ArgumentError: halo must be ≤ size in ocean_simulation on coarse TripolarGrid

[glwagner]

The reason is that the default free_surface does not work when there are fewer than 70 points in a direction.

julia> using ClimaOcean, Oceananigans, OrthogonalSphericalShellGrids
Precompiling ClimaOcean finished.
  4 dependencies successfully precompiled in 21 seconds. 220 already precompiled.

julia> grid = TripolarGrid(CPU(); size=(90, 45, 2), halo=(7, 7, 7), z=(-1000, 0))
90×45×2 OrthogonalSphericalShellGrid{Float64, Periodic, Oceananigans.Grids.RightConnected, Bounded} on CPU with 7×7×7 halo and with precomputed metrics
├── centered at (λ, φ) = (71.8913, 1.8042)
├── longitude: Periodic  extent 360.16 degrees                           variably spaced with min(Δλ)=0.146465, max(Δλ)=4.21462
├── latitude:  Oceananigans.Grids.RightConnected  extent 174.524 degrees variably spaced with min(Δφ)=0.223895, max(Δφ)=4.61174
└── z:         Bounded  z ∈ [-1000.0, 0.0]                               regularly spaced with Δz=500.0

julia> ocean = ocean_simulation(grid)
┌ Warning: Are you totally, 100% sure that you want to build a simulation on
│
│ 90×45×2 OrthogonalSphericalShellGrid{Float64, Periodic, Oceananigans.Grids.RightConnected, Bounded} on CPU with 7×7×7 halo and with precomputed metrics
│
│ rather than on an ImmersedBoundaryGrid?
└ @ ClimaOcean.OceanSimulations ~/Projects/ClimaOcean.jl/src/OceanSimulations/OceanSimulations.jl:132
┌ Warning: `ImmersedBoundaryCondition` is experimental.
└ @ Oceananigans.ImmersedBoundaries ~/Projects/Oceananigans.jl/src/ImmersedBoundaries/immersed_boundary_condition.jl:57
┌ Warning: `ImmersedBoundaryCondition` is experimental.
└ @ Oceananigans.ImmersedBoundaries ~/Projects/Oceananigans.jl/src/ImmersedBoundaries/immersed_boundary_condition.jl:57
┌ Warning: `ImmersedBoundaryCondition` is experimental.
└ @ Oceananigans.ImmersedBoundaries ~/Projects/Oceananigans.jl/src/ImmersedBoundaries/immersed_boundary_condition.jl:57
┌ Warning: `ImmersedBoundaryCondition` is experimental.
└ @ Oceananigans.ImmersedBoundaries ~/Projects/Oceananigans.jl/src/ImmersedBoundaries/immersed_boundary_condition.jl:57
ERROR: ArgumentError: halo must be ≤ size for coordinate y
Stacktrace:
  [1] validate_halo(TX::Type, TY::Type, TZ::Type, size::Tuple{Int64, Int64, Int64}, halo::Tuple{Int64, Int64})
    @ Oceananigans.Grids ~/Projects/Oceananigans.jl/src/Grids/input_validation.jl:87
  [2] validate_rectilinear_grid_args(topology::Tuple{…}, size::Tuple{…}, halo::Tuple{…}, FT::Type, extent::Nothing, x::Tuple{…}, y::Tuple{…}, z::Nothing)
    @ Oceananigans.Grids ~/Projects/Oceananigans.jl/src/Grids/rectilinear_grid.jl:289
  [3] RectilinearGrid(architecture::CPU, FT::DataType; size::Tuple{…}, x::Tuple{…}, y::Tuple{…}, z::Nothing, halo::Tuple{…}, extent::Nothing, topology::Tuple{…})
    @ Oceananigans.Grids ~/Projects/Oceananigans.jl/src/Grids/rectilinear_grid.jl:264
  [4] RectilinearGrid (repeats 2 times)
    @ ~/Projects/Oceananigans.jl/src/Grids/rectilinear_grid.jl:250 [inlined]
  [5] (OrthogonalSphericalShellGrid{…} where {…})(arch::CPU, FT::DataType; size::Tuple{…}, southernmost_latitude::Int64, halo::Tuple{…}, radius::Float64, z::Tuple{…}, north_poles_latitude::Int64, first_pole_longitude::Int64)
    @ OrthogonalSphericalShellGrids ~/.julia/packages/OrthogonalSphericalShellGrids/LLaMu/src/tripolar_grid.jl:135
  [6] OrthogonalSphericalShellGrid
