To use WRF through WRF4G2, you need to configure the app variable in the experiment.wrf4g file. By default, WRF4G2 provides a bundle with all you need to simulate a WRF experiment, which is configured as follows :
app = wrf_all_in_one | bundle | /home/user/wrf4g/repository/apps/WRF/WRFbin-3.4.1_r2265_gfortran.tar.gz
This bundle can be used on x86_64 Linux systems, and contains binaries, libraries and configuration files for WRF-3.4.1 build against gfortran-4.4.7, openmpi-1.4.3 and netcdf-4.1.3, which are included in the bundle too (hence, all in one).
This configuration is quite useful, because you do not need to install WRF on every computing resource.
However, if you'd like to use your own WRF, installed on your machine or on a cluster, you can do so by updating the app variable. There are several ways to configure this variable depending on your WRF configuration. Let's see some examples:
app = wps | command | module load wps/4.5.1 wrf | command | module load wrf/4.5.1
We assume that each command is configuring other pieces of software, such as openmpi, netcdf and so on.
app = netcdf_bin | command | PATH=/home/user/netcdf-4.1.3/bin/:$PATH netcdf_lib | command | LD_LIBRARY_PATH=/home/user/netcdf-4.1.3/lib/:$LD_LIBRARY_PATH wrf_bin | command | PATH=/home/user/WRF4.5.1/:$PATH
As you might have probably noticed, we didn't mention heavy input files required by some WRF configurations (e.g. CAM_ABS_DATA, CAM_AEROPT_DATA, co2_trans, etc). In order to make these available for a specific WRF experiment, it is highly recommended to create a directory tree under the wrf4g_files directory for that experiment.