If you've made it to this point, WRFDA should be fully compiled on your system. If you haven't already, you should make an environment variable pointing to the directory where you unpacked WRFDA (this should be the directory that you downloaded the WRFDA_V3.5.tar.gz file in).
A test data set is available on the WRFDA website:
This test case is a continental US domain case at 12Z on February 5, 2008. This was the morning of the so-called Super Tuesday Tornado Outbreak, which resulted in more than 50 deaths and over $1 billion in damage across the central and southeastern United States.
Download the test data package to the same folder you downloaded the WRFDA package to, then unpack it in the same way you unpacked the WRFDA package.
> cd $WRFDA_DIR
> tar -xvf WRFDAV3.5-testdata.tar.gz
You should now see two .tar.gz files and four directories:
> ls -al
total 234892
drwxr-xr-x 6 user 4096 Oct 29 14:39 .
drwxr-xr-x 19 user 4096 Oct 29 11:57 ..
drwxr-xr-x 2 user 4096 Mar 18 2012 be
drwxr-xr-x 9 user 4096 Mar 12 2012 ob
drwxr-xr-x 4 user 4096 Mar 9 2012 rc
drwxr-xr-x 17 user 4096 Oct 29 12:25 WRFDA
-rw-r--r-- 1 user 47726138 Oct 29 11:58 WRFDA_V3.5.tar.gz
-rw-r--r-- 1 user 192531933 Oct 29 14:31 WRFDAV3.5-testdata.tar.gz
The "WRFDA" directory is the directory where we just compiled the WRFDA system. The other three directories contain the sample data for our test case:
"be" contains our background error data in a binary file named "be.dat". This file must be generated on a case-by-case basis. The utility used to generate background error data is known as "gen_be"; instructions on how to use this will be covered in the advanced tutorial.
"ob" contains our observation data. This folder contains many different types of data, but the file we will be using for this basic tutorial is named "ob.ascii". This is an ASCII text file, and contains a number of different observation types (including balloon soundings, surface METAR, ships, and buoy observations.
"rc" contains the first guess and boundary files. These files are generated by WPS and WRF; you can find more information in the User's Guide and the WRF/WPS Online Tutorial.
In order to avoid confusion and possibly overwriting important files, you should run every experiment in a separate working directory. Create a directory named "3dvar" and enter it; we will run the test case in this directory:
> mkdir 3dvar
> cd 3dvar
Step 8: Run the test case
The first and most important file for running WRFDA is the namelist.input file. A sample namelist is included in the WRFDA distribution; copy it into your working directory:
> cp ../WRFDA/var/test/tutorial/namelist.input .
You will also need to link several other files to your working directory. This can be done using the ln command:
The -sf option specifies that this should be a symbolic link (analogous to a "shortcut" in the Windows operating system) and that the command should overwrite any link by the same name which might already exist (in this case, not a problem, but it is a good habit to develop). These files are all important for running WRFDA:
"wrfinput_d01" is a netCDF file which was created by WPS and WRF. This is our "first guess" file (and so is linked as "fg"); the meteorological analysis that we start with before adding observations.
"LANDUSE.TBL" is a tabular file that describes the physical properties of the different land use categories which are specified in the first-guess file.
"be.dat" is the aforementioned background error file.
"ob.ascii" is the aforementioned observation file.
And finally, "da_wrfvar.exe" is our WRFDA executable file. This is the program that runs the WRFDA system.
Let's take a closer look at the namelist file. You can open it using the text editor of your choice (vi, emacs, etc.). You will see a large number of namelist sections, but only the ones listed below are important:
&wrfvar18
analysis_date="2008-02-05_12:00:00.0000", This option sets the full date of the analysis
/
...
&wrfvar21
time_window_min="2008-02-05_11:00:00.0000", Beginning time for the assimilation window
/
&wrfvar22
time_window_max="2008-02-05_13:00:00.0000", Ending time for the assimilation window
/
...
&time_control
start_year=2008, The start parameters should be the same as the analysis time
start_month=02,
start_day=05,
start_hour=12,
end_year=2008, The end parameters are ignored by WRFDA, so they can be set to anything
end_month=02,
end_day=05,
end_hour=12,
/
...
&domains
e_we=90, The number of horizontal east-west grid spaces in the domain
e_sn=60, The number of horizontal north-south grid spaces in the domain
e_vert=41, The number of vertical levels in the domain
dx=60000, The east-west grid spacing in meters
dy=60000, The north-south grid spacing in meters
hypsometric_opt=2, Vertical integration method. This value must match the first guess file!
/
...
&physics Starting with version 3.5, it is not necessary to set these options;
mp_physics=3, the necessary fields are read directly from the first guess file.
ra_lw_physics=1,
ra_sw_physics=1,
radt=60,
sf_sfclay_physics=1,
sf_surface_physics=1,
bl_pbl_physics=1,
cu_physics=1,
cudt=5,
num_soil_layers=5,
...
