How to Compile?

How to Run?

WRF Namelist

Known Problems and Fixes

Graphic Tools





WRF Model Version 3.7: UPDATES

WRF Model Update

The WRF model tar file has been updated to Version 3.7 on April 17, 2015. For known problems in V3.7, please click here.

The WRF Pre-Processing System (WPS) has been updated to Version 3.7.

Note: V3.7 WRF can use wrfinput and wrfbdy files from V3.6 version. Please note that the VEGPARM.TBL in run/ is updated, and must be used with 3.7 code.

As always, be cautious when using new options.

If you are interested in seeing how V3.7 has been tested, click here.

New in Version 3.7:


- Cumulus parameterization:

* New Tiedtke scheme (cu_physics=16): This version is similar to that used in REGCM4 and ECMWR cy40r1. The main differences from the old Tiedtke schemes are:

(a) New trigger functions for deep and shallow convections;
(b) Non-equilibrium situations are considered in the closure for deep convection;
(c) New convection time scale for deep convection closure
(d) New entrainment and detrainment rates for all convection types
(e) New formula for conversion from cloud water/ice to rain/snow
(f) Different way to include cloud scale pressure gradients
This option is experimental. (Jakob and Siebesma 2003; Bechtold et al. 2004, 2008, 2014; Sundqvist 1978; Gregory et al. 1997; Wu and Yanai 1994; IFS documentation - cy33r1, cy37r2, cy38r1, cy40r1) (Thanks to Chunxi Zhang of Univ. of Hawaii)

* Multi-scale Kain-Fritsch scheme (cu_physics=11, note that this option works only with YSU PBL in 3.7):

(a) With diagnosed deep and shallow KF cloud fraction;
(b) Scale-dependent Dynamic adjustment timescale for KF clouds;
(c) Scale-dependent LCL-based entrainment methodology;
(d) Scale-dependent fallout rate;
(e) Scale-dependent stabilization capacity;
(f) Estimation and feeback of updraft vertical velocities back to gridscale vertical velocities;
(g) new Trigger function based on Bechtold method.
(Zheng et al. 2015, MWR) (Thanks to Kiran Alapaty and Jerry Herwehe of EPA)

- PBL:

* Shin-Hong PBL (bl_pbl_physics = 11): scale-dependency for vertical transport in convective PBL while the vertical mixing in stable PBL and free atmosphere follows YSU (Shin and Hong, 2015, (Thanks to Hailey Shin of NCAR/ASP)
* A top-down mixing option in YSU driven by radiative cooling (ysu_topdown_pblmix = 1). (Thanks to Travis Wilson of UCLA).

- Land-surface models:

* New hydrological processes are added to single-layer UCM: (details)

(a) Oasis effect;
(b) Urban irrigation;
(c) Anthropogenic latent heat;
(d) Evaporation over impervious surface;
(e) Multi-layer green roof
(Thanks to Jiachuan Yang, Zhi-Hua Wang of Arizona State Univ and Shiguang Miao of IUM/CMA, China)

* PX LSM: added impervious surface physics. (Thanks to EPA)

- Radiation and cloud effect:

* A compute-performance optimized option for RRTMG radiation (ra_lw_physics, ra_sw_physics = 24): On conventional and Intel Xeon Phi processors, no other action is needed. For GPU, the code must have been configured for “PGI accelerator compiler” and CUDA shared object libraries must be accessible at run time. Results agree with the original RRTMG to within floating point roundoff but are not bit-for-bit identical. Certain coupling capabilities with microphysics are not supported. (Thanks to Mike Iacono and David Berthiaume of AER and John Michalakes of NCEP). In one test on yellowstone using Intel compiler, a speedup of 8% was seen. However the improvement will be configuration and application dependent.

* A new cloud fraction option (icloud = 3). This is a RH-based scheme following Sundqvist et al. (1989, MWR). (Read more.) (Thanks to Greg Thompson of NCAR/RAL)

- Others

* 9 sec, NLCD 2006/2011 landuse data. (Thanks to EPA)
9 sec impervious surface and canopy fraction datasets derived from NLCD2011 landuse dataset (Thanks to EPA)

Improvements and Bug fixes:


Atmospheric Radiation:

- Diffused, direct and direct normal irradiance calculation is extended to CAM and FLG radiation options. (Thanks to A. Montornes and B. Codina, University of Barcelona, Spain)
- RRTM: fixes a diagnostic calculation
- RRTMG: bug fix, and thread-safe change. The default ozone input optoin has been changed to CAM (o3input = 2). However, one must download a fix to allow this to happen.


