Notebooks¶

The notebooks directory contains supplementary examples and research artifacts for CryoCast. They are not required to use the package.

Setup¶

Start with the CryoCast CLI demo_pipeline.ipynb notebook, using the notebooks environment:

uv sync --group notebooks
cd notebooks
uv run jupyter lab

Some of these notebooks will not run immediately as they require you to set up credentials for downloading datasets. Details can be found in the notebooks themselves.

Main notebook set¶

Notebook	Purpose	Prerequisite
`demo_pipeline.ipynb`	CLI walk through. CryoCast walkthrough covering account-free synthetic data, training/evaluation artifacts, model architecture and persistence, Hydra configuration, multimodality, and an optional real-data route.	The default route uses generated synthetic data and local-file logging; there is also an option to use CDS/W&B for real data.
`layer_diagnostics.ipynb`	Model exploration. Activation-capture investigation for the current UNet/`quick_test` model and multimodal real-data path.	Supply a compatible checkpoint and existing real datasets; the notebook uses local-file logging for evaluation.
`ARGO_data.ipynb`	Data exploration. Download, inspect and grid ARGO float observations for the non-gridded data path.	Independent of the demo pipeline; requires network access and its geospatial/data dependencies.
`case_study_whale_corridors.ipynb`	Demo videos. Produce whale-corridor and shipping visualisations.	Requires the case-study data/configuration and is not part of the default CLI walkthrough.

Optional synthetic non-gridded workflow¶

The following pair can be used to create a synthetic non-gridded dataset, using nongriddedenv.yaml:

extract_anomalies.ipynb creates gridded ERA5 pressure anomalies.
degrid_and_visualise.ipynb samples those anomalies into synthetic station and buoy observations.

Maintenance¶

When adding, removing, or renaming a notebook, update this page. The notebooks are supplementary and are not part of the automated test suite.