This setup will result more realistic sea ice simulations. One year of NCEP forcing for latm5 is included with this CCSM2.0 release. This setup is the D configuration discussed in section 3.3, with the latm5 data atmosphere model and the ocean mixed layer model within the ice model. The ocean mixed layer is a simple slab model which computes an ocean surface temperature and allows for ice formation due to ocean supercooling. More information on the formulation of this model can be found in the Scientific Document. latm5 runs on the T62 grid; the GUI can be used to set up the scripts for this configuration. If you are not using the GUI, first make the changes listed in section 3.3 for the D configuration. The following additional changes should be made in the main script:
setenv GRID T62_gx1v3 # T42_gx1v3 or T31_gx3 set SETUPS = ( latm dlnd ice docn cpl ) # setup script name
The following lines
if ($GRID == T42_gx1v3 ) then set NTASKS = ... set NTHRDS = ... else if ($GRID == T42_gx3 ) then set NTASKS = ... set NTHRDS = ... . . endif
will need to be changed to
if ($GRID == T62_gx1v3 ) then set NTASKS = ... set NTHRDS = ... else if ($GRID == T62_gx3 ) then set NTASKS = ... set NTHRDS = ... . . endifso that $NTASKS and $NTHRDS are set. For multi-year simulations, set
# @ wall_clock_limit = 21600
In ice.setup.csh, set
set OCEANMIXED_ICE = .true.
In dlnd.setup.csh, set
data_year0 = 16 data_nyear = 16so that land data from the T62 grid will be used. When these values are set to 5, data from the T42 grid is used.
In the latm.setup.csh namelist, data_nyear determines how many years of forcing data will be used; the default value is one. In this case, one year of data will be used repeatedly. ncpl specifies how many times per day latm5 exchanges information with the coupler. There are other parameters in the latm namelist that may be of interest. data_oyear is the coupler year matched to data_year0. This configuration runs at about 15 minutes wall clock time per model month using 16 processors for the ice model, and about 30 minutes wall clock time per model month using 8 processors. Both simulations were done on the T62_gx1v3 grid. This setup does not run as efficiently as the B or F configurations, since the buffer sent to and received from the coupler is not packed as efficiently.