Truly horizontal meshes are not readable

[jploveless]

Crust-mantle interface meshes may need to be truly horizontal if they intersect neatly with a subduction zone mesh, rather than having a slight tilt like jackan_mock_cmi.msh.

These meshes fail to be read because of the logic used in _compute_mesh_edge_elements. I think this can be addressed with some new conditionals, and I'll look into this sometime soon.

[brendanjmeade]

Sounds interesting @jploveless! Anything I can do to be helpful at this time?

[jploveless]

I don't think I'll touch this right now, but it won't be too long before it becomes important, and I'll let you know if I need help when I revisit it!

I don't know why I'm hung up on wanting the CMI to neatly intersect a contour of a subduction mesh, but I am, and a tilting CMI can't do that. This spring, one of my students confirmed that horizontal elements can follow a set slip sign convention, and a current student is trying to functionalize making a CMI that intersects a subduction zone at a constant but arbitrary depth, and I think these are both important steps in using CMIs more readily in celeri models.

[brendanjmeade]

I don't know why I'm hung up on wanting the CMI to neatly intersect a contour of a subduction mesh, but I am

Is it an issue if a CMI mesh deviates from some depth contour by some small amount to enable tilt? Say 10 meters? That should make a negligible difference in terms of elastic deformation.

and a tilting CMI can't do that. This spring, one of my students confirmed that horizontal elements can follow a set slip sign convention, and a current student is trying to functionalize making a CMI that intersects a subduction zone at a constant but arbitrary depth, and I think these are both important steps in using CMIs more readily in celeri models.

This sounds like a more robust solution!

[brendanjmeade]

It's not quite vibe coding, but Claude has a bunch of suggestions, and the MeshLab (and PyMeshLab) ones look very easy.

MeshLab:
Loading Meshes:

1. Open MeshLab
2. File → Import Mesh → Select your first .msh file
3. File → Import Mesh → Select your second .msh file (this adds it to the same project)

Performing Boolean Operations:

1. Go to Filters → Mesh Layer → Boolean Union/Intersection/Difference

2. Choose your operation:

   • Boolean Difference: Clips mesh A with mesh B (A - B)
   • Boolean Intersection: Keeps only overlapping parts
   • Boolean Union: Combines both meshes

3. Select which mesh is the "First Mesh" and which is the "Second Mesh"

4. Click Apply

PyMeshLab:

import pymeshlab

# Create a new MeshSet
ms = pymeshlab.MeshSet()

# Load both GMSH meshes
ms.load_new_mesh('mesh1.msh')
ms.load_new_mesh('mesh2.msh')

# Perform Boolean difference (mesh1 clipped by mesh2)
ms.apply_filter('generate_boolean_difference', 
                first_mesh=0,    # Index of first mesh
                second_mesh=1)   # Index of second mesh

# Save the result
ms.save_current_mesh('clipped_result.ply')

# Optional: Clean up the result
ms.apply_filter('meshing_remove_duplicate_vertices')
ms.apply_filter('meshing_remove_unreferenced_vertices')
ms.apply_filter('meshing_merge_close_vertices', threshold=1e-6)

CGAL is supposed to be very good for this, but it can be challenging to install.

[jploveless]

Is it an issue if a CMI mesh deviates from some depth contour by some small amount to enable tilt? Say 10 meters? That should make a negligible difference in terms of elastic deformation.

The deviation will depend on the applied CMI dip as well as the curvature of the subduction mesh contour. For a dip of 0.1º, the depth deviation would be about 0.17 km per degree of contour position range. For example, the longitudinal range of the Japan Trench is about +6º from south to north, and so a CMI that dips east or west would differ in the depth at which it intersects the subduction zone by about 1 km, so not huge (and reducing the CMI dip would make it smaller), but it's something, and perhaps more complicated is how to deal with overlaps/gaps where the two meshes imperfectly meet.

I love the MeshLab and PyMeshLab suggestions! I'll have my student compare these against the function she's put together.

[brendanjmeade]

The deviation will depend on the applied CMI dip as well as the curvature of the subduction mesh contour. For a dip of 0.1º, the depth deviation would be about 0.17 km per degree of contour position range. For example, the longitudinal range of the Japan Trench is about +6º from south to north, and so a CMI that dips east or west would differ in the depth at which it intersects the subduction zone by about 1 km, so not huge (and reducing the CMI dip would make it smaller), but it's something, and perhaps more complicated is how to deal with overlaps/gaps where the two meshes imperfectly meet.

Let's discuss this when we next chat. I certainly don't understand the details and would like to.

I love the MeshLab and PyMeshLab suggestions! I'll have my student compare these against the function she's put together.

Let me know how it goes. I'm hoping there's a good solution with well-known tools!

[jploveless]

Sounds good! I'll keep you posted!

[brendanjmeade]

@jploveless Do we want to keep this open? For "flat" meshes, I've just been adding a slight (1 meter over ~1000 km) dip and that works fine.

[jploveless]

I think it's fine to close this. There are still some complexities in applying a uniform tilt to meshes with nonlinear edges, but I can always reopen this as I dig more into meshes that abut subduction zones if I can't get the tilting to work.

[brendanjmeade]

@jploveless Let's leave this open if there are outstanding issues. Having it open is a good reminder!