[MITgcm-devel] changes to exf_init_runoff.F

[Dimitris Menemenlis]

Patrick,

> What I am confused about is that period=0 used to be reserved for monthly
> forcing. Now that has changed? If so, not a good idea to do that
> "in-the-middle-of-nowhere" (no tag, etc.) because its impact is similar to
> changes of units, but probably I'm just confused and all is ok.

No you are not confused.  For runoff, and also I think for all the other fields, 
we need (and should have) the option to force with a time-mean climatology.  The 
way I have done this right now is to reserve "period=0" for this time-mean case 
and "period=-12" for the monthly forcing.

Note that period=0 was reserved for monthly forcing only in the exf_clim 
routines and that I have changed nothing in that part of the code until we have 
time to discuss and to agree on best way to proceed.  Eventually, with proper 
tags, we should make exf and exf_clim consistent and get rid of more of the 
redundant code ... but probably not before the ECCO2 meeting.

> I am actually not updating code currently since I am still struggling
> with what's there now (prior to your latest exf changes).

Do you want me to revert to pre-exf changes?  I need the monthly mean runoff 
capability for some hi-res cube sphere tests but I can put these changes in 
MITgcm_contrib area for time being.  I have checked and all the same 
forward-model functionality as existed prior to this change for runoff is still 
there.  That is, the default, without changing anything in input files, remains 
a time-mean runoff field.

> At GFDL I am still not able to run v3 with newest code (NaN after few timesteps).

... but runs OK at Ames?

> And newest code with v2 (i.e. no seaice, no bulk forcing) crashes after a
> while (~ 2years or so) with cal error.

On columbia, I am presently running with code checked out on 2006/12/05.  Is 
your code that causes above crash newer?

> That same crash I now also get running arctic40km with latest code
> (i.e. stop in cal package). No clue why.

I will try to look at this later today if I can.

> Re. your error message below,
> seems simply that netcdf library is not found at compile time,

Thanks.  That explains it.  I also thought it was netcdf-related but was 
confused since that experiment does not use netcdf.

Let me know how you would like me to proceed re runoff change ... and sorry to 
be adding to stress.  D.