[GRASS-user] Buffering large grid maps

[Glynn Clements]

Moritz Lennert wrote:

> >> Thanks for your reply. In fact I need to generate a (uncontroled) distance
> >> map from all my values=1.
> >> I confess that I don´t understood how r.grow will help me on this task.
> >> I need to build a for looping and grow and grow many times the new generated
> >> map?
> >> Or is there a a way of simple generate a distance map on one command?
> > 
> > "r.grow ... in=1 out=2" has the same effect as r.buffer with a single
> > distance, except that r.grow's radius= option is measured in cells
> > while r.buffer's distance= option is measured in physical units (e.g. 
> > metres).
> > 
> > You can get the same effect as r.buffer with multiple distances by
> > running r.grow once for each distance value, then overlaying the
> > results, from largest to smallest buffer size.
> > 
> > r.grow will be quite a bit slower, particularly for large distances,
> > but it doesn't require that the map fits into memory.
> > 
> > Alternatively, you could reduce the resolution, so that the map fits
> > into memory. Or you could run r.buffer on smaller sections of the
> > data, then merge the results with "r.series method=minimum".
> 
> Not sure I understand correctly what you are trying to do, but maybe you 
> could use r.cost on a map with cell value = resolution and with your 
> binary map as starting points (start_rast) ?

Note that r.cost will produce polygonal buffers rather than circles,
and it's also likely to be quite slow.

More significantly, I don't know whether it will handle a map that
large. It relies upon the segment library, which doesn't appear to
have large file support enabled.

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