Zarr-compatible data access from ECMWF/Datastore using Earthkit

[Lizzy-LE]

Hello Earthkit team,

First of all, thank you for developing this very promising Python ecosystem — it’s great to see a modular and modern approach to ECMWF data access and processing.

I have a question regarding data access patterns:

At present, I understand that datasets from ECMWF’s Datastore (CDS, marine, atmosphere, emergency services, etc.) can be retrieved via API as NetCDF files. I am wondering whether Earthkit can already serve as a Zarr access point, or if there are plans to support this.

The motivation is the need for rapid, web-friendly access to subsets of large datasets, as well as integration with GPU-based processing pipelines, so that data can be visualized and analyzed almost on the fly, without downloading full files.

Would Earthkit support or plan to support a workflow where a dataset from the ECMWF Datastore can be accessed lazily as a Zarr array, directly via its Python API? If so, can you provide a minimal working example?

Thank you for your time and any guidance on this.

Best regards,
Lizzy