WRF/WRFDA cycling on a nested domain
1. You will need a global analysis for initial conditions (for instance, the GFS-based FNL Analysis (http://rda.ucar.edu/datasets/ds083.2/) or the NCEP/NCAR global reanalysis (http://rda.ucar.edu/datasets/ds090.0/)) at every time step you wish to do assimilation on, and run this through WPS to produce a set of met_em* files. For example: if you wish to do 24-hour forecasts with assimilation at each 12Z point, starting on January first of this year, you must have (using FNL as an example) the files "fnl_20120101_12_00_c", "fnl_20120102_12_00_c", etc. through the end of the time you would like to run your cycle.
2. Run WPS as you normally would for a nested case, with "interval_seconds = 86400" in the "&share" section of "namelist.wps" so that you will get one met_em.d01* and one met_em.d02* file every 24 hours from your initial conditions.
3. Run real.exe to get wrfbdy_01, wrfinput_01 and wrfinput_02
4. Run your data assimilation of choice, using wrfinput_02 as your first guess file ("fg")
5. Overwrite the wrfinput_02 file with your wrfvar_output file.
6. Run your 24-hour WRF forecast as usual.
7. Run real.exe again for the next time period.
8. Copy your final wrfout file (wrfout_d02_2012-01-02_12:00:00 in this example) and give the copied file the name "fg" (this will be your first guess file for the second cycle of assimilation)
9. Run update_bc.exe with the following parame.in settings:
&control_param
da_file = './fg'
wrf_input = './wrfinput_02'
update_lateral_bdy = .false.
update_low_bdy = .true.
/
This step will update the lower boundary conditions of the "fg" file before assimilation.
10. Run your data assimilation method of choice.
Repeat steps 5-10 until you are done cycling. There are a few more steps if you want to do assimilation on the outer domain, or if you would like to use multiple nests.