init_atmosphere crashed when trying to generating L60 conus12km init.nc from RRFSv1

[guoqing-noaa]

My init_atmosphere crashed when trying to generating init.nc for conus12km L60 from RRFSv1 grib2 products.
Here is the crash message:

----------------------------------------------------------------------
Beginning MPAS-init_atmosphere Error Log File for task      79 of     160
    Opened at 2025/12/05 09:23:56
----------------------------------------------------------------------

ERROR: extrap_type == 2 not implemented for target_z >= zf(1,nz)
ERROR: *****************************************************************
ERROR: Error in interpolation of t(k,iCell) for k=  34, iCell=         1
CRITICAL ERROR: *****************************************************************
Logging complete.  Closing file at 2025/12/05 09:23:56

Her is my rundir:
/scratch3/BMC/wrfruc/gge/rrfs2/regressionTests/rrfs12km/stmp/20251119/rrfs_ic_00_v2.1.2/det.crash/ic_00

I checked the related code snippet in mpas_init_atm_cases.F:

 9734       if (target_z >= zf(1,nz)) then
 9735          call mpas_log_write('bad z target $r zf top $r zf top-1 $r',realArgs=(/target_z,zf(1,nz),zf(1,nz-1)/))
 9736          if (extrap_type == 0) then
 9737             vertical_interp = zf(2,nz)
 9738          else if (extrap_type == 1) then
 9739             slope = (zf(2,nz) - zf(2,nz-1)) / (zf(1,nz) - zf(1,nz-1))
 9740             vertical_interp = zf(2,nz) + slope * (target_z - zf(1,nz))
 9741          else if (extrap_type == 2) then
 9742              call mpas_log_write('extrap_type == 2 not implemented for target_z >= zf(1,nz)', messageType=MPAS_LOG_ERR)
 9743              if (present(ierr)) ierr = 1
 9744              return
 9745          end if
 9746          return
 9747       end if

It looks like I should found the message bad z target... somewhere. It is odd that I cannot find this line in any output files.

[guoqing-noaa]

FWIW, we can successfully generate conus3km L60 init.nc from RRFSv1 grib2 files (only conus12km crashes).

[clark-evans]

Hi @guoqing-noaa - thanks for raising this issue. Just for our awareness, did it only occur in this one test, or is it more general to multiple tests (different cases, etc.) and thus reproducible?

[guoqing-noaa]

@clark-evans Yes, it is reproducible.
@sudeer also got a similar crash.

[clark-evans]

The error message here is at k=34, which is a relatively low altitude for this to crash. @guoqing-noaa - do you know about what altitude that corresponds to in the L60.txt file? And, just for documentation, what pressure levels are represented in the RRFS input GRIB2 data?

[guoqing-noaa]

@clark-evans k=34 is at 5723.531 km.
The input RRFS GRIB2 data should be fine as I can create an conus3km init.nc using the same vertical levels and GRIB2 data.

[keenaneure]

I am getting a similar issue in my runs, but on the LBC task instead.

[guoqing-noaa]

@keenaneure Could you provide more detailed information here and also save a standalone case so people can repeat it?

I suspect there might be a related bug somewhere with init_atmosphere. tag @clark-evans

[clark-evans]

This might be a good issue to open in the authoritative repo, particularly if it can be replicated with NCAR's code. (And if it can't, that'd be useful for us to know as well - though I know it's not an easy swapout for a test.)

[guoqing-noaa]

This might be a good issue to open in the authoritative repo, particularly if it can be replicated with NCAR's code. (And if it can't, that'd be useful for us to know as well - though I know it's not an easy swapout for a test.)

Thanks for the suggestion. We will think about it.