Brian Jewett's Weather Research and Forecasting Model (WRF) page
WRF Post-processing with RIP
The RIP (Read-Interpolate-Plot) package is available from the
downloads page at NCAR; documentation is also there, at
this address.
Alternatively, go to the developer's (Mark Stoelinga)
page
at the University of Washington.
I have compiled RIP and built a script called rip_plot.csh to run the preprocessor (ripdp_wrf), run RIP,
and to create GIF images from the RIP output -- for
idealized or real-data history files from WRF. The script
and accompanying tar file (with needed executables and data files) are available below for
platforms on which I have it implemented:
To use, create a directory, copy or link in the wrfout files you want plotted, and run rip_plot.csh.
It expects to find rip_plot.tar in the same directory. It uses NCAR graphics; the local definition for
NCARG_ROOT is set in the script for each machine. Feel free to contact me with questions.
Examples of the output from RIP are available in
this 2km triggered-WRF page.
WRF 2.2
WRF version 2.2 was released 12/22/2006.
It is available from the NCAR download page.
Notes:
- bugs were found in the new Thompson (v2) microphysics; a new release is available
from the NCAR WRF "Known Problems and Fixes" page.
- many people have encountered compile problems with some of the larger top-level routines; these are often
avoided with use of -O0 or -O1 optimization directives in the configure.wrf file. More comments are on the
"Known Problems" page.
- Compiled versions of WRF 2.2 are available on NCSA teragrid machines
tungsten, mercury, cobalt and copper at
~bjewett/WRF/WRF2.2 ... I also have scripts I use to run the various steps (run_wps.csh, run_wrf.csh ...) in
the same directory. The versions on tungsten and mercury are most up-to-date.
WRF Preprocessing System (WPS)
Information on compilation and use are here.
Optimization of WRF
I have begun optimization tests of WRF 2.2 across NCSA's teragrid platforms.
- Xeon cluster Tungsten: -O2 -axN -auto -unroll0 best; configure.wrf
here.
- SGI Altix Cobalt: -O3 -unroll0 best; configure.wrf
here.
- IBM p690 Copper: optimization tests not yet begun; a working configure.wrf is
here.
- IBM IA-64 cluster Mercury: optimization tests not yet begun; a working configure.wrf is
here.
More information is available here.
Impact of changes in parameterizations in 2.2
My short 48h Katrina simulation revealed large differences in intensity between runs
with WRF 2.1.2 and with WRF 2.2. The change appears to be in the Grell parameterization.
Simulations under WRF 2.2 with Grell cumulus resulted in a simulated
central pressure 20+ mb weaker than identical runs with 2.1.2, and simulations with
WRF2.2 with Grell are 35 mb weaker after 48 hours than those with KF.
(I need to run 2.1.2 with KF to verify that those large differences are new).
More information is available here; a subset of plots follow.
| Model | Run | Configuration | 48h stats |
Pressure/wind | Wind speed | 700mb RH |
WRF
2.1.2 |
kat0b | Grell Cumulus
WSM6 microphys
2/2 diffusion (TKE)
Timing: 14:11 |
943 mb |
 |
 |
 |
WRF
2.2
RSL |
2b | Grell Cumulus
WSM6 microphys
2/2 diffusion (TKE) |
968 mb |
 |
 |
 |
| 3a | KF Cumulus
WSM6 microphys
1/4 diffusion (Smag) |
936 mb |
 |
 |
 |
Brian F. Jewett -- home page