[mpas-developers] Vector Reconstruction

Wed Mar 24 14:15:33 MDT 2010

Hi, Todd.

I wonder whether there's something odd with how I've added the new
module_vector_reconstruction.F code and Registry entry. I had been
testing with branches/swmodel for simplicity, and the changes are
really trivial. If you wouldn't mind, could you give a try with
the tar file I've been testing with? I've placed this at
http://www.mmm.ucar.edu/people/duda/files/swmodel_vectors.tar.gz.
The namelist is set up to run just 5 time steps, and the input
files are all included, so it should just be a matter of compiling
and running. Also included in the tar file is the output file that
I generated for SWTC2; doing an ncdump or using ncview, you can
see the large values in uReconstructZ. 

If you're able to get correct results with my tar file, I'd not
object to committing the new module_vector_reconstruction.F; I can
propose to add this in a new shared directory 'operators', to be
created in trunk/mpas/src with appropriate changes to the
Makefiles. However, if you can reproduce the large values of
uReconstructZ with my tar file, I can look for any differences if
you could supply the code that you were able to successfully run
with. Does that sound like a reasonable way forward?

Cheers,
Michael

On Wed, Mar 24, 2010 at 12:06:19PM -0600, Todd Ringler wrote:
> 
> Hi Michael,
> 
> I can not reproduce your finding. Below is the rbf reconstruction  
> (vectors) and z-component of velocity (colors with colorbar, units m/ 
> s). This is SWTC#2 on the quasi-uniform 2562 grid. I am happy to  
> followup over the phone. Have to run, back at 2 pm.
> 
> Cheers,
> Todd
> 

> 
> On Mar 23, 2010, at 11:56 AM, Michael Duda wrote:
> 
> >Hi, Xylar.
> >
> >Thanks for the quick reply. In SWTC2, the maximum zonal velocity
> >is around 38 m/s at the equator, so 10 m/s for uReconstructZ does
> >seem out of place. Just to make sure I understand how to interpret
> >uReconstruct{X,Y,Z}, these are the components of the velocity
> >vector in Cartesian space, with the z-axis positive through the
> >north pole of the sphere, right? I had been testing on a fairly
> >coarse grid (10242 cells), so that could be a factor; I'll do a
> >little more testing at higher resolution.
> >
> >Cheers,
> >Michael
> >
> >
> >On Tue, Mar 23, 2010 at 11:41:18AM -0600, Xylar S. Asay-Davis wrote:
> >>Hi Michael,
> >>
> >>I don't think your attachment made it through.
> >>
> >>I'll take a look at shallow water test case 2 to see what might be  
> >>the
> >>problem there. The value of uReconstructZ won't machine zero for a  
> >>purely
> >>zonal flow because the reconstruction in its current form will only
> >>guarantee exact reconstruct of a constant flow in the tangent plane  
> >>of the
> >>cell, where as a zonal flow has components slightly out of the  
> >>tangent
> >>plane.  In my tests, I was seeing uReconstructZ smaller than  
> >>uReconstructX
> >>and uReconstructY by a factor of around 1e-6.
> >>
> >>-Xylar
> >>
> >>
> >>>Hi, Xylar.
> >>>
> >>>The module looks good to me -- very clean code. The only changes
> >>>I'd suggest are to the comments at the beginning of the
> >>>init_reconstruct() routine, and a few changes to indentation, as
> >>>in the attached module_vector_reconstruction.F.
> >>>
> >>>This is most likely a problem in how I'm performing the
> >>>computation given that both you and Todd have checked the
> >>>reconstructed vectors, but I'm having trouble getting a zonal and
> >>>meridional wind from the reconstructed vector in (x,y,z). Even
> >>>plotting the magnitude of the reconstructed vector, sqrt(x^2 + y^2
> >>>+ z^2), I'm getting some unexpected values in the reconstructed
> >>>wind speed field for shallow water test case 2 (with alpha=0).
> >>>Given that we should have purely zonal flow for this test case,
> >>>I'm suprised to see values in uReconstructZ of order 10. Do either
> >>>you or Todd have any thoughts or suggestions?
> >>>
> >>>Cheers,
> >>>Michael
> >>>
> >>>
> >>>On Mon, Mar 22, 2010 at 03:30:39PM -0600, Xylar S. Asay-Davis wrote:
> >>>>Attached is my proposed vector reconstruction code.  I have run  
> >>>>several
> >>>>sanity checks and it seems to work fine.  However, I haven't been  
> >>>>able
> >>>>to
> >>>>get a full ocean core run with this code.  It is not clear to me  
> >>>>if this
> >>>>has to do with this code or with recent changes that Mark has  
> >>>>made to
> >>>>the
> >>>>ocean core relating to 3D density.  I'll keep working with Mark  
> >>>>to try
> >>>>to
> >>>>figure that out, but in the mean time it would be great to have  
> >>>>Todd
> >>>>and/or Michael test out this code to see if it will work for you.
> >>>>
> >>>>A registry entry for the coeffs_reconst is required as follows  
> >>>>(already
> >>>>in
> >>>>the ocean core):
> >>>>var real    coeffs_reconstruct ( R3 maxEdges nCells ) -
> >>>>coeffs_reconstruct
> >>>>- -
> >>>>
> >>>>Registry entries for  matrix_reconstruct, normal and  rbf_value  
> >>>>are no
> >>>>longer required by this code.
> >>>>
> >>>>-Xylar
> >>>
> >>>>_______________________________________________
> >>>>mpas-developers mailing list
> >>>>mpas-developers at mailman.ucar.edu
> >>>>http://mailman.ucar.edu/mailman/listinfo/mpas-developers
> >>>
> >>>_______________________________________________
> >>>mpas-developers mailing list
> >>>mpas-developers at mailman.ucar.edu
> >>>http://mailman.ucar.edu/mailman/listinfo/mpas-developers
> >>>
> >>
> >>
> >>***********************
> >>Xylar S. Asay-Davis
> >>E-mail: xylar at lanl.gov
> >>Phone: (505) 606-0025
> >>Fax: (505) 665-2659
> >>CNLS, MS B258
> >>Los Alamos National Laboratory
> >>Los Alamos, NM 87545
> >>***********************
> >>
> >_______________________________________________
> >mpas-developers mailing list
> >mpas-developers at mailman.ucar.edu
> >http://mailman.ucar.edu/mailman/listinfo/mpas-developers
> 