[GRASS-dev] r.sim.water example

[Helena Mitasova]

Soeren,

I am taking this out of bug tracker as these comments are not directly
relevant to the r.sim.water problem on Windows.

>Helena can you please provide a simple spearfish example for
>r.sim.sediment, so i can test my changes with this module?

I don't use spearfish anymore but here are couple ready to use examples
for the nc_spm_08 data set:
http://courses.ncsu.edu/mea582/lec/001/GIS_anal_grass/GIS_Anal_grsimwe.html

the data set is on the GRASS website and also here:
http://courses.ncsu.edu/mea582/common/media/01/nc_spm_08.zip
and the result of the first example should look something like this:
http://skagit.meas.ncsu.edu/~helena/wrriwork/lakewh/flowsimwedem.jpg

>BTW: Is it possible to add the water depth in every step to the elevation
>map and additionally  computing the direction derivatives dx and dy to
>avoid infinite water accumulation in sinks and pits? I suggest this will
>make the solution of the partial differential equation non linear, but
>maybe the time steps are already small enough to implement this as an
>explicit step?

the diffusion term prevents infinite water accumulation in deperessions,
the water depth parameter hmax controls when the diffusion term increases
to get water out of pits, how much it increases is controled by halpha.
So water will flow out of depressions and over dams (see the video below)
but it could still grow too fast and create spikes or bumps so it is not
the real diffusive wave. If you are getting spikes in depressions
for Spearfish you need to lower the value of hmax and increase halpha.
But you are right on target that the elevation should be updated with water level 
and new gradient computed so that gradient of surface water is used for routing.
That would allow the depressions to be filled more accurately without artificial spikes.
The time step is estimated in such a way that for each time step the particle 
does not travel through more than one cell so the water depth can be added and 
gradient computed as an explict step. I did some experiments in that direction -
the gradient needs to be computed from smooth enough approximation
to prevent artifacts that would get amplified after each time step.

If you could help implement the water depth updates that would be fantastic,
let me know. 
We use simwe a lot for the Tangible GIS project - see e.g., here:
http://courses.ncsu.edu/mea582/lec/001/GIS_anal_lecture/tg1bak2_ed4_1min640.mov

Thanks for the update to grass7,

Helena


Helena Mitasova
Associate Professor
Department of Marine, Earth and Atmospheric Sciences
North Carolina State University
Raleigh, NC 27695
hmitaso at unity.ncsu.edu



On Jan 27, 2010, at 5:55 PM, GRASS GIS wrote:

> #617: r.sim.water crashes on WinGrass
> ---------------------------+------------------------------------------------
>  Reporter:  hamish        |       Owner:  grass-dev at lists.osgeo.org
>      Type:  defect        |      Status:  new                      
>  Priority:  normal        |   Milestone:  6.4.0                    
> Component:  Raster        |     Version:  6.4.0 RCs                
> Resolution:                |    Keywords:  wingrass, r.sim.water    
>  Platform:  MSWindows XP  |         Cpu:  x86-32                   
> ---------------------------+------------------------------------------------
> Comment (by huhabla):
> 
> I have hardly tried to find the error using valgrind, but was
> unsuccessful.
> Neither the grass6.4 version nor the grass7 version showing any suspicious
> behavior on 32bit Linux.
> So i don't know if my changes may have solved the problem, but it would be
> great, if anybody may
> compile the latest grass7 version on MacOS or Windows for testing, running
> the speafish60 example.
> 
> Helena can you please provide a simple spearfish example for
> r.sim.sediment, so i can test my changes with this module?
> 
> BTW: Is it possible to add the water depth in every step to the elevation
> map and additionally  computing the direction derivatives dx and dy to
> avoid infinite water accumulation in sinks and pits? I suggest this will
> make the solution of the partial differential equation non linear, but
> maybe the time steps are already small enough to implement this as an
> explicit step?
> 
> Soeren
> 
> -- 
> Ticket URL: <https://trac.osgeo.org/grass/ticket/617#comment:5>
> GRASS GIS <http://grass.osgeo.org>