Refactor (from scratch) ClimaLandSimulations

[AlexisRenchon]

ClimaLandSimulations is a high level API to run ClimaLand simulations.
For example, running a a one column (fluxnet site) simulation via ClimaLandSimulations is as simple as

run_fluxnet("US-MOz")

using default for model config etc, which can be changed via option kwarg.
We expect ClimaLandSimulations to be the main package used by most users (i.e., research scientists), and also for the calibration framework.

To discuss/address

• Too many dependencies
• lib or ext?
• (improve speed! 1 column is slow... implicit solver will improve it?)
• Add PFTs params (that can be adjusted)
• Use gap-filled data (our current met data gapfilling is bad, slow and annoying)
• Add global run in ClimaLandSimulations
• Use diagnostics for outputs
• Something else to read inputs, etc.

[AlexisRenchon]

Deps:

• ClimaComms
• ClimaLand
• ClimaUtilities
• ClimaParams
• ClimaTimeSteppers
• SciMLBase
• Julia

abstract type LandSimulation end

struct SingleColumnSimulation <: LandSimulation
    floating_point_type  # Redundant
    context # Redundant
    computational_domain
    land_model
    output_dir
    params
    timestepper
    drivers
    mutable callbacks
    mutable hooks?
    mutable diagnostics
    mutable output_writers
    mutable time_info
    mutable verbose
    config_id
    job_id
    gap_fill_data
    initalized
    private _integrator
end

function FluxnetSimulation(site_id = "US-MOz";
                           time_info = (; t_init = FT(0), t_end = FT(30 * 86400), dt = FT(3600) )
                           job_id = site_id,
                           FT = Float64,
                           output_dir = OutputPathGenerator.generate_output_path(joinpath("output", job_id)),
                           context = ClimaComms.SingletonCommsContext(ClimaComms.CPUSingleThreaded())
                           computational_domain = ClimaLand.Domain.Column(; context),
                           params = ClimaParams(),
                           timestepper = (),
                           drivers = Drivers(site_id),
                           model = ClimaLand.LandModel(....)
                           callbacks = (),
                           hooks = (),
                           output_writers = DictWriter()
                           diagnostics = ClimaLand.default_diangostics(model; dict_writer),
                           verbose = false
                           config_id = FluxnetSimulation()
                           )
    return SingleColumnSimulation
end

SingleColumnSimulation( ...., params = ClimaParams() )

struct MultiColumnSimulation <: LandSimulation end

struct RegionalSimulation <: LandSimulation end

struct GlobalSimulation <: LandSimulation
    ...
    drivers -> (AtmosDriver -> ERA5Driver(), LAI)
    ...
    params -> ClimaParams() + maps -> CLMMaps()
end

function run!(simulation::LandSimulation)
    if more_init?
        do_more_init()
        # download data
        # preprocess data
        # prepreare output dir
    end

    _integrator
    solve
end

# As extension
function plot(simulation) end

function plot(simulation.config_id, simulation.ouput_dir) end