Add `ParcelModel` mode

[glwagner]

Like SingleColumnModel mode for grid with topology = (Flat, Flat, Bounded), it will also be nice to have a "parcel model mode" for grids with topology = (Flat, Flat, Flat). This would be a 0D simulation of only microphysics. Somehow in this model we will want to be able to change the environmental conditions somehow, which will require a bit of thinking about what the inputs and outputs of the model are.

[glwagner]

@kaiyuan-cheng can you describe a typical workflow when you do a parcel simulation? I'm in particular curious how you prescribe the thermodynamic state and what is varied; ie I think we want to simulate the evolution of a parcel along some trajectory (perhaps prescribing the evolution of energy density / related variable and simulating microphysics only?)

[glwagner]

@bischtob could have some insights too

[bischtob]

What this requires is probably just a AdiabaticAscentModel or something and the tendencies will involve heating etc. but an advective velocity may be prescribed. So a value of w or some profile/function may be attached to the struct.

[glwagner]

We can design an abstraction that mutates the state consistent with an adiabatic change in altitude... ? Then the user can evolve the trajectory from the outside, eg something like

for z in heights
    adjust_adiatbatically!(model, z)
end

We can develop a way to store the current height inside a 0D grid. We have a similar abstraction for column models that need to live at some lat/lon.

[glwagner]

Eg for a column model with a non-trivial horizontal "location":

# Ocean station papa location
location_name = "ocean_station_papa"
λ★, φ★ = 35.1, 50.1

grid = RectilinearGrid(size = 200,
                       x = λ★,
                       y = φ★,
                       z = (-400, 0),
                       topology = (Flat, Flat, Bounded))

from https://github.com/CliMA/ClimaOcean.jl/blob/main/examples/single_column_os_papa_simulation.jl

[glwagner]

Ok but now I am realizing that the microphysics has history, so we need to know the time-step and vertical velocity actually, so maybe more like

for n in 1:10
    displace_adiabatically!(model, w, dt) # eg dz = w * dt
end

or something.

but it would be even better if we could treat this like a normal Simulation and run! it. Ok, I think the structure is starting to come into view for me.

[bischtob]

I think we can make "z" part of the state stored in the model struct. Then we update it during the time evolution. @kaiyuan-cheng, how is this currently done in your parcel model?

[kaiyuan-cheng]

The parcel starts with its initial properties, including microphysics, thermodynamics, and any initial forcing (such as updraft or buoyancy). It then updates its position based on the forcing and the timestep. After that, it reads the environmental pressure and temperature (from a sonding profile), computes the microphysics and thermodynamics, and determines the new forcing, and so on. The parcel’s position can also be prescribed, e.g., a constant updraft or acceleration.