Sphedron: Polyhedral meshes on the sphere

A python package for creating refinable polyhedral meshes on the sphere. The refinement is either rectangle or triangle based.

This was developed as a component of Graph Neural Networks (GNN) for Geospatial ML projects, and it's designed with that in mind.

The package implements:

• Icosphere
• Octasphere
• Cubesphere
• Uniform Latitude/Longitude

It is also straightforward to extend it to include other triangular/rectangular meshes.

You can find code examples in the examples/ folder.

To understand the inner workings of a mesh and how it is refined, refer to the Icosphere tutorial.

Requirements

• python >= 3.10
• numpy
• scipy (nearest neighbor query)
• trimesh (radius query)

Optional dependencies:

• cartopy (plots)
• shapely (land mask feature)

Install

pip install .

Using notebooks in ReadTheDocs

This project's Sphinx docs can render Jupyter notebooks via nbsphinx.

• Notebook sources live under docs/source/notebooks/ (copied from examples/ as needed).
• The docs build defaults to not executing notebooks (it renders the saved inputs/outputs).
• To execute notebooks locally during a Sphinx build:
  • SPHEDRON_DOCS_EXECUTE_NOTEBOOKS=auto