think you are right that it has to do with scale and resolution.  One
thing that helps visualize this is to zoom in on a small spot of your
raster and then superimpose the vector and a grid (with d.grid if your
region boundaries start on whole units equal to your resolution) to see
where the gaps are and why.   If the digitized polygon takes up less than
1/2 of a grid square it does not show up.  If the feature is less than the
resolution, you can not deal with it.

I have had similar needs and problems dealing narrow polygons, in my case
with roads and narrow spots along meadows.  I digitized them as areas
because I wanted a raster with representative road widths and I want my map
to reflect the connectivity.  I tried finer resolutions, eventually it
works, but that is a pain, I don't really want to work with a 2m x 2m
raster.  One compromise is to make the features labeled lines in v.digit,
then rasterize those vectors and r.patch the narrow raster on the other
map.  Often the only connectivity is at the corners, so it's not really one
continuous polygon, but it's better than nothing.  Also, of course, your
feature is exaggerated, but depending on your question, that may not be a
big problem.  I believe the only alternative is to work with a finer
resolution, or vectors.  (I used the latter, when trying to calculate
landscape metrics.)
Good luck.

Erin O'Doherty
Laramie, WY, USA



                                                                                             
                    "J.c.m. van der                                                          
                    Kwast"                  To:     grasslist at baylor.edu                     
                    <jkwast at home.nl>        cc:                                              
                    Sent by:                Subject:     [GRASSLIST:3247] Re: Rastering      
                    owner-GRASSLIST@                                                         
                    baylor.edu                                                               
                                                                                             
                                                                                             
                    02/27/02 09:20                                                           
                    AM                                                                       
                                                                                             
                                                                                             




Hi Karine,

Yes, I've checked region settings, there's no problem at all. I ddin't
mention it before, but I've also tried to buffer the cells (r.grow (1
cell) and r.buffer, and with mapcalc tried to define them as areas again
( in mapcalc use some like newfile = file + buffered_file). This does'nt
seem to work as well, unless the buffer is like 5 m which would
exaggerate the features enormously. Also the neighborhood module, has
been tried. Same story, you must exaggerate enormously (e.g. buffer 9
cells up to 13 cells). I've been trying afterwards to thin the map with
r.thin leaving me the result I've started out with. I think the problem
has more to do with scale and resolution. Well I'll keep on trying this
and that, thanks anyway,

Greetings,

Sjors




Karine ZUERCHER wrote:

>Sjors,
>
>Have you checked the region settings (g.region -p) just before doing the
>v.to.rast?
>You should have the nsres & ewres equal to 1 &  rows & cols 5000.
>
>Hope that may help
>In Kindness
>
>Karine
>
>On Wednesday 27 February 2002 15:43, J.c.m. van der Kwast wrote:
>
>>Hi everybody,
>>
>>For some time now I'm wrestling with the rasterization of thin
>>linear/area objects. The case is as follows, xy projection, area 5000 x
>>5000 m (or units), grid 1 x1. I've got thin areas of 3 m wide and trying
>>to raster them. I would expect to see them as thin linear objects in a
>>raster map of the whole area. But the areas are broken. When zooming in
>>extremely then the areas seem to form lines or areas again, but are they
>>continuous? Maybe it has something to do with the scale?
>>When given a z value and patched with a dem, the dem is correctly
>>displayed but the thin areas are broken into pieces (as they are in the
>>whole raster map). Does it mean that I should divide the map up into
>>small portions with a desirable resolution, raster them and patch them
>>back together in the desired resolution? Personnally I don't think it
>>will help as the areas formed do not seem to be continuous.
>>Some time ago I wrote a mail with the title landscape. The problem is
>>much the same, only with rastering linear features. As my background is
>>not GIS (but agriculture) it could be I'm trying to do something which
>>isn't possible at all. But looking at it logically, a feature of 1 cell
>>in a grid of 1 x1 should at least take the value I've given it. If it
>>doesn't show up at the screen I can live with it. Problem is, it does
>>sometimes will and sometimes won't show up.
>>I've been playing around with a lot of functions and manuals but still
>>this puzzles me greatly. If anyone has any answers, I'd be most grateful,
>>
>>Warm regards,
>>
>>Sjors
>>
>