Initializing experiments

Previous step in Weakly-Coupled OSSE: generating synthetic observations

Initial directory creation / ensemble members

For the initial background, the ensemble members are chosen from either the CFSR data or a nature run, from a year different than the initial date of our experiments.

The following command will generate the initial ensemble members for a 10 member experiment, starting Jan 1, 2010, from a previously generated nature run:

./run/init_cycle 2010010100 DATA/exp 10 --nature DATA/nature

Run the init_cycle command with the -h option to see the full documentation for the script

It should be noted though that this will pick the 40 members from January and February of 2009 of the nature run. This will probably result in an ocean ensemble that has too small of a spread as it does not take into account any interannual variations. As a result the spin-up time might take a little longer. A more appropriate script would take ensemble members from December and January of multiple years (I'll post that script if I ever get around to that. - Travis)

The following directories will be created by this script DATA/exp:

• /anal - The analysis
• ./XXX - A folder for each ensemble member, these are cleared when they are no longer needed in order to keep from using too much data
• ./mean - Analysis mean
• ./sprd - Analysis spread
• /gues - The background guess, i.e the forecast made from the previous analysis
• ./XXX - A folder for each ensemble member, cleared out as well when no longer needed
• ./mean
• ./sprd
• /obsop - The output from the observation operators goes here, so that they can be shared between the ocean LETKF and atmosphere LETKF. Old dates are automatically deleted by the run scripts.

It should be noted that the initial ensemble members are copied into the anal directory, not the gues directory as one might expect. This is because we need to run a forecast first before doing data assimilation. With the 4D LETKF, output is required every hour at hours 3 through 9 for the atmosphere. Because of this, eventhough our initial date is 2010-01-01 00Z, the first observations are not assimilated until 2010-01-01 06Z

LETKF configuration

In addition to the initial ensemble members and directory structure, the init_cycle script will also create the letkf.nml file. This file is responsible for defining the runtime configuration parameters used by both the MOM-LETKF, GFS-LETKF, and respective observation operators. It defines parameters such as localization radius and inflation methods. The default values are probably fine to use, but examine the in-file comments, as well as the documentation at: LETKF configuration

Next step in Weakly-Coupled OSSE: running DA cycles