Guides
Remote data
Visualize remote Zarr and Icechunk stores.
Browzarr fetches chunks over HTTP(S) directly from the store's host, exactly like a zarr.open(...) from Python would — only in your browser.
Opening a remote store
- Open the dataset modal (
databaseicon) - Pick an entry from the curated catalog, or switch to the remote tab and paste a URL
- Select a variable and plot
Supported remote backends:
- Zarr — any HTTP/S3-hosted store (
.zarr/zarr.json layout) - Icechunk — Icechunk repositories over S3/HTTP
CORS
Because the browser (not a server) performs the requests, the remote host must send permissive CORS headers:
Access-Control-Allow-Origin: *
# or restrict to the Browzarr origin
Access-Control-Allow-Origin: https://browzarr.io
Access-Control-Allow-Methods: GET, HEAD, OPTIONS
Access-Control-Expose-Headers: Content-Length, Content-Range
If a store loads in Python but not in Browzarr, missing
Access-Control-Allow-Origin on the bucket is almost always the reason. Configure the bucket (S3, GCS, or your static host) to allow the Browzarr origin.Known-good public stores
The curated catalog entries are verified to work (they already ship correct CORS settings) and double as configuration examples, e.g.:
- SeasFire Cube —
https://s3.bgc-jena.mpg.de:9000/misc/seasfire_rechunked.zarr - NASA MUR SST —
https://mur-sst.s3.us-west-2.amazonaws.com/zarr-v1/ - CMIP6 stores on Google Cloud Storage
Tips
- Prefer stores with consolidated metadata (
.zmetadata/zarr.json) — the initial parse is much faster - Chunk sizes of 1–10 MB per chunk fetch work best
- Use the variable sliders to limit the first load of very large cubes