tripolar + conservative - updated in ext

[juliasloan25]

Purpose

Yet another version of #1679. I split that PR up into #1715 and this one since it was getting big.

To do

• debug

Content

• add ConservativeRegridding dependency
• extend Interfacer.remap! for Oceananigans fields using CR.jl
• remove interpolate! calls
• add a function construct_remappers to make the remapper object and allocate scratch space
• switch to TripolarGrid + JLD2 diagnostics

---

• I have read and checked the items on the review checklist.

[juliasloan25]

change to 25 after debugging

[juliasloan25]

move to remapping.jl

[juliasloan25]

We can't use Float32 as the simulation float type when we use ConservativeRegridding. Even if we set it correctly to Float32 for the ocean grid and boundary space, there's a mismatch in source/dest manifolds because of the way GO.Spherical is called. Here for Oceananigans and here for ClimaCore

We need the radius for the Oceananigans grid to agree with defaults.FloatType, or to otherwise use this FT to construct the Spherical object.

Error

julia> remapping = ClimaCouplerCMIPExt.construct_remappers(grid, boundary_space)
ERROR: Destination and source manifolds must be the same.  Got GeometryOpsCore.Spherical{Float32}(6.371f6) and GeometryOpsCore.Spherical{Float64}(6.371e6).
Stacktrace:
 [1] error(s::String)
   @ Base ./error.jl:35
 [2] ConservativeRegridding.Regridder(dst::ClimaCore.Spaces.SpectralElementSpace2D{…}, src::Oceananigans.Grids.OrthogonalSphericalShellGrid{…}; kwargs::@Kwargs{…})
   @ ConservativeRegridding ~/.julia/packages/ConservativeRegridding/TJLZf/src/regridder/regridder.jl:116
 [3] construct_remappers(grid_oc::Oceananigans.ImmersedBoundaries.ImmersedBoundaryGrid{…}, boundary_space::ClimaCore.Spaces.SpectralElementSpace2D{…})
   @ ClimaCouplerCMIPExt ~/ClimaCoupler.jl/ext/ClimaCouplerCMIPExt/oceananigans.jl:225

julia> OC.defaults.FloatType
Float32

julia> CC.Spaces.undertype(boundary_space)
Float32

[simone-silvestri]

We need the radius for the Oceananigans grid to agree with defaults.FloatType, or to otherwise use this FT to construct the Spherical object.

We can use 0.104.5 that should solve this issue

[juliasloan25]

change substeps = 150 in the free surface constructor

[juliasloan25]

move these to a file that's included later so we can keep the Simulation struct definitions in the corresponding files

[simone-silvestri]

we should try removing sea ice dynamics while debugging:

sea_ice_simulation(grid, ocean; dynamics=nothing)