clarification of restart files produced and required for CAM hybrid runs

[mudryk@cita_utoronto_ca]

Hello,

I'm running the full CESM model (component set B) with transient 20th century forcings. I would like to initialize several different ensemble members starting on January 1, 1955, with the lnd, ice, ocn states from the parent run, but with differing atmospheric states.

In order to accomplish this, I've tried to produce daily restarts from a 2 month parent run in December 1954/January 1955 with the $REST_N variable in env_run.xml set to 1 day. The .r. and .rs. files produced look fine (there is one for each day in December and January), however only a single initialization file is produced for each calendar year, dated 01-01. Instead there are .rh0. files are produced for each day.

My issue is that if I then try to start a hybrid run using an atmospheric state from the .r. files (say parent_run.cam2.r.1954-12-29-00000.nc) the model complains that it wants an initialization file from 12-29, rather than the 01-01 date.

1) Is the .i. file required for hybrid restarts, branch restarts or both?

2) Can you clarify which types of restarts (hybrid and branch) require which of the CAM restart files produced (there are .r., .i., .rs. and .rh0. files that I've seen). Are the .h0. and .h1. files only required to recompute the monthly averages? I haven't been able to find this information clearly specified in any of the online user guides.

3) Is there a better way to restart a run with differing atmospheric states than the one I'm attempting to use?

Thanks for your help, Lawrence.

 

[eaton]

Full disclosure: I've never actually done a hybrid run using the CESM scripts, so I'm not sure about all the details of doing this. Hopefully the following comments will be helpful.

A hybrid run is set up like an initial run for both the CAM and CLM components. So both those components need to have initial files specified, while the pop and cice components will start from restart files. That said, since we converted the restart files to the netcdf format, CAM is now able to use a restart file in place of an initial file. But whether and initial or restart file is used, that file still needs to be specified using CAM's ncdata namelist variable since that's how you specify the CAM initial file.

An initial run for CAM only requires the file specified by ncdata (which can be either an initial ".i" file or a restart ".r" file. None of the other files generated from a previous run are needed for a CAM initial run.

I think the procedure that you're attempting to use should work, assuming that you get ncdata specified correctly. As an aside it would be possible to do the 2 month run generating initial files at the end of each day by setting the CAM namelist variable inithist='DAILY' in that run. But I think you should be able to use the restart files that you've already generated for the same purpose.