    @ ~/.julia/packages/OrthogonalSphericalShellGrids/LLaMu/src/tripolar_grid.jl:58 [inlined]
  [7] with_halo
    @ ~/.julia/packages/OrthogonalSphericalShellGrids/LLaMu/src/with_halo.jl:15 [inlined]
  [8] maybe_extend_halos
    @ ~/Projects/Oceananigans.jl/src/Models/HydrostaticFreeSurfaceModels/SplitExplicitFreeSurfaces/split_explicit_free_surface.jl:287 [inlined]
  [9] materialize_free_surface(free_surface::SplitExplicitFreeSurface{…}, velocities::@NamedTuple{…}, grid::Oceananigans.Grids.ZRegOrthogonalSphericalShellGrid{…})
    @ Oceananigans.Models.HydrostaticFreeSurfaceModels.SplitExplicitFreeSurfaces ~/Projects/Oceananigans.jl/src/Models/HydrostaticFreeSurfaceModels/SplitExplicitFreeSurfaces/split_explicit_free_surface.jl:171
 [10] HydrostaticFreeSurfaceModel(; grid::Oceananigans.Grids.ZRegOrthogonalSphericalShellGrid{…}, clock::Clock{…}, momentum_advection::VectorInvariant{…}, tracer_advection::@NamedTuple{…}, buoyancy::SeawaterBuoyancy{…}, coriolis::HydrostaticSphericalCoriolis{…}, free_surface::SplitExplicitFreeSurface{…}, tracers::Tuple{…}, forcing::@NamedTuple{}, closure::Oceananigans.TurbulenceClosures.TKEBasedVerticalDiffusivities.CATKEVerticalDiffusivity{…}, timestepper::Symbol, boundary_conditions::@NamedTuple{…}, particles::Nothing, biogeochemistry::Nothing, velocities::Nothing, pressure::Nothing, diffusivity_fields::Nothing, auxiliary_fields::@NamedTuple{}, vertical_coordinate::Oceananigans.Models.HydrostaticFreeSurfaceModels.ZCoordinate)
    @ Oceananigans.Models.HydrostaticFreeSurfaceModels ~/Projects/Oceananigans.jl/src/Models/HydrostaticFreeSurfaceModels/hydrostatic_free_surface_model.jl:202
 [11] ocean_simulation(grid::Oceananigans.Grids.ZRegOrthogonalSphericalShellGrid{…}; Δt::Float64, closure::Oceananigans.TurbulenceClosures.TKEBasedVerticalDiffusivities.CATKEVerticalDiffusivity{…}, tracers::Tuple{…}, free_surface::SplitExplicitFreeSurface{…}, reference_density::Int64, rotation_rate::Float64, gravitational_acceleration::Float64, bottom_drag_coefficient::ClimaOcean.OceanSimulations.Default{…}, forcing::@NamedTuple{}, biogeochemistry::Nothing, timestepper::Symbol, coriolis::ClimaOcean.OceanSimulations.Default{…}, momentum_advection::VectorInvariant{…}, equation_of_state::SeawaterPolynomials.BoussinesqEquationOfState{…}, boundary_conditions::@NamedTuple{}, tracer_advection::FluxFormAdvection{…}, vertical_coordinate::Oceananigans.Models.HydrostaticFreeSurfaceModels.ZCoordinate, verbose::Bool)
    @ ClimaOcean.OceanSimulations ~/Projects/ClimaOcean.jl/src/OceanSimulations/OceanSimulations.jl:190
 [12] ocean_simulation(grid::Oceananigans.Grids.ZRegOrthogonalSphericalShellGrid{…})
    @ ClimaOcean.OceanSimulations ~/Projects/ClimaOcean.jl/src/OceanSimulations/OceanSimulations.jl:81
 [13] top-level scope
    @ REPL[5]:1
Some type information was truncated. Use `show(err)` to see complete types.

I recommend changing the default so that the number of substeps is computed based on the situations. In particular, the key assumption we have to make regards the time-step. We should encode this assumption explicitly, and then compute substeps based on g, H. Then we should limit these substeps by Nx, Ny. In the case of a limitation, we should print a warning that instability may occur if dt > something because the CFL will be violated for the free surface.

[simone-silvestri]

right, there is a "heuristic" solution in #276.