ensemble_slice_model_simulation is in order

[navidcy]

Similarly to ensemble_column_model_simulation,

https://github.com/CliMA/OceanTurbulenceParameterEstimation.jl/blob/f7086e11e82e72f3bb7e533c6553ef9cac86031c/src/EnsembleSimulations.jl#L7

it'll be useful to add an ensemble_slice_model_simulation.

simulation = ensemble_slice_model_simulation(observations;
                                             Nensemble = Nensemble,
                                             architecture = architecture,
                                             tracers = (:b, :e, :c),
                                             free_closures = gent_mcwilliams_diffusivity,
                                             additional_closures = closures::Tuple
                                             )

There is an additional difficulty here: the slice simulations usually need to include more than one closures. That is, in the case of the mesoscale eddy fluxes closure we want to be able to calibrate for the gent_mcwilliams_diffusivity (free_closures) but allow for the model ensembles to also include other closures, same for all ensemble members (the additional_closures).

cc @glwagner

[glwagner]

I don't understand what's meant by free_closures?

I think you just want one kwarg, closures. Every member of the ensemble is an identical setup, right?

[navidcy]

By free_closures I meant those whose parameters we calibrate (following the naming for "free parameters") and by additional closures I meant those which are the same for all particles.

But I didn't pay too much thinking to the name -- there may be better choices.

[glwagner]

Oh that's interesting. How about a type Free or Calibrated to distinguish?

closure = (Calibrated(gent_mcwilliams_diffusivity), anisotropic_diffusivity)

[glwagner]

kwargs work too for sure