[GRASS-dev] Re: grass-dev Digest, Vol 33, Issue 37

[Markus Metz]

Helena Mitasova wrote:
>
> On Jan 17, 2009, at 12:07 AM, Michael Barton wrote:
>
>> Markus,
>>
>> I compiled last night and was able to try this on my Mac this 
>> evening. No problem at all with the North Carolina elev_lid792_1m 
>> DEM. Everything ran very fast, of course--even though it runs at 
>> 32bit on the Mac.
>>
>> As expected, the MFD results look much more realistic than the SFD 
>> results. Much less linearity and more sinuosity. I can post the 
>> displays if anyone wants to see them.
>>
>> No errors from the command.
>>
>> No errors in compiling.
>>
>> Seems to work fine on the Mac.
>>
>> I'm not sure why this can't be backported to develbranch_6 since 
>> 6.4RCx is already out. Does it change any of the r.watershed 
>> arguments or behavior using the pre-MFD arguments?
Yes, MFD is default, so using the same arguments produces different 
results. See also my reply to Helena:
http://lists.osgeo.org/pipermail/grass-dev/2009-January/041965.html

I would prefer to stay out of the discussion about backporting and leave 
this decision to the long-time developers.
>
> Michael
> - it needs to be thoroughly tested on wide range of data
> before replacing it. I am trying to get to it too and Markus M has 
> done already quite a bit of testing himself. If you can ask Isaac 
> and/or whoever has grass7 and is around to test it with different data
> (wide range of resolutions, data sources, combinations of parameters, 
> integer or float DEMs, latlong, different size and type of depressions 
> etc.) that would help.
I fully agree. BTW (repeating myself), the results improve a lot if 
float DEMs are multiplied with 100 or 1000 before used as input.

Markus M