Need additional kwargs in observation map function when wrapped into CAL.calibrate

[Julians42]

When running calibrations using one of the new methods like:

calibrate(WorkerBackend, 
        ekiobj, 
        ensemble_size,
        n_iterations,
        prior,
        output_dir
)

the observation map function is inaccessible for passed configuration arguments that are needed to extract observations correctly. I can load them in by hardcoding the config_dict read into the function but would be nice to be able to stick with the methods already used in ClimaAtmos.jl e.g., https://github.com/CliMA/ClimaAtmos.jl/blob/8b6f11d86ce53fcadf126d41ff5ce8c500b43621/calibration/experiments/gcm_driven_scm/observation_map.jl#L9

Perhaps I'm missing something and this is already there - if so feel free to delete this :)

[nefrathenrici]

What are some of the passed configuration arguments you want to support? Are these related to the EKP.Observations module?

[Julians42]

Not related to EKP.Observations, this is more related to how to process the data. E.g., in order to create the G_ensemble matrix we need to know the height of the column (e.g., number of levels), the variables we want to extract, and the batch size. We also need to specify which section of the simulation we're processing for our observations, e.g., hours 60-72, for example. I think it's easiest to pass a dict here (here's my workflow):

function observation_map(iteration; config_dict::Dict)

    full_dim =
        config_dict["dims_per_var"] *
        length(config_dict["y_var_names"]) *
        config_dict["batch_size"]
    
    G_ensemble =
        Array{Float64}(undef, full_dim..., config_dict["ensemble_size"])

    iter_path = path_to_iteration(config_dict["output_dir"], iteration)
    eki = JLD2.load_object(joinpath(iter_path, "eki_file.jld2"))
    for m in 1:config_dict["ensemble_size"]
        member_path =
            path_to_ensemble_member(config_dict["output_dir"], iteration, m)
        try
            G_ensemble[:, m] .= process_member_data(
                member_path,
                eki;
                y_names = config_dict["y_var_names"],
                t_start = config_dict["g_t_start_sec"],
                t_end = config_dict["g_t_end_sec"],
                z_max = config_dict["z_max"],
                norm_factors_dict = config_dict["norm_factors_by_var"],
                log_vars = config_dict["log_vars"],
            )
        catch err
            @info "Error during observation map for ensemble member $m" err
            G_ensemble[:, m] .= NaN
        end
    end
    return G_ensemble
end