search

ESCOMP / POP2-CESM

Parallel Ocean Program (POP2) in CESM
http://www.cesm.ucar.edu/models/cesm2/ocean/
4 stars 24 forks source link

Update dt_count for RCP extensions #27

Closed mnlevy1981 closed 3 years ago

mnlevy1981 commented 4 years ago

Description of changes:

Provide support for new SSP534EXT and SSP585EXT compsets -- the main goal was to provide a mechanism to set dt_count=48 by default in the SSP585 extension.

Testing:

I've provided branch tags to Mike Mills et al, who are testing the new compsets and think everything looks okay

Test case/suite: None Test status: bit-for-bit

Fixes [POP2 Github issue #] N/A

User interface (namelist or namelist defaults) changes?

I have two concerns relating to the robustness of this PR:

  1. Setting dt_count via

    <dt_count ocn_grid="gx1v7"  ocn_transient="ssp585ext">48</dt_count>

    seems clunky, in the sense that it ignores other variables that might affect dt_count. In fact, I am confused why this value for dt_count wins out over

    <dt_count ocn_ncpl="24" time_mix="robert"  ocn_grid="gx1v7">24</dt_count>

    The former matches 2 out of 2 attributes, while the latter appears to matches 3 out of 3 and yet dt_count is set to 48 rather than 24. I think this is indicative of a bug in how time_mix is being passed to namelist defaults, and that fixing said bug will change the behavior in the new compset.

  2. The variables that depend on ocn_transient="ssp585" are limited to some options in the abio tracer module (abio_atm_d14c_filename, abio_atm_co2_opt, abio_atm_d14c_opt), some options in the ciso tracer module (ciso_atm_d13c_opt, ciso_atm_d14c_opt, ciso_atm_d13c_filename, and ciso_atm_d14c_filename), and ndep settings in the ecosys tracer module. It looks like ssp534 is not used at all in the namelist defaults file, so I don't know how the 2015-2100 portion of that run was configured.

    I have not done anything to support abio or ciso in this new compset, and set ndep_data_type="driver" for both of the ssp extension compsets. This is a departure from SSP585, which uses shr_stream to read in a dataset containing 88 years of data.

klindsay28 commented 4 years ago

I don't think this should be setting ndep_data_type for the extension compsets to driver. It should be shr_stream, so that we get shr_stream for non-WACCM extension compsets, which are in the works.

The setting of ndep_data_type='driver' in WACCM compsets is accomplished via the %NDEP modifier for POP in the compset longname. This triggers OCN_NDEP_DRIVER in env_run.xml, which triggers setting ndep_data_type='driver' in build-namelist.

mnlevy1981 commented 4 years ago

I don't think this should be setting ndep_data_type for the extension compsets to driver. It should be shr_stream, so that we get shr_stream for non-WACCM extension compsets, which are in the works.

The setting of ndep_data_type='driver' in WACCM compsets is accomplished via the %NDEP modifier for POP in the compset longname. This triggers OCN_NDEP_DRIVER in env_run.xml, which triggers setting ndep_data_type='driver' in build-namelist.

Good catch! I've verified that those two lines are not necessary and have removed them. I'll add the non-WACCM support in the next commit

klindsay28 commented 3 years ago

I'm noticing that this PR adds support for ssp534ext (2100-2300) (e.g. ndep), but we don't have support for ssp534 (2015-2100). That seems kinda pointless. I've sent an email to Keith Oleson to see if ndep is available for ssp534 (2015-2100).

klindsay28 commented 3 years ago

Keith Oleson reports that no one has asked for non-WACCM ssp534, so the ndep dataset has not been created. We'll do that if someone requests it.