# QGSW Pytorch PyTorch implementation of multi-layer quasi-geostrophic (QG) and shallow-water (SW) models. ## Requirements ```torchnumpymatplotlib```For the realistic configurations, the following libraries are also required ```scipynetCDF4scikit-image``` Tested with Intel CPUs and NVIDIA GPUs (RTX A3000 Laptop, RTX 2080Ti, V100, A100). ## Structure The source code which is in the `src/` directory has is composed of the following Python scripts - `finite_diff.py`: implements finite difference calculations. - `reconstruction.py`: implements reconstruction, i.e. interpolation for finite volume quantities. - `flux.py`: implements flow calculations for finite volume methods. - `helmholtz.py`: implements a classical Helmholtz equation solver ($\Delta f - \lambda f = r$) using pseudo-spectral methods. - `helmholtz_multigrid.py`: implements a generalized Helmholtz equation solver ($\nabla \cdot \left( c \nabla f \right) - \lambda f = r$) by multigrid method. - `masks.py`: implements mask logic for domains with non-rectangular geometries. - `sw.py`: implements Shallow-water solver - `qg.py`: implements Quasi-geostrophic solver ## Examples Examples are included in `examples/` directory. - `vortexshear.py` : a vortex shear instability to compare QG and SW solutions starting from the same initial condition `python vortexshear.py` - `doublegyre.py` : idealized double-gyre configuration to run with QG or SW `python doublegyre.py` - `natl.py` : simplified north atlantic configurations with realistic costal geometies (but without inflows/outflows across northern and southern boundaries, with flat bottom) `python natl.py`