WRFv4 on ARC3: User documentation

Running WRFv4 on ARC3 for undergraduate teaching

Chris Dearden, CEMAC

These notes describe how to run WRFv4 on ARC3 or ARC4, from your account using the centrally installed WRF and WPS and stored test case data.

For those interested in the compilation methodology, the exact details can be found on this page. If you are a University of Leeds researcher and are interested in running WRFv4 on ARC3, it is recommended that you follow the compilation guide to build a local version of WRFv4 under your own ARC3 account instead of running from the central build.

To date, the centrally installed WRF has been successfully used in teaching for four specific case studies – the two single domain cases from the on-line WRF tutorial (the January 2000 Nor’easter cyclone and Hurricane Katrina) as well as dummyfc and the forecast case from 2017.

Instructions

First, log in to ARC3 using.
```
   ssh -X <username>@arc3.leeds.ac.uk
```
We now need to run a simple shell script to configure our environment for running WRF. Enter the following at the command prompt:
```
  . /nobackup/WRF/Teaching/WRFv4/WRFv4_setup_env.sh
```

Note that the . at the start, followed by a space, is crucial since it ensures that the settings are applied within the current shell session. This script does two things – firstly, it sets the paths to the pre-compiled WRF and WPS executables, along with some essential libraries and other dependencies that are required during run-time. Secondly, it creates two directories within /nobackup/$USER/WRFv4, namely WRFv4_Katrina_test_env and WRFv4_dummyfc_test_env, one for each case study. These are the directories from which you will execute the model, and where the output files will be generated. They can be thought of as self-contained local test environments for configuring and running the model. Note that if a user wants to deviate away from either of these pre-defined case studies and create their own, this can be achieved by simply taking a copy of either WRFv4_Katrina_test_env or WRFv4_dummyfc_test_env, re-naming it as appropriate, and then modifying the content of the namelist files as necessary.

NB for first session of the course check WRFv4_Katrina_test_env or WRFv4_dummyfc_test_env do not already exist else fresh instances will not be initiated and you will need to remove the old WRFv4 folder

Listing the contents of WRFv4_Katrina_test_env should reveal two sub-directories: WRF and WPS. This mimics the directory structure of a full WRF installation, but they only contain copies of the namelists and other essential files that are required during run-time. The same sub-directories also exist within WRFv4_dummyfc_test_env, the only difference being the settings used in the namelist files, which are case-dependent. No changes are required to either namelist.wps or namelist.input in order to run the case studies – they have already been tested for numerical stability and should run ‘out of the box’.
Choose one of the two test case directories, and change into the WPS sub-directory. From here, you can proceed to run the test case by following the steps in the on-line tutorial as a guide (http://www2.mmm.ucar.edu/wrf/OnLineTutorial/CASES/SingleDomain/index.html). Before doing so, please note the following important points:

Once you have run the shell script in step #2, the WPS executables (geogrid, ungrib and metgrid) and WPS utilities (g1print and rd_intermediate) will be available to you from the command line, and are invoked simply by typing the executable name without the . at the beginning, e.g.

	geogrid.exe

In most cases, it is OK to run the WPS executables interactively on the login nodes, since they run very quickly and do not require much memory. The exception is the 2017 case study, where the relatively high resolution of the GRIB input files requires ungrib.exe to be run as a serial batch job.

The WRF executables, real.exe and wrf.exe, should be run using the batch submission scripts run_real.sh and run_wrf.sh, that reside within WRF/test/em_real for each case study.

GRIB input data is available on ARC3 for each of the case studies at nobackup/WRF/Teaching/WRFv4/WRF_case_studies/<case-study>/GRB/. You can link to these GRIB files using the link_grib.csh script within the WPS sub-directory.

You can then run ncl scripts, e.g. from within the WPS sub-directory, type: