Tested changes for efficient SMC grid coupling #16
Conversation
|
These are the weight files to run the new regtest. |
|
Hi Juan, |
|
Attaching the result of the coupling regtests. There are small rounding differences in the OASACM and OASOCM tests, which I assume they are due to changing the order of some lines in w3oacpmd.f90 . |
|
In fact the differences look like they are caused by different OASIS restart files, although supposedly I am using the same. |
|
ok I supposed the differences are due to the bug NOAA-EMC#259 |
Sure, I will do once the bug fix is merged to the staging branch, and I merge this branch to pick the bug fixes. |
|
After updating the branch to merge the latest coupling bug fix, the restests show no differences in the original coupling tests - although there are some differences in the new OASACM2-5. |
ukmo-ccbunney
force-pushed
the
staging_oct20
branch
from
e698b48 to
2ae6342
Compare
There was a problem hiding this comment.
These changes look good to me @ukmo-juan-castillo other than the one comment I have made regarding the DLAT/DLON comment.
model/ftn/w3oacpmd.ftn
Outdated
| !/SMC ! ---------------------------------- | ||
| !/SMC NNODES = NSEA | ||
| !/SMC ! | ||
| !/SMC ! Distinguish between 2 and 4-level SMC models |
There was a problem hiding this comment.
Only a small thing:
I think this comment is a bit misleading, the DLON and DLAT values are the smallest grid cell increments (i.e. the tier 1 cells). The number of tiers could (theoretically!) be set to any value, although 2, 3 or 4 are the likely to be the most common.
ukmo-juan-castillo
force-pushed
the
fb_SMC_coupling
branch
from
9ef5b43 to
7711ce6
Compare
* Fb oasis t+0 (#13) In the original code a coupling lag had to be used, coupling took place at the last time step but not at the first, and the coupling fields had to be written in an oasis restart file. With this ticket, the program can run in a different way by not using a coupling lag, coupling taking place at the first time step but not the last, and the coupling fields are written in the wave restart file - no extra oasis restart file needed. * Fb uprstr inp (#15) Enhancement to improve the way in which input data is read in and logged by the ww3_uprstr program: * read in variables specific to the update process selected * output the values provided in the ww3_uprstr.out log file * update the .inp template file and regtests to improve clarity and work with the changes * add capability to read inputs from a namelist (ww3_uprstr.nml) file * Bf unconforming where in coupling routines (#17) * Fix non-conforming WHERE statements in coupled routines * Stop comparing history lines in OASIS rmp files * Changes for efficient SMC grid coupling (#16) * Changes for efficient SMC grid coupling * Ensure consistency between SMC coupled test nml and inp files Co-authored-by: Juan Manuel Castillo Sanchez <48921434+ukmo-juan-castillo@users.noreply.github.com> Co-authored-by: Andy Saulter <48921142+ukmo-ansaulter@users.noreply.github.com>
The problem with the weight files has been traced to the grid file that OASIS generates. The grid file is perfectly fine, but because the domain used for the regtest is quite small and the grid dimensions do not contain a 'units' attribute, the ECMF program used to generate the weight files interprets the lat and lon units as radians, when they are in fact degrees. This should not be a problem when running with larger domains, or in the worst case, one can add the unites to the grid file generated by OASIS by hand. |
Modifications to tackle issue 250:
NOAA-EMC#250
This branch was successfully tested using the Met Office amm15 and indsl2 configurations, and I made sure that the OASIS grid file contained the same information as the operational version of the code, although with a different format. Jian-Guo provided the SMC grid that was added in the regtests, and he successfully run a propagation test.
The remap files from the toy model to the wave smc grid have to be added to data_regtests, as for some reason they could not be generated on the fly by OASIS and I had to generate them offline - I did not have this problem with the amm15 and indsl2 grids, though. I will attach them as separate files.
I will later provide a new commit with an update of the files ww3_grid_OASACM6.nml and ww3_shel_OASACM6.nml so that they are consistent with their *.inp versions.