3.2 Uncoupled Run Script

The run script for the uncoupled model is called csim_run and is located in /ccsm3/models/ice/csim4/src. Its purpose is to coordinate setting the batch system options, the environment variables, executing the CSIM setup script, setting up the stdout file, and submitting the model to run.

At the top of the run script, the settings for the IBM SP and the SGI Origin 2000 batch queue environments are set. These commands are machine and site dependent. Following this, the variables for the run environment are defined. These variables are listed in Table 1.

$CASE is a character string that identifies a particular model run. It can be up to 16 characters long, but it is best kept short, since it is used as part of the restart, history, and initial filenames. $CASESTR is a longer string that describes a model case.

$RUNTYPE is a character string that specifies the state in which the model is to begin a run. startup and continue are the supported run types. A startup run can be initialized by reading input from a file or from initial conditions set within the ice model. This option is controlled by the environment variable $RESTART in the setup script (see Section 6.1 ). Continue runs are described in Section 6.2.

$NCAT is an integer that sets the number of ice thickness categories. The default value is 5 categories. If you are considering changing $NCAT to values other than 3 or 10, read Section 7. This is an involved process that deserves its own section.

3.2.1 Using the Ocean Mixed Layer Model within CSIM

$OCEANMIXED_ICE is a logical variable, if .true., is used to implement the slab ocean mixed layer model in the ice model. It is a simple model that is forced using output from a POP ocean simulation. More details on the physics of the ocean mixed layer model can be found in the Physics Documentation. It can be run with the gx3v5 or the gx1v3 grid. To use the mixed layer model, set

    setenv \$OCEANMIXED\_ICE .true.

in csim_run.

3.2.2 Changing Grid Resolution

$GRID is a character string used to specify the horizontal grid. Presently, two resolutions are supported for the ice model: gx3v5 and gx1v3. In both of these grids, the North Pole has been displaced into Greenland. gx3v5 is the coarser grid, with longitudinal resolution of 3.6 degrees. The latitudinal resolution varies, with a resolution of approximately 0.9 degrees near the equator. gx1v3 is the finer resolution grid, with a longitudinal resolution of approximately one degree. Its latitudinal resolution is also variable, with a resolution of approximately 0.3 degrees near the equator.

3.2.3 Changing the Number of Processors

$NX and $NY are the number of processors used by the ice model for internal parallelization. Currently, $NX and $NY MUST divide evenly into the grid dimensions. There are checks for this in the setup script and in the ice source code; the model will stop if these criteria are not met. $NLAT and $NLON are used for this purpose. For load balancing purposes, $NY should be $<$=2. If it is greater than this, the processors assigned subdomains near the equator will not be doing much work.

For the gx3v5 grid, the ice model is typically run on 8 tasks, with NX=4, NY=2. Running the ice model with NX=8 and NY=1 tasks on the gx3v5 grid wil result in an error, since 8 does not divide evenly into the 100 longitude points. When this happens, the model will stop with an error message written to the log file.

If you are submitting the model to a batch queue with the number of processors modified from the default, you will also have to modify the batch queue environment information at the top of the script. The default setting for the IBM is:

  # @ total_tasks = 8
  # @ node = 1

These two lines request a total of eight MPI processes, on one 8-way node. For the NCAR SGI, the default setting is:

  # QSUB -l mpp_p=8           # request 8 processors

and for other SGI's it may be

  # BSUB -n 8   .