CMAQ failing after rrtmg_sw is called on two separate simulation months

WRF-CMAQ
1

Hi all,

I am running the two-way coupled WRF v4.5.1 – CMAQ v5 using emissions from the EPA 2022v1 emission modeling platform for four seasonal months. Three of the four simulations (April 2023, Summer 2023 baseline, and Fall 2023) run successfully for both nested domains (4 km and 1.3 km). However, I am encountering issues with the January 2023 simulation.

For January 2023, I can run the December spin-up period (Dec 22–Dec 31) without any problems. But when starting January 1st, the model fails almost immediately. The rsl output shows that the model stops right after entering the shortwave radiation routine:

d01 2023-06-23_00:00:00 CALL rrtmg_sw

d01 2023-06-23_00:00:00  RRTMG SW CLWRF interpolated GHG values year:        2023  julian day:   173.000000

d01 2023-06-23_00:00:00    co2vmr:    4.2318935102095279E-004  n2ovmr:   3.3441521645392862E-007  ch4vmr:   1.9421490433496905E-006  cfc11vmr:   2.0992038831436712E-010  cfc12vmr:   4.8090092426630067E-010

Although the failure occurs inside rrtmg_sw, I am not convinced the radiation scheme is the root cause, because the same setup works for the other months.

Additionally, I am running a Summer 2023 sensitivity simulation in which the only change from the baseline is the emissions (all other inputs and namelists are identical). This case completes the first spin-up day successfully, and the diagnostic output looks normal. However, when the model proceeds into the second day (still within spin-up), it again fails immediately after calling rrtmg_sw, similar to the January case.

This pattern suggests that the emission files may be contributing to the failure, even though both inline and 2D emission files are complete, time stamps match the modeling period, file sizes and global attributes match the baseline files, and no obvious erroneous values.

Does anyone have suggestions on how to resolve the error? I’m also happy to share any relevant files or logs that may help in identifying the cause of this error.

Thank you in advance for your time and help!!

2

Hi Vlang,

When you said “running the two-way coupled WRF v4.5.1 – CMAQ v5”, I assume you are running the coupled model with feedback turned on. If not, please correct me. Please provide all the rsl files so I can get a better picture of the crash. One thing you can try is to add -traceback option (I assume you are using ifort) in the configure.wrf, and recompile your code by typing “compile em_real” after removing all *.o and *.mod files in the phys sub-directory. When it crashes again, hopefully you will see additional crashing related information in the rsl.error* files. By the way, I am curious how you put CO2 and other GHG in the WRF model.

Cheers, David

3

Thank you for your reply, David!

A quick update to this previous post. After updating biogenic inputs, the model now has a slightly different error, but it still occurs within the first ~60 seconds of the run. I have enabled additional diagnostic flags, but so far, I am not seeing anything obvious in the logs that clearly explains the failure. The model advances through initialization and begins stepping, but aborts shortly thereafter. For reference, I have rechecked all inputs (emissions, ICBC, surface files, meteorology) and all files are present, non-empty, have expected time stamps, and are consistent with inputs from other successful runs.

I am not able to upload the rsl output directly to the forum, but I am hosting the files on my GitHub page for reference.

If anyone can provide suggestions or additional insight, I would greatly appreciate the guidance!

-VL