Updating lateral boundary conditions for FV3 Limited Area Model (LAM) cases

JEDI
1

Hi,

I would like to run 3Dvar for a limited area domain (LAM) and to use the JEDI output as the updated input for UFS. For a similar process, if we use WRFDA and WRF, after running “da_wrfvar.exe” to assimilate observations, it is necessary to run “da_update_bc.exe” to update the lateral boundary conditions. It is mentioned here.

Now, I am wondering about updating lateral boundary condition in JEDI. Is it something that hasn’t been developed yet for regional cases, or the data structure of the initial and boundary condition is somehow different such that “updating boundary conditions” is not necessary when we use JEDI?

For example, here are the UFS input files for a restart run of a regional case.

C3337_grid.tile7.halo3.nc
coupler.res
fv_core.res.nc
fv_core.res.tile1.nc
fv_srf_wnd.res.tile1.nc
fv_tracer.res.tile1.nc
gfs_bndy.tile7.000.nc
gfs_bndy.tile7.003.nc
gfs_bndy.tile7.006.nc
gfs_bndy.tile7.009.nc
gfs_bndy.tile7.012.nc
gfs_ctrl.nc
gfs_data.nc
grid_spec.nc
grid.tile7.halo4.nc
oro_data_ls.nc
oro_data.nc
oro_data_ss.nc
oro_data.tile7.halo4.nc
phy_data.nc
sfc_data.nc

not sure if all of the files are necessary for running UFS; however, gfs_bndy_tile7.hhh.nc files are lateral boundary conditions.

For the the test “fv3jedi_test_tier1_3dvar_lam_cmaq” available in “jedi-bundle”, the JEDI output files are:

YYYYMMDD.HHHHH.coupler.res
YYYYMMDD.HHHHH.fv_core.res.nc
YYYYMMDD.HHHHH.fv_tracer.res.nc

So, it sounds like that fv3jedi_var.x doesn’t generate updated lateral boundary conditions.

Any advice is appreciated.

2

Hi,

I don’t believe the LAM BC’s have received careful attention in JEDI yet. As far as I know, the code should have the capability to compute an increment/analysis at BC grid points, but if any BC-specific treatments need to be applied that would be missing from the code (probably at the model-interface level).

Folks like @TingLei-NOAA who work on the regional applications would know more about this… perhaps they can say more?

This wasn’t very detailed, if you have more questions we’ll have to find an expert…

Cheers,
Francois

3

Thanks to @fmahebert for looping me in discussion. Yes, the jedi would read in and process background files and for regional fv3, they don’t include “LBC” points in the LBC files. So, it is difficult to update lbc files according to analysis.
Maybe it could be explored that the lbc files are also read in and after some fv3-specifc conversion to be treated as other inner data points and, then, after analysis, be used to update lbc files (again, with model specific conversion).