Exact, time-averaged heat content diagnostics missing

[rmholmes]

For all the runs in the documentation paper, we are saving monthly averages of temp and dzt. However the exact, spatially-resolved, time-averaged heat content cannot be reproduced exactly from these because of missing correlations between the two. The missing 3D diagnostic is temp_rhodzt. Of course, errors are likely small (I'm not sure how small...).

For future runs, can we please add the 2D diagnostic temp_int_rhodz? As we are saving all of the vertically-integrated heat fluxes and the surface heat flux, this should allow exact closure of the vertically-integrated heat budget.

[rmholmes]

Oh wait, we still could not close the budget exactly without saving snapshots (of which we have some in the restarts). I think the 2D diagnostic would still be useful though.

[rmholmes]

Just adding to this with another request: It looks like dzt is saved only annually, not monthly in most of the runs. I would suggest saving monthly-averages of dzt.

[aekiss]

OK I'll make dzt monthly in the new configs.
Is temp_int_rhodz the right name? It doesn't show up in a search but I guess it's one of those programmatically-generated names. And I guess you'd like that monthly too?

[rmholmes]

temp_int_rhodz is indeed the name (https://github.com/rmholmes/MOM5/blob/master/src/mom5/ocean_tracers/ocean_tracer.F90#L1211-L1216). It's a 2D field so monthly would be great (and it'd make it super easy to plot vertically-integrated heat content anomalies). Thanks.

[russfiedler]

dzt can always be calculated exactly from knowledge of eta_t so doesn't need to be stored. You need the time invariant fields depth, dstlo and dstup. The latter 2 are in all ocean_thickness.res.nc restarts. We do this for BRAN. In fact you only need the sum of dstlo and dstup for this calculation so you could have that saved once.

[rmholmes]

Fair enough. I've never done that calculation but I'm sure it's easy enough. That would save a 3D variable at the cost of a bit more faffing for analysis.

[russfiedler]

Yeah, I've got the Fortran code. Just a few lines the calculation is pretty simple and quick. Once you get the 3D array for dstlo + dstup loaded it should fly through. Have a look at /scratch/v45/raf599/restarts/make_restart_thick_tripolar_decomp.f90 for one version. Lines 64 and 129 do the work.

[aekiss]

@rmholmes I've added monthly temp_int_rhodz and dzt to the new defaults but let me know if you want me to remove dzt.

[rmholmes]

You can remove dzt.

[aekiss]

@russfiedler, @rmholmes we are storing sea_level, not eta_t. These are the same with the current configuration (max_ice_thickness=0) but if that changes they'll differ, which will complicate the calculation of dzt under ice. I guess we'll worry about that if/when it happens.

[russfiedler]

eta_t can always be added if that happens and we can probably calculate it pretty accurately from the ice mass if that is saved at the same frequency.