@andreas-h the PR lgtm, but I'd like to have in the comments a reference to why these variables are needed. Do you think you can add that?

1. WPS: I wasn't able to figure out how to tell WPS the correct path to netCDF-Fortran otherwise. Build was working properly using intel-2018b, but with foss I was running into issues where netcdff wasn't found. This was the only way I could fix that.

2. the HDF5 env variable probably isn't strictly needed, but ./configure was complaining about it not being set:

   HDF5 not set in environment. Will configure WRF for use without.

3. PnetCDF is an optional dependency. Setting the PNETCDF env variable enables output using ParallelNetCDF.

4. EM_CORE and NMM_CORE variables control which dynamical model core is being built. You can only choose one of them. If nothing is explicitly set, EM_CORE is being built.
   I agree that optimally, one would make this configurable and add (yet) another version suffix to indicate the choice.However, I have no experience at all with NMM_CORE and so didn't want to do that. I'm working on a PR to optionally build the WRF-Chem chemistry code inside the wrf EasyBlock and that one only works with EM_CORE anyways.

Now that I think about it, maybe it'd be best to not set the _CORE variables explicitly here. I would then only do that in the wrf-chem stuff I'm developing, since there it's not optional any more. Apart from that I don't think that any additional comments help. I can add some if you think it's needed.