Community Dataset Catalog
A curated collection of high resolution (sub-hourly, km scale) observational datasets designed for meteorological machine learning workflows. Analysis Ready and Cloud-Optimized (ARCO), and directly accessible via Python thanks to the European Weather Cloud (EWC).
6
Countries
~97
Cumulative Years
7.4M+
Time Steps
5 min
Best Cadence
Option A: Python Package
Install our stable mlcast-datasets package from PyPI for consistent versioning.
import mlcast_datasets
cat = mlcast_datasets.open_catalog()
Option B: Direct Catalog Access
Use the most recent version of the catalog directly from our source repository.
import intake
cat = intake.open_catalog("https://raw.githubusercontent.com/mlcast-community/mlcast-datasets/main/src/mlcast_datasets/catalog/catalog.yml")
Available precipitation datasets
Every dataset in the catalog is versioned, documented, and ready to load. Each entry includes the source description, driver, and direct access path.
cloud_upload ARCO Standard
We adopt the ARCO (Analysis-Ready, Cloud-Optimized) pattern. Our data is stored in Zarr format, enabling partial reads and streaming directly from the cloud without massive local storage.
storage EWC Infrastructure
Hosted on the European Weather Cloud, our datasets benefit from high-throughput connectivity to meteorological supercomputers, minimizing egress latency.
code Code-Addressable
Data is not a static file; it's an object in your Python pipeline. Using intake, every dataset is queryable, version-controlled, and ready for training.