Dataset for paper Pavel Perezhogin, Laure Zanna, Carlos Fernandez-Granda "Generative data-driven approaches for stochastic subgrid parameterizations in an idealized ocean model" submitted to JAMES.

The dataset consists of the directory tree of .zarr archives. See Github repository for the description of the dataset.

The directory tree is:

├── eddy
│ ├── 48
│ │ ├── gauss
│ │ ├── hires-gauss
│ │ ├── hires-sharp
│ │ ├── lores
│ │ └── sharp
│ ├── 64
│ │ ├── gauss
│ │ ├── hires-gauss
│ │ ├── hires-sharp
│ │ ├── lores
│ │ └── sharp
│ ├── 96
│ │ ├── gauss
│ │ ├── hires-gauss
│ │ ├── hires-sharp
│ │ ├── lores
│ │ └── sharp
│ └── hires
├── jet
│ ├── 48
│ │ ├── gauss
│ │ ├── hires-gauss
│ │ ├── hires-sharp
│ │ ├── lores
│ │ └── sharp
│ ├── 64
│ │ ├── gauss
│ │ ├── hires-gauss
│ │ ├── hires-sharp
│ │ ├── lores
│ │ └── sharp
│ ├── 96
│ │ ├── gauss
│ │ ├── hires-gauss
│ │ ├── hires-sharp
│ │ ├── lores
│ │ └── sharp
│ └── hires

Every individual dataset is a .zarr archive

eddy/jet - configuration of the pyqg; eddy is default; See Ross2022 for description

hires.zarr - high-resolution simulation at 256x256 grid

48/64/96 - resolution of the coarse models

lores.zarr - low-resolution simulation

gauss.zarr, sharp.zarr - training datasets for prediction of subgrid forcing obtained with Gaussian or Sharp filters

hires-gauss.zarr, hires-sharp.zarr - high-resolution simulation projected onto coarse grid with Gaussian or Sharp filters

The directory tree is split into small tar.gz files each representing a separate .zarr archive. Download any required parts of the dataset and unpack with:

tar -xf *.tar.gz 

The directory tree will be restored automatically!