Meteorology and Climate Center

Running the WRF Model

Experiments
ECMWF coarse-grid data and scripts to run an initial test forecast are provided.
When you are familiar with WRF, you can edit the run scripts to perform additional forecasts.
For example:

Change the domain area, resolution, time step
Experiment with the different integration schemes
Run COMSO and WRF and compare the two forecasts
Experiment with the Physical Parameterization schemes
Recompile to assess how optimization may affect speed & accuracy

PRACTICAL TIPS:

Try to make just one change at a time
Experiment with small grids and coarse spatial resolution first
Don't perform long forecasts until you have tested the configuration
Make back-up versions of the reference files before changing them

Compiling and Running the WRF Model

Connect to Tyndall
ssh -X username@tyndall.ucd.ie

USERNAME: mccNN
PASSWORD: ********
Copy and extract model code
export username='your user name here'

cd [Change to your home directory]

cp /home/pnolan/WRFV3.2.tar.gz . [Note the final dot]

tar -zxvf WRFV3.2.tar.gz
Compile WRF 3.2
cd /home/${username}/WRFV3.2/WRFV3

Edit "save_configure.wrf" to change optimization level etc (not required)

./intel_build_wrf.sh to compile WRF (see file for comments)

Choose option 9: "Linux x86_64 i486 i586 i686, ifort compiler with icc (serial)"

Choose option 1 for basic nesting.

tail -f compile.out to check progress of compilation

Depending on optimization levels, the compilation will take ~5-10mins

When code is compiled check 'compile.out' for errors

If successful the executables should appear in /home/${username}/WRFV3.2/WRFV3/main/
Compile WPS 3.2
cd /home/${username}/WRFV3.2/WPS

Edit "save_configure.wps" to change optimization level etc (not required)

./intel_build_wps.sh to compile WPS (see file for comments)

Choose option 2: PC Linux x86_64, Intel compiler serial

tail -f compile.out to check progress of compilation

When code is compiled check 'compile.out' for errors
Run a forecast of Hurricane Katrina using NCEP boundary data
Here we run a 6 hour forecast of hurricane Katrina using NCEP boundary data

The forecast is a 2-way nested run (including SST as input)

See "http://www.mmm.ucar.edu/wrf/OnLineTutorial/CASES/index.html" for further details

cd /home/${username}/WRFV3.2/Run_WRF/

Run the forecast: ./run_wrf_katrina_NCEP.sh

The WRF Input files are stored in WRF_Input_NCEP

The WRF Ouput files are stored in WRF_Output_NCEP

The data can be viewed using ncview

For example: ncview WRF_Output_NCEP/wrfout_d02_2005-08-28_0* &
To run the forecast using ECMWF 1 degree data as input run:
./run_wrf_katrina_ECMWF_1deg.sh

The WRF Input files are stored in WRF_Input_ECMWF_1deg

The WRF Ouput files are stored in WRF_Output_ECMWF_1deg
To run the forecast using the high resolution ECMWF 25km data as input:
./run_wrf_katrina_ECMWF_25km.sh

The WRF Input files are stored in WRF_Input_ECMWF_25km

The WRF Ouput files are stored in WRF_Output_ECMWF_25km

Note that since this run is high resolution over a large area it will take longer
Further Exercises: (See WRF webpage and online tutorial for help)

Run the forecast for a longer period.
Use WRF post-processing tools to analyze the data.
Add a third nest.
Run a forecast over Ireland.
Run a restart simulation.
Prepare graphics of the output using NCL (already installed on tyndall).
Recompile the code using higher optimization. Comment on potential speed-up.
Investigate the vortex tracking option of WRF (requires recompiled code).
Research a number of severe storms which affected Ireland (1958 - present).
Compare WRF-NCEP with WRF-ECMWF output.