Sampling Error Table

gen_sampling_err_table

In order to apply Sampling Error Correction, we use gen_sampling_err_table to compute the statistics described in Anderson (2012). 1

The table with the error correction values is stored in the netCDF file sampling_error_correction_table.nc in:

${DARTROOT}/assimilation_code/programs/gen_sampling_err_table/work/

However, the table typically only contains values for ensemble sizes up to 200 ensemble members. Thus we edit input.nml in the above directory thus we edit the following namelist to add ensemble sizes of 250, 500 and 1000:

&gen_sampling_error_table_nml
  ens_sizes = 250, 500, 1000
  debug = .false.
/

We then run quickbuild.csh and submit.csh the latter of which simply submits a batch job that runs gen_sampling_err_table on a compute node.

The program required 01:53:14 on 8 CPUs to complete and add the correction values for the additional ensemble sizes.

DART_config.template

Finally, we need to edit Line 325 of the DART_config.template script to reflect the fact that we are now able to run ensemble sizes of larger than 200. The template is here:

$ vim ${DARTROOT}/models/cam-fv/shell_scripts/cesm2_1/DART_config.template

And we change 200 to 1000 here:

if ( $num_instances < 3 || $num_instances > 1000 ) then

References

1

Anderson, Jeffrey L. (2012) “Localization and Sampling Error Correction in Ensemble Kalman Filter Data Assimilation.” Monthly Weather Review 140: 2359-2371, doi: 10.1175/MWR-D-11-00013.1.