[GRASSLIST:5172] Re: resolution problems

[Glynn Clements]

bfdi533 wrote:

> I have a mapset that is a global lat/long region.  I have created some raster
> maps from sites.  Eash site has 1 point in it for a region of interest.  When
> the raster maps were created, they were created at a resolution of 00:01:00.  I
> then take these raster maps and create a poly from them and then convert these
> to vector using the procedure here:
> 
> 
> r.buffer units=miles input=country_us output=buffer1 distances=983.788895393345
> r.poly -l -b input=buffer1 output=buffer1
> ..
> (repeats 7 more times for buffer2 - buffer8)
> ..
> v.patch input=buffer1,buffer2,buffer3,buffer4,buffer5,buffer6,buffer7,buffer8
> ou
> tput=buffer_2002-11-26_top10
> v.support buffer_2002-11-26_top10
> 
> The problem is that this takes a very long time and since the output will be a
> postscript printout on a letter-size page, the resolution does not need to be
> this big.
> 
> It was my understanding that if I change the resolution then the maps created
> would be resampled as needed.  However, when I change the resolution to
> 00:05:00 the commands run very quickly but produce blank output layers.  There
> is nothing in any of the bufferx files nor is there anything in the vector
> files or even the patched output.

You need to decrease the resolution *before* converting the sites to
rasters. If you create the original rasters at 1 minute resolution,
then change the resolution so that they are resampled at 5 minute
resolution, the resampled version will only include 1 cell out of
every 25 cells in the orignal raster (the resampled version will
consist of every fifth column of every fifth row).

Statistically, you would expect to see 1/25th of the sites; however,
if the site coordinates are actually rounded to a precision which
isn't substantially smaller than the resampling resolution, you may
find that they always fall in the rows or columns which are discarded.

Basically, the "resampling" process is exactly that: a discrete
sampling of the base map, with no aggregation.

-- 
Glynn Clements <glynn.clements at virgin.net>