Can't use the standard GHA environments since ESMF takes a long time to build, especially with MPI support. Instead, can create and use custom Docker image with ESMF already built, similar to what GEOS-Chem does. Ultimately could create a build matrix with different compilers, etc., but probably will start with Ubuntu 22.04 LTS + GFortran v11, NetCDF, etc. via apt.

Sample config with HTAP PM2.5 and PM10 only. Data files lossily compressed with NCO total ~ 500 MB. But the initial version could just check for successful NEXUS build. Then, run the sample case (Barry suggested running for 6 hours). Then, check output is consistent with reference output.

Currently you can't build NEXUS without MPI:

However, HEMCO doesn't require MPI, so in principle it should be fine to build/run without MPI.

From my experiments, currently you can run with non-MPI ("mpiuni") ESMF, though NEXUS fails in debug mode since ESMF_FieldWrite doesn't work with mpiuni ESMF.

CMake supports this but currently we have that setting the CMAKE_Fortran_COMPILER env var is required:

Currently, the driver for NEXUS is a pure ESMF driver. The initial move to a NUOPC component is to create the NUOPC CAP and a single model driver. This would allow us to have 1) a future capability to run NEXUS offline and 2) the first step to the NEXUS component for the UFS-weather-model.

The forecast task 'run_fcst' of Online-CMAQ fails with the latest hash of NEXUS.
The current hash of NEXUS ("c05ee48") works well in Online-CMAQ.
You can find the log files:

• On wcoss2 (w/o point source): /lfs/h2/emc/lam/noscrub/chan-hoo.jeon/online-cmaq-nexus-test/expt_dirs/aqm_nexus_test/log
• On Hera (w/ point source): /scratch1/NCEPDEV/fv3-cam/Chan-hoo.Jeon/online-cmaq-nexus-test/expt_dirs/aqm_nexus_test/log

The ESMF logfile (PET000.ESMF_LogFIle) says:
20221029 141252.612 ERROR PET000 aqmio.F90:1238 - Field CO not defined in NetCDF data set (strerr=NetCDF: Variable not found, ncerrToCheck=-49)
20221029 141252.613 ERROR PET000 aqmio.F90:1056 - Passing error in return code
20221029 141252.613 ERROR PET000 aqmio.F90:812 - Passing error in return code
20221029 141252.613 ERROR PET000 aqm_emis_mod.F90:1055 - Passing error in return code
20221029 141252.613 ERROR PET000 aqm_comp_mod.F90:189 - Failed to update emissions
20221029 141252.613 ERROR PET000 aqm_cap.F90:596 - Passing error in return code
20221029 141252.613 ERROR PET000 AQM:src/addon/NUOPC/src/NUOPC_ModelBase.F90:2226 - Passing error in return code
20221029 141252.613 ERROR PET000 UFS Driver Grid Comp:src/addon/NUOPC/src/NUOPC_Driver.F90:3340 - Phase 'RunPhase1' Run for modelComp 2 did not return ESMF_SUCCESS
20221029 141252.613 ERROR PET000 UFS Driver Grid Comp:src/addon/NUOPC/src/NUOPC_Driver.F90:3625 - Passing error in return code
20221029 141252.613 ERROR PET000 UFS Driver Grid Comp:src/addon/NUOPC/src/NUOPC_Driver.F90:3255 - Passing error in return code
20221029 141252.613 ERROR PET000 UFS.F90:403 - Aborting UFS
20221029 141252.613 INFO PET000 Finalizing ESMF

Should we change any configurations in Online-CMAQ?

I checked out the latest NEXUS code on Cactus.

Here the arl_nexus code log information
commit c7b9af5
Merge: 6a7a994 72c51f5
Author: Barry Baker [email protected]
Date: Thu Jun 29 15:34:48 2023 -0400
Merge pull request #35 from noaa-oar-arl/feature/add_soilnox
Feature/add soilnox

However, the nexus jobs failed with the following error message.

**Traceback (most recent call last):
File "/lfs/h2/emc/physics/noscrub/jianping.huang/nwdev/packages/aqm.v7.0.82a/sorc/arl_nexus/utils/python/nexus_gfs_bio.py", line 428, in
raise SystemExit(main(parse_args()))
File "/lfs/h2/emc/physics/noscrub/jianping.huang/nwdev/packages/aqm.v7.0.82a/sorc/arl_nexus/utils/python/nexus_gfs_bio.py", line 171, in main
assert len(files) >= 2, "need at least 2 for time interp and time calcs"
AssertionError: need at least 2 for time interp and time calcs

More details can be found at
/lfs/h2/emc/ptmp/jianping.huang/emc.para/output/20200815 > vim nexus_emission_2020081506_s00.id_1688123510.log

It seems that the timezone file specified in NEXUS_Config.rc is not the recommended one. The recommended timezones_voronoi_1x1.nc allegedly does better for coastal grid cells that have ocean overlap. Has this been tried in NEXUS?