Conversation
….yaml Co-authored-by: Doruk Ardağ <38666458+Dooruk@users.noreply.github.com>
….yaml Co-authored-by: Doruk Ardağ <38666458+Dooruk@users.noreply.github.com>
| # --------------------------------- | ||
| if npx_proc is not None and npy_proc is not None: | ||
| self.jedi_rendering.add_key('gsibec_configuration', self.config.gsibec_configuration()) | ||
| if gsibec_configuration is not None: |
There was a problem hiding this comment.
I changed this to allow for future background error covariance methods beyond gsibec. This conditional wasn't really appropriate. This now ensures flexibility for other FV3 configurations like GEOS-CF to adopt different B matrix flavors
|
@mranst I am not sure I have access to the tier1 test src. Let me know if you need me to add the test there. |
jeromebarre
changed the title
| qd.check_for_obs(False) | ||
| ], | ||
| geos_cf=[ | ||
| qd.cycle_times(['T18']), |
There was a problem hiding this comment.
I would remove this. This is redundant with start_cycle_point and final_cycle_point. If you keep this and change to final_cycle_point to run a longer experiment, you will need to add other cycle_times there too: T00 , T06, T12.
There was a problem hiding this comment.
yes good catch I am not sure why this line is here...
| qd.model_components(['geos_cf']), | ||
| qd.check_for_obs(False) | ||
| ], | ||
| geos_cf=[ |
There was a problem hiding this comment.
Why not put everything under questions?
There was a problem hiding this comment.
I followed the logic in place in the other suites like 3d_var marine. @Dooruk would like to tell why this structure between questions vs geos_<application> and _3dvar_<app>_cycle_tier1 vs _3dvar_<app>_cycle is preferrable?
|
|
||
| # -------------------------------------------------------------------------------------------------- | ||
|
|
||
| _3dvar_cf = QuestionList( |
There was a problem hiding this comment.
Minor suggestion: I defined it the otherway for hofx_cf. First, the experiment and then experiment_tier1 = experiment. It makes more sense to me, logically.
There was a problem hiding this comment.
|
|
||
| varincrement1 = { | ||
| 'write increment': True, | ||
| 'increment': { |
There was a problem hiding this comment.
This works fine. I'm just curious, what's the difference between writing the increment using these keys under online diagnostics vs using increment to structured grid?
There was a problem hiding this comment.
I wonder that too...
I followed what was done with 3dvar_atmos
2 participants
This PR adds support for 3D-Var data assimilation on the GEOS-CF
As a first step it uses Identity for background error. Subsequent PRs will add BUMP and diffusion.
More obs will be added over time too.