Using restart files to initialzed the UFS S2S model

[bnepal22]

I am attempting to initialized the UFS (FV3_GFS_v15p2) model as a cold start run based on the Restart output of previous run. But when checking/comparing the output of chgres_cube.sh and the RESTART outputs, except for sfc_ data.tile[1-6] other file varies significantly (even the sfc_data from chgres_cube and restart files varies slightly with additional variables in it):

RESTART Files

Initial conditions from chgres_cube

In this situation what would be the appropriate way to initialized the UFS model with RESTART files?

[jdkublnick]

Hello,

We have received your question and have reached out to the appropriate SMEs. You will have a reply shortly.

Thanks,

Josh

[gspetro-NOAA]

Hi @bnepal22 ,

I was speaking to my colleagues about your question, and we are a bit unclear on what you are asking. When the RESTART files are processed using chgres_cube, the output should be different than the original--this is the point of pre-processing. Have you tried to run the model with the processed files? If so, what was the outcome?

In general, we need a bit more information to help you effectively. For example, understanding what system you are running on, what version of the WM you are using (develop branch or a specific tag/release?), what specific items you think you should see in the pre-processed output versus what you are actually seeing, and why it seems like a problem. In general, providing text-based input instead of screenshots is easier for us. The screenshots attached are quite small and hard to read.

Thanks,
Gillian Petro

[bnepal22]

@gspetro-NOAA ,
Sorry for not making it clear earlier, however I will try to elaborate my issue here:

Actually, I want to:

• Run the UFS model for 12 hours starting from an initial condition at 2025-03-24 1200 UTC, generating RESTART files at 2025-03-25 0000 UTC.
• Re-initialize (cold start) the UFS model at 2025-03-25 0000 UTC using these RESTART files for a longer integration (one month).

But when comparing the RESTART files (generated after the 12-hour run) and the initial condition files (from chgres_cube.sh), I observed inconsistencies in coordinate variable names as mentioned below:

RESTART Files

1. Physics Restart Files (e.g., phy_data.tile[1-6].nc):
   Coordinates: xaxis_1, yaxis_1, zaxis_1

2. Dynamics Core Restart Files (e.g., fv_core.res.tile[1-6].nc):
   Coordinates: xaxis_1, xaxis_2, yaxis_1, yaxis_2, zaxis_1

INITIAL CONDITION FILES (from chgres_cube.sh)

1. Atmospheric Data (e.g., Gfs_data.tile[1-6].nc):
   Coordinates: geolon, geolat, geolon_s, geolon_w, geolat_s, geolat_w

2. Surface Data (e.g., sfc_data.tile[1-6].nc):
   Coordinates: xaxis_1, yaxis_1, zaxis_1, time, geolon, geolat

[jdkublnick]

Hello @bnepal22,

The difference in coordinate variable names between the RESTART files and the initial condition files generated by chgres_cube is not expected to cause any issues unless the model fails to run properly. According to our internal SMEs, this is a minor inconsistency between how chgres_cube writes the initial conditions and how the UFS model handles restart input and output. As long as the model reads the initial condition and restart files without errors and runs successfully, these naming differences can be safely ignored. If the model fails to initialize or produces unexpected results during the restarted run, then it may be worth investigating further. Sorry for the late response, I hope this can still help!

Thanks.

Josh Kublnick