CMAQ v5.5 crashes on Day 2 with OPACONC error and “Invalid file name argument CTM_CONC_1

WRF-CMAQ
1

Hi everyone,

I am running CMAQ v5.5 on an HPC cluster. The simulation period is April 2026.

I originally had a MEGAN-related issue, but after fixing the restart configuration the model is now able to complete Day 1 successfully and start Day 2.

The output files from Day 1 are generated correctly, including:

CCTM_CGRID_v55_gcc_D03_Apr2026_20260401.nc

CCTM_CONC_v55_gcc_D03_Apr2026_20260401.nc

CCTM_MSOILOUT_v55_gcc_D03_Apr2026_20260401.nc

Day 2 starts normally:

Processing 2026-04-02

and a Day 2 concentration file is also created:

CCTM_CONC_v55_gcc_D03_Apr2026_20260402.nc

However, the model crashes before completion.

The batch log ends with:

MPI_ABORT was invoked on rank 3 in communicator MPI_COMM_WORLD
ERROR: CMAQ failed on 2026-04-02

The processor-specific logs show:

*** ERROR ABORT in subroutine OPACONC on PE 001
*** ERROR ABORT in subroutine OPACONC on PE 002
*** ERROR ABORT in subroutine OPACONC on PE 003

Immediately before the OPACONC error I see:

Maximum current record number         4

>>--->> WARNING in subroutine DESC3
Invalid file name argument "CTM_CONC_1"

*** ERROR ABORT in subroutine OPACONC on PE 001

My run script contains:

setenv CTM_CONC_1 $OUTDIR/CCTM_CONC_${CTM_APPL}.nc

and the output directory exists and is writable.

Has anyone encountered an OPACONC crash associated with:

WARNING in subroutine DESC3
Invalid file name argument "CTM_CONC_1"

Any suggestions for additional diagnostics would be greatly appreciated.

Thank you.

2

Is your second day run failing right at the beginning, or does it run a few hours into the simulation before it fails?

Just to make sure, if you have existing output files for 20260402 when you try to restart the model on 20260402, make sure to delete the 20260402 output files first before restarting.

If you still see the error after this, please post the error message and maybe 10-20 lines immediately above it from one of the CTM_LOG files.

3

OPACONC calls OPEN3 on CTM_ACONC_1. When the previous run was scancel 'd mid-write, it leaves a partial ACONC file (truncated header). On restart, depending on your run script’s disposition setting, OPEN3 it either refuses to clobber an existing file or fails the header-match check, and that surfaces as the ERROR ABORT in subroutine OPACONC. I believe if any partial file or stale log of 2026-04-02 is still there in your output directory, you can clean it and rerun.