Hi @junruy, @jbnaliboff
I think John’s points are very insightful.
A few months ago, I encountered similar challenges with free surface treatment and boundary traction loading, albeit in 3D spherical shell models. The issues mainly revolved around numerical convergence and differences in how free surface and traction were handled. The discussions were centered around these community posts:
-
[AMG vs. GMG Solver Inconsistency in 3D Elasticity Models in Spherical Domains](AMG vs. GMG Solver Inconsistency in 3D Elasticity Models in Spherical Domains - #23 by jbnaliboff)
-
[Questions Regarding Free Surface and Tractions in Spherical Shell Models](Questions Regarding Free Surface and Tractions in Spherical Shell Models)
These issues were eventually addressed through a series of GitHub PRs in ASPECT and the underlying deal.II library:
ASPECT issues & PRs:
- [#6248](Velocity and Strain Rate Anomalies at Mesh Refinement Boundaries in 3D Spherical Models with Solver Differences · Issue #6248 · geodynamics/aspect · GitHub),
- [#6284]([WIP] Improve free surface accuracy by gassmoeller · Pull Request #6284 · geodynamics/aspect · GitHub)
- [#6285](Fix traction ascii data by tiannh7 · Pull Request #6285 · geodynamics/aspect · GitHub)
deal.II PR:
If you base your work on the current main branch and incorporate these PRs (#6284 for ASPECT, #18396 for deal.II), it should help resolve the issues related to free surface that you are facing, I’ve been doing this lately without any problems.
From my experience, the ALE and sticky air methods generally produce consistent large-scale lithospheric stress patterns, though differences at the surface stress details are to be expected. In particular, the surface stresses at top from free surface models and stresses at the lithosphere for sticky air models should not be expected to be identical, as the numerical approximations and physical assumptions behind these methods differ.
Best,
Ninghui Tian