Save vertical fluxes at selected depth.

[plazzari]

Is it possible to activate a diagnostic in GOTM to save vertical fluxes at a selected depth, including the sinking term of variables from a biogeochemical model?

[jornbr]

FABM does create internal (hidden-by-default) diagnostics for the total sinking velocity of biogeochemical tracers, which are called <TRACERNAME>_w_tot. You can add these to output, e.g., by adding - source: <TRACERNAME>_w_tot in the output section of gotm.yaml. You can further let GOTM interpolate these velocities to chosen depth(s), if desired, by something like for a file under output:

  operators:
    - type: interp
      dimension: z
      coordinates:
        - -100.0

If you need the actual flux (tracer/area/time), you would additionally save the tracer values in the same file (thus using the same interpolation, if any), so you can then take te product of the two in postprocessing.

Two caveats though:

1. Taking the product in postprocessing will produce the correct value only if all values - velocity, tracer, flux - are instantaneous in time (as opposed to temporally-averaged)
2. The local product of velocity and tracer is not exactly equal to what is transported over that depth interface in the model, as there the specifics of the advection scheme that implements the sinking come into play.

Whether these caveats are a problem will depend on your application. It would be possible to reconstruct the exact advective transports over the cell interfaces by examining the tracer values pre- and post-advection, but that is not currenty implemented. (It is for the diffusion operator though - those fluxes are available as TRACERNAME_diff.

[plazzari]

Many thanks, so as an example, for the tracer with name P1_11_c, I coud set in the gotm.yaml output section:

output:
result:
time_method: point
....
....
- source: P1_11_c
- source: P1_11_c_w_tot
- source: P1_11_c_diff
....
....
and in my postproc to get the instaneous flux J(P1_11_c) I will compute at a given model discretized depth

J(P1_11_c) = P1_11_c * P1_11_c_w_tot + P1_11_c_diff

[jornbr]

That sounds fine. Just note that P1_11_c_diff will be defined at layer interfaces while P1_11_c and P1_11_c_w_tot will be defined at layer centers. Therefore, to sum the advective and diffusive contributions, one of the two will need interpolating.