search

gridapapps / GridapGeosciences.jl

Gridap drivers for geoscience applications
MIT License
35 stars 3 forks source link

Adding darcy cubed sphere #2

Closed amartinhuertas closed 3 years ago

amartinhuertas commented 3 years ago

Hi @santiagobadia, @fverdugo,

in this PR I add the necessary code to perform a systematic convergence study of a Darcy problem on a sphere, i.e., a 2D manifold embedded in 3D space.

To discretize the sphere, the code uses the so-called cubed sphere mesh, i.e., a quasi-uniform quadrilateral mesh of the sphere which is ubiquitous in geophysical flows codes (because it enjoys a number of desirable properties).

In particular, I have implemented two variants of the cubed sphere mesh:

  1. click here Arbitrary degree polynomial approximation of the cubed sphere. I consider this implementation to be clean in the sense that it re-uses the data types in Gridap.Geometry, and perfectly fits the software architecture.
  2. click here Analytical mapping cubed sphere with Jacobian calculated with AD. While functional, I consider this implementation to be odd/dirty. I think this is caused by the current design of Gridap.Geometry somehow assuming a polynomial approximation of the geometry of the domain.

Can you please take a look and let me know your thoughs about this? Thanks!

fverdugo commented 3 years ago

Feel free to merge to this repo!

amartinhuertas commented 3 years ago

Feel free to merge to this repo!

Ok, thanks! I am currently committing the Manifest.toml file.

Once we have the PR https://github.com/gridap/Gridap.jl/pull/638 merged into master and released a new Gridap patch release, I will be able remove the Manifest.toml file.