Regridding a constant field from TripolarGrid to CubedSphere introduces extrema

[juliasloan25]

When we regrid a constant, non-zero field from the Oceananigans TripolarGrid to ClimaCore CubedSphere, new extrema are introduced. When we run without RightFaceFolded, there are zeros along the southernmost latitudes and new maxima near the northern latitudes. When we run with RightFaceFolded, the southern zeros remain but we don't see the unusual pattern in northern latitudes.

I'm running this on a branch with the fixes from both #60 and #56

Regridding a field of ones without RightFaceFolded

Dest field

To reproduce

using ConservativeRegridding
using ConservativeRegridding: Trees
import GeometryOps as GO, GeoInterface as GI, LibGEOS

using ClimaCore:
    CommonSpaces, Fields, Spaces, RecursiveApply, Meshes, Quadratures, Topologies, Domains, ClimaComms
using Oceananigans

const ClimaCoreExt = Base.get_extension(ConservativeRegridding, :ConservativeRegriddingClimaCoreExt)

    tripolar_grid = TripolarGrid(size=(360, 180, 1))
    cubedsphere_space = CommonSpaces.CubedSphereSpace(;
        radius = tripolar_grid.radius,
        n_quad_points = 2,
        h_elem = 32,
    )

    # Set the source field to a constant, non-zero value
    src_tripolar = Field{Center, Center, Nothing}(tripolar_grid)
    set!(src_tripolar, src_tripolar + 1)

    dst_cubedsphere = zeros(cubedsphere_space)
    ones_cubedsphere = ones(cubedsphere_space)
    cubed_sphere_vals = zeros(Meshes.nelements(cubedsphere_space.grid.topology.mesh))

    regridder = ConservativeRegridding.Regridder(cubedsphere_space, tripolar_grid)

    ConservativeRegridding.regrid!(cubed_sphere_vals, regridder, vec(interior(src_tripolar)))

    ClimaCoreExt.set_value_per_element!(dst_cubedsphere, cubed_sphere_vals)
    @show extrema(dst_cubedsphere) # Should be (1.0, 1.0) but we get (0.0, 1.707097145678124)

    # using ClimaCoreMakie, Makie, CairoMakie
    # fig = Figure();
    # ax = Axis(fig[1, 1])
    # hm = fieldheatmap!(ax, dst_cubedsphere)
    # Colorbar(fig[:, 2], hm)
    # fig

Regridding a field of ones with RightFaceFolded

Dest field

To reproduce

using ConservativeRegridding
using ConservativeRegridding: Trees
import GeometryOps as GO, GeoInterface as GI, LibGEOS

using ClimaCore:
    CommonSpaces, Fields, Spaces, RecursiveApply, Meshes, Quadratures, Topologies, Domains, ClimaComms
using Oceananigans

const ClimaCoreExt = Base.get_extension(ConservativeRegridding, :ConservativeRegriddingClimaCoreExt)

    tripolar_grid = TripolarGrid(size=(360, 180, 1), fold_topology = RightFaceFolded)
    cubedsphere_space = CommonSpaces.CubedSphereSpace(;
        radius = tripolar_grid.radius,
        n_quad_points = 2,
        h_elem = 16,
    )

    # Set the source field to a constant, non-zero value
    src_tripolar = Field{Center, Center, Nothing}(tripolar_grid)
    set!(src_tripolar, src_tripolar + 1)

    dst_cubedsphere = zeros(cubedsphere_space)
    ones_cubedsphere = ones(cubedsphere_space)
    cubed_sphere_vals = zeros(Meshes.nelements(cubedsphere_space.grid.topology.mesh))

    regridder = ConservativeRegridding.Regridder(cubedsphere_space, tripolar_grid)

    ConservativeRegridding.regrid!(cubed_sphere_vals, regridder, vec(interior(src_tripolar)))

    ClimaCoreExt.set_value_per_element!(dst_cubedsphere, cubed_sphere_vals)
    @show extrema(dst_cubedsphere) # Should be (1.0, 1.0) but we get (0.0 1.000000000000009)

    # using ClimaCoreMakie, Makie, CairoMakie
    # fig = Figure();
    # ax = Axis(fig[1, 1])
    # hm = fieldheatmap!(ax, dst_cubedsphere)
    # Colorbar(fig[:, 2], hm)
    # fig

[juliasloan25]

cc @simone-silvestri @asinghvi17

[simone-silvestri]

I think it might be connected with the extra half-row that #60 is removing via a treeify function.

[asinghvi17]

Yeah, that sounds very possible. I'm not yet convinced I did the tree splitting correctly though...will take a look tonight or tomorrow morning.