coupled

MITgcm can be an atmosphere-ocean coupled model using its own coupler. There is a verification called cpl_aim+ocn, and it is a coupled version of atmospheric and oceanic model on cubed sphere grid that we did so far. The instruction below can be found in README file in cpl_aim+ocn to run the coupled simulation.

README.md
Atmosphere-Ocean coupled set-up example “cpl_aim+ocn”
using simplified atmospheric physics (AIM), in realistic configuration (orography & continent) with land and seaice component, on cubed-sphere (cs-32) grid.

overview

Uses “in-house” MITgcm coupler (pkg/atm_ocn_coupler, pkg/compon_communic, pkg/atm_compon_interf, pkg/ocn_compon_interf ) with each component config and customized src code in: code_cpl, code_atm, code_ocn ; and input parameter files in: input_cpl, input_atm, input_ocn.

Atmos set-up and parameter is similar to “aim_5l_cs/” experiment
Ocean set-up and parameter is similar to “global_ocean.cs32x15/” experiment

Requires the use of MPI; as default, use 1 proc for each component.

cleaning

If you compile the code for the first time, you do not need to do this. But if you want to redo compilation, then it is good to clean up the existing files.

To clean everything:

  ../../tools/run_cpl_test 0

compile

We need to do one thing to avoid a compiling error that might come from the compiler version issue. Hover the curser here and do right-click, then click Save link as.... Save this file in code_cpl as setdir.c.

Now, we will compile three models: coupler, ocean model and atmospheric model with the following line.

  ../../tools/run_cpl_test 1 -of ../../tools/build_options/linux_amd64_gfortran

integration with default configuration

To run primary setup, thermodynamic seaice only (no seaice dynamics):

  ../../tools/run_cpl_test 2
  ../../tools/run_cpl_test 3

integration with modification

Step 2 above copies input files and directories, step 3 runs the coupled model.

To run secondary test (with seaice dynamics as part of ocean component), using input parameter files in: input_cpl.icedyn, input_atm. icedyn, input_ocn.icedyn:

  ../../tools/run_cpl_test 2 icedyn
  ../../tools/run_cpl_test 3

use more CPUs

The default setting uses total 3 CPUs (1 for each), which may not give you the best speed of integration. We can use more than 1 CPUs for each component to speed up. Let’s first check the number of CPUs we can use. In terminal,

sysctl -n hw.ncpu

For me, I have 8 CPUs. So, I am going to use 4 CPUs for the atmospheric component, 2 CPUs for the oceanic component and 1 for the coupler. Then, you can edit run_cpl_test to specify the total number of CPUs available (Npr), and the number of CPUs for the ocean model (NpOc). After this edit, you can repeat the previous procedure to run the model.

model results

Results are written in rank_{0,1,2} dir, for coupler, ocean and atmos comp. respectively

analysis

Note:
To check the results, monitor output could be compared to reference (in results/) using “run_cpl_test”, step 4.
For primary set-up:

  ../../tools/run_cpl_test 4

and for secondary test:

  ../../tools/run_cpl_test 4 icedyn

but this requires, in your path, a simple comparison script “comp_res” (which is not provided here but could be found in: http://wwwcvs.mitgcm.org/viewvc/MITgcm/MITgcm_contrib/jmc_script/ ), along with some other files found in this archive.