[GRASS-user] Re: understanding r.watershed

[Georg Kaspar]

> ... in the basins output I assume. If you really want to treat lakes as
> real depressions, i.e. water is not supposed to leave these lakes, then
> there will always be NULL areas around the lakes unless you set the
> threshold to something much smaller than 100, but then the stream
> segments and basins become meaningless because there are too many...

ok, so I only need to specify depressions that do not loose water!? 
Artificial depressions like postholes or pits for example!? What would be 
the difference between purging the DEM from those depressions and 
providing them as binary input for r.watershed?