py-SWE

The project is ongoing and is not available yet!

Overview¹

Similar Works

There are plenty of SWE solvers outhere, however, they are either easy to use but overly pedagogical or highly complicated. For instance the spectral based pySWE

• Primarily solves linear Matsuno-Gill like problems
• Hinges on spectral methods hence less strict conservation
• Not parallelizable

Or in the complicated to use case, for instance TRiSK based programs which contains sophisticated mesh-generation algorithms.

Main Feature

py-SWE aims to provide a middle ground with

1. Capability to simulate non-linear SWEs
2. Finite-volume based algorithm to ensure conservation
3. Rectangular domain with simple square mesh
4. Full parallelized implementation in CUDA with interface in pure python
5. Integrated viuslization pipeline

The package is ideal to be deployed on clusters like NCSA Delta or Keeling at UIUC’s department of ATMS as well as on laptops/PC with cuda compatible GPUs. It is suitable for generating training data for machine learning models or be used as the backbone of more complex idealistic models.

Algorithm

Available soon!