Help outputting spectral diagnostics

[AndrewILWilliams]

Hi there, me again!

I'm trying to implement a simple, spectral radiation scheme in Isca. I'm doing a spectrally-resolved calculation of LW optical depth/fluxes, then integrating over wavenumber before passing heating rates onto the rest of the model.

I'd like to be able to output spectrally-resolved diagnostics, to check that everything I'm doing is working correctly. It's not obvious how I should do this, so I was wondering if you had any tips? I've found some code in the socrates interface which seems promising (copied below)?

! Need to actually give bins arrays values
id_soc_bins_lw = diag_axis_init('soc_bins_lw', soc_bins_lw, 'cm^-1', 'n', 'socrates lw spectral bin centers', set_name='socrates_lw_bins')

! spectral diagostic fields
id_spectral_olr = &
      register_diag_field ( soc_mod_name, 'soc_spectral_olr',(/ axes(1:2), id_soc_bins_lw/) , Time, &
      'socrates substellar LW OLR spectrum', &
      'watts/m2', missing_value=missing_value )

So if I already have an array of wavenumbers, I can just pass it to diag_axis_init and then use that id in register_diag_field? Is there more i have to do?

Another question: can register_diag_field handle data with >3 dimensions? It looks like there might be some hard-coded limits there? It would be nice to be able to output arrays of the shape lat, lon, phalf, wavenumber.

Cheers
Andrew

[gkvallis]

Brett McKim might have something to say about this, but I'm not sure he checks this channel or is on GitHub.

[ntlewis]

@daw538 (Dan) and Ryan Boukrouche (who I can’t tag, but will point to this) should also be able to help with this :)

[Ryan-Boukrouche]

Not sure I'll be of great help, but it looks like a good start to emulate the socrates interface with the pipeline id _X --> register_diag --> send_data --> runscript DiagTable. I did try to output 4D spectral fluxes a while ago, but it never worked because the code misses the corresponding routine, we would have to add send_data_4d in src/shared/diag_manager/diag_manager.F90 among other things I think

[AndrewILWilliams]

Yeah it seems like the main bottleneck is having a send_data_4d routine—not sure if anyone has written that yet?

[daw538]

I'm not aware of anyone having written a specific routine for a 4D+time variable. Isca's diag manager is currently configured for 0D to 3D fields. The diag manager directory (src/shared/diag_manager) would be a good place to start in general. Here for example is where the 3D routine is set up in diag_manager.F90.

(should've read Ryan's reponse properly when writing mine, I'm basically saying the same thing)