Wind-forced constricted channel example

[dkytezab]

cc @emmomp @glwagner @jlk9, see #5700. Ported from GB. Pared down some pieces, feel free to change text

[dkytezab]

Hitting this at model construction:

Invalid operands for select instruction!
  %1310 = select <2 x i1> %1297, double %1308, double %1309
Invalid operands for select instruction!
  %1314 = select <2 x i1> %1297, double %1312, double %1313
Invalid operands for select instruction!
  %1320 = select <2 x i1> %1297, double %1318, double %1319
Invalid operands for select instruction!
  %1404 = select <2 x i1> %1297, double %1402, double %1403
Invalid operands for select instruction!
  %1535 = select <2 x i1> %1528, double %1402, double %1534
Invalid operands for select instruction!
  %1538 = select <2 x i1> %1528, double %1537, double %1319
Invalid operands for select instruction!
  %1543 = select <2 x i1> %1528, double %1312, double %1542
Invalid operands for select instruction!
  %1626 = select <2 x i1> %1528, double %1625, double %1309

[navidcy]

No disrespect to Ryan, but were they really the first that set up a reentrant channel to deserve naming after therm? I think it's much better if we call this example reentrant_channel or something along those lines? Of course, cite Ryan's et al. paper that this was inspired from. What do you think?

[glwagner]

This is also quite a bit different than the channel in Ryan's paper due to the bathymetry 😂 . I propose navid_channel.jl

[navidcy]

I don't like navid_channel.jl either :D

[navidcy]

We can do this (or something very close to this) with ReferenceToStretchedDiscretization, right?

[navidcy]

julia> using Oceananigans

julia> z = ReferenceToStretchedDiscretization(extent=2000, stretching = PowerLawStretching(1.029), constant_spacing=10, constant_spacing_extent=10); z.faces
33-element Vector{Float64}:
 -2108.88
 -1842.41
 -1614.75
 -1419.38
 -1251.0
 -1105.27
  -978.63
  -868.14
     ⋮
   -87.29
   -71.9
   -57.65
   -44.43
   -32.14
   -20.69
   -10.0
     0.0

Which is very similar?

[navidcy]

I suggest

Suggested change

	underlying_grid = RectilinearGrid(architecture,
	topology = (Periodic, Bounded, Bounded),
	size = (Nx, Ny, Nz),
	halo = (4, 4, 4),
	x = (0, Lx),
	y = (0, Ly),
	z = z_faces)
	## We construct a vertical discretization for a domain of depth around 2000 meters via `ReferenceToStretchedDiscretization`. We use a 10-meter vertical spacing at the surface and increase the spacing at depth via a power law.
	z = ReferenceToStretchedDiscretization(extent=2000,
	stretching = PowerLawStretching(1.029),
	constant_spacing = 10,
	constant_spacing_extent = 10)
	Nz = length(z)
	Lz = abs(z.faces[1]) # if this is used elsewhere
	underlying_grid = RectilinearGrid(architecture;
	topology = (Periodic, Bounded, Bounded),
	size = (Nx, Ny, Nz),
	halo = (4, 4, 4),
	x = (0, Lx),
	y = (0, Ly),
	z)

and remove the definition of Lz and Nz above along with the geometric spacing construction.

[navidcy]

Suggested change

	g = 9.8061 # [m s⁻²] gravitational acceleration
	g = Oceananigans.defaults.gravitational_acceleration

[navidcy]

Suggested change

	ΔB = 8 * α * g, # surface buoyancy contrast [m s⁻²]
	ΔT = 8, # surface temperature contrast [K]
	ΔB = α * g * ΔT, # surface buoyancy contrast [m s⁻²]

[navidcy]

If I'm not mistaken, that 10 in Qᵇ = 10 / (ρ * cᵖ) * α * g is a heat flux. May I suggest:

Suggested change

	Qᵇ = 10 / (ρ * cᵖ) * α * g, # buoyancy flux magnitude [m² s⁻³]
	Q = 10, # heat flux magnitude [W m⁻²]
	Qᵇ = Q / (ρ * cᵖ) * α * g, # buoyancy flux magnitude [m² s⁻³]

[navidcy]

Suggested change

	cᵖ = 3994.0 # [J K⁻¹] heat capacity
	ρ = 999.8 # [kg m⁻³] reference density
	cᵖ = 3994.0 # specific heat capacity [J K⁻¹ kg⁻¹]
	ρ = 999.8 # reference density [kg m⁻³]

[navidcy]

Suggested change

	coriolis = BetaPlane(f₀ = -1e-4, β = 1e-11)
	coriolis = BetaPlane(latitude = -60)