[GRASSLIST:5462] Re: r.watershed

[Markus Neteler]

On Mon, Feb 03, 2003 at 09:32:19PM -0500, Gustavo Alcides Concheiro Perez wrote:
> Hi all!
> 
> I am having terrible problems with r.watersheds.  In short: I can't
> get it to work!  I am trying to estimate the confluence of present-day
> basins if the sea level was lower.  I'm using the ETOPO2 DEM
> restricted to Panamá.  The program seems to run, until it reaches step
> 6 (``closing maps'').  Then it goes on forever---I've let it run for
> days, and always had to abort it.  The machine is not so fast: PIII
> running Linux.
> 
> Is there any way to fix this?  Is there any parameter which value
> would cause this?  How long is the program supposed to run?  Can it
> really be ``closing maps'' for 10 days?

Yes, that is possible. r.watershed does not work well with large
data sets.

There is a replacement in the CVS: r.terraflow.
http://www.cs.duke.edu/geo*/terraflow/

You either have to compile yourself or wait for 5.0.2 (AFAIK it will
be included) [1].

Best regards

 Markus Neteler

[1] There is an inofficial GNU/Linux binary version of the recent CVS HEAD:
    http://grass.itc.it/pipermail/grass5/2003-January/004647.html
    including r.terraflow (FLOAT version) and r.terraflow.short (Integer
    version).
    Note that this compilation is inofficial and may not be 100% identical
    to the upcoming 5.0.2 release. If you feel confused, please ignore
    this suggestion :-)