software · Global

FourCastNet (NVlabs)

Initial public release of code, data, and model weights for FourCastNet

Claim this listing

Represent FourCastNet (NVlabs)? Send a claim for review so we can discuss listing updates. Claims do not automatically grant ownership or verification.

Explore paid sponsorship →

What's in it

Data offered docs-verified: global 0.25° weather forecasts · surface wind speed forecasts · precipitation forecasts · atmospheric water vapor forecasts · pre-processed ERA5 training data (HDF5) · trained model weights (backbone.ckpt, precip.ckpt) · precomputed normalisation statistics (npy) · precipitation diagnostic model
Coverage llm-inferred: Global · ERA5 files 1979–2018 (train 1979–2015, test 2016–2017, out-of-sample 2018)
Resolution docs-verified: 0.25° spatial (global); temporal forecast step not stated in DOCS
Formats docs-verified: HDF5 · netCDF (ERA5 input) · NPY · CKPT (model weights)
Access docs-verified: Open to all via Globus and web download links; the training data needs a Globus account and login. Licence not stated in DOCS.

Useful for

Use cases llm-inferred: Precipitation and storm-context screening for drainage or flood risk studies using global ML forecasts as a research benchmark · Benchmarking ML forecast skill (RMSE/ACC) against ERA5 for climate-data QA on a custom interval · Short-range (up to about a week) weather scenario context for construction or site-works planning, with validation before use
Audience llm-inferred: flood consultants · hydrologists · drainage engineers · GIS analysts · meteorologists and climate scientists

Limitations

Constraints llm-inferred: Research code and weights; initial release v0.0.0 only · Training data is about 5TB and requires a Globus account · Training and pre-processing workflows assume an HPC/Slurm environment and GPU resources · Custom-interval inference needs manual ERA5 download and HDF5 pre-processing · Global 0.25° model; not a local or high-resolution catchment model · Licence not stated in DOCS; check the repository before commercial use

Licence: Other
Authentication: Unknown — review source docs
GitHub stars: 703
Last commit: 02/10/2023
Last successful page check: 14/10/2026
Page health: healthy
Package: Not recorded
Verification: Not independently verified

Auto-extracted, unverified fields: tools, auth. Check upstream documentation.

Provenance & evidence

Verified via GitHub API (2026-10-15). Tools/install/auth auto-extracted from README.

Metadata observed 14/10/2026 · research sweep: manual-2026-10-15 (Chris sources (NVIDIA Earth-2))