[MITgcm-support] Offline Model Help

[Stephanie Dutkiewicz]

Ryan -

the offline pkg does a linear interpolation between
fields. (This is same procedure as is done for the
forcing fields in the online model).
(if you are curious see offline_fields_load.F)

you don't need to rename your fields. there is a
seperate "deltaToffline" which will keep track of
when (and what timestamp to look for) to read the
fields. From your error message I'm guessing that
your deltaToffline isn't correct.

Here is an example from verification/tutorial_cfc_offline/input/data.off


  &OFFLINE_PARM02
   offlineIter0=4248000,
   deltaToffline=43200.,
   offlineForcingPeriod=2592000.,
   offlineForcingCycle=31104000.,
  &

offlineIter0 is the number of the first file to read.
deltaToffline is the timestep of the original online run
    used to create the fields.
I think what you've done for offlineForcingPeriod, offlineForcingCycle
should work okay.

Note though, that the first timestep of the offline run will
read in the last and first file of your online fields - and
interpolate between them to give the fields used.

Hmmm - complicated to explain - come see me if this
doesn't make much sense!

stephanie


On Wed, 15 Aug 2007, Ryan Abernathey wrote:

> Hello Wise GCC Users,
>
> I'm hoping someone can help me understand the inner workings of the offline 
> package... This is a slightly complicated question, so please disregard 
> this message if you don't care to wade through it.
>
> I have a 2 year long set of velocity fields that I want to use to advect a 
> tracer. I want to be sure I understand how the offline fields are imported 
> and interpolated in the model.
>
> I don't want any cycling of these fields, just a simple 2 year run. 
> Therefore, I made all these parameters the same:
> data
> periodicExternalForcing =.TRUE.,
> externForcingPeriod = 62208000.,
> externForcingCycle = 62208000.,
> data.off
> offlineForcingPeriod=62208000.,
> offlineForcingCycle=62208000.,
>
> It is less clear to me how the input files are mapped to actual times 
> inside the model. From reading the documentation 
> (http://mitgcm.org/r2_web_testing/latest/online_documents/node172.html) I 
> concluded that the input files need to be number with suffixes that 
> correspond to the offline model timestep. But the documentation doesn't 
> mention interpolation, even though I am fairly sure it is happening.
>
> I am using a different deltat in my advection run (1200) than was used to 
> generate the offline files (900), so I have a script to re-number the 
> offline files accordingly. (i.e. .0000000480 becomes .0000000360 etc.) But 
> I do not have a file for every timestep. I just have one for every 360 
> timesteps. If I change the value of offlineForcingPeriod etc., the gcm 
> sometimes crashes with the following sort of error:
> (PID.TID 0000.0001)  MDS_READ_FIELD: filename: uVeltave.0000000518.data
> (PID.TID 0000.0001)  MDS_READ_FIELD: File does not exist
> ...with the actual timestep depending on the value. I don't understand why. 
> Why does it require certain offline timesteps but not others? What logic is 
> used when searching for offline data files? This makes me worried that I 
> don't understand how the timesteps are being mapped, and my results also 
> suggest that maybe it is not using the full two years but rather doing 
> something strange and mysterious.
>
> In general, I am not confident that the offline model is doing what I want 
> and would be very grateful (i.e. buy you a beer grateful) for someone to 
> explain it to me thoroughly.
>
> Cheers,
> __________________
> Ryan Abernathey
> rpa at mit.edu
> MIT Ph.D. Student
> Program in Oceans Atmospheres & Climate
> http://web.mit.edu/rpa/
>
>
>