[MITgcm-support] Different output shortly after restart

[Jean-Michel Campin]

Hi Roland,

I did not see any answer on the list, sorry if it's little bit late now
but will try to answer your 2 questions:

> As I have not been saving .data pickup files in my longer runs, I was wondering 
> whether a script exists to convert the mnc pickup files to meta/data pickup files? 
> Otherwise I may need to restart the longer runs I have already partially completed. 
I would recommend to just provide "new initial conditions" files:
 hydrogThetaFile=
 hydrogSaltFile =
 pSurfInitFile=     (for SSH "Eta" field)
 uVelInitFile =
 vVelInitFile =
that you can extract from the MNC pickup files (e.g., using matlab rdmnc.m)
and write into separated files (1 for each field).

This way requires less fields than when restarting from a pickup and is also
more tolerent (in case one field is missing).

> Also, I wondered whether a script exists to convert a model state with a particular tile 
> decomposition to another decomposition (e.g. from 8x16 to 8x8). Or is the best way to do 
> that to generate binary start files for each field from the original decomposition and 
> use them as initial conditions via the &PARM05 namelist?
With binary files, can use rdmds.m to read-in tiled files (output or pickup) and then just 
write a global file that the model can read. 
However, this does not work as well with NetCDF files since (and this is one of the MNC 
limitation) the model is not able to read-in "global" NetCDF file (i.e. full domain in one file).

Cheers,
Jean-Michel

On Sat, Dec 13, 2014 at 01:55:59PM +0000, Roland Young wrote:
> Hi Jean-Michel,
> 
> I tried your suggestion and I think it has fixed it.
> 
> I ran four short runs:
> 
> One for 1d then another 1d, with mnc pickup I/O off
> A second for 2d with mnc pickup I/O off
> 
> One for 1d then another 1d, with mnc pickup I/O on
> A second for 2d with mnc pickup I/O on
> 
> In the second pair of runs I found the same thing I found in my longer runs - the %MON diagnostics at the restart point were slightly different in the angular momentum quantities, and then the restarted run began to diverge from the 2d run after one time step.
> 
> In the first pair of runs all the diagnostic output was identical after the restart, including the am_ diagnostics and the cg2d diagnostic.
> 
> So not using the mnc pickup I/O looks like it fixes it.
> 
> As I have not been saving .data pickup files in my longer runs, I was wondering whether a script exists to convert the mnc pickup files to meta/data pickup files? Otherwise I may need to restart the longer runs I have already partially completed. 
> 
> Also, I wondered whether a script exists to convert a model state with a particular tile decomposition to another decomposition (e.g. from 8x16 to 8x8). Or is the best way to do that to generate binary start files for each field from the original decomposition and use them as initial conditions via the &PARM05 namelist?
> 
> Best wishes,
> 
> Roland
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support