SEEDS: GEFS Reforecast and ERA5 Cubed-Sphere Zarr Stores for Diffusion Weather-Ensemble Emulation

Provider
Google Research (Li, Carver, Lopez-Gomez, Sha, Anderson)
License
CC-BY-4.0 (source)
Format
Zarr (in sampled files)
Stored at
Google Cloud Storage, gs://gresearch/seeds (us)
Access
Free and public. No account with the data provider is needed.
DOI
10.1126/sciadv.adk4489
Category
Climate & Weather
Source
https://doi.org/10.1126/sciadv.adk4489

About this dataset

Inputs, evaluation labels and trained models for SEEDS, the diffusion-model emulator of weather forecast ensembles of Li et al. (Science Advances, 2024), which generates thousands of GEFS-like members at a fraction of the cost of running the physics-based system. data/gefs_forecast_2022_cubedsphere.zarr holds the 2022 GEFS v12 operational forecasts regridded to a cubed-sphere grid as the model input; measured 2026-11-15 its anomaly array is (time 365, number 31, step 5, field 8, values 13824), that is 31 ensemble members, 5 lead times and 8 fields on 13,824 cells, with latitude and longitude coordinate arrays. data/era5_2022_cubedsphere.zarr is the matching ERA5 reanalysis used as the evaluation label and data/climatology_cubedsphere.zarr the static climatology that forms part of the model input. model_checkpoints/ carries the 22 models reported in the paper as TensorFlow SavedModels. Measured 2026-11-15: 1,892 objects and about 47 GB under data/. Open the stores with xarray.open_zarr and follow the demo colab in the seeds folder for the regridding from operational GEFS. Datasets in the google-research repository are CC BY 4.0; cite Li et al. 2024.

How to access this dataset in Python

import gcsfs
fs = gcsfs.GCSFileSystem(token='anon')
fs.ls('gresearch/seeds/', detail=False)[:10]

Lists the files with anonymous access. Runs on any machine, with no cloud account.

Data structure

File types in sampled files: Zarr

Files

Example file: seeds/data/climatology_cubedsphere.zarr/.zattrs

Analyze this dataset without downloading it

On Scigantic, SEEDS: GEFS Reforecast and ERA5 Cubed-Sphere Zarr Stores for Diffusion Weather-Ensemble Emulation opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.

This page is also available as Markdown and as JSON.

Related Climate & Weather datasets