Convert the ocean output from a B-type simulation to force an F-type simulation

Convert SST and Sea Ice from POP output to DOCN input

Sources

• NCAR has a description to do this. I adapted it for CDO instead of ncdump.
• Jan Sedlacek had his own version of this conversion from which I took the code for grid_fill.ncl.

Warning

Linear Interpolation

DOCN uses the monthly mean fields of SST and Sea Ice and linearly interpolates between them. The interpolation causes the monthly means to be unequal (i.e. the original monthly means and the interpolated monthly means). In the NCAR description they mention a way to avoid this, however, the necessary files are not available.

Sea Ice

The interpolation is not very good for sea ice, which can become 0. The linearly interpolated and the original data may look very different (especially the onset and decay of the sea ice). I do not think that there is a way around this.

Spatial Interpolation

The output of POP is on a rotated grid which we here convert to a regular grid. After reading from the regular grid DOCN regrids this again to the rotated grid. This is of course not very clever, one should instead keep the original grid and tell POP to use this.

Discrepancy

When I compare the B and F simulation I get a temperature anomaly in Northern High latitudes, most likely due to a Sea Ice effect (probably due to the linear interpolation). I have not looked in to that.

Usage

The main script is create_SST_forcing_example.sh. You need to change the parameters in this file (and I recommend to rename it):

• casename: name of the simulation
• root: folder where the simulation is stored
• dest: folder to write the output to
• first: first year to use
• last: last year to use

For first and last I recommend to use all years in the B simulation, even if you don't want to simulate all years in the F simulation.

Run .\create_SST_forcing_example.sh and wait...

Further Hints

Tell CESM to use your input file (must be an F-type simulation):

# year when your simulation starts (and not the start of your SST dataset!)
start_year=1950
# year when your simulation starts (and not the start of your SST dataset!)
end_year=2006
# name of the B simulation
CONTROL=B...

# set the SST and ICE forcing file
./xmlchange -file env_run.xml -id SSTICE_DATA_FILENAME -val /path/to/file/SST_ICE_${CONTROL}.nc

./xmlchange -file env_run.xml -id SSTICE_YEAR_ALIGN -val ${start_year}
./xmlchange -file env_run.xml -id SSTICE_YEAR_START -val ${start_year}
./xmlchange -file env_run.xml -id SSTICE_YEAR_END -val ${end_year}

Conversion

The conversion proceeds as follows:

• Sea Surface Temperatures
  • extract TEMP
  • select the uppermost level
  • rename TEMP to SST_cpl
  • regrid to 1° x 1°
  • interpolate to continents with grid_fill.ncl
• Sea Ice
  • extract aice
  • convert from percent to fraction
  • rename aice to ice_cov
  • regrid to 1° x 1°
  • interpolate to continents with grid_fill.ncl
• Concatenate all files
• Merge SST and Sea Ice file
• Change time axis with change_time.py

Helper Files

The main script - create_SST_forcing_example.sh - calls generate_SST_forcing.sh, which in turn executes the CDO commands and calls the python and ncl script.

griddes_1x1.txt

CDO grid description of the usual 1° x 1° ocean input file at cesm/inputdata/atm/cam/sst

cdo griddes sst_HadOIBl_bc_1x1_1850_2012_c130411.nc

grid_fill.ncl

Fills the continents with interpolated data. This is important for costal regions if the ocean has another resolution than the inputfile (which it usually does).

change_time.py

POP uses days since 0-01-01 00:00:00 as netCDF time string. However, DOCN can not read this time string, so we use python to shift the time axis to 1850.