Dataset layouts

The anemoi-datasets package supports three types of data organisation (called “layouts”), all of which are designed to efficiently store and access large datasets.

• Gridded datasets are typically model fields featuring a (possibly unstructured) spatial grid that is regular in time.

• Tabular datasets are typically observations which are unstructured both in time and space.

• Trajectories datasets store forecast fields indexed by a base date (model-run time) and a forecast step, with a 5-D on-disk array.