[Wrf-users] incorrect diffwrf output for standard tests

[Verdi March]

Hi,

I encounter problems for WRF-ARW v2.1.1 with two benchmark dataset,
conus12km and conus2.5km, which are frequently used in various
WRF-related publications (http://box.mmm.ucar.edu/wrf/WG2/bench/).

After messing around with various permutations of compiler flags and
systems, I suspect that my problems might be due to netcdf version
(mine is v3.6.2). Is there a "recommended" netcdf for WRF v2.1.1, or
at least netcdf which has been shown to be well tested?

I've done very extensive testings on two systems, (i) Solaris + Sun
compilers 12 + Sun CT 7.1, and (ii) Linux x64 (SLES 10) + PGI 7.1 +
Scali MPI. Both systems show consistent behaviors when diffwrf-ing
the reference output (wrf_reference) and my output, even without any
compiler optimizations:

- conus2.5km:
      NetCDF error: NetCDF: Unknown file format
      NetCDF error: NetCDF: Unknown file format
      NetCDF error in wrf_io.F90, line 926
      NetCDF error in wrf_io.F90, line 926
      error opening wrfout_d01_2005-06-04_06:00:00
      Status =  -1021

  The code snippet around line 926 in external/io_netcdf/wrf_io.F90:
      stat = NF_OPEN(FileName, NF_NOWRITE, DH%NCID)
      call netcdf_err(stat,Status)
      if(Status /= WRF_NO_ERR) then
        write(msg,*) 'NetCDF error in ',__FILE__,', line', __LINE__
        call wrf_debug ( WARN , TRIM(msg))
        return
      endif

- conus12km: diffwrf doesn't report two lines which are started with PHB
  and PB. For the other lines, my diffwrf's output result is quite
  accurate compared to the reference diffwrf's output
(http://www.mmm.ucar.edu/wrf/WG2/bench/conus12km_2001/diffwrf_conus12.txt).

Since these two data are widely used (I've even seen recent publications
dated in 2007), I really believe there's a good explanation somewhere
out there on my particular problems.


Regards,
Verdi