[MITgcm-support] cubed sphere for Atmospheric models

[Yuan Lian]

Hi,

I just figured out the problem when using cubed sphere. The tile*.mitgrid
files created by 'convertMITgrid.m' didn't work well even they had the
same file sizes as those in experiment hs94.cs-32x32x5.

I did the following steps to create cubed sphere:

1.set parameters in gengrid_fn.m as:
 nx=33
 nratio=1
 meth='conf'
 ornt='c'

 write tiles as:
 write_tile([dir 'LONG'], 180/pi*lonG )
 write_tile([dir 'LATG'], 180/pi*latG )
 write_tile([dir 'LONC'], 180/pi*lonC )
 write_tile([dir 'LATC'], 180/pi*latC )

 write_tile([dir 'DXG'], Rsphere*expand(dxg) )
 write_tile([dir 'DYG'], Rsphere*expand(dyg) )
 write_tile([dir 'DXF'], Rsphere*expand(dxf) )
 write_tile([dir 'DYF'], Rsphere*expand(dyf) )
 write_tile([dir 'DXC'], Rsphere*expand(dxc) )
 write_tile([dir 'DYC'], Rsphere*expand(dyc) )
 write_tile([dir 'DXV'], Rsphere*expand(dxv) )
 write_tile([dir 'DYU'], Rsphere*expand(dyu) )
 write_tile([dir 'RA'], Rsphere^2*expand(Ec) )
 write_tile([dir 'RAW'], Rsphere^2*expand(Eu) )
 write_tile([dir 'RAS'], Rsphere^2*expand(Ev) )
 write_tile([dir 'RAZ'], Rsphere^2*expand(Ez) )
 these give the all *.bin files.
2.used 'convertMITgrid.m' to convert *.bin files to *.mitgrid files, I
 set "nx=32; ieee='b' (linux system)", then I got all six tile files.

All tile*.mitgrid files in experiment hs94.cs-32x32x5 worked without any
problem. But there were extreme values when I used tile*.mitgrid created
by matlab codes 'gengrid_fn.m' and 'convertMITgrid.m' from MITgcm_contrib.

I am wondering how the tile*.mitgrid in experiment hs94.cs-32x32x5 were
generated. Were they created by the same matlab codes 'gengrid_fn.m' and
'convertMITgrid.m'?

Thanks a lot for any help!

-Yuan

On Mon, 20 Dec 2004, Yuan Lian wrote:

> Hi, Chris,
>
> I am doing an atmospheric model for uranus. I did change the radius when I
> generate the cubed sphere.
>
> rotationPeriod=6.2064E4,
> gravity=11.0,
> rSphere=24766.E3,
> atm_cp=13000
>
> I keep the others are the same as Earth parameters.
>
> Now I guess I didn't generate the grids properly. Here are what I did to
> generate cubed sphere:
>
> 1. use gengrid_fn.m to generate all neccessary *.bin files.
> 2. use ocnvertMITgrid to generate *.mitgrid files (I used 6 tiles).
> 3. use driver.m to generate  w2_e2setup.F and W2_EXCH2_TOPOLOGY.h, then I
> copied these two files to test/code directory.
>
> I compiled the code and ran, it turns out that the code won't work well
> with all the files I generated. hs94.cs-64x64x5 can run but gives some
> weird results which I didn't notice.
>
> I will keep working on this and see if I can figure out where I did wrong.
> And I will post a detailed description here if I couldn't work it out.
>
> Thanks a lot!
>
> -Yuan
>
> On Mon, 20 Dec 2004, chris hill wrote:
>
> > Yuan,
> >
> >  Which atmospheric physics are you are using.
> >
> > Thanks,
> >
> > Chris
> > On Mon, 2004-12-20 at 18:02, Yuan Lian wrote:
> > > Hi,
> > >
> > > I had some problems to make cubed sphere working properly for atmospheric
> > > models. I can run hs94.cs-32x32x5 without any problems, even
> > > hs94.cs-64x64x5 will work. I tried to run another atmospheric model which
> > > works fine with spherical coordinates. However extreme values came out at
> > > second time step if I use cubed sphere. The resolution I used is 32X32X25.
> > >
> > > I switched spherical coordinates to cubed sphere since I thought cube
> > > sphere will handle the polar regions better. I am wondering if anyone
> > > had the same experience. Should the cubed sphere have less numerical
> > > probelms than spherical coordinates?
> > >
> > > Thanks for all the help!
> > >
> > > -Yuan
> > >
> > > _______________________________________________
> > > MITgcm-support mailing list
> > > MITgcm-support at mitgcm.org
> > > http://dev.mitgcm.org/mailman/listinfo/mitgcm-support
> >
> > _______________________________________________
> > MITgcm-support mailing list
> > MITgcm-support at mitgcm.org
> > http://dev.mitgcm.org/mailman/listinfo/mitgcm-support
> >
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://dev.mitgcm.org/mailman/listinfo/mitgcm-support
>