Closed DeniseWorthen closed 2 months ago
I have an initial implementation working for downscaling a 1/4 deg ocean and ice restarts to 1 deg. I haven't tested other destination resolutions, and some work is still needed on the ice side.
Using the DATM configuration, I was able to generate warm-start files for the ocean and ice from the 1/4deg restart file. I ran out the warm-start case for 3 days. The ocean model fields appear reasonable when compared to the 1/4 deg control case. This is the first 6h-mean SSH for the 1/4deg (left) and 1deg (right) warmstart
After 3 days, SSH for the the same 1-deg configuration started from initial Z-level T,S (left) as is done currently vs. the warm-start case (right)
The following shows the CICE6 restart vectors mapped from the 1/4 deg tripole (top) to the 1 deg tripole (bottom)
The vector speed at 1000m for the 1/4deg ocean (left) and the warm-start 1-deg ocean (right) after ~3 days. The same field when starting from initial T,S only is also shown.
Using restart files obtained from https://noaa-ufs-gefsv13replay-pds.s3.amazonaws.com/2021/03/2021032206/
provided by @LydiaStefanova-NOAA, I ran the fully coupled model at both 1deg and 1/2 deg using the new warmstarts. Both models completed w/o issue and comparisons like those above (which were using a DATM configuration) indicate that the warmstarts are working as expected.
Using the replay ICs as 1deg warm starts for CICE show a significantly different thickness field compared to what is used currently in the RTs. Below is the ice thickness after 3-d for the fully coupled model for the current ICs (left) vs the new warm start ICs (right):
Depending on resolution, the ICs for the MOM6 and CICE6 models in the coupled configurations in the UWM RTs have different sources.
For low resolution, I believe the MOM6 ICs are provided as T,S fields on Z-levels from CFSR MOM4 (5?) restarts. (This should be verified.) For the 1/4 deg resolution, the ICs are used from a Travis Sluka's DA run using MOM6.
CICE6 ICs are generated from the CPC reanalysis at 1/4deg. This is a SIS/SIS2 model, which is contains velocities on the C-grid staggers (N,E faces). CICE6 is a B-grid model for dynamics (velocities on the corners). A NCL script, modified from one originally written by Dave Baily (NCAR) and Xingren Wu (EMC) is used to create CICE6 ICs at 1/4deg and lower resolutions. It was recently found that this procedure produces incorrect velocities for the CICE6 model ICs. See https://github.com/ufs-community/ufs-weather-model/issues/2163. While the model starts w/ incorrect ICs, after about 24 hours the velocities have recovered and restarts produced as the model integrates are correctly oriented.
ICs are now available from the marine DA system, but are only available at 1/4deg and 5deg. A new utility is needed to down-sample the MOM6 and CICE6 restarts from 1/4deg to generate warmstarts files lower resolutions.
For background, the following shows the existing issue w/ the CICE6 ICs that were produced from the CPC SIS2 restarts. The figure shows the 1/4deg IC velocities used by the DATM configurations for 2011-10-01. Vectors within the marked areas show regions of opposite vector orientations on adjacent rows. These regions also appear in the down-scaled warmstart files created using the current (NCL) procedure. The NCL procedure is not a viable alternative, even if it were converted to a F90 routine. Therefore, a new utility is required.