[MITgcm-support] Re: non-MDSIO input files
Dmitri Leonov
dleonov at u.washington.edu
Thu Aug 10 14:22:36 EDT 2006
Hello Ed,
Thanks for the info. I can turn on option 1 or 3. The only question is
how to generate the input files (using matlab or otherwise).
Could you point me to some examples?
For a serial 1-tile run I can modify the wrong 'grid.t001.nc' file and
restart the model with readgrid_mnc turned on. Will that work? How do I
use readgrid_mnc with multiple tiles?
Thank you
Dmitri
Ed Hill wrote:
> On Wed, 2006-08-09 at 14:42 -0700, Dmitri Leonov wrote:
>
>> Hi Matthew,
>>
>> I think my question was about arrays of sizes (Nx+1,Ny), (Nx,Ny+1) and
>> (Nx+1,Ny+1). I couldn't find consistent way of initializing them using mds.
>>
>
> Hi Dmitri,
>
> Actually, the mdsio routines *can* handle most of the "+1" cases that
> you describe above. Specifically, given the "standard" cube orientation
> that we use, all values on U,V, and Mass points ("staggerings") are
> written and read using the MDSIO routines. The exception is values at
> vorticity (corner staggerings) which omit two (or is it three?) of the
> corner values. This is an annoyance but not a huge problem for initial
> conditions or pickup/checkpoint files.
>
> For the "+1" and other non-MDSIO I/O needs, there are a number of
> routines including:
>
> 1) READSYMTILE_RS in ini_curvilinear grid for the grid info
> 2) the read facilities within the EXF package for forcing data
> 3) the entire MNC package for netCDF IO (MNC is capable of
> almost any shape that you'd care to define).
>
> Of the three, the MNC package was designed to gracefully handle all the
> "+1" cases (including vorticity staggerings) so you might want to look
> into it.
>
> Ed
>
>
More information about the MITgcm-support
mailing list