[MITgcm-devel] monthly, inter-annually varying forcing with exf

[Oliver Jahn]

Thanks, Martin.  I thought of negative RepCycles too.  I wasn't sure how
to handle positive values.  They would have to be seconds which are then
converted to months using a "mean month length" (similar to calendar
dumps).  This might get ambiguous for very long cycles.  I guess one
just has to be clear on how long a "mean month" is.  It would be more
consistent.  Or one could just through an error if RepCycle is positive
and period = -12.

Oliver


On 2019-11-18 7:40 a.m., Martin Losch wrote:
> Hi Oliver,
> 
> sounds good to me. 
> 
> It may be a little confusing that in general *RepCylce is in seconds, except for this particular case when it is in months. I can see that this is as confusing (or not confusing) as the *period = -12., but in that case the negative sign makes it clear that -12 is a special value. I am not sure if that should really be a concern, but would negative *RepCycles make sense in this context?
> 
> Martin
> 
> 
> 
>> On 15. Nov 2019, at 15:54, Oliver Jahn <jahn at mit.edu> wrote:
>> 
>> Hi Developers,
>> 
>> I have some code for extending the exf package to read calendar-monthly
>> forcing fields that vary from year to year.  I think other people could
>> find this useful too, see for instance
>> 
>> http://mailman.mitgcm.org/pipermail/mitgcm-support/2019-October/012213.html
>> 
>> I'd like to get some input on how to enable this feature.  Right now the
>> magic value of *period = -12 is used for monthly climatologies.  I have
>> talked to Jean-Michel about this and we think the most logical extension
>> would be to use the same value for inter-annually varying monthly fields
>> and distinguish by *RepCycle.  A *RepCycle of 0 would be for
>> non-repeating forcing, and 12 could be for climatology.  One could also
>> support monthly forcing that repeats after a different number of months
>> (well, years, usually), and *RepCycle could be in units of months - as
>> suggested by *period = -12.  So, *RepCycle = 120 would be monthly
>> forcing that repeats every 10 years.
>> 
>> We do not want to break existing setups, of course, so we were thinking
>> of defaulting *RepCycle to 12 when *period is -12 and to repeatPeriod in
>> all other cases.  This would be accomplished by an UNDEF default value
>> that is then changed to either 0 or repeatPeriod once *period has been set.
>> 
>> Any thoughts on this would be very welcome!  Maybe this breaks someone's
>> use case we didn't think of.  Please let us know!
>> 
>> Cheers,
>> Oliver
>> _______________________________________________
>> MITgcm-devel mailing list
>> MITgcm-devel at mitgcm.org
>> http://mailman.mitgcm.org/mailman/listinfo/mitgcm-devel
> 
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mailman.mitgcm.org/mailman/listinfo/mitgcm-devel
>