Transform NetCDF to Zarr¶
Convert a NetCDF dataset into a Zarr store in your EDITO S3 storage, or — as an override — on the EPT service filesystem.
POST /v1/conversions/netcdf/zarr (Bearer auth required).
Standard usage: remote NetCDF to your S3 storage¶
The easiest call names only a source and a target folder. The Zarr store is
named after the source file (without its .nc extension) and written into
target inside your personal EDITO bucket oidc-<username> (derived from
your bearer token):
curl --fail-with-body \
-X POST "${EPT_API_URL}/v1/conversions/netcdf/zarr" \
-H "Authorization: Bearer ${ACCESS_TOKEN}" \
-H "Content-Type: application/json" \
-d '{
"source_uri": "https://data.example.org/datasets/observations-2025.nc",
"target": "datasets"
}'
This writes the store to s3://oidc-<your-username>/datasets/observations-2025.zarr.
Supply the S3 connection fields returned by POST /v1/auth as flat request
fields (endpoint_url, access_key_id, secret_access_key, optional
session_token). When they are omitted, the EPT service falls back to its own
S3_* environment configuration.
Overrides¶
bucket— write into a different bucket you can access (e.g. a project bucketproject-<projectid>; browse your buckets at https://datalab.dive.edito.eu/file-explorer). Defaults to your personaloidc-<username>bucket.output_name— store name override; defaults to the source name without its.nc/.nc4/.cdfextension, with a.zarrsuffix (appended when missing).local_output— directory on the EPT service filesystem to write the Zarr store into instead of S3 (the store name is appended the same way). When set,bucketandtargetmust be null, and no S3 connection fields are needed for a purely local conversion.source_uri— local path on the EPT service filesystem,s3://URI, or HTTP(S) URL including OPeNDAP/DODS, GRIDDAP, and THREDDS endpoints. Remote HTTP(S) and S3 files are downloaded once into a local cache before opening (setEPT_NC_CACHEto change the cache directory).engine— xarray backend engine override.chunks— xarray chunk configuration (int, per-dimension dict, or"auto"); requires a chunk manager such as Dask.decode_cf,mask_and_scale— CF decoding overrides passed toxarray.open_dataset.consolidated(defaulttrue) — write consolidated Zarr metadata.overwrite(defaultfalse) — replace an existing target store. For S3 targets this deletes all existing objects under the target prefix first.zarr_format—2or3; defaults to v2 for S3 output.
Response¶
{
"source_uri": "https://erddap.example.org/erddap/griddap/observations-2025",
"bucket": "oidc-myusername",
"target_uri": "s3://oidc-myusername/datasets/observations-2025.zarr",
"status": "completed"
}
bucket is null for local output. target_uri is a full https:// URL
when endpoint_url is provided, an s3:// URI otherwise. The operation
completes synchronously; when overwrite is false an existing target
store is rejected.