The analysis date option specifies the analysis date and time, which should be the same date and time as the first guess file. The time_window_min and time_window_max options specify the beginning and end times of the so-called "assimilation window"; WRFDA will only assimilate observations in the observation file that fall within this window. Starting with version 3.5, the span class="style6" id="Font12">&domains and &physics namelists can be ignored, as the information necessary to run WRFDA is now read directly from the first guess file.
Prior to version 3.5, the parameters in the &domains and &physics namelists should match those in the first guess file. You can check the values of these parameters in the first guess file by issuing the command ncdump -h fg.
When all these files are linked properly, you can run WRFDA by issuing the command:
> ./da_wrfvar.exe >& wrfda.out
This command runs the WRFDA executable (da_wrfvar.exe) and records the output to a log file (wrfda.out, though again, you can name this file whatever you wish). This tutorial case should take less than a minute to run, though depending on your system setup it may take up to five minutes.
Once the executable is done running, let's take a look at what our working directory looks like:
> ls -al
total 34688
drwxr-xr-x 2 user 4096 Nov 1 14:48 .
drwxr-xr-x 7 user 4096 Oct 30 19:12 ..
lrwxrwxrwx 1 user 38 Oct 29 21:26 be.dat -> /home/USER/WRFDA341/be/be.dat
-rw-r--r-- 1 user 0 Nov 1 14:46 buddy_check
-rw-r--r-- 1 user 0 Nov 1 14:46 check_max_iv
-rw-r--r-- 1 user 378 Nov 1 14:46 cost_fn
lrwxrwxrwx 1 user 58 Oct 29 21:40 da_wrfvar.exe -> /home/USER/WRFDA341/WRFDA/var/build/da_wrfvar.exe
lrwxrwxrwx 1 user 55 Oct 29 21:39 fg -> /home/USER/WRFDA341/rc/2008020512/wrfinput_d01
-rw-r--r-- 1 user 349 Nov 1 14:46 grad_fn
-rw-r--r-- 1 user 6084853 Nov 1 14:46 gts_omb_oma_01
-rw-r--r-- 1 user 6084853 Nov 1 14:46 gts_omb_oma_01.0000
-rw-r--r-- 1 user 2758 Nov 1 14:46 jo
lrwxrwxrwx 1 user 50 Oct 29 21:25 LANDUSE.TBL -> /home/USER/WRFDA341/WRFDA/run/LANDUSE.TBL
-rw-r--r-- 1 user 1075 Oct 29 16:08 namelist.input
-rw-r--r-- 1 user 92879 Nov 1 14:46 namelist.output.da
lrwxrwxrwx 1 user 51 Oct 29 21:38 ob.ascii -> /home/USER/WRFDA341/ob/2008020512/ob.ascii
-rw-r--r-- 1 user 10205 Nov 1 14:46 qcstat_conv_01
-rw-r--r-- 1 user 28690 Nov 1 14:46 rej_obs_conv_01.000
-rw-r--r-- 1 user 29280 Nov 1 14:46 statistics
-rw-r--r-- 1 user 2725792 Nov 1 14:46 unpert_obs
-rw-r--r-- 1 user 2725792 Nov 1 14:46 unpert_obs.0000
-rw-r--r-- 1 user 7489 Nov 1 14:46 wrfda.out
-rw-r--r-- 1 user 17624520 Nov 1 14:46 wrfvar_output
Many of these files can be safely ignored for now. The details of the important files are listed below:
cost_fn
Details the various components of the cost function for each outer iteration
grad_fn
Details the various components of the gradient function for each outer iteration
gts_omb_oma_01
Detailed statistics about all observations.
gts_omb_oma_01.0000
Detailed statistics about the observations used on each processor. For parallel runs, there will be one of these files for each processor used.
jo
Cost function statistics for each observation type.
namelist.output.da
Those namelist options not specified in the "namelist.input" file will be assigned default values. This file lists ALL namelist options selected for your run, including all default options.
qcstat_conv_01
Details quality control information for each individual variable for each observation type.
rej_obs_conv_01.000
Lists each rejected observation
statistics
Lists useful statistics for each observation type
wrfda.out
This is the file where you sent WRFDA's standard output. This should contain run-time information, as well as any errors which may have occurred.
wrfvar_output
Finally, the most important file! This is the analysis produced by WRFDA in netCDF format.
More information on these files and the information they contain can be found in this tutorial presentation. Depending on the options selected for your run, there may be more or fewer of these files. These options will be covered in more detail in other tutorials.
You have now successfully run the WRFDA-3DVAR test case!
Return to main menu
Note: This website is currently under construction. If you notice any problems with the website, including but not limited to broken links, formatting errors, or incorrect information, please submit an email to the or wrfhelp@ucar.edu so that we can address the problem. We appreciate your feedback!
NOTE: These pages makes use of Java Script. Enable Java on your computer to make use of all the functionality.