- NSSL 2-moment scheme: Updated (Thanks to Ted Mansell of NSSL)

(a) Improvents for stratiform development in mesoscale convective systems. Produces snow particles (instead of graupel) from sufficiently small frozen drops.
(b) Fixes a problem with droplet issue for nssl_2mom.
(c) Adds terms for wet scavenging used by WRF-CHEM, also turns off droplet nucleation for wrf-chem when progn = 1
(d) Adds a radar bright-band enhancement for melting snow
(e) Changes default primary ice nucleation to Phillips/DeMott (lower ice nuclei concentration), also fixes an error in the Cooper parameterization.
(f) Improvements to drop freezing by ice crystal capture.
(g) Reduced droplet collection efficiency by graupel and hail, resulting in slightly lower precipitation efficiency
(h) Enabled sedimentation of cloud droplets, which has a noticable effect on rain autoconversion
(i) Changes default snow fall speed to lower values as used in Ferrier (1994)
(j) Fixes stray tiny amounts of droplet mass at temperatures below -40 C
(k) passing effective radii to RRTMG.

- Milbrandt: updated, and more robust. User is advised to avoid using the version in 3.6.
- Thompson: updated, fixed the restart problem for the aerosol version; and improved initialization. Look-up tables will be computed and written out in a first run, and can be used in the future runs.
- WSM, WDM schemes: effective radii from cloud water/ice and snow will be computed and passed to RRTMG radiation (Thanks to Sooya Bae of KIAPS, S. Korea)


- Tiedtke (cu_physics=6): rewritten in modern Fortran, and fixes a bug related to initial cloud base wind calculation (small effect), and changes adjustment time from 1800 sec to 2400 sec (more appropriate for grid sizes of 15 - 40 km)

Planetary boundary layer and surface layer physics:

- Revised and original MM5 surface layer schemes (sf_sfclay_physics = 1, 91): heat and moisture exchange coefficients over water is changed from Carlson-Boland to COARE 3. This helps to reduce fluxes over ocean. A term proportional to 1/u* is added to represent roughness of ocean at very low wind speed.
- Revised MM5 surface layer scheme: the code is sped up, and will run in comparable time as with the original scheme.
- ACM2 PBL: updated.
- Old QNSE (bl_pbl_physics=94): removed.

Land Surface physics:

- NoahMP: There are a number of improvements in this release

(a) option to use the WRF microphysics frozen precipitation partitioning
(b) addition of precipitation advected heat to the surface energy budget
(c) dynamic vegetation bug fixes and initialization of leaf carbon mass through wrfinput LAI
(d) allow for patchy-snow surface temperature in sensible heat fluxes
(Thanks to Mike Barlage of RAL/NCAR)

- Noah: fixes bugs to prevent discountinous behavior for soil ice melting, and add a dew term to QFX over snow cover. Sets minimum LAI. For mosaic option, fields are now interpolated to the nest and allowed to fed back to parent domain.

- RUC LSM: updated:

(a) use "mosaic" approach for treating grid cells partially covered with snow.
(b) ensures water and energy conservation, and provides water and energy budget components.
(c) small changes to remove inconsistencies when the Lake Model is turned on.
(d) update parameters in VEGPARM.TBL

- PX LSM: updated to work with fractional seaice.

- SSiB: updated.

WRF-Hydro 3.0: (more info)

Version 3 of the WRF-Hydro system includes numerous enhancements to model componentry and to model computational performance. The main modification made include:

(a) Numerous updates to the NoahMP land surface model;
(b) Parallelization of the NoahMP land model driver;
(c) Parallelization of I/O to improve model computational performance for large domains and very high resolution runs including large netcdf file support with netcdf4;
(d) Improved memory management for large domain/high resolution runs;
(e) Reformulation of the exponential bucket model to permit fill/spill capability in addition to normal exponential model discharge;
(f) Support for gfortran compilation on MacOS;
(g) Expanded ArcGIS pre-processing support for WRF-Hydro
(h) R-based statistical evaluation package with USGS streamflow data support;
(i) Expanded script library for pre-/post-processing of WRF-Hydro inputs/outputs;
(j) Expanded test case library.


