[GRASS-user] edges in basin map from r.watershed

[Markus Metz]

Christian Schwartze wrote:
> Dear GRASS users,
>
> with r.watershed I get strange basin boundaries for some areas und I'm not able
> to give account of it. Attached you can find that part of the basin map which
> looks curiously. I means the sharp-edged regions...
> Whats the reason?
>   
This is most probably a flat area (no slope). Flow direction, flow 
accumulation, stream segments and basins can not reasonably be 
calculated for flat areas, these are regarded as missing information and 
some assumption has been made by the algorithm.
What could help is to use a raster DEM as input that is *not* filled, 
some would say not hydrologically correct, but r.watershed works better 
with the raw, not filled DEM.
What could also help, if this does not work or it really is a flat area, 
is r.watershed of grass7 with multiple flow direction. Note that the 
result may look nicer, but it still holds true that drainage direction 
(and therefore all other output) has to be estimated for flat areas. The 
A * Search of r.watershed is doing a pretty good job, and multiple flow 
accumulation can improve it a bit more, within limits.
> Basis is an Arc Info .adf raster file for DEM data.
>   
I think Arc Info wants a depressionless, filled, hydrologically correct 
DEM. r.watershed does explicitely not want such a DEM, it wants a raw 
DEM with depressions, not filled, and not hydrologically correct.

I hope that helps,

Markus M