How to open ECMWF IFS S2S reforecasts stored as Icechunk in a Jupyter notebook
Open the Icechunk repository, take a read-only session on main, and pass its store to xr.open_zarr with a group name. In a Scigantic notebook the repository root is mounted at /mnt/archive, so icechunk.local_filesystem_storage opens it as a local path and xarray reads only the chunks you select.
The code on this page runs in a Scigantic hosted notebook, where the dataset is mounted read-only at /mnt/archive. The same path is in the SCIGANTIC_MOUNT_PATH environment variable and linked from ~/archive. The public source is s3://planette-ifs-46-day/reforecasts/ (AWS us-east-2, public, no account needed).
Open Planette ECMWF IFS S2S reforecasts on SciganticOpen the sample notebook: One forecast, ten possible futures
Code
Open the repository and one group as an xarray Dataset
import os, icechunk, zarr, xarray as xr
MOUNT = os.environ['SCIGANTIC_MOUNT_PATH'] # the Icechunk repo root; nothing under it is a readable file
storage = icechunk.local_filesystem_storage(MOUNT)
repo = icechunk.Repository.open(storage) # prints a harmless LocalFileSystem concurrent-commit WARN
print(list(repo.list_branches()), list(repo.list_tags())[:10])
session = repo.readonly_session("main")
root = zarr.open_group(session.store, mode="r")
print(root.tree()) # pressure/1p5latx1p5lon, single/1p5latx1p5lon, single/1platx1plon, each with a status subgroup
ds = xr.open_zarr(session.store, group='single/1p5latx1p5lon', consolidated=False, chunks={})
print(ds) # number 10, init_time 3592 (2006-01-03 .. 2025-06-01), lead 46, lat 121, lon 240
The starter cell from this dataset's schema card. It ran without error on 2026-10-15 in the Scigantic notebook image against the mounted data.
What the files are
Planette's cube of ECMWF IFS CY49R1 subseasonal-to-seasonal reforecasts in one Icechunk (Zarr v3) repository: 10 ensemble members, 3,592 initialization times from 2006-01-03 to 2025-06-01, and 46 daily lead times.
There are three grid groups: pressure/1p5latx1p5lon (q, t, u, v, z on 10 levels), single/1p5latx1p5lon (24 surface and land variables) and single/1platx1plon (sea ice thickness, sea surface height and salinity).
How the files are laid out
| Path under the mount | What it holds |
|---|---|
| repo | The repository file with branch references. Icechunk reads it; you do not. |
| chunks/, manifests/, snapshots/, transactions/ | Icechunk internal storage under opaque IDs. Nothing here is a readable data file, and the names say nothing about variable or date. |
| (logical) single/1p5latx1p5lon | The group opened in the code: dimensions number 10, init_time 3,592, lead 46, lat 121, lon 240. |
Size, format and license
| Format | Icechunk repository (Zarr v3), opened with icechunk and xarray |
|---|---|
| Size | 268,797 objects, 7,995 GiB stored; 15.5 TB as float32 |
| License | CC BY 4.0 |
| Provider | Planette.ai (ECMWF IFS CY49R1 reforecasts), on AWS Open Data |
| Public source | s3://planette-ifs-46-day/reforecasts/ (AWS us-east-2, public, no account needed) |
| Notebook image | Weather and climate forecast (icechunk, zarr and xarray preinstalled) |
Things to know before you read it
- Open the repository, never the files under it.
- Pass
consolidated=False: an Icechunk store has no consolidated metadata. - Opening the repository prints a concurrent-commit warning for local file storage. It does not affect reading.
- Select before you load. One pressure-level variable is about 1.9 TB.
Run it
Start a notebook from the Planette ECMWF IFS S2S reforecasts archive page. The data is mounted when the notebook opens, so the code above runs as written and nothing is downloaded first. For a longer worked example, open the sample notebook One forecast, ten possible futures.
Related guides
- How to open NEXRAD Level II radar data in a Jupyter notebook without downloading it
- How to read NEXRAD Level III products in Python in a Jupyter notebook
- How to open OME-Zarr volumes in a Jupyter notebook
- How to open the Genomes to Fields (G2F) maize genotype by environment data in a Jupyter notebook
Last checked 2026-11-21. All dataset guides.