WRF-NMM Users Page

WRF Modeling System Updates

The WRF model tar file has been updated to Version 3.0 on April 4, 2008

The WRF Pre-Processing System (WPS) has replaced SI in Version 3.

Note: The WRFV3 is NOT backward compatible with old input files. One must rerun WPS to generate data for real_nmm/wrf.

WRF Version 3.0 new features, updates and bug fixes

WRF-NMM specific updates

  • Added capability to use WPS output with NMM
  • Divergence damping was enhanced only for the external mode.
  • Horizontal diffusion had not been allowed to act at all over much of the elevated terrain; now it is allowed to operate on more of the terrain where the slopes are quite steep.

General updates


  • Microphysics
    • Ferrier scheme
      • During melting precipitation ice particles are assumed to have the same mean diameter (1 mm) as at the freezing level.
      • Two changes intended to increase the presence of supercooled liquid water and improve forecast products for use in aircraft icing algorithms:
      • The temperature at which small amounts of supercooled liquid water, if present, are assumed to be glaciated to ice was lowered from -30C to -40C.
      • The temperature at which ice nucleation is allowed to occur was lowered from -5C to -15C based on aircraft icing observations.
    • WSM6
      • Use snow/graupel combined fall speed.
    • WSM5
      • Performance improvement. Will change result within roundoff.
  • Cumulus
    • BMJ
      • Triggering of deep and shallow convection is considered only for grid points with positive cape throughout a parcel's ascent; the search for parcel instability is extended to include not only whether the most unstable (highest theta-e) parcel can support convection, but also whether parcels originating at higher levels become positively buoyant when lifted to their LCL. Convective adjustments are made with respect to the parcel associated with the greatest instability (largest CAPE )
      • The search for the most unstable parcel is extended from the lowest twenty percent of the atmosphere to the lowest 40 percent of the atmosphere.
      • Water loading effects are now included in assessing the buoyant instability of parcels from which a revised (lower) cloud top is determined to be at the highest level of positive buoyancy.
      • The latent heat of vaporization used to calculate equivalent potential temperatures during model integration is made to be consistent with the value used in generating the initial lookup tables.
      • When a grid point fails the entropy check for deep convection but still has positive CAPE , changes in temperature and moisture by shallow convection are then considered at these so-called "swap" points. The first-guess estimate for the top of shallow convection is based on the highest level where the parcel remains positively buoyant (this is more restrictive than positive CAPE ), and the vertical extent of shallow convection is not to exceed 0.2 times the atmospheric pressure depth (e.g., 200 hPa for a surface pressure of 1000 hPa). A final adjustment is made to the top of shallow convection in which it can extend to higher altitudes if the mean ambient relative humidity (RH) in the cloud layer exceeds a threshold RH while remaining positively buoyant (i.e. CAPE greater than 0). The threshold RH is based on the RH at cloud base that is consistent with a deficit saturation pressure of 25 mb (usually near 90%). (The maximum cloud top height for shallow convection remains limited to 450 hPa.)
      • The first-guess reference temperatures in the upper-half of shallow convective clouds are limited to be no more than 1 deg C colder than the ambient temperature.
  • Land Surface Physics
    • Unified Noah
      • Improved emissivity treatment over snow. New capability for WRF-NMM.
    • RUC LSM
      • minor update. Improved initialization from NAM data.
    • LANDUSE.TBL: some changes to emissivity values.
  • Surface Layer Physics
    • MM5 surface layer scheme (sf_sfclay_physics = 1) now uses convective velocity following Wyngaard (old MM5 method) to help in zero wind case.
  • PBL
    • YSU: improvement for stable PBL. New capability for WRF-NMM.


  • No need to add all new namelist record in the namelist file. The code will not abort because of missing namelist records. A warning will be printed for information.
  • One namelist variable is removed: dyn_opt


  • Streamlined infrastructure and memory utilization
    • Dynamic core selection now done at compile time
    • Reduced compile-time expansion of complex routines allowing more of WRF to be compiled with full optimization
    • Memory for physics packages conditionally allocated at run time
    • New, scalable data structures for lateral boundary conditions
    • Simplified build unified over WRF models, Chemistry, and Var
  • New or enhanced support for computer systems (compilers):
    • IBM Blue Gene and Power series (Xlf)
    • Linux x86, x86_64, and ia64 (PGI, Intel, Pathscale, g95, gfortran)
    • WinCCS (PGI)
    • Intel Mac (PGI, Intel, g95, and gfortran compilers)
    • Cray XT3/4 (PGI, Pathscale)
    • SGI Altix and Origin (Intel)
    • HP
    • NEC SX series
    • MPI support: MPICH (various), vendor versions, and Intel OpenMPI
  • Miscellaneous:
    • Single executable ensemble support
    • Parallel NetCDF and enhanced Quilting
    • ESMF component capability
    • Full RSL_LITE support for nesting, transposes, periodic BC's, etc. RSL communication package has been retired in V3
    • GPU accelerated WSM5 microphysics (experimental)