Initialize gaussian_data in gaussian_sfcanl

[CatherineThomas-NOAA]

What is wrong?

In the gaussian_sfcanl utility, the variable gaussian_data is used before it is initialized to zero (Line 273):

  gaussian_data%vfrac(row(i))  = gaussian_data%vfrac(row(i)) + s(i)*tile_data%vfrac(col(i))

Sometimes the resulting values are unreasonable (for example 1E+200).

What should have happened?

The gaussian_data variable should be initialized to zero after allocation.

What machines are impacted?

All or N/A

Steps to reproduce

• Clone and build the global-workflow
• Run the WCDA CI test at C48mx500 resolution
• Examine the gaussian sfcanl nc file and ncdump variable veg

Additional information

More details are in global-workflow issue #2669.

Do you have a proposed solution?

The gaussian_data variable should be initialized to zero after allocation.

[GeorgeGayno-NOAA]

I approve this solution. Thanks for catching my bug.

[CatherineThomas-NOAA]

@GeorgeGayno-NOAA: Here is a RUNDIR for my C48 test on Hera:

/scratch1/NCEPDEV/da/Catherine.Thomas/git/global-workflow/c48test/TMP/RUNDIRS/c48test/analcalc.410225

Let me know if you need me to run anything else.

[GeorgeGayno-NOAA]

Case set up on Hera: /scratch1/NCEPDEV/da/George.Gayno/global-workflow.git/gaussian_sfcanl

[GeorgeGayno-NOAA]

Started another test branch. Added some compiler options (576a55c) that will initialize variables to a very large value that will hopefully cause a run time failure.

The test case was rerun using the new branch. A floating overflow was detected:

 - WRITE VARIABLE alnsf
forrtl: error (72): floating overflow
Image              PC                Routine            Line        Source
gaussian_sfcanl.x  00000000004423FB  Unknown               Unknown  Unknown
libpthread-2.28.s  000014CE08968CF0  Unknown               Unknown  Unknown
gaussian_sfcanl.x  0000000000433B37  get_netcdf_var_           915  gaussian_sfcanl.f90
gaussian_sfcanl.x  000000000042C4CE  write_sfc_data_ne         888  gaussian_sfcanl.f90
gaussian_sfcanl.x  0000000000412A14  MAIN__                    374  gaussian_sfcanl.f90

[GeorgeGayno-NOAA]

All variables were initialized to zero (289d090) and the test case was re-run. It completed normally:

 - WRITE VARIABLE zc

 - NORMAL TERMINATION


     ENDING DATE-TIME    JUL 10,2024  15:47:24.617  192  WEN   2460502
     PROGRAM GAUSSIAN_SFCANL HAS ENDED.