- Topographical gravity wave drag: updated and includes flow blocking (Thanks to Hyun-Joo Choi of KIAPS, S. Korea)

- Topo shading: fixes a bug for shading effect when no terrain slope exists (Thanks to Jeff Mirocha of LLNL). Also unified the diffused fraction calculation.

- SKEBS: updated.

(a) The ability to manufacture a random pattern (random perturbation) that is able to be used by any physics or dynamics scheme;
(b) The results are bit-for-bit with a restart;
(c) The code uses a more robust initialization of the random number seeds (some compilers were previously ignoring the seeds);
(d) Cleaned up user interface, where the user requests which major option to activate: skebs or random perturbation;
(e) perturbation BC is added for chemistry variable;.
(f) Major options are now independent and give reproducible patterns regardless of other selected major options.

Moving nests:

- A limitation is noted in moving nest: the number of vertical levels needs to be less than 55.


- improved specified bdy for long simulations (Thanks for the report from Heimo Truhetz, Sebastian Knist and colleagues at Univ Graz, Austria)
- check p_top against data in input (start_em)
- prevent using too many processors in real which may break the code to create LBCs
- fixed a problem with restarting in the last period when LBC is available.
- LAI, if processed in geogrid, will always be in the input
- if water/ice friendly aerosols are present from another WRF model run and converted to grib data, it can be processed in program real
- recompute density and pressure just like that in the model in program real to make pressure (and density) fields the same in wrfinput and 0 h wrfout file
- minor correction to cubic-spline vertical interpolation
- makes T in output in grib 1 output T+300 (potential temp)
- friendly message to catch model top higher than soundings in idealized case (seabreeze, quarter_ss)
- add capability to input SST in les and convrad cases
- adds option to define land and wate for idealized cases, ideal_xland
- fixes large nest ratio problem (though not recommended)
- DFI: turn off other advection except for Qv
- prevents DFI to be activated in a model restart run
- trajectory code: fixes a dimension error.


- AFWA diagnostics: updated
- Climate diagnostics: fixes a bug related to restart. Restart needs to occur at the end of climate diagnostics accumulation period.
- NWP diagnostics: added maximum hail diameter diagnostics that are consistent with each of the microphysics in history output intervals (Thanks to Greg Thompson of NCAR/RAL)


- Vertical nesting (Thanks Katie Lundquist of LLNL) - use with care.

New idealized cases:

- Adds a new ideal case: convective-radiative equilibrium case (or em_convrad)
- Adds a new shallow convection LES test case (in em_les directory)


- Adds an option to include Qv and QC tendencies in sovler similar to the use of h_diabatic term (option use_q_diabatic). With this option the model will produce correct symmetric solution in a "moist benchmark" case where the flow is saturated and (strong) mean wind exists. (Thanks to George Bryan of NCAR, ) Real-data testing indicates that smaller time step may be needed when this option is used.
- Fixes a problem in LES runs when mean wind and strong gradient of moisture are present (option use_theta_m = 1) (Thanks to PNNL for reporting the problem.) Note that this doesn't work with nest yet.
- Prevents idealized case to use hypsometric_opt = 2 option.
- Use of nba_mij array to store stress term even if sfs_opt isn't used.


- Changed the format to choose compile options; if build is successful, one gets a message at the end; more information on model version, computer type, and compiler version is included in the begining of the compile
- More CF compliant - if one writes output files one per time period
- Support less-than 4-digit year
- Adds option to compile on ppc64le
- Supports real*8 integration on Intel/PGI
- Memory reduction (for a typical run, the memory could be down 30%)
- Improves fft for polar filtering in global application
- Changed 'call flush' to 'flush'
- Bug fixes.


Previous Update

WRF V3.0.1
WRF V3.0.1.1


Home -- Model System -- User Support -- Doc / Pub -- Links -- Download -- WRF Real-time Forecast