Proposal: boundary-topography contribution in gravity point values (chunk/shell), and 2D support

Proposal: boundary-topography contribution in gravity point values (chunk/shell/sphere), and 2D support

Hi all, I’m Bernard Ngu (Baylor University), working on regional mantle-flow models of the Cameroon Volcanic Line. I would like to work on an extension of the gravity postprocessor.

Motivation. For regional models with free-slip top/bottom boundaries, the topography is diagnosed by the dynamic topography postprocessor, but gravity point values only integrates the interior density. The mass associated with the deflected surface and bottom boundary is missing, and it is typically of comparable magnitude and opposite sign to the interior contribution. The geoid postprocessor includes these terms, but only for 3D spherical shells. As a result, gravity predicted for chunk models can’t be compared directly with observed free-air/Bouguer anomalies.

Proposed changes (as separate PRs):

  1. Optional surface and bottom topography contributions in gravity point values, modeled as thin mass sheets σ = Δρ·h using DynamicTopography::topography_vector() and its density above/below. Off by default; skipped automatically for mesh-deformation boundaries (already included through the deformed mesh). Interior / surface / bottom contributions written as separate columns.
  2. 2D support (line-mass kernel), currently execute() is only implemented for dim == 3.

Verification plan: analytic uniform-shell case, comparison against the geoid postprocessor in a 3D shell, chunk vs. full-shell consistency, and the Bouguer-slab limit.

Questions:

  • Is anyone already working on this or on related gravity/geoid changes?
  • Would you prefer extending gravity point values as above, or a shared helper that both geoid and gravity point values use?
  • Any concerns about the treatment of compressible models (the postprocessor currently asserts incompressible)?

Thanks, and happy to adjust the design based on feedback.

The separate interior, surface, and bottom contributions sound useful, especially for checking where differences between models come from.

One thing I’d be careful about is keeping the sign conventions consistent between gravity point values and geoid, particularly when switching between chunk and spherical-shell geometries.

A shared helper might make sense for that part, even if the two postprocessors still handle their outputs differently.

I also wonder whether the verification could include a case where the boundary displacement is zero, just to confirm that enabling the option doesn’t change the existing gravity result.

That seems like a useful baseline alongside the analytic shell and Bouguer-slab tests you’ve already proposed.

Thanks for the suggestions, they are very useful.

Sign conventions. I would follow the convention of dynamic topography and geoid: the surface term as a thin sheet σ = (ρ_below − ρ_above)·h with h positive upward, and the bottom term using the density contrast across the bottom boundary in the same way. To make sure the two postprocessors stay consistent, I would add two checks as tests: in a 3D shell, the ratio of the boundary contributions to the interior contribution should match what geoid gives for the same model, and the same buoyant anomaly in a chunk and in a full shell should give the same gravity anomaly once the mean along a profile is removed.

Shared helper. I agree the part worth sharing is the definition of the mass sheet: which densities above and below are used, and the sign. The rest differs, since geoid works with the spherical-harmonic expansion of the topography while gravity point values would integrate the sheet face by face at arbitrary points. One option would be a small helper that returns the surface mass density σ at the boundary quadrature points, used by gravity point values in the first PR, with geoid switching to it in a follow-up so the first PR stays easy to review. I’m happy to go the other way if the maintainers prefer.

Zero-displacement baseline. Good idea, I’ll add it. A model without lateral density variations has no flow and no boundary deflection, so switching the new options on should leave the gravity output unchanged. With the options off (the default), the interior calculation would not be touched, so the existing gravity tests should not change either.

Thanks again,