Skip to main content

Banner image of dust, lightening and irrigation

NU-WRF Version 8 "Bjerknes" (v8-wrf371-lis71rp7)

NU-WRF Version 8 ("Bjerknes") was released on 23 June 2016.

WHAT CHANGED:

  • Upgrade to WRF 3.7.1, WPS 3.7.1, PREP_CHEM_SOURCES 1.5, RIP 4.6.7, UPP 3, and MET 5.1, plus community patches released through May 2016.
  • Upgrade to LIS 7.1rp7 and LDT 7.1rp2, including fixes for Mercator map projection.
  • Upgraded Intel compilers to 15.0.3.187.  Default MPI implementation on both Discover and Pleiades is SGI MPT.  Intel MPI 5.1.2.150 is also supported on Discover (for legacy Sandy Bridge nodes).  GFORTRAN 4.9.0 and Open MPI 1.7.3 are temporarily supported for debugging (specifically with VALGRIND).  Intel MPI 4 support on Pleiades is retired.
  • Removed -xSSE4.2 hardware specific Intel compiler flag.  Tests show no consistent performance improvement.
  • Modified build system to require MPI for WRF, WPS, and UPP (no serial builds permitted).
  • External libraries are now packaged in separate NUWRFLIB collection.  This NU-WRF release links against NUWRFLIB 8r1 on Discover and Pleiades.  This includes new dependencies (CAIRO, FREETYPE, GHOSTSCRIPT, and PIXMAN) and updates to older dependencies (e.g, NETCDF).
  • Added sample namelist, lis.config, and ldt.config files for multiple test cases.  The LIS cases include use of SRTM 30-sec topography and "slope-aspect" correction which improves the spin-up of soil moisture.
  • Updated SST2WRF processing:  Changed output mask to separately indicate "low confidence" and "high confidence" SST points.  "Low confidence" SST data are now included in output (originally were rejected, but this approach has a high false alarm rate in the C3VP test case).
  • Updated METGRID processing:  SST is now interpolated with parabolic, bi-linear, weighted 4-pt, or weighted 16-pt average (tried in order until one works).  When processing TAVGSFC (from AVG_TSFC preprocessor), METGRID tries parabolic before bi-linear interpolation.  This helps reduce data voids and discontinuities in (blended) SST field.
  • Updated REAL preprocessing:
    1. Preserves LIS greenness and albedo data but processes time-varying LAI data from GEOGRID and METGRID.
    2. Changed SST processing to check for obviously bad ocean SSTs when 30-sec MODIS+Lakes land data are used.  Also, disabled use of TAVGSFC as a replacement for bad ocean SST (it seems better to just use skin temperature in this case).
  • Added option to skip compiling WRF CLM4 code.  This saves about 10 minutes of build time.
  • WRFV3 files module_first_rk_step_part1.F and module_comm_dm_4.f90 are now on the "do not optimize" list.  This saves about 10 minutes of build time.
  • Multiple bug fixes to prevent memory corruption wiht arrays allocated with size 1 instead of the WRF "memory" dimensions.
  • Updated GEOGRID.TBL files to make GEOGRID.TBL.ARW_CHEM_NUWRF a strict superset of GEOGRID.TBL.ARW_CHEM, with the latter now a strict superset of GEOGRID.TBL.ARW.
  • Updated dynamic dust emissions.  The temperature of the second soil layer is now used instead of skin temperature to detect frozen soil and shut off emissions (this fixes a diurnal cycle bias).  Also, the threshold for snow depth is raised to 0.01 m (was zero).
  • Increased geopotential height threshold to discriminate between oceans and lakes in sea salt emission calculation.
  • Connected the GOCART sulfate chemistry with the RADM2 gas phase chemistry to account for the heterogeneous conversion of SO2 to sulfate.  Previously such conversion was not sufficently represented in RADM2.
  • Redistribution of BC and OC emissions into BC1, BC2, OC1, and OC2 aerosol species.  This makes the NU-WRF aerosol algorithm consistent with the global GOCART version.
  • Added handling of both cumulative diabatic heating (K) and "snap-shot" diabatic heating rates (K/s) from 3ICE and 4ICE microphysics schemes.  Cumulative heating is now written to restart file.
  • 4ICE microphysics updates:
    1. Change to "melting" diabatic heating rate to account for wet hail accretion.
    2. Adjusted intercept scaling factors are now used to calculate reflectivity.
  • 3ICE microphysics update:  Bug fix for warm rain code.
  • Upgraded G-SDSU to version 3.5.1.  Changes include:
    1. Add support for Thompson microphysics scheme.
    2. Visible-IR simulator is more stable (less run-time crashes).
    3. New statistical model for CRM diagnosis.
    4. Updated G-SDSU User Guide
  • CASA2WRF updated to add bug fixes for the following:
    1. CO2 flux output files now has CO2 flux for time t and flux tendency between time t and time t+1.
    2. Ocean CO2 and fossil fuel data is on the corner of the grid instead of the middle point of the grid.  All other components are at the midpoint of the grid.
  • Added symbolic link to GODDARDRAD_SSLUT in WRFV3/run directory.
  • Added fetch_rss_sst.sh script to download RSS SST data (more fault tolerant than Run_SST.csh)
  • Added batch scripts to run WPS utility AVG_TSFC.
  • Updated Discover SLURM batch scripts to remove references to sp1 constraint (no longer exists).
  • Multiple bug fixes to initialize variables

KNOWN BUGS:

  • Grell-Freitas cumulus scheme (cu_physics=3) is sensitive to CPU count (inherited from WRF 3.7.1).  No fix.
  • LIS only supports up to 3 grids.
  • CASA2WRF is not using netCDF4/HDF5 compression in output.
  • Intel MPI hangs on Discover for large (> 2000 cpu) jobs.  Use SGI MPT instead for big jobs.
  • Bad TEMF PBL restarts (inherited from WRF 3.7.1).  No fix.
  • Unphysical results if urban physics only used on some grids (inherited from WRF 3.7.1).  Use urban physics on all grids or on none.

 

Jump